{"id": "2606.17165", "title": "Statistical Foundations of LLM-based A/B Testing: A Surrogacy Framework for Human Causal Inference", "abstract": "Organizations and researchers show increasing interest in using large language models (LLMs) in place of human participants in A/B tests, in the hope of experimenting faster and at lower cost. We study when a treatment effect estimated on LLM outcomes can recover the effect that would have been measured on the human population of interest. Distributional equivalence between LLM and human outcomes would make any standard estimator valid but is unrealistic. We therefore develop a statistical framework that adapts surrogate endpoint theory to LLMs, showing that calibrating LLM outcomes to human outcomes identifies the average treatment effect under surrogacy and comparability conditions that are jointly weaker than distributional equivalence. We present a falsification test for surrogacy and a bound on the worst-case bias from limited overlap between the LLM and human samples. We further show that the stochasticity inherent to LLMs can weaken surrogacy for identification while also introducing bias and variance during estimation, but that using an average over multiple LLM draws per unit as the surrogate mitigates these issues. Simulations validate the results, and an empirical application to A/B tests on Upworthy headlines shows that raw LLM predictions recover only 39\\% of the human treatment effect while nonparametric calibration closes the gap. A central takeaway is that A/B testing on LLMs yields correct results only by assumption, whereas A/B testing on humans is correct by design, and that the required assumptions are hardest to justify precisely where A/B testing on LLMs promises the greatest benefit. We discuss the role of LLM choice, prompting, and temperature as design variables, the compounded challenge posed by long-term outcomes, and how to size human pilot studies for validation.", "date": "2026-06-15", "categories": []}
{"id": "2606.19361", "title": "Computational Identifiability", "abstract": "Identification conditions describe the computability of a target query or parameter of interest as a function of the type and amount of information available. In causal identification, this information is often expressed in the form of a causal graph, and data are observed or collected for some subset of variables in the graph. Target queries may be for a single effect alone or for a class of effects in a given model. The derivation of an identification algorithm then defines mathematically the process by which the desired causal effect(s) can be uniquely determined, theoretically, in expectation. Identifiability in expectation, or 'theoretical identifiability,' generally assumes asymptotic properties, infinite data, or other mathematically idealized conditions. In this paper, we explore a fundamental distinction between this theoretical, idealized notion of identifiability and a proposed alternative that is computation-bound. The framework we propose - 'computational identifiability' - is to instead define a finite computational search procedure for an empirical estimator. If this process finds an estimator empirically, within a desired error tolerance, then identifiability is satisfied, conditional on the specified assumptions of the search (i.e., a prior distribution over the parameters) and conditional on the search procedure itself. Through several experiments, we demonstrate how this framework allows us to answer fine-grained, practical identification questions, such as identification with small finite samples, with ambiguous graphical criteria, with mixed observational-interventional data, and across counterfactual data and estimands. Code is available at https://github.com/lbynum/metadentify.", "date": "2026-06-08", "categories": []}
{"id": "2606.19594", "title": "Unsupervised Causal Abstractions Discovery", "abstract": "Causal abstractions formalize when a high-level structural causal model (SCM) captures the interventional behavior of a lower-level SCM. Existing applications of this notion largely follow a hypothesis-testing paradigm: an expert proposes a candidate high-level model and then evaluates if the low-level system implements it. We study the complementary problem of learning a high-level model directly from low-level measurements. Our contributions leverage hypotheses from low-rank causal discovery, and can be summarized as follows: (1) we show that observations generated by a low-rank graph induce latents that form a causal abstraction, (2) we provide identifiability results about these latents, and (3) we propose a practical objective to learn this high-level SCM.", "date": "2026-06-17", "categories": []}
{"id": "2606.19610", "title": "Latent Confounded Causal Discovery via Lie Bracket Geometry", "abstract": "Recent work on Kan-Do-Calculus (KDC) has established that the boundary between passive observation and active intervention in causal inference is a category-theoretic bi-adjunction, with interventions modeled by left Kan extensions and conditioning by right Kan extensions. This paper introduces two causal discovery algorithms under latent confounding, building on the information-geometric and categorical consequences of KDC. In smooth statistical settings, Radon-Nikodym derivatives between observational and interventional measures induce local causal vector fields; failures of these fields to close under Lie brackets become computable Frobenius residuals, which we interpret as witnesses of failed visible integrability and possible latent or unmodeled structure. Our first algorithm, BRIDGE (Bracket Residuals for Interventional Discovery and Geometric Estimation), combines an interventional density or Radon-Nikodym-ratio engine with a geometric screen that proposes a high-recall family of admissible arrows, identifies non-closing visible pairs as latent-obstruction candidates, and passes the reduced family to downstream score-based or differentiable discovery routines. The second algorithmic contribution, Spectral Kan-Do Flow Matching (SKFM), learns amortized intervention fields and factors latent curvature spectrally, exposing the direct Lie-space endpoint toward which BRIDGE points. A detailed set of experiments show that both algorithms are capable of discovering causal models with latent confounders while collapsing the super-exponential space of possible DAGs by many orders of magnitude. This paper introduces a new paradigm in causal discovery, where latent structure is inferred directly from the geometry of intervention-induced flows.", "date": "2026-06-17", "categories": []}
{"id": "2505.15215", "title": "Clustering and Pruning in Causal Data Fusion", "abstract": "Data fusion, the process of combining observational and experimental data, can enable the identification of causal effects that would otherwise remain non-identifiable. Although identification algorithms have been developed for specific scenarios, do-calculus remains the only general-purpose tool for causal data fusion, particularly when variables are present in some data sources but not others. However, approaches based on do-calculus may encounter computational challenges as the number of variables increases and the causal graph grows in complexity. Consequently, there exists a need to reduce the size of such models while preserving the essential features. For this purpose, we propose pruning (removing unnecessary variables) and clustering (combining variables) as preprocessing operations for causal data fusion. We generalize earlier results on a single data source and derive conditions for applying pruning and clustering in the case of multiple data sources. We give sufficient conditions for inferring the identifiability or non-identifiability of a causal effect in a larger graph based on a smaller graph and show how to obtain the corresponding identifying functional for identifiable causal effects. Examples from epidemiology and social science demonstrate the use of the results.", "date": "2025-05-21", "categories": []}
{"id": "2606.18281", "title": "A Guide to Estimating Conditional Average Treatment Effects in Competing Risks Settings", "abstract": "Conditional average treatment effects (CATEs) are central to treatment decision-making in personalized medicine. In competing risks settings, estimating CATEs from survival data allows for patient-specific assessments of treatment effectiveness for a specific event of interest while properly accounting for alternative event types. This distinction is essential in the presence of comorbidities, where competing causes of death may otherwise confound the therapeutic benefit. Focusing on right-censored survival times with binary treatment, we examine CATEs defined as covariate-conditional differences in the absolute risk for the event of interest at a fixed time. To this end, we study meta-learners which adapt machine learning algorithms for CATE estimation in competing risks scenarios. We systematically compare six meta-learners, combining Cox regression or random survival forests for risk modeling with elastic net regression or random forests for direct CATE modeling. To provide practical guidance on model selection, we evaluate their performance in multiple simulation settings, that differ in hazard complexity, treatment heterogeneity, treatment assignment, event type distribution and censoring. To facilitate applied use, we provide the R package, crsurvlearners, which implements all considered approaches.", "date": "2026-06-08", "categories": []}
{"id": "2606.18506", "title": "Beyond AHI: An Interpretable Causal-Discovery-Guided Framework for Sleep Recovery in Connected Health", "abstract": "Objective sleep assessment relies on polysomnography (PSG), yet clinical impact is often better reflected in patient-reported outcomes (PROs) such as sleepiness and fatigue. Existing summary indices, including the Apnea-Hypopnea Index (AHI), provide limited insight into the multidomain physiology underlying functional recovery. We propose an interpretable, causal-discovery--guided framework for deriving a hierarchical Sleep Recovery Score (SRS) from multimodal PSG. Using two large population cohorts (MESA: n=1540; MrOS: n=825), we apply directed acyclic graph (DAG) learning to identify candidate physiological drivers spanning respiratory burden, hypoxic burden, sleep fragmentation, sleep architecture, and autonomic regulation. Although derived from clinical PSG, these domains map naturally to sensing streams increasingly available in connected health technologies, including wearable ECG, oximetry, and sleep-stage estimation devices. To preserve mechanistic plausibility, we introduce a two-stage screening process that combines physiology-based constraints with constrained LLM-assisted auditing to identify and remove structural confounders and construct-overlapping variables. Across cohorts, these five domains emerge as recurrent physiological domains associated with recovery, and the resulting SRS shows up to 2.5$\\times$ stronger alignment with perceived recovery than AHI. By linking multimodal sleep physiology to patient-centered outcomes through an interpretable, bias-aware, and domain structured framework, this work provides a practical foundation for recovery modeling across both clinical sleep studies and emerging smart and connected health settings.", "date": "2026-06-16", "categories": []}
{"id": "2606.18535", "title": "Shrinkage priors for Bayesian Substitute Confounders", "abstract": "Multi-cause observational studies contain information about unmeasured confounding through the dependence structure among causes. However, literal imputation of the unobserved confounder is often more complex than learning a lower-dimensional substitute score that preserves the shared assignment variation needed for stable causal adjustment. The deconfounder (Wang and Blei, 2019) and related substitute confounder methods exploit this idea, but flexible assignment models can fit the joint distribution of the causes while producing scores that over-encode the treatment vector, collapse overlap, or capture single-cause variation. We develop a Bayesian factor assignment framework for learning sparse substitute confounders that retain coarse multi-cause dependence with shrinkage priors. The theory is stated at the level of posterior concentration, factor score contraction, and overlap-preserving assignment geometry and therefore does not rely on a particular shrinkage prior. Under these conditions, the proposed regression-adjusted estimators are consistent for mean potential outcomes when the corresponding latent variable identification assumptions hold. Shrinkage priors provide a natural tool for latent structural learning: they favour low-dimensional factors supported by multiple causes, discourage effectively single-cause factors, and induce an ordering of the latent factors through progressive shrinkage. Synthetic experiments illustrate the roles of signal strength, outcome validity, and geometry-aware regularization. In an Alzheimer's Disease Neuroimaging Initiative (ADNI) baseline analysis, sparse substitute scores recover much of the adjustment obtained by directly conditioning on invasive cerebrospinal-fluid biomarkers, while collapse diagnostics identify when fitted factors reduce to individual observed measurements.", "date": "2026-06-16", "categories": []}
{"id": "2606.18834", "title": "Identifying Structural Biases from Causal Mechanism Shifts", "abstract": "Causal discovery methods commonly assume that all data is independently and identically distributed (i.i.d.) and that there are no unmeasured variables affecting the system. In practice, these assumptions are often violated, leading to inaccurate inference. In this paper, we study how to identify hidden confounding and selection biases from causal mechanism shifts. In particular, we show that structural biases lead to dependent mechanism shifts. That is, by considering for which variables the mechanisms change given data from different environments, we can tell which variables are unbiased, which are subject to hidden confounding, and which are undergoing selection bias. We formalize this into an empirically testable criterion based on mutual information, and show under which conditions it identifies structural biases. To tell which nodes are subject to what kind of bias, we introduce the StruBI algorithm. Experiments on synthetic and real-world data show that StruBI works well in practice, accurately recovering affected variable sets and types of biases, outperforming the state-of-the-art by a wide margin.", "date": "2026-06-17", "categories": []}
{"id": "2606.19117", "title": "Wasserstein Policy Learning for Distributional Outcomes", "abstract": "Offline policy learning has received growing attention in causal inference. The primary objective is to learn a policy (individualized treatment rule) as a mapping from covariates to treatment that maximizes the empirical welfare defined as the mean of scalar-valued potential outcomes. In this paper, we study offline policy learning with distribution-valued outcomes, where each potential outcome is a probability measure on $\\mathbb{R}$ and the reward is defined through a utility functional applied to the Wasserstein barycenter of induced outcome distributions. We establish statistical guarantees for the policy learning framework based on both Inverse Probability Weighting (IPW) and Doubly Robust (DR) estimators. By handling the challenging uniform deviation over the product of the combinatorial policy class and the infinite-dimensional quantile domain, we prove that the finite-sample regret has leading dependence $\\widetilde{\\mathcal{O}}(\\sqrt{\\mathrm{N\\text{-}dim}(\\Pi)/N})$. In the one-dimensional Wasserstein setting and under the stated regularity conditions, the leading regret rate is still governed by the policy-class complexity. Moreover, we provide a minimax lower bound establishing the sharpness of the leading dependence on $N$ and $\\mathrm{N\\text{-}dim}(\\Pi)$.", "date": "2026-06-17", "categories": []}
{"id": "2507.11178", "title": "A Gradient-based Causal Discovery Framework with Applications to Complex Industrial Processes", "abstract": "With the advancement of deep learning technologies, various neural network-based Granger causality models have been proposed. Although these models have demonstrated notable improvements, several limitations remain. Most existing approaches adopt the component-wise architecture, necessitating the construction of a separate model for each time series, which results in substantial computational costs. In addition, imposing the sparsity-inducing penalty on the first-layer weights of the neural network to extract causal relationships weakens the model's ability to capture complex interactions. To address these limitations, we propose Gradient Regularization-based Neural Granger Causality (GRNGC), which requires only one time series prediction model and applies $L_{1}$ regularization to the gradient between model's input and output to infer Granger causality. Moreover, GRNGC is not tied to a specific time series forecasting model and can be implemented with diverse architectures such as KAN, MLP, and LSTM, offering enhanced flexibility. Numerical simulations on DREAM, Lorenz-96, fMRI BOLD, and CausalTime show that GRNGC outperforms existing baselines and significantly reduces computational overhead. Meanwhile, experiments on real-world DNA, Yeast, HeLa, and bladder urothelial carcinoma datasets further validate the model's effectiveness in reconstructing gene regulatory networks.", "date": "2025-07-15", "categories": []}
{"id": "2508.04492", "title": "Learning Robust Intervention Representations with Delta Embeddings", "abstract": "Causal representation learning has attracted significant research interest during the past few years, as a means for improving model generalization and robustness. Causal representations of interventional image pairs (also called ``actionable counterfactuals'' in the literature), have the property that only variables corresponding to scene elements affected by the intervention / action are changed between the start state and the end state. While most work in this area has focused on identifying and representing the variables of the scene under a causal model, fewer efforts have focused on representations of the interventions themselves. In this work, we show that an effective strategy for improving out of distribution (OOD) robustness is to focus on the representation of actionable counterfactuals in the latent space. Specifically, we propose that an intervention can be represented by a Causal Delta Embedding that is invariant to the visual scene and sparse in terms of the causal variables it affects. Leveraging this insight, we propose a method for learning causal representations from image pairs, without any additional supervision. Experiments in the Causal Triplet challenge demonstrate that Causal Delta Embeddings are highly effective in OOD settings, significantly exceeding baseline performance in both synthetic and real-world benchmarks.", "date": "2025-08-06", "categories": []}
{"id": "2602.08939", "title": "CausalT5k: Diagnosing Refusal and Failure Modes in Trustworthy Causal Reasoning Across Causal Rungs", "abstract": "Large language models increasingly produce fluent causal explanations, yet they often fail in ways aggregate accuracy cannot diagnose: confusing association with intervention, abandoning correct judgments under pressure, over-refusing valid claims, or answering when evidence is underdetermined. We introduce CTK, a diagnostic benchmark of 5,147 cases and growing, across 10 domains and all three levels of Pearl's Ladder of Causation. Unlike benchmarks that only score correctness, CTK reveals why a model failed by annotating causal rung, trap type, pressure sensitivity, refusal quality, and Utility-Safety tradeoffs. Its Sheep/Wolf taxonomy separates valid causal designs from inferential traps; paired neutral/pressure variants measure sycophantic drift through Bad Flip Rate; and Wise Refusal fields test whether a model identifies the missing information needed before endorsing a claim. CTK exposes failure modes hidden by aggregate accuracy: the Skepticism Trap, Rung Collapse under scaling, pressure-induced drift, Detection-Correction gaps, and counterfactual error modes. Rather than prescribing a correction method, it provides the diagnostic substrate for studying causal-reasoning failure profiles.", "date": "2026-02-09", "categories": []}
{"id": "2603.02159", "title": "Instrumental and Proximal Causal Inference with Gaussian Processes", "abstract": "Instrumental variable (IV) and proximal causal learning (Proxy) methods are central frameworks for causal inference in the presence of unobserved confounding. Despite substantial methodological advances, existing approaches rarely provide reliable epistemic uncertainty (EU) quantification. We address this gap through a Deconditional Gaussian Process (DGP) framework for uncertainty-aware causal learning. Our formulation recovers popular kernel estimators as the posterior mean, ensuring predictive precision, while the posterior variance yields principled and well-calibrated EU. Moreover, the probabilistic structure enables systematic model selection via marginal log-likelihood optimization. Empirical results demonstrate strong predictive performance alongside informative EU quantification, evaluated via empirical coverage frequencies and decision-aware accuracy rejection curves. Together, our approach provides a unified, practical solution for causal inference under unobserved confounding with reliable uncertainty.", "date": "2026-03-02", "categories": []}
{"id": "2603.20775", "title": "Evaluating Uplift Modeling under Structural Biases: Insights into Metric Stability and Model Robustness", "abstract": "In personalized marketing, uplift models estimate the incremental effect of an intervention by modeling how customer behavior would change under alternative treatments using counterfactual analysis. However, real-world marketing data often exhibit various biases, such as selection bias, spillover effects, measurement error, and unobserved confounding. These biases can adversely affect both the accuracy of uplift estimation and the validity of evaluation metrics. Despite the importance of bias-aware assessment, there remains a lack of systematic studies evaluating how different models and metrics perform under such biased conditions. To bridge this gap, we design a systematic benchmarking framework. Unlike standard predictive tasks, real-world uplift datasets inherently lack counterfactual ground truth. This limitation renders the direct validation of evaluation metrics infeasible and prevents the precise quantification of biases. Therefore, a semi-synthetic approach serves as a critical enabler for systematic benchmarking. This approach effectively bridges the gap by retaining real-world feature dependencies while providing the ground truth needed to isolate structural biases. Our investigations reveal that (i) uplift targeting and prediction can manifest as distinct objectives, where proficiency in one does not ensure efficacy in the other; (ii) while many models exhibit inconsistent performance under diverse biases, TARNet shows notable robustness, providing insights for subsequent model design; (iii) the stability of evaluation metrics is linked to their mathematical alignment with the ATE, suggesting that ATE-approximating metrics yield more consistent model rankings under structural data imperfections. These findings suggest the need for more robust uplift models and evaluation metrics under real-world data imperfections.", "date": "2026-03-21", "categories": []}
{"id": "2606.10703", "title": "From Observation to Intervention: A Causal Audit of Expert Importance in Mixture-of-Experts Models", "abstract": "Interpretability methods routinely use population-level summary statistics over observed model behaviour to license claims about the effects of targeted interventions on specific computations; in Pearl's terms, they treat rung-1 associational evidence as if it supported rung-2 interventional conclusions, a move whose validity is rarely tested. We examine one concrete instance: the use of routing statistics in Mixture-of-Experts (MoE) pruning, where utilization rates, activation norms, and routing weight distributions are treated as predictors of which experts can be removed without functional cost. A token-level interventional audit across three high-redundancy MoE architectures (OLMoE-1B-7B-0924, Qwen1.5-MoE-A2.7B, DeepSeek-V2-Lite) finds no observational metric predicts causal expert importance in any model: across all 60 metric-layer combinations effect sizes stay below Cohen's $d = 0.23$, and no metric is reliably positive under our corrected, dual-test criterion. A per-token routing weight control, run with identical $n$, rules out insufficient power, recovering a signal whose CI excludes zero at OLMoE's final MoE layer ($d = +0.231$, 95\\% CI $[+0.09, +0.37]$, $p = 0.0013$). Existing pruning methods succeed in this regime not by identifying dispensable experts but because early-layer redundancy renders most selection criteria interchangeable. Our results provide an explicit counterexample to the common inferential step from population-level observational summaries to token-level interventional claims about expert importance, and illustrate how interventional audits can calibrate the evidential standards for interpretability claims.", "date": "2026-06-09", "categories": []}
{"id": "2606.12623", "title": "Estimating Individualized Treatment Effects in Acute Ischemic Stroke with Causal Transformation Models (TRAM-DAG): A Multi-Centre Observational Study with External RCT Validation", "abstract": "Personalized medicine in acute ischemic stroke requires moving beyond average treatment effects (ATE) to individualized treatment effect (ITE) estimates to support treatment decisions. In acute ischemic stroke, mechanical thrombectomy has been shown to be more effective on average than lysis in randomized controlled trials (RCTs), such as the MR CLEAN study. We aim to identify which individual patients benefit most from mechanical thrombectomy compared to lysis. The outcome of interest is the modified Rankin Scale (mRS) at three months, an ordinal measure of functional disability (0: no symptoms, 6: death). We demonstrate that causal transformation models on directed acyclic graphs (TRAM-DAG) can be used for ITE estimation after being fitted on observational MAGIC multi-center stroke patient data. To ensure comparability with the MR CLEAN population, which we use for validation, we train the TRAM-DAG on a MAGIC sub-population with NIHSS at admission >= 6, corresponding to one inclusion criterion of MR CLEAN. The fitted model is then used to estimate ITEs for stroke patients in the MR CLEAN population. While these ITE estimates cannot be confirmed experimentally, we show that their average is consistent with the trial's reported ATE. Furthermore, the ITE estimates correctly rank trial patients by their observed frequency of a good outcome (mRS at three months <= 2). These findings support the use of causal models like TRAM-DAG for personalized decision-making in stroke care and highlight their ability to bridge the gap between observational evidence and clinical trials.", "date": "2026-06-10", "categories": []}
{"id": "2606.14438", "title": "CADET: Physics-Grounded Causal Auditing and Training-Free Deconfounding of End-to-End Driving Planners", "abstract": "End-to-end (E2E) autonomous-driving planners trained by imitation are prone to statistical shortcuts: they associate scene elements that merely co-occur with expert actions (a roadside object, a building facade) with driving decisions, rather than the variables that causally determine them. Such causal confusion silently compromises reliability in long-tail scenarios, and it is difficult to detect, because prevailing open-loop metrics (L2 displacement and collision rate) are dominated by ego status and do not indicate whether a planner depends on spurious cues. Existing remedies based on causal-intervention training require retraining large models and cannot audit a planner that is already deployed. We present CADET, a training-free framework that audits, benchmarks, and repairs spurious reliance in pretrained E2E planners without any parameter update.", "date": "2026-06-12", "categories": []}
{"id": "2606.17516", "title": "FoundCause: Causal Discovery with Latent Confounders from Observational Data", "abstract": "Causal discovery from observational data remains challenging due to the need to recover directed structure and latent confounding without interventions. We propose FoundCause, an amortized causal discovery model trained entirely on synthetic data that maps datasets directly to causal graphs in a single forward pass. By learning from large collections of simulated structural causal models, FoundCause captures transferable statistical patterns that generalize beyond individual datasets. The architecture incorporates several key inductive biases for causal discovery. It uses a permutation-invariant transformer encoder with alternating attention over samples and variables to jointly model cross-variable dependence and per-variable distributions. Pairwise statistical features derived from classical asymmetry measures are injected through statistics-conditioned attention, guiding the model toward known causal signals. A factorized decoder separates edge existence from direction, while a triangular refinement module enables reasoning over higher-order causal motifs such as chains and colliders. In addition, a dedicated confounder module based on learnable latent tokens explicitly models hidden common causes, and the model explicitly handles missing data via its masked input representation. To our knowledge, FoundCause is the first amortized causal discovery approach to explicitly model latent confounding. FoundCause outperforms 11 classical non-amortized methods (e.g., PC, GES, NOTEARS-style optimization) and 4 amortized causal discovery methods on 15 real-world datasets, achieving +9.6% improvement in $F_1$, +1.2% in AUROC, and an 18.9% reduction in structural Hamming distance relative to the strongest non-amortized methods, while performing inference in a single forward pass.", "date": "2026-06-16", "categories": []}
{"id": "2606.18011", "title": "Fast Nonparametric Conditional Independence Testing via Two-Stage Regression", "abstract": "Constraint-based causal discovery relies on repeated conditional independence tests, but fast nonparametric tests often sacrifice calibration, especially when variables depend on the conditioning set through nonlinear relationships. We introduce BLITZ (Broad-to-Local Independence Testing via residualiZation), a nonparametric conditional independence test designed to run well under a second while maintaining the accuracy needed for the thousands of queries performed by constraint-based causal discovery algorithms. BLITZ first removes broad smooth dependence on the conditioning set using low-order polynomial regression, then applies a small nonlinear feature map and residualizes those features with shallow tree regressions. The resulting statistic tests residual cross-covariance, with a moment-matched chi-square approximation to the null distribution. We show theoretically that the two-stage design reduces the effective complexity faced by the tree residualizers, allowing shallow trees to control residual conditional-mean bias while avoiding excessive overfitting. In simulations, BLITZ provides better null calibration than fast kernel, random-feature, and regression-based competitors while remaining among the fastest methods tested. In causal discovery experiments on synthetic graphs and flow-cytometry data, BLITZ yields more reliable endpoint orientations among retained adjacencies and competitive structural recovery. These results suggest that broad-to-local residualization is a practical route to calibrated, scalable nonparametric conditional independence testing for causal discovery.", "date": "2026-06-16", "categories": []}
{"id": "2606.18074", "title": "Tensor-based second-order causal discovery", "abstract": "Causal discovery seeks to uncover the causal dependencies among variables. For this purpose, we propose an algorithm called Tensor-based Second-order Causal Discovery (TSCD). Its input is a tensor obtained from the covariance matrices of observational and interventional data. Assuming the causal dependencies follow a linear structural equation model on a directed acyclic graph (DAG), TSCD outputs the DAG and the functions on its edges, requiring only that the noise variables are uncorrelated. We also implement a version of the approach for nonlinear models. Our focus on second-order statistics (via the covariance matrices) is motivated by their statistical and computational efficiency relative to higher-order moments, their identifiability relative to first-order statistics, and that they work regardless of whether the variables are Gaussian. We show that TSCD has identifiable causal order and parameters from a number of interventions that is logarithmic in the number of variables. Experiments show that TSCD is robust to noise, competitive with existing methods, and scales to hundreds of variables.", "date": "2026-06-16", "categories": []}
{"id": "2602.02819", "title": "Causal Evaluation of Membership Inference Attacks", "abstract": "Membership Inference Attacks (MIAs) aim to distinguish training points (members) from unseen data (non-members), and are widely used to quantify memorization and assess privacy risks. Standard MIA evaluation requires repeated retraining, which is computationally costly for large models. One-run (single training with randomized data inclusion) and zero-run (post hoc evaluation) methods are often used instead, but their statistical validity remains unclear. We address this gap by framing MIA evaluation as a causal inference problem, defining \\emph{memorization as the causal effect of including a data point in the training set}. This novel formulation reveals and formalizes key sources of bias in existing protocols: one-run methods suffer from interference between jointly included points, while zero-run evaluations are additionally confounded by distribution shift between member and non-member evaluation data. We derive causal analogues of standard MIA metrics and propose practical estimators for multi-run, one-run, and zero-run regimes with non-asymptotic consistency guarantees. We validate our approach in several settings, including pretrained and fine-tuned LLMs, showing that it enables reliable measurement of MIA performance without retraining and under distribution shift. Overall, our framework provides a principled foundation for privacy evaluation in modern AI systems.", "date": "2026-02-02", "categories": []}
{"id": "2605.09169", "title": "Prediction Bottlenecks Don't Discover Causal Structure (But Here's What They Actually Do)", "abstract": "A Mamba state-space model trained only for next-step prediction appears to recover Granger-causal structure through a simple readout $S = |W_{out} W_{in}|$, with early experiments suggesting the phenomenon generalized across architectures and benefited from interventional data at $p < 10^{-5}$. We package the protocol used to test that claim -- standardized synthetic generators (VAR/Lorenz/CauseMe-style), three intervention semantics ($do(X=c)$, soft-noise, random-forcing), edge-provenance cards on three real datasets, and size-matched control arms -- as a reusable falsification benchmark, and walk the claim through it in five stages. The method-level claim does not survive: (i) a plain linear bottleneck does as well or better; (ii) tuned Lasso beats the bottleneck on synthetic CauseMe-style benchmarks, and on Lorenz-96 (the only real benchmark with unambiguous ground truth) classical PCMCI and Granger lead a tight cluster in which the bottleneck trails; (iii) the headline intervention advantage is roughly 60% a sample-size confound, and the residual disappears under standard $do(X=c)$ interventions, surviving only under a non-standard random-forcing scheme; (iv) even that residual reproduces, with a larger effect, in classical bivariate Granger -- the effect is method-agnostic. What survives is a narrow characterization result; the benchmark is the lasting artifact, and each stage above is one of its control arms.", "date": "2026-05-09", "categories": []}
{"id": "2606.05692", "title": "Benchmarking Counterfactual Prediction in Epidemic Time Series with Time-Varying Interventions", "abstract": "Deep learning has enabled significant advances in time-series causal inference, yet progress remains constrained by the lack of realistic benchmarks with observable counterfactual outcomes. Existing datasets either rely on real-world observations without ground-truth counterfactuals or on simplified simulations that fail to capture complex causal dynamics. To address this gap, we develop a large-scale benchmark for counterfactual prediction in epidemic time series under dynamic interventions. Unlike existing benchmarks, it supports static and time-varying treatments, as well as both single-policy and multi-policy intervention settings, enabling evaluation of causal inference methods across a broad range of causal inference scenarios. Leveraging a calibrated agent-based model grounded in real-world demographic, mobility, epidemiological, and policy data, we generate realistic counterfactual trajectories across more than 150 U.S. counties. Using this benchmark, we evaluate widely used and state-of-the-art causal inference methods, revealing substantial performance differences and highlighting the challenges of realistic time-series causal reasoning.", "date": "2026-06-04", "categories": []}
{"id": "2606.14892", "title": "Relational Structural Causal Models", "abstract": "An artificial intelligence must have a model of its environment that is causal, supporting reasoning about interventions and counterfactuals, and also combinatorial, supporting generalization to unseen combinations of objects. In this work, we formally study when and how such a model can be learned. We develop relational structural causal models, extending structural causal models (Pearl 2009) to settings where objects and their relations vary. First, we show how answers to not only causal but also observational queries about unseen combinations of objects can not be identified without further assumptions. To enable such identification--including in the presence of unobserved confounding--we define relational causal graphs and derive symbolic identification criteria. Finally, we propose relational neural causal models, a provably correct approach that outperforms non-relational baselines on simulated traffic scenes with varying cars, signals, and pedestrians.", "date": "2026-06-12", "categories": []}
{"id": "2606.15257", "title": "AI for Social Good: An Investigation of the Causal Relationship Between Environmental Regulations and Their Effects on Air Pollution in London, UK", "abstract": "Air pollution regulation is central to urban public health governance, but estimating its effects is difficult because policies are implemented non-randomly and pollution trajectories are shaped by meteorology, socioeconomic change, temporal trends, and overlapping interventions. This study develops an uncertainty-aware Bayesian deep learning framework to estimate the aggregate effect of air pollution regulations on PM$_{2.5}$ concentrations in London from 2010 to 2020. The framework integrates daily PM$_{2.5}$ observations from Inner London monitoring stations, meteorological covariates, annual socioeconomic indicators, month-of-year and day-of-week indicators, and daily regulation status data for 32 policy measures. A Bayesian LSTM captures temporal dependencies in environmental and socioeconomic covariates, Bayesian embedding layers represent temporal and regulation status inputs, and a regulation status prediction branch supports propensity score-based adjustment for non-random policy implementation. Regulatory effects are estimated by comparing observed PM$_{2.5}$ concentrations with counterfactual predictions under a hypothetical no-regulation scenario, with uncertainty summarized across repeated Bayesian training runs and bootstrap resampling. Results show that London's regulations were associated with an average PM$_{2.5}$ reduction of 1.88 $\\mu$g/m$^3$, a relative reduction of 12.35%, with a 95% confidence interval of 1.64-2.12 $\\mu$g/m$^3$. Estimated effects were limited before 2013, became clearer from 2013 to 2017, and were strongest in 2018 and 2019. The findings suggest that sustained and cumulative regulatory interventions contributed to measurable improvements in London's air quality. This study demonstrates how uncertainty-aware causal AI can support environmental accountability, public health protection, and evidence-based governance for environmental decision-making.", "date": "2026-06-13", "categories": []}
{"id": "2606.15559", "title": "SDVDiag: Multimodal Causal Discovery for Online Diagnosis in Software-defined Vehicles", "abstract": "The transition toward software-defined vehicles concentrates an increasing share of vehicle functionality into distributed software services, where failures propagate through service dependencies and the surface symptom is often several causal hops away from the underlying defect. Existing approaches to causal root-cause analysis in such systems address this only partially: they typically reason over a single observability modality and operate in an offline, operator-driven mode that does not match the demands of continuous vehicle operation. This paper presents SDVDiag, a multimodal causal-discovery pipeline that fuses log-based and metric-based service representations into a shared embedding space before graph construction, coupled with an anomaly-driven trigger that converts the diagnostic platform from a manually operated batch tool into a continuously running online system. Evaluation on an Autonomous Valet Parking testbed shows that the multimodal pipeline produces sparser causal graphs than a metrics-only baseline (134 vs. 182 edges on average) and consistently outperforms it in edge-weighted reward against an expert knowledge graph at every stage of human-feedback refinement, showing a 2.4-fold improvement over the baseline after 60 feedback queries. An end-to-end fault-injection scenario further demonstrates that the integrated trigger correctly recovers a true root cause located two causal hops upstream of the observable symptom.", "date": "2026-06-14", "categories": []}
{"id": "2606.15756", "title": "From Correlation to Causation in Lane Change Prediction for Automated Driving: A Causal Explanation Framework", "abstract": "Lane-change prediction is a central task in intelligent vehicles, where early maneuver anticipation can support safer decision-making. However, many existing approaches mainly learn statistical associations between observed driving variables and future maneuvers, while overlooking the causal dependencies among the input variables themselves. This limits interpretability, especially when physically related variables such as longitudinal gap, relative longitudinal velocity, and Time-To-Collision (TTC) are treated as independent flat inputs. This article presents a causal-inference-based framework for lane-change prediction and explanation. The proposed approach combines linguistic feature construction, expert-constrained causal discovery, deep structural causal modeling with Deep End-to-end Causal Inference (DECI), intervention-based effect analysis, refutation testing, and recursive causal-chain explanation. The objective is not only to predict the future maneuver, but also to identify candidate variables that directly contribute to the prediction, the upstream factors influencing them, and the causal chains through which these effects propagate. The framework achieves average F1-scores above 95% during the first three seconds before the lane-marking crossing event. Beyond prediction accuracy, the framework uses intervention-based effect analysis to distinguish influential from weakly influential variables under the learned causal structure. It further distinguishes candidate direct contributors from mediated effects and generates contrastive causal-chain explanations that clarify why the predicted maneuver is favored and why the alternative maneuvers are less supported. The main contribution is therefore a mechanism-aware lane-change prediction pipeline that moves beyond correlation-based classification toward more interpretable causal reasoning for maneuver prediction.", "date": "2026-06-14", "categories": []}
{"id": "2606.15779", "title": "Faithful Action-unit Causal Reasoning for Counterfactually Faithful Emotion Explanations", "abstract": "Multimodal models can name the action units (AUs) behind a facial emotion, but their AU->emotion rationales are typically plausible rather than faithful: nothing forces the AUs a model invokes to be the AUs that actually drive its prediction. We cast AU->emotion reasoning as a counterfactual-consistency problem between the rationale, the label, and a structural AU->emotion causal graph G, and propose FACR, which grounds the reasoner in an independently induced, polarity-aware G and trains a counterfactual-faithfulness objective: a do-intervention on an AU that G marks causal for a class must move the prediction, while one it marks irrelevant must leave it unchanged. Faithfulness is thereby both trainable and measurable through a matching interventional metric, which we evaluate against a known causal structure, the PSPI pain-AU composition, as no existing affective-reasoning benchmark allows. We are explicit that this metric tests fidelity to the supplied structure rather than its rediscovery: it asks whether the trained reasoner invokes the AUs the structure marks causal, on held-out subjects and a second dataset. Under subject-independent evaluation on UNBC-PAIN, the objective raises the agreement between the invoked AUs and the PSPI composition from a no-objective baseline of 0.08 to 0.57, at a small detection cost; an unfaithfulness control attributes the gain to the objective. On a cross-dataset emotion transfer, the objective likewise raises fidelity to G on a seven-class task (0.50 to 0.84). Finally, we attach a language verbalizer and extend the audit to the generated text: biasing each action unit's emission by its latent activation makes the rationale faithful by construction, so that ablating an AU removes it from the explanation, a property that transfers to a second language-model backbone, whereas a freely generated rationale is unfaithful.", "date": "2026-06-14", "categories": []}
{"id": "2606.15784", "title": "Bayesian Networks with Latent Time Embedding for Stage-Aware Causal Modeling of Alzheimer's Disease Progression", "abstract": "Alzheimer's disease (AD) progression is often described through the amyloid-tau-neurodegeneration, or AT(N), cascade. However, most longitudinal models represent this cascade either as a fixed sequence of biomarkers or as a black-box forecasting task. This makes it difficult to determine when biologically guided biomarker relationships influence future regional pathology. In this study, we introduce Bayesian Networks with Latent Time Embedding (BN-LTE), a Bayesian structural framework for stage-aware modeling of AD progression. BN-LTE estimates disease pseudotime from baseline biomarker profiles and constrains directed dependencies according to biologically plausible AT(N) ordering. Posterior spline-varying structural equations are then used to link initial multimodal measurements with future annualized regional tau-PET change. Across repeated subject-disjoint evaluations using ADNI data, BN-LTE shows strong spatial reconstruction of tau progression compared with the included forecasting baselines. Beyond spatial reconstruction, BN-LTE recovers posterior stage-varying AT(N)-constrained effects and identifies a mid-pseudotime window of amyloid sensitivity. This window is supported by model-implied g-formula contrasts, root-adjusted AIPW, mechanism-sensitive ablations, and robustness analyses across spline and prior specifications. Overall, these findings position BN-LTE as a Bayesian structural framework for forecasting tau progression while examining stage-dependent AT(N)-cascade mechanisms in observational longitudinal neuroimaging data. Our code is available at https://github.com/danleneurocom/BN-LTE.", "date": "2026-06-14", "categories": []}
{"id": "2606.15940", "title": "Causal-Privacy Audit Workflow for Synthetic and Distilled Data in Dropout Support", "abstract": "Synthetic and distilled student data are increasingly used to enable privacy-conscious learning analytics, yet their suitability for decision-facing institutional support remains uncertain. In dropout support, generated data must preserve not only predictive utility or distributional resemblance, but also the financial-status evidence used to guide advising, payment-plan assistance, and scholarship-related decisions. Method: This study introduces CaP-Eval, a decision-facing causal-privacy audit workflow for evaluating generated student data under a fixed estimand, timing-aware adjustment design, estimator set, and empirical privacy-governance screen. The workflow compares original, distilled, adversarial synthetic, statistical synthetic, and DPGNet privacy-oriented generated data on predictive utility, treatment-effect fidelity, robustness to alternative estimators, and local training-record proximity. Results: DPGNet and distilled data preserved the original financial-status treatment-effect structure more reliably than the adversarial and Gaussian Copula baselines. DPGNet preserved full direction and rank agreement across epsilon levels; epsilon = 10 produced the smallest non-original IPW and DML deviations, while epsilon = 1 and epsilon = 5 amplified several financial-status contrasts. Distilled data remained highly faithful but retained the strongest local training-record proximity signal. TabularGNet preserved qualitative directions with moderate attenuation, and Gaussian Copula compressed effect magnitudes. Conclusions: Predictive utility, privacy orientation, empirical disclosure signals, and causal fidelity diverged; generated student data require joint audits of direction, magnitude, overlap, and release-governance risk before decision use.", "date": "2026-06-14", "categories": []}
{"id": "2606.16944", "title": "A Causal Model of Theory of Mind in Conflict for Artificial Intelligence", "abstract": "Theory of mind (ToM), the capacity to ascribe mental states to others and use those ascriptions for prediction and inference, is widely assumed to be essential for effective human-machine integration. Existing AI-ToM models address \\emph{how} to mentalize, but leave the question of when largely unaddressed. The central question is: under what situational and agent-level conditions is ToM engagement causally warranted in conflict? This paper presents a structural causal model formalized as a directed acyclic graph (DAG), treating ToM as a mechanism activated by situational and agent-level conditions rather than as an always-on capacity. The model specifies four exogenous variables capturing situational and agent-level conditions, five endogenous mediators, and a mechanistic ToM node producing engagement states through three distinct causal pathways: a tractability pathway, a reasoning-depth pathway, and an enabling-cause pathway. The primary outcome is epistemic accuracy, which decouples social reasoning from behavioral policy and generalizes across social phenomena beyond conflict. The framework gives AI systems a principled, resource-rational decision procedure for mentalizing, with implications for efficiency, trust, and the development of robust artificial social intelligence. Simulation validation, empirical human-machine teaming studies, and ethical considerations arising from conflict-optimized mentalizing are discussed.", "date": "2026-06-15", "categories": []}
{"id": "2606.17010", "title": "From Tokens to Policy: Causal and Interpretable Heterogeneous Treatment Effects Identification", "abstract": "Heterogeneous Treatment Effect (HTE) identification is crucial to explain the impact of an intervention and optimize our policies accordingly. Existing approaches trade expressivity for interpretability, but, if some active heterogeneity drivers are unmeasured, methods at both ends of this spectrum allow for spurious HTE characterization with no causal reading. In this work, we focus on controlled experiments and argue that an oracle HTE causal characterization via the latent interactors is now within reach, thanks to (i) more extensive pre-treatment measurements, i.e., multi-modal and multi-view, and (ii) scalable representations with minimal human supervision. We then re-frame HTE identification as a Markov-blanket discovery problem on a sufficient and aligned pre-treatment representation, and introduce Neural EXposure Interaction Search (NEXIS), an iterative procedure with provable and empirically validated consistent selection. We deploy NEXIS on two anti-poverty programs in Africa, augmenting each with satellite imagery capturing previously unmeasured environmental effect modifiers, leading to novel, interpretable and prescriptive guidelines to optimize the programs' next iterations.", "date": "2026-06-15", "categories": []}
{"id": "2504.20908", "title": "MOSIC: Model-Agnostic Optimal Subgroup Identification with Multi-Constraint for Improved Reliability", "abstract": "Current subgroup identification methods typically follow a two-step approach: first estimate conditional average treatment effects and then apply thresholding or rule-based procedures to define subgroups. While intuitive, this decoupled approach fails to incorporate key constraints essential for real-world clinical decision-making, such as subgroup size and propensity overlap. These constraints operate on fundamentally different axes than CATE estimation and are not naturally accommodated within existing frameworks, thereby limiting the practical applicability of these methods. We propose a unified optimization framework that directly solves the primal constrained optimization problem to identify optimal subgroups. Our key innovation is a reformulation of the constrained primal problem as an unconstrained differentiable min-max objective, solved via a gradient descent-ascent algorithm. We theoretically establish that our solution converges to a feasible and locally optimal solution. Unlike threshold-based CATE methods that apply constraints as post-hoc filters, our approach enforces them directly during optimization. The framework is model-agnostic, compatible with a wide range of CATE estimators, and extensible to additional constraints like cost limits or fairness criteria. Extensive experiments on synthetic and real-world datasets demonstrate its effectiveness in identifying high-benefit subgroups while maintaining better satisfaction of constraints.", "date": "2025-04-29", "categories": []}
{"id": "2505.17961", "title": "Federated Causal Inference from Multi-Site Observational Data via Propensity Score Aggregation", "abstract": "Causal inference typically assumes centralized access to individual-level data. Yet, in practice, data are often decentralized across multiple sites, making centralization infeasible due to privacy, logistical, or legal constraints. We address this problem by estimating the Average Treatment Effect (ATE) from decentralized observational data via a Federated Learning (FL) approach, allowing inference through the exchange of aggregate statistics rather than individual-level data.\n  We propose a novel method to estimate propensity scores via a federated weighted average of local scores using Membership Weights (MW), defined as probabilities of site membership conditional on covariates. MW can be flexibly estimated with parametric or non-parametric classification models using standard FL algorithms. The resulting propensity scores are used to construct Federated Inverse Propensity Weighting (Fed-IPW) and Augmented IPW (Fed-AIPW) estimators. In contrast to meta-analysis methods, which fail when any site violates positivity, our approach exploits heterogeneity in treatment assignment across sites to improve overlap. We show that Fed-IPW and Fed-AIPW perform well under site-level heterogeneity in sample sizes, treatment mechanisms, and covariate distributions. Theoretical analysis and experiments on simulated and real-world data demonstrate clear advantages over meta-analysis and related approaches.", "date": "2025-05-23", "categories": []}
{"id": "2512.19805", "title": "Guardrailed Uplift Targeting: A Causal Optimization Playbook for Marketing Strategy", "abstract": "This paper introduces a marketing decision framework that optimizes customer targeting by integrating heterogeneous treatment effect estimation with explicit business guardrails. The objective is to maximize revenue and retention while adhering to constraints such as budget, revenue protection, and customer experience. The framework first estimates Conditional Average Treatment Effects (CATE) using uplift learners, then solves a constrained allocation problem to decide whom to target and which offer to deploy. It supports decisions in retention messaging, event rewards, and spend-threshold assignment. Validated through offline simulations and online A/B tests, the approach consistently outperforms propensity and static baselines, offering a reusable playbook for causal targeting at scale.", "date": "2025-12-22", "categories": []}
{"id": "2602.12379", "title": "Deep Doubly Debiased Longitudinal Effect Estimation with ICE G-Computation", "abstract": "Estimating longitudinal treatment effects is essential for sequential decision-making but is challenging due to treatment-confounder feedback. While Iterative Conditional Expectation (ICE) G-computation offers a principled approach, its recursive structure suffers from error propagation, corrupting the learned outcome regression models. We propose D3-Net, a framework that mitigates error propagation in ICE training and then applies a robust final correction. First, to interrupt error propagation during learning, we train the ICE sequence using Sequential Doubly Robust (SDR) pseudo-outcomes, which provide bias-corrected targets for each regression. Second, we employ a multi-task transformer with a covariate simulator head for auxiliary supervision, regularizing representation learning, and a target network to stabilize training dynamics. For the final estimate, we discard the SDR correction and instead use the uncorrected nuisance models to perform Longitudinal Targeted Minimum Loss-Based Estimation (LTMLE) on the original outcomes. This second-stage, targeted debiasing ensures robustness and optimal finite-sample properties. Comprehensive experiments demonstrate that our model, D3-Net, robustly reduces bias and variance across different horizons, counterfactuals, and time-varying confoundings, compared to existing state-of-the-art ICE-based estimators.", "date": "2026-02-12", "categories": []}
{"id": "2605.26759", "title": "Time Series Causal Discovery via Context-Conditioned and Causality-Augmented Pretraining", "abstract": "Causal discovery from time series is critical for many real-world applications, such as tracing the root causes of anomalies. Existing approaches typically rely on dataset-specific optimization, making it difficult to transfer their causal discovery capabilities to new time series governed by diverse causal mechanisms. In this paper, we propose \\textbf{PTCD}, a novel \\textbf{P}retraining framework for \\textbf{T}ime-series \\textbf{C}ausal \\textbf{D}iscovery, which improves cross-task generalization through context-conditioned modeling and transferable causal augmentation. To model complex temporal causal dependencies, PTCD employs a dual-scale iterative attention mechanism to capture window-level causal relationships, and a Gaussian mixture with a context-level routing mechanism to handle heterogeneous exogenous distributions. To further address distribution shifts across causal graphs, PTCD adopts a pretraining paradigm on synthetic datasets that integrates intervention-based learning and a causal mixup strategy, promoting stable causal discovery and stronger generalization. Extensive experiments on multiple real-world out-of-distribution (OOD) datasets demonstrate that PTCD excels in both causal discovery and root cause identification.", "date": "2026-05-26", "categories": []}
{"id": "2606.02231", "title": "Identifiable Markov Switching Models with Instantaneous Effects and Exponential Families", "abstract": "Temporal systems often exhibit non-stationary behaviour, such as seasonal climate variation or glucose fluctuations in patients with type-1 diabetes. One way to model non-stationarity is through discrete latent regimes, i.e., stationary segments of time. Such systems induce a Markov Switching Model (MSM), a class of Hidden Markov Models with autoregressive dependencies among latent regimes and observed variables. Identifying latent regimes is challenging in the presence of frequent regime switches and nonlinear and non-Gaussian dynamics, particularly when there are instantaneous effects between the variables, e.g., due to slow rates of measurements. In this work, we establish the identifiability of both latent regimes and regime-dependent causal structures under temporal regime dependencies, nonlinear lagged and instantaneous effects, and independent noise from the exponential family. Our identifiability theory subsumes non-temporal mixtures of causal models. Furthermore, we introduce FlowMSM, a regime detection framework that can be paired with any stationary causal discovery method to recover regime-dependent causal structures. Experiments on synthetic benchmarks and a financial economics dataset demonstrate the effectiveness of our approach to detect latent regimes and discover causal structures from non-stationary time series.", "date": "2026-06-01", "categories": []}
{"id": "2606.13821", "title": "Attention-Based Estimation of the Individual Treatment Benefit Probability under Dose Variation", "abstract": "Estimating the probability that a treatment outperforms a control for an individual patient, called the Individual Probability of Treatment Benefit (IPTB), offers a clinically intuitive alternative to population-average metrics. However, existing methods for IPTB estimation are largely confined to binary treatment settings, despite the prevalence of dose-varying interventions in clinical practice. We propose a general framework for IPTB estimation with ordinal outcomes under discrete dose assignments, called Dose-AIPTB (Dose Attention-based IPTB). Our approach recasts the problem as binary classification over the unobserved sign of the individual treatment effect, constructing pseudo-labels from covariate-similar pairwise comparisons and aggregating them via attention mechanisms or Nadaraya-Watson kernel regression. This formulation naturally accommodates multiple discrete dose levels, extending beyond the binary treatment paradigm. Through numerical experiments on real-world and synthetic data under covariate shift, varying sample sizes, and heterogeneous outcomes, we demonstrate that attention-based aggregation consistently outperforms kernel alternatives. The framework provides a foundation for personalized dose selection grounded in individual-level benefit probabilities. Codes implementing the model are publicly available at https://github.com/NTAILab/AIPTBDose.", "date": "2026-06-11", "categories": []}
{"id": "2606.13884", "title": "Capability Minimization as a Safety Primitive: Risk-Aware Causal Gating for Least-Privilege LLM Agents", "abstract": "Modern decision systems increasingly rely on learned components whose outputs may be confident yet wrong, exposing downstream actions to costly errors. We introduce Risk-Aware Causal Gating (RACG), a framework that decides whether to act on, defer, or abstain from a model's prediction by combining causal effect estimation with calibrated risk control. RACG models the causal pathway from candidate actions to outcomes and gates each decision according to an estimated counterfactual risk rather than raw predictive confidence. To make gating reliable, we derive distribution-free bounds on the probability of acting under high-risk conditions and show how these bounds translate into operating thresholds that satisfy user-specified safety constraints. We further propose an adaptive gating policy that adjusts to distribution shift by monitoring discrepancies between predicted and realized outcomes, tightening the gate when causal assumptions appear violated. Across simulated interventions and real-world decision benchmarks, RACG reduces high-cost errors substantially while preserving most of the utility of an ungated policy, and it outperforms confidence-based and selective-prediction baselines at matched abstention rates. Our results indicate that explicitly separating causal risk from predictive uncertainty yields decision systems that are both safer and more transparent, offering a principled mechanism for trustworthy automation in high-stakes settings.", "date": "2026-06-11", "categories": []}
{"id": "2606.14636", "title": "Graph Diffusion Residuals for Control-Function Instrumental Variables", "abstract": "Control-function instrumental variable estimators need a first-stage residual, not merely a first-stage prediction. High-capacity first stages can interpolate treatment and leave too little residual information for the outcome equation. We study Adaptive Anisotropic Instrumental Heat Flow (A-IHF), a deterministic graph-diffusion residual extractor for flexible control functions. A-IHF treats treatment as a signal on a graph of first-stage features, uses pilot diffusion to detect large treatment jumps, attenuates conductance across those jumps, and computes the generated control with a sparse graph resolvent. Its observational selection rule uses only $(Z,X)$, combining graph generalized cross-validation, roughness, residualized-treatment relevance, and graph-admissibility filtering. The analysis decomposes error into structural leakage, residual attenuation, and residualized treatment variation, yielding finite-sample bounds, graph-admissibility rates under latent piecewise-smooth geometry, and finite-path selection calibration. Across 54 synthetic benchmark cells with tuned graph, kernel, tree, boosting, series, and neural control-function baselines, guarded observational A-IHF has the lowest average structural-response MSE; the A-IHF family beats the best non-A-IHF baseline in 32 cells. Performance is strongest when the graph captures piecewise-smooth first-stage structure.", "date": "2026-06-12", "categories": []}
{"id": "2410.00903", "title": "Causal Inference with Generative Artificial Intelligence: Application to Texts as Treatments", "abstract": "In this paper, we demonstrate how to enhance the validity of causal inference with unstructured high-dimensional treatments like texts, by leveraging the power of generative Artificial Intelligence (GenAI). Specifically, we propose to use a deep generative model such as large language models (LLMs) to efficiently generate treatments and use their internal representation for subsequent causal effect estimation. We show that the knowledge of this true internal representation helps disentangle the treatment features of interest, such as specific sentiments and certain topics, from other possibly unknown confounding features. Unlike existing methods, the proposed GenAI-Powered Inference (GPI) methodology eliminates the need to learn causal representation from the data, and hence produces more accurate and efficient estimates. We formally establish the conditions required for the nonparametric identification of the average treatment effect, propose an estimation strategy that avoids the violation of the overlap assumption, and derive the asymptotic properties of the proposed estimator through the application of double machine learning. Finally, using an instrumental variables approach, we extend the proposed GPI methodology to the settings in which the treatment feature is based on human perception. The GPI is also applicable to text reuse where an LLM is used to regenerate existing texts. We conduct simulation and empirical studies, using the generated text data from an open-source LLM, Llama 3, to illustrate the advantages of our estimator over state-of-the-art causal representation learning algorithms.", "date": "2024-10-01", "categories": []}
{"id": "2603.14483", "title": "Disentangling Dynamical Systems: Causal Representation Learning Meets Local Sparse Attention", "abstract": "Parametric system identification methods estimate the parameters of explicitly defined physical systems from data. Yet, they remain constrained by the need to provide an explicit function space, typically through a predefined library of candidate functions chosen via available domain knowledge. In contrast, deep learning can demonstrably model systems of broad complexity with high fidelity, but black-box function approximation typically fails to yield explicit descriptive or disentangled representations revealing the structure of a system. We develop a novel identifiability theorem, leveraging causal representation learning, to uncover disentangled representations of system parameters without structural assumptions. We derive a graphical criterion specifying when system parameters can be uniquely disentangled from raw trajectory data, up to permutation and diffeomorphism. Crucially, our analysis demonstrates that global causal structures provide a lower bound on the disentanglement guarantees achievable when considering local state-dependent causal structures. We instantiate system parameter identification as a variational inference problem, leveraging a sparsity-regularised transformer to uncover state-dependent causal structures. We empirically validate our approach across four synthetic domains, demonstrating its ability to recover highly disentangled representations that baselines fail to recover. Corroborating our theoretical analysis, our results confirm that enforcing local causal structure is often necessary for full identifiability.", "date": "2026-03-15", "categories": []}
{"id": "2603.15158", "title": "Point-Identification of a Robust Predictor Under Latent Shift with Imperfect Proxies", "abstract": "Addressing the domain adaptation problem becomes more challenging when distribution shifts across domains stem from latent confounders that affect both covariates and outcomes. Existing proxy-based approaches that address latent shift rely on a strong completeness assumption to uniquely determine (point-identify) a robust predictor. Completeness requires that proxies have sufficient information about variations in latent confounders. For imperfect proxies the mapping from confounders to the space of proxy distributions is non-injective, and multiple latent confounder values can generate the same proxy distribution. This breaks the completeness assumption and observed data are consistent with multiple potential predictors (set-identified). To address this, we introduce latent equivalent classes (LECs). LECs are defined as groups of latent confounders that induce the same conditional proxy distribution. We show that point-identification for the robust predictor remains achievable as long as multiple domains differ sufficiently in how they mix proxy-induced LECs to form the robust predictor. This domain diversity condition is formalized as a cross-domain rank condition on the mixture weights, which is substantially weaker assumption than completeness.\n  We introduce the Proximal Quasi-Bayesian Active learning (PQAL) framework, which actively queries\n  a small, targeted set of diverse domains that satisfy this rank condition. PQAL can recover the point-identified predictor, demonstrates robustness to varying degrees of shift and outperforms previous methods on synthetic data and semi-synthetic dSprites, IHDP, ACS Folktables datasets.", "date": "2026-03-16", "categories": []}
{"id": "2606.12680", "title": "How Useful is Causal Invariance for Domain Adaptation in Finite-Sample Settings?", "abstract": "Machine learning models often degrade when they are deployed on a target distribution that differs from the source distributions they were trained on. Recent work in causality-based domain generalization has shown how shared causal structure between domains can induce invariant predictors, e.g., models on a subset of features which have stable risk across structured domain shifts. However, the extent to which such population-level causal invariances can lead to gains in finite-sample settings remains underexplored. In particular, in practice we often have access to a few labeled target samples, a setting called supervised domain adaptation (sDA). In this paper, we explore when (full or partial) causal knowledge can provably improve supervised domain adaptation.\n  As a first step, we study linear regression, where full or partial causal knowledge specifies a collection of invariant or possibly invariant feature subsets, each yielding a source-trained candidate predictor. We derive matching upper and lower bounds showing that finite-sample gains are governed by the target-risk margins separating the candidates, together with the finite-source estimation error. When these margins are sufficiently large relative to $n_Q$, an adaptive aggregation procedure can match the best candidate predictor while avoiding negative transfer relative to target-only learning. On the other hand, when the margins are too small, no algorithm can reliably exploit the candidate collection to obtain faster finite-sample rates. We further connect these margins to structural shift magnitude in linear SCMs and validate the theory on real-world causal benchmarks.", "date": "2026-06-10", "categories": []}
{"id": "2606.12892", "title": "Prediction-Powered Causal Inference by Automatic Debiased Machine Learning and Semi-Supervised Riesz Regression", "abstract": "This study investigates semiparametric efficient estimation of causal and structural parameters in a semi-supervised setting. In our setting, unlabeled auxiliary regressors are available in addition to labeled observations consisting of outcomes and regressors. Our goal is to construct estimators of causal and structural parameters whose asymptotic variances are smaller than those of estimators constructed using only labeled data. We refer to this framework as prediction-powered causal inference (PPCI). We first derive the efficient influence function and the efficiency bound, which imply that the use of auxiliary regressors can attain a smaller asymptotic variance than the efficiency bound attainable from labeled observations alone. Then, by combining the efficient influence function with the debiased machine learning (DML) framework, we propose methods that we call DML-PPCI. If we construct an estimating-equation estimator, we refer to the method as EE-DML-PPCI; if we construct a targeted-learning estimator, we refer to the method as TMLE-DML-PPCI. The asymptotic variances of both estimators match our derived efficiency bound. In the construction of the estimators, estimation of the efficient influence function plays an important role. In our study, the efficient influence function is also a Neyman orthogonal score, which depends on the Riesz representer and the regression function. For Riesz representer estimation, we develop semi-supervised generalized Riesz regression with convergence rate guarantees.", "date": "2026-06-11", "categories": []}
{"id": "2606.13024", "title": "CausalMoE: A Billion-Scale Multimodal Foundation Model for Granger Causal Discovery with Pattern-Routed Heterogeneous Experts", "abstract": "Granger Causal Discovery (GCD) is fundamental for analyzing temporal dependencies in complex systems. However, existing neural GCD methods predominantly rely on a \"one-size-fits-all\" paradigm, struggling to capture distribution shifts and dynamic regime changes inherent in real-world time series. This often leads to entangled representations and spurious causal graphs. In this paper, we propose CausalMoE, a billion-scale multimodal Granger causal foundation model that explicitly models patch-level heterogeneity. CausalMoE introduces a Pattern-Routed Mixture of Heterogeneous Experts, which dynamically identifies latent temporal patterns and routes patches to specialized domain experts, effectively decoupling regime-specific mechanisms from shared dynamics. To ensure interpretable graph recovery, we design a Causality-Aware Self-Attention mechanism operating across variables, yielding sparse Granger causal graphs via proximal optimization. Furthermore, CausalMoE is the first to integrate LLMs and VLMs to align numerical signals with textual and visual priors, regularizing causal estimation in complex scenarios. Extensive experiments demonstrate that CausalMoE establishes a new state-of-the-art on fully supervised benchmarks, while effectively generalizing to few-shot settings where traditional methods fail.", "date": "2026-06-11", "categories": []}
{"id": "2606.13532", "title": "Graphical Causal Reasoning for Root Cause Analysis in Cloud Networks", "abstract": "Cloud-computing relies on large-scale networks which are inherently complex systems. In this paper, we present a novel approach to root cause analysis (RCA) of cloud network incidents, leveraging graph-based causal discovery techniques. Our method addresses the limitations of rule-based automation by introducing a spatiotemporal grouping strategy and an automation ontology to reduce the dimensionality of the problem. We construct a causal graph from binary time series data using bivariate Granger causality and conditional independence tests. For inference, we introduce a probabilistic method that assigns edge-specific conditional probabilities as a function of time lag, allowing for interpretable, time-aware root cause scoring via causal graph traversal.\n  We evaluated the system using a labeled dataset of 35 production incidents from a major cloud provider. The model successfully recalled the correct root cause in 85.7% of incidents and produced an exact match in 74.3%. In production, the deployed system has been used in over 800 real-world incidents, with positive qualitative feedback from network engineers. These results highlight the practicality of a data-driven, causal approach to RCA in dynamic and large-scale operational environments.", "date": "2026-06-11", "categories": []}
{"id": "2606.13543", "title": "NetCause: Counterfactual Learning for Root Cause Analysis in Large-Scale Networks", "abstract": "Can a learned model capture how faults propagate through a large-scale network and use this knowledge to causally attribute customer impact to its underlying root cause? Existing root cause analysis techniques often rely on static rules, correlation heuristics, or topology-local reasoning, which struggle to generalize in dynamic environments where faults propagate across complex physical and logical dependencies.\n  We present NetCause, a self-supervised learning-based framework that models network incidents as graph-temporal processes and uses counterfactual simulation to rank candidate root causes. This approach produces an interpretable ranking of root cause hypotheses and integrates naturally with operator-defined mitigation and remediation actions.\n  We train the model on over 1,500 incidents collected over six months from a leading cloud provider's production network and evaluate it on 31 expert-labeled incidents. NetCause consistently improves root cause ranking quality in the regime most relevant to operational decision-making, achieving a 16.1% accuracy improvement over a rule-based heuristic baseline. While training is computationally intensive, inference is lightweight, requiring only seconds of GPU runtime per incident (well below typical telemetry collection latencies).", "date": "2026-06-11", "categories": []}
{"id": "2411.10959", "title": "Program Evaluation with Remotely Sensed Outcomes", "abstract": "We study causal inference in experiments and quasi-experiments, where the economic outcome is imperfectly measured by a remotely sensed variable. The remotely sensed variable is low-cost, scalable, and predictive of the economic outcome in observational data; examples include satellite imagery and mobile phone activity. We model the remotely sensed variable as post-outcome: variation in the economic outcome causes variation in the remotely sensed variable. For example, changes in environmental quality cause changes in satellite imagery, not vice versa. Under this assumption, we propose a formula to nonparametrically identify the causal parameter by combining experimental and observational data. We develop a method for n^{-1/2} inference that is robust to misspecification and that does not restrict the algorithms used to process remotely sensed variables.", "date": "2024-11-17", "categories": []}
{"id": "2602.23545", "title": "Planning under Distribution Shifts with Causal POMDPs", "abstract": "In the real world, planning is often challenged by distribution shifts. As such, a model of the environment obtained under one set of conditions may no longer remain valid as the distribution of states or the environment dynamics change, which in turn causes previously learned strategies to fail. In this work, we propose a theoretical framework for planning under partial observability using Partially Observable Markov Decision Processes (POMDPs) formulated using causal knowledge. By representing shifts in the environment as interventions on this causal POMDP, the framework enables evaluating plans under hypothesized changes and actively identifying which components of the environment have been altered. We show how to maintain and update a belief over both the latent state and the underlying domain, and we prove that the value function remains piecewise linear and convex (PWLC) in this augmented belief space. Preservation of PWLC under distribution shifts has the advantage of maintaining the tractability of planning via $\\alpha$-vector-based POMDP methods.", "date": "2026-02-26", "categories": []}
{"id": "2512.17629", "title": "SCOPE: Sequential Causal Optimization of Process Interventions", "abstract": "Prescriptive Process Monitoring (PresPM) recommends interventions during running business processes to optimize key performance indicators (KPIs). In realistic settings, interventions are rarely isolated: organizations need to align sequences of interventions to jointly steer the outcome of a case. Existing PresPM approaches only partially address this challenge. Many focus on a single intervention decision, while others treat multiple interventions independently, ignoring how they interact over time. Methods that do address these dependencies depend either on simulation or data augmentation to approximate the process to train a Reinforcement Learning (RL) agent, which may create a reality gap and introduce bias. We introduce SCOPE (Sequential Causal Optimization of Process Interventions), a PresPM approach that learns aligned sequential intervention recommendations. SCOPE employs backward induction to estimate the effect of each candidate intervention action, propagating its impact from the final decision point back to the first. By leveraging causal learners, our method can utilize observational data directly, unlike methods that require constructing process approximations for RL. Experiments on both an existing synthetic dataset and a new semi-synthetic dataset show that SCOPE consistently outperforms state-of-the-art PresPM techniques in optimizing the KPI. The novel semi-synthetic setup, based on a real-life event log, is provided as a reusable benchmark for future work on sequential PresPM.", "date": "2025-12-19", "categories": []}
{"id": "2602.01135", "title": "Your Autoregressive Model Already Reveals the Causal Graph", "abstract": "Autoregressive models trained via next-token prediction implicitly learn the conditional independence structure of their data-generating process. We exploit this observation to perform scalable causal discovery from a single observed sequence of discrete events -- without any task-specific retraining. Such single-stream settings arise naturally in vehicle diagnostics, manufacturing systems, and patient trajectories, yet they remain largely unsolved: the absence of repeated samples, massive event vocabularies, and long-range temporal dependencies render existing methods either inaccurate or computationally intractable. We introduce TRACE, a framework that repurposes any pretrained autoregressive model as a density estimator for conditional mutual information, the fundamental primitive for conditional independence testing. By constructing parallelized CI tests on GPUs, TRACE recovers both the sample-level time causal graph and its summary projection, scaling linearly with the vocabulary size while naturally handling delayed causal effects. Crucially, we prove that minimizing the standard cross-entropy pretraining loss directly minimizes an upper bound on the causal identification error, establishing a duality between sequence prediction and causal discovery. On nonlinear SCMs (|X| = 8000) and real-world vehicle diagnostic logs (|X| = 29100), TRACE is the first applicable method at this scale, outperforming the strongest baseline by over 20 F1 points.", "date": "2026-02-01", "categories": []}
{"id": "2606.09892", "title": "LMT: A Bayesian Framework for Causal Discovery from Textual Alarm Records in Manufacturing Systems", "abstract": "Textual event records, such as alarm logs, have become an increasingly common data source in engineering and manufacturing systems. Beyond identifying correlations or recurring patterns, engineers are often interested in understanding which types of events causally trigger or influence other events during system operation. Textual event descriptions may contain semantic clues about such causal relationships, and recent large language models (LLMs) provide a promising tool for extracting these signals. However, relying solely on LLM-encoded textual information is insufficient for accurate causal discovery, since semantic patterns do not directly reveal causal mechanisms and may confuse causation with correlation or frequent sequential patterns. To address these challenges, we propose \\textbf{LMT}, a Bayesian causal discovery framework for engineering event data that jointly leverages textual descriptions and timestamps. Specifically, LMT first uses LLMs to extract semantic causal signals from event descriptions and constructs a prior distribution over causal graphs among event types or event clusters. It then incorporates temporal evidence through a Poisson-process-based likelihood, allowing the LLM-informed prior to be refined by timestamp-based statistical evidence. By integrating the textual and temporal information, LMT produces a causal graph that is both interpretable and data-supported. Simulation studies show that the proposed framework is effective across different settings and is especially advantageous in small-sample alarm-event scenarios.", "date": "2026-06-03", "categories": []}
{"id": "2606.10607", "title": "Causal Ensemble Agent: Hierarchical Causal Discovery with LLM-guided Expert Reweighting", "abstract": "Causal discovery aims to uncover causal structures from observational data, which is crucial for real-world decision-making. However, different causal discovery algorithms can produce divergent results that conflict with each other, complicating the identification of accurate causal graphs. Traditional approaches rely on numerical values and statistical assumptions, often ignoring rich domain-specific information, such as feature descriptions, which could also help structure learning. While recent works explore using Large Language Models (LLMs) to infer causal relations via direct queries, such methods can be unreliable due to a lack of alignment with the actual data. To address these limitations, we propose Causal Ensemble Agent (CEA), a novel framework that aggregates structural insights from statistical discovery experts across different graph levels via linear opinion pooling, and uses an LLM as a meta-referee to dynamically reweight experts when the aggregated confidence is close to the decision boundary, thereby composing an improved and more complete causal graph. Extensive experiments on both synthetic and real-world datasets demonstrate that CEA achieves the strongest overall performance across a wide range of causal discovery methods, highlighting the effectiveness of using LLMs for meta-analysis in causal discovery.", "date": "2026-06-09", "categories": []}
{"id": "2606.10934", "title": "WorldKernel: A World Model is the Coupling Kernel of Admissible Possible Worlds", "abstract": "A common assumption holds that enough observational and interventional data, given to a strong enough predictor, suffices. We report a failure mode that contradicts it. Across hundreds of structural causal models, on identified quantities a strong predictor and a Bayesian baseline both succeed, but on unidentified quantities (the couplings between counterfactual worlds) the predictor collapses to a point, on 28% of models to one no valid model can produce, while the truth is an admissible interval more data never narrows. The gap is structural: prediction cannot represent uncertainty over counterfactual couplings. We cast a world model as a single positive semidefinite coupling kernel K(T,T') over admissible worlds, whose diagonal is the ordinary posterior (what a predictor recovers) and whose off-diagonal is the cross-world coupling it cannot, which every counterfactual reads. The paper is the theory of that off-diagonal. It is real: two states with identical posteriors differ on a cross-world query, and the off-diagonal is the coupling that fixes counterfactuals. It can be bounded: positive semidefiniteness is partial-identifying information the marginals lack, and enforcing it bounds counterfactuals in polynomial time where the exact response-type program is intractable. Logical structure sharpens it: ontology axioms tighten the bound by up to a third, propagating to couplings they never touch. It can be acquired: targeted scars, constraints learned from encountered infeasibilities, close the gap several times faster than untargeted ones. Its full reconstruction is approximate counting of the admissible worlds, tractable below the Sly-Sun threshold and inapproximable above; we do not claim to beat the worst case.", "date": "2026-06-09", "categories": []}
{"id": "2405.17151", "title": "Smoke and Mirrors in Causal Downstream Tasks", "abstract": "Machine Learning and AI have the potential to transform data-driven scientific discovery, enabling accurate predictions for several scientific phenomena. As many scientific questions are inherently causal, this paper looks at the causal inference task of treatment effect estimation, where the outcome of interest is recorded in high-dimensional observations in a Randomized Controlled Trial (RCT). Despite being the simplest possible causal setting and a perfect fit for deep learning, we theoretically find that many common choices in the literature may lead to biased estimates. To test the practical impact of these considerations, we recorded ISTAnt, the first real-world benchmark for causal inference downstream tasks on high-dimensional observations as an RCT studying how garden ants (Lasius neglectus) respond to microparticles applied onto their colony members by hygienic grooming. Comparing 6 480 models fine-tuned from state-of-the-art visual backbones, we find that the sampling and modeling choices significantly affect the accuracy of the causal estimate, and that classification accuracy is not a proxy thereof. We further validated the analysis, repeating it on a synthetically generated visual data set controlling the causal model. Our results suggest that future benchmarks should carefully consider real downstream scientific questions, especially causal ones. Further, we highlight guidelines for representation learning methods to help answer causal questions in the sciences.", "date": "2024-05-27", "categories": []}
{"id": "2505.08908", "title": "Statistical Decision Theory with Counterfactual Loss", "abstract": "Many researchers apply classical statistical decision theory to evaluate treatment choices and learn optimal policies. However, because this framework relies solely on realized outcomes under chosen actions and ignores counterfactuals, it cannot assess the quality of a decision relative to feasible alternatives at the unit level, which is an important requirement in some settings. For example, in pretrial bail decisions, a judge must balance crime prevention upon release against the risk of imposing unnecessary burdens on arrestees. A central challenge in this framework is identification: since only one potential outcome is observed per unit, counterfactual risk is typically not identifiable. We show that, under strong ignorability, counterfactual risk is identifiable if and only if the loss is additive in the potential outcomes. We further demonstrate that additive counterfactual losses can yield treatment recommendations that differ from those based on standard losses when more than two treatment options are available. We show that additive counterfactual losses capture not only decision accuracy but also decision difficulty, whereas standard losses reflect accuracy alone. Finally, we introduce a symbolic linear inverse program that determines whether a given counterfactual loss yields an identifiable risk, without requiring data.", "date": "2025-05-13", "categories": []}
{"id": "2509.01916", "title": "Causal Representation Learning from Network Data", "abstract": "Causal disentanglement from soft interventions is identifiable under the assumptions of linear interventional faithfulness and availability of both observational and interventional data. Prior work has focused on unstructured observations without leveraging known relational context among measured entities. In many scientific applications, however, the measured variables come with an observed interaction network that provides structured context, such as protein-protein interactions and pathway-gene membership. We propose GraCE-VAE, a graph-aware causal discrepancy variational autoencoder that treats pathway-level information as an auxiliary view of the latent causal programs. The graph neural network encoder conditions on this auxiliary pathway view and the biological graph to improve amortized inference, while the causal decoder remains a latent SCM with soft interventions. Assuming samples are i.i.d. within each intervention regime, we show that GraCE-VAE inherits the identifiability guarantees of causal discrepancy VAEs and identifies the latent causal graph and intervention targets up to the standard equivalence class. Experiments on three CRISPR perturbation datasets demonstrate that leveraging structured biological context improves prediction of interventional outcomes, including unseen perturbation combinations.", "date": "2025-09-02", "categories": []}
{"id": "2510.27544", "title": "TempoBench: Evaluating Temporal Causal Reasoning in Large Language Models", "abstract": "Temporal reasoning involves understanding how systems evolve over time through input-driven state transitions. A key aspect is temporal causal reasoning, causally reasoning about what prior inputs were necessary in causing an observed outcome. While large language models (LLMs) perform well at forward simulation, predicting outputs from inputs, they struggle to identify the minimal causal inputs of outcomes. To study this distinction, we define two tasks: \\textit{trace simulation} (SIM), which requires models to simulate system execution, and \\textit{minimal causal attribution} (MIN), which identifies the minimal set of inputs necessary for a given outcome. We introduce \\textsc{TempoBench}, the first formally verified benchmark for temporal causal reasoning, built from synthesized Mealy machines with controllable complexity and provably correct causal labels. Across frontier models, we observe that despite achieving up to 96\\% accuracy on the SIM task, performance on the causal attribution MIN task drops below 25\\%; models fail to reason about causal necessity. Over 94\\% of causal errors involve overspecification, where models perform retrieval and list all possible inputs rather than reasoning about the minimal causal subset. Fine-tuning on \\textsc{TempoBench} training corpus improves causal reasoning and generalizes better than math, code, or instruction training, with gains across standard reasoning benchmarks.", "date": "2025-10-31", "categories": []}
{"id": "2601.21221", "title": "Causal Discovery for Explainable AI: A Dual-Encoding Approach", "abstract": "Understanding causal relationships among features is fundamental for explaining machine learning model decisions. However, traditional causal discovery methods face challenges with categorical variables due to numerical instability in conditional independence testing. We propose a dual-encoding causal discovery approach that addresses these limitations by running constraint-based algorithms with complementary encoding strategies and merging results through majority voting. Applied to the Titanic dataset, our method identifies causal structures that align with established explainable methods.", "date": "2026-01-29", "categories": []}
{"id": "2601.22736", "title": "UA-DCM: Uncertainty-aware Causal Decision Making via Effect Bound Decomposition", "abstract": "Causal inference from observational data can provide strong evidence for finding the best action in a decision-making scenario without having to perform expensive randomized trials. The causal effect of an action is often not pointwise identifiable even with infinite data due to unobserved confounding factors. Furthermore, having only finitely many samples adds another layer of uncertainty to causal effect estimation. Several existing methods can be used to obtain upper and lower bounds to the causal effect, ranging from symbolic methods to the more recent neural network-based approaches, which implicitly incorporate both sources of uncertainty. However, these methods do not inform whether collecting more samples may or may not help identify the best action from observational data, leaving experts in the dark about their data collection strategies. We address this problem with a novel framework that can distinguish the range of causal effect values that might be eliminated by collecting more samples from the range of values that, with high probability, cannot be eliminated with more observational samples. We show that this partitioning can be obtained by solving max-min and min-max optimization problems. We leverage neural causal models to approximately recover this decomposition in practice. We demonstrate via experiments on synthetic and real-world datasets that our algorithm can determine when collecting more samples will not help determine the best action. Our framework can help practitioners decide when to resort to non-observational studies or seek to measure some of the unmeasured confounders for optimal decision-making.", "date": "2026-01-30", "categories": []}
{"id": "2602.16061", "title": "Partial Identification under Missing Data Using Weak Shadow Variables from Pretrained Models", "abstract": "Estimating population quantities such as mean outcomes from user feedback is fundamental to platform evaluation and social science, yet feedback is often missing not at random (MNAR): users with stronger opinions are more likely to respond, so standard estimators are biased and the estimand is not identified without additional assumptions. Existing approaches typically rely on strong parametric assumptions or bespoke auxiliary variables that may be unavailable in practice. In this paper, we develop a partial identification framework in which sharp bounds on the estimand are obtained by solving a pair of linear programs whose constraints encode the observed data structure. This formulation naturally incorporates outcome predictions from pretrained models, including large language models (LLMs), as additional linear constraints that tighten the feasible set. We call these predictions weak shadow variables: they satisfy a conditional independence assumption with respect to missingness but need not meet the completeness conditions required by classical shadow-variable methods. When predictions are sufficiently informative, the bounds collapse to a point, recovering standard identification as a special case. In finite samples, to provide valid coverage of the identified set, we propose a set-expansion estimator that achieves slower-than-$\\sqrt{n}$ convergence rate in the set-identified regime and the standard $\\sqrt{n}$ rate under point identification. In simulations and semi-synthetic experiments on customer-service dialogues, we find that LLM predictions are often ill-conditioned for classical shadow-variable methods yet remain highly effective in our framework. They shrink identification intervals by 75--83\\% while maintaining valid coverage under realistic MNAR mechanisms.", "date": "2026-02-17", "categories": []}
{"id": "2606.00568", "title": "On the Recoverability of Causal Relations from Bulk Gene Expression Data", "abstract": "Bulk gene expression profiling, which aggregates pooled RNA across cells within a biological sample, remains important in the single-cell era because it is typically less noisy, more sensitive, and more cost-effective than single-cell assays. Accordingly, a growing body of computational methods seeks to recover causal relations among genes from bulk expression data. However, aggregation is a lossy, non-invertible coarsening of the underlying cellular system, and it remains unclear whether and under what conditions causal relations are recoverable from aggregated bulk gene expression data. To answer this, we formalize recoverability under aggregation through two notions of consistency: functional-form consistency and conditional-independence consistency. We then derive necessary and sufficient conditions for recoverability, showing that these properties are preserved only under linear aggregations (e.g., sum/mean) coupled with affine structural equations. To assess the practical plausibility of these conditions, analyses of four bulk and four single-cell gene expression datasets further reveal that the estimated pairwise regulatory functions among genes deviate from linearity in both data types, providing limited empirical support for the linearity assumptions required for recoverability. Together, these results caution against recovering causal relations from aggregated bulk expression data without strong additional assumptions.", "date": "2026-05-30", "categories": []}
{"id": "2606.04421", "title": "Trivium: Temporal Regret as a First-Class Objective for Causal-Memory Controllers", "abstract": "Many current agentic systems and LLM pipelines correct mistakes by optimizing outcome reward. This addresses only the what of failure: when an outcome diverges from prediction, the why and when of the mismatch are not systematically logged, reviewed, or corrected, so the same error can recur episode after episode. We argue that this is a structural problem, not merely a model-capacity one. We propose long-horizon temporal regret as a first-class objective alongside outcome regret and epistemic regret over the working causal model. Temporal regret captures when failure persists: how long a miscalibrated causal model is tolerated before correction. Epistemic regret captures why failure persists: residual uncertainty or error in the working causal model. Together, the three regrets give a falsifiable account of what, why, and when a long-lived agent can fail. Modeling the agent as a stream of E episodes, we prove three conditional results under explicit causal-probing, persistence, and detectability assumptions. First, under observationally equivalent confounding, outcome-only learning cannot distinguish causal from spurious structure without an intervention channel, so temporal miscalibration can persist linearly even after outcome regret is driven to zero. Second, with a persistent causal log and budgeted probes, total probe complexity is logarithmic in the episode horizon, inducing O(log E) temporal regret. Third, under K detectable change-points, the rate extends to O(K log E). We instantiate Trivium and pre-register five falsifiable predictions. On CausalBench-Seq, Trivium follows the predicted logarithmic envelope while outcome-only baselines grow linearly. A pilot real-LLM stream provides preliminary external-validity evidence across one full E = 500 run and three E = 100 frontier-model pilots. Self-learning here means revising an external causal model, not retraining LLM weights.", "date": "2026-06-03", "categories": []}
{"id": "2606.05797", "title": "Causal Longitudinal Prior-Fitted Networks for Counterfactual Outcome Prediction", "abstract": "Longitudinal treatment decisions from multivariate time-series data require predicting potential outcomes under future treatment sequences in the presence of time-varying confounding, heterogeneous patient dynamics, and limited domain-specific data. Existing longitudinal causal estimators typically address this problem by training a new model for each cohort or simulator. We introduce Causal Longitudinal Prior-Fitted Networks (CausalLongPFN), a prior-fitted network for time-series causal inference in longitudinal treatment-response data and zero-shot in-context counterfactual outcome prediction. The model is pretrained entirely on synthetic episodes sampled from a broad prior over temporal structural causal models, exposing it to treatment-confounder feedback, latent heterogeneity, nonlinear state evolution, delayed effects, and cumulative treatment responses. At test time, CausalLongPFN remains frozen and is used zero-shot: it conditions on support trajectories, a query history, and a planned future treatment sequence, and returns a predictive distribution over future outcomes without gradient updates or propensity-model fitting. Multi-step predictions are obtained by recursively applying the one-step predictor under the specified treatment sequence. We evaluate the model on branchable cancer, HIV, and warfarin benchmarks with ground-truth counterfactual labels, and on factual-only rolling-origin prediction in MIMIC-III ICU trajectories. CausalLongPFN is competitive with domain-trained longitudinal baselines on counterfactual benchmarks and performs strongly on factual MIMIC-III prediction, suggesting that broad synthetic causal pretraining can provide a frozen, amortized alternative for zero-shot longitudinal treatment-response prediction when repeated domain-specific training is costly or impractical.", "date": "2026-06-04", "categories": []}
{"id": "2606.07525", "title": "Implicit Causal Graph Construction in Text via Chain Discovery", "abstract": "Causal graphs in text are typically populated by observable, predefined events. In contrast, we study implicit causal graph construction from text by treating each described cause-effect pair as the begin- and endpoint of an underlying latent causal graph and using large language models (LLMs) to infer intermediate causal events. We compare end-to-end graph construction with methods that frame the task as causal chain discovery. In the latter, graphs are built either by aggregating inferred chains or by progressively expanding partial chains through an iterative search process. We further explore Wisdom of the Crowd extensions that access causal knowledge from multiple LLMs in post-hoc aggregation and collaborative inference settings. We analyze trade-offs among these approaches and evaluate the validity of inferred causal relations using a manually curated database of 1,560 scientifically validated causal pairs. This database-based evaluation is proposed as reliable, resource-efficient, and transferable to settings where ground-truth graphs are unavailable.", "date": "2026-04-20", "categories": []}
{"id": "2606.07693", "title": "Transfer learning for causal forest", "abstract": "Transfer learning addresses the challenge of transfering knowledge from one domain to another. Traditional transfer learning focuses on adapting models trained on a source domain (with a lot of observations) to improve performance on a target domain (with few observations). In this work we consider the case of a model shift and we focus on the transfer learning applied to a causal forest namely HTERF. This causal forest aims to estimate the Conditional Average Treatment Effect (CATE). The approach considered is the offset method presented by Wang (2016) adapted to a causal context. This method relies on the use of intermediate models in order to estimate the offset between source and target distributions. Our main result is a bound on the CATE error of HTERF on target depending on the error of the intermediate models. Simulation studies show the good performances of this approach in different settings on simulations and on a real-world dataset.", "date": "2026-06-05", "categories": []}
{"id": "2606.07865", "title": "Instrumented data for causal scientific machine learning", "abstract": "Scientific machine learning is limited less by model size than by the data it is trained on. Observational data records what happened but not why; template synthetic data has a known generating process but only for the simulator's template, not the case a user faces. We argue a third option is now operationally feasible: instrumented data, in which every datum carries the mechanistic model that produced it, an explicit uncertainty over that model, and an executable family of counterfactuals. Verification-and-validation (V&V) instrumented image-to-simulation pipelines are one realisation: a sensor observation becomes a fully specified, solver-backed simulation with explicit, editable parameters and a propagated aleatoric/epistemic uncertainty. The substrate is case-specific, mechanistically supervised, and supports causal interventions through Pearl's do-operator. Near-term consequences for validation, auditing, and surrogate training span computational biology, climate, materials, fluid mechanics, and medical imaging; a longer-term, falsifiable implication concerns foundation models for scientific reasoning.", "date": "2026-06-05", "categories": []}
{"id": "2606.08196", "title": "Beyond Additivity: Causal Discovery in Location-Scale Noise Models with Hidden Variables", "abstract": "We study causal discovery from observational data when some variables are hidden and the data-generating process follows a location-scale noise model (LSNM). Existing methods that handle hidden confounders typically assume additive noise, but in practice, causes often modulate not just the mean but also the variance of their effects. We prove that acyclic directed mixed graphs (ADMGs) satisfying a bow-free condition are identifiable under LSNM with hidden variables, establishing the first identifiability result for causally insufficient models beyond noise additivity. We further provide sufficient conditions for identifying causal direction even when the bow-free assumption is violated. Our two-stage algorithm, LSNM-UV, is sound and complete, and experiments demonstrate improved performance over additive baselines on heteroscedastic data.", "date": "2026-06-06", "categories": []}
{"id": "2606.08275", "title": "Causal Agent Replay: Counterfactual Attribution for LLM-Agent Failures", "abstract": "When an LLM agent fails -- issues a refund it should not have, calls the wrong tool, leaks data -- existing tooling answers what happened (observability) or whether it passed (evaluation), but not which step caused the failure. The obvious heuristics are wrong: the step that executes the harmful action is usually not the step that decided on it, and LLM-judge attribution is correlational and unreliable (state-of-the-art step-level accuracy on the Who&When benchmark is about 14%). We present Causal Agent Replay (CAR), which answers the question by intervention: it models an agent run as a structural causal model, applies a do-operation to a step, and re-executes the trajectory forward under the same stochastic policy, measuring the shift in the outcome distribution. We define an intervention algebra over agent steps, a single-step contrastive estimator whose point-of-commitment rule resolves a confound specific to stochastic run-forward, and a budget-bounded Monte-Carlo Shapley estimator that splits credit across interacting steps. Every effect is reported with confidence intervals. We validate against synthetic structural causal models with planted ground truth: the contrastive estimator recovers the pivotal step, and Shapley recovers a two-step interaction (0.44, 0.45, ~0; efficiency sum 0.909 versus the analytic 0.91). CAR is open source and runs on hosted or free local models.", "date": "2026-06-06", "categories": []}
{"id": "2606.08305", "title": "MEC-Cox: Machine-Learning-Assisted Generalized Entropy Calibration for ATT Marginal Hazard-Ratio Estimation", "abstract": "Externally controlled survival trials are increasingly used when concurrent randomized controls are infeasible, particularly in oncology and rare-disease settings with time-to-event endpoints. We target an average-treatment-effect-on-the-treated (ATT)-type marginal hazard-ratio estimand, comparing treatment with counterfactual control in the treated trial population, and estimate it using inverse-probability-weighted (IPW) Cox regression. Valid inference is challenging because IPW Cox regression depends on the weights through both event contributions and risk-set averages, making flexible machine-learning nuisance estimation difficult to incorporate directly. Building on machine-learning-assisted generalized entropy calibration (MEC) by Lee and Kim (2026), we propose MEC-Cox for ATT-weighted IPW Cox regression. The method begins with normalized source-propensity-score odds weights for external controls and then applies Bregman calibration to balance cross-fitted prognostic summaries between external controls and treated trial patients. The calibration basis may include control-survival predictions, Cox linear predictors, penalized-survival-model predictions, or other prognostic-score summaries. MEC-updated weights therefore play a dual role as source-transport and prognostic-score balancing weights. We establish consistency, characterize a calibration-induced efficiency gain, and develop a stacked sandwich variance estimator. Simulations show that MEC-Cox can reduce bias, increase efficiency, and improve coverage through flexible machine-learning-assisted adjustment.", "date": "2026-06-06", "categories": []}
{"id": "2606.08941", "title": "Estimate Collapsibility of Causal Effects in Completed Partial DAGs via Strong d-Convex Hulls", "abstract": "This paper proposes a collapsible method for estimating causal effects that maintains the estimator's consistency before and after marginalization over some variables in completed partially directed acyclic graphs (CPDAGs). We first introduce the estimate collapsibility for CPDAGs and characterize the minimal collapsible sets as strong d-convex hulls. An efficient algorithm is devised to obtain such sets in DAGs and is generalized to CPDAGs. Then, we combine the graph reduction procedure with the IDA framework. Finally, experiments and empirical analysis show the effectiveness of the collapsibility for causal estimations in CPDAGs. Code is available at https://github.com/Jamyang-D/strongly-convex.", "date": "2026-06-08", "categories": []}
{"id": "2606.09181", "title": "Counterfactual Reasoning for Fine-Grained Evidence Disentanglement in VideoQA", "abstract": "Recent advances in video multimodal models have significantly improved VideoQA performance. However, these systems often rely on spurious statistical correlations rather than answer-relevant causal evidence, resulting in unfaithful and brittle reasoning, especially in complex real-world scenarios. Existing methods either rely on cross-modality correlations, costly curated training resources, or insufficient causal assumptions and constraints, and typically operate at the time-interval level. As a result, they fail to explicitly disentangle causal visual cues from confounders and provide limited fine-grained evidence localization. To address this issue, we propose a Counterfactual Reasoning framework for fine-grained Evidence Disentanglement (CREDiT). CREDiT formulates the VideoQA process using a structural causal model and learns cross-modality representations that are explicitly decomposed into causal and non-causal components under independence and minimality constraints. To facilitate faithful disentanglement, we introduce feature-level causal interventions and construct counterfactual inputs that approximate causal effects while suppressing non-causal correlations. Extensive experiments on NExT-GQA, SportsQA, and SPORTU-video demonstrate that CREDiT consistently improves answer accuracy and reasoning reliability across both generic and complex sports scenarios, leading to more trustworthy VideoQA systems.", "date": "2026-06-08", "categories": []}
{"id": "2606.09822", "title": "Causally Evaluating the Learnability of Formal Language Tasks", "abstract": "Language models, as multi-task learners, acquire a wide range of abilities during training. A fundamental question is how much task-specific data is needed to learn a given task. Answering this for natural language is difficult: tasks are hard to delineate and can confound one another. To rigorously investigate the relationship between data frequency and learnability, we turn to a controlled setting using formal languages induced from probabilistic finite automata. These serve as a methodological testbed to demonstrate that standard correlational evaluation practices are inherently flawed. To enable causal analysis, we introduce the binning semiring, an algebraic object that lets us control how often a targeted property occurs in a sampled corpus. We formulate the experimental pipeline as a causal graphical model and derive decomposed Kullback-Leibler divergence metrics to measure the learnability of specific sub-tasks. Our experiments show that evaluating learnability without causal intervention leads to incorrect conclusions due to confounders in correlational analysis, and serve as a warning about correlational pitfalls in natural-language settings.", "date": "2026-06-08", "categories": []}
{"id": "2502.21123", "title": "Causality Is Key to Understand and Balance Multiple Goals in Trustworthy ML and Foundation Models", "abstract": "Ensuring trustworthiness in machine learning (ML) systems is crucial as they become increasingly embedded in high-stakes domains. This paper advocates for integrating causal methods into machine learning to navigate the trade-offs among key principles of trustworthy ML, including fairness, privacy, robustness, accuracy, and explainability. While these objectives should ideally be satisfied simultaneously, they are often addressed in isolation, leading to conflicts and suboptimal solutions. Drawing on existing applications of causality in ML that successfully align goals such as fairness and accuracy or privacy and robustness, this paper argues that a causal approach is essential for balancing multiple competing objectives in both trustworthy ML and foundation models. Beyond highlighting these trade-offs, we examine how causality can be practically integrated into ML and foundation models, offering solutions to enhance their reliability and interpretability. Finally, we discuss the challenges, limitations, and opportunities in adopting causal frameworks, paving the way for more accountable and ethically sound AI systems.", "date": "2025-02-28", "categories": []}
{"id": "2604.27011", "title": "Automatic Causal Fairness Analysis with LLM-Generated Reporting", "abstract": "AutoML, intended as the process of automating the application of machine learning to real-world problems, is a key step for AI popularisation. Most AutoML frameworks are not accounting for the potential lack of fairness in the training data and in the corresponding predictions. We introduce \\textsc{FairMind}, a software prototype aiming to automatise fairness analysis at the dataset level. We achieve that by resorting to the assumptions of the \\emph{standard fairness model}, recently proposed by Ple\\v{c}ko and Bareinboim. This allows for a sound fairness evaluation in terms of causal effects, based on \\emph{counterfactual} queries involving the target, possibly confounders and mediators, and the different values of an input feature we regard as \\emph{protected}. After the necessary data preprocessing, the tool implements a closed-form computation of the effects. LLMs are consequently exploited to generate accurate reports on the fairness levels detected in the training dataset. We achieve that in a zero-shot setup and show by examples the expected advantages with respect to a direct analysis performed by the LLM. To favour applications, extensions to ordinal protected variable and continuous targets and novel decomposition results are also discussed.", "date": "2026-04-29", "categories": []}
{"id": "2606.06715", "title": "Does Topic Sentiment Cause Perceived Ideology? Comparing Human and LLM Annotations in Political News Articles", "abstract": "We ask whether topic sentiment has a causal effect on perceived political ideology, and whether the answer depends on who assigns the ideology label. Using articles from AllSides, paired with shared sentiment annotations from Llama-3.3-70b-versatile, we compare ideology labels from expert human annotators, GPT-4o-mini (baseline and finetuned), and Llama-3.3-70B. We apply Double Machine Learning (DML) and community-level mediation analysis across all four annotation paradigms. Human annotations yield no significant causal effects at the community level. Fine-tuned GPT-4o-mini achieves the highest classification accuracy (F1=72.48) and is the only annotator paradigm that produces significant community-level treatment effects and significant natural direct effects (NDEs) in mediation. We interpret this as evidence of shortcut learning: fine-tuning on ideology-labeled data causes the model to internalise a spurious sentiment--ideology coupling not operative in human judgment for this task. This coupling is structurally invisible to F1-based evaluation, with implications for the use of LLM annotations as silver labels and as proxies for human judgment in downstream causal analyses.", "date": "2026-06-04", "categories": []}
{"id": "2606.06908", "title": "polyDAG: Polynomial Acyclicity Constraints for Efficient Continuous Causal Discovery in Visual Semantic Graphs", "abstract": "Modern image-analysis pipelines often convert images into structured semantic variables, such as facial attributes, object concepts, and scene descriptors. Learning directed dependencies among these variables can produce interpretable visual semantic graphs, but continuous directed acyclic graph learning is limited by the cost of enforcing acyclicity. We present polyDAG, a polynomial acyclicity framework for efficient continuous causal discovery in visual semantic graphs. polyDAG replaces the matrix-exponential acyclicity constraint with a finite polynomial trace constraint and proves that the new constraint is zero exactly for acyclic graphs. We further derive a geometric-series implementation that avoids the explicit summation loop while preserving the same acyclicity condition. Experiments on synthetic Erdos-Renyi graphs and CelebA facial visual attributes show that polyDAG improves efficiency and structure recovery. Averaged over the revised synthetic protocol with d in {100, 200, 500}, polyDAG reduces mean structural Hamming distance from 318.4 to 285.4 and improves mean F1 score from 0.725 to 0.756. At 100 nodes, the geometric variant runs in 3.44 seconds compared with 5.16 seconds for the exponential baseline, corresponding to a 33.4 percent speedup. Code and data are publicly available at https://github.com/wenhaoz-fengcai/polyDAG.", "date": "2026-06-05", "categories": []}
{"id": "2606.07399", "title": "Automatic, Debiased, and Invariant Counterfactual Generation under General Interventions", "abstract": "Generative models for counterfactual outcomes have great potential to support decision-making under complex interventions, but existing approaches are limited by unstable estimation, poor generalization across environments, and bias from nuisance model misspecification. We introduce ADIGen, a framework for automatic, debiased, and invariant counterfactual generation under general interventions, including high-dimensional interventions and outcomes. ADIGen combines Riesz regression to avoid unstable density-ratio estimation, causal invariance to improve generalization under distribution shift, and orthogonal statistical learning to obtain doubly robust guarantees against nuisance model misspecification. We provide excess-risk bounds showing that ADIGen controls counterfactual risk under general interventions, with a product-bias nuisance remainder and an invariant risk bound across environments.", "date": "2026-06-05", "categories": []}
{"id": "2412.11800", "title": "Scalable Temporal Anomaly Causality Discovery in Large Systems: Achieving Computational Efficiency with Binary Anomaly Flag Data", "abstract": "Extracting anomaly causality facilitates diagnostics once monitoring systems detect system faults. Identifying anomaly causes in large systems involves investigating a broader set of monitoring variables across multiple subsystems. However, learning graphical causal models (GCMs) comes with a significant computational burden that restrains the applicability of most existing methods in real-time and large-scale deployments. In addition, modern monitoring applications for large systems often generate large amounts of binary alarm flags, and the distinct characteristics of binary anomaly data -- the meaning of state transition and data sparsity -- challenge existing causality learning mechanisms. This study proposes an anomaly causal discovery approach (AnomalyCD), addressing the accuracy and computational challenges of generating GCMs from temporal binary flag datasets. The AnomalyCD presents several strategies, such as anomaly data-aware causality testing, sparse data and prior link compression, and edge pruning adjustment approaches. We validate the performance of the approach on two datasets: monitoring sensor data from the readout-box system of the Compact Muon Solenoid experiment at CERN, and a public dataset from an information technology monitoring system. The results on temporal GCMs demonstrate a considerable reduction of computation overhead and a moderate enhancement of accuracy on the binary anomaly datasets. Code: https://github.com/muleina/AnomalyCD .", "date": "2024-12-16", "categories": []}
{"id": "2507.12257", "title": "Robust Causal Discovery in Real-World Time Series with Power-Laws", "abstract": "Exploring causal relationships in stochastic time series is a challenging yet crucial task with a vast range of applications, including finance, economics, neuroscience, and climate science. Many algorithms for Causal Discovery (CD) have been proposed; however, they often exhibit a high sensitivity to noise, resulting in spurious causal inferences in real data. In this paper, we observe that the frequency spectra of many real-world time series follow a power-law distribution, notably due to an inherent self-organizing behavior. Leveraging this insight, we build a robust CD method based on the extraction of power-law spectral features that amplify genuine causal signals. Our method consistently outperforms state-of-the-art alternatives on both synthetic benchmarks and real-world datasets with known causal structures, demonstrating its robustness and practical relevance.", "date": "2025-07-16", "categories": []}
{"id": "2603.14169", "title": "Beyond Means: Topological Causal Effects under Persistent-Homology Ignorability", "abstract": "Average treatment effects (ATE) and conditional average treatment effects (CATE) are foundational causal estimands, but they target changes in expected outcomes and can miss treatment-induced changes in the shape of outcome distributions. A canonical failure mode occurs when control outcomes are unimodal, treated outcomes become bimodal, and both distributions have the same mean. In such cases mean-based causal estimands are zero even though the geometry and topology of the outcome law change substantially. This paper develops a topological causal framework based on persistent homology. We formalize a persistent-homology ignorability condition, define topological analogues of CATE and ATE, and prove that these estimands are identifiable up to an explicit error bound under approximate topological ignorability. We also clarify a subtle but important point: a marginal persistence-diagram effect is not identified from conditional topological ignorability alone because persistent homology does not in general commute with mixtures over covariates. To preserve the original intuition while ensuring scientific correctness, we retain the marginal effect as a motivating quantity, but place the mathematically sound conditional estimands at the center of the theory. A synthetic experiment with mean-preserving topology change shows that mean-based causal estimands remain near zero while the proposed topological effect increases sharply and remains recoverable after adjustment for confounding.", "date": "2026-03-15", "categories": []}
{"id": "2603.16475", "title": "Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures", "abstract": "In schema-guided reasoning (SGR) pipelines, LLMs produce explicit intermediate structures -- rubrics, checklists, or verification queries -- before committing to a final decision. SGR is increasingly adopted because it promises controllability: practitioners expect to inspect, edit, and override these structures to steer the outcome. But does the promise hold? We introduce a causal evaluation protocol to measure it: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across 12 models and 4 benchmarks, models appear self-consistent with their own intermediate structures but fail to update predictions after intervention -- revealing that apparent faithfulness is fragile once the intermediate structure changes. When derivation of the final decision from the structure is delegated to an external tool, this fragility largely disappears; stronger prompting yields only limited improvements, while preference optimization substantially improves intervention faithfulness. Overall, intermediate structures in schema-guided pipelines function as influential context rather than stable causal mediators.", "date": "2026-03-17", "categories": []}
{"id": "2603.20980", "title": "From Causal Discovery to Dynamic Causal Inference in Neural Time Series", "abstract": "Time-varying causal models provide a powerful framework for studying dynamic scientific systems, yet most existing approaches assume that the underlying causal network is known a priori - an assumption rarely satisfied in real-world domains where causal structure is uncertain, evolving, or only indirectly observable. This limits the applicability of dynamic causal inference in many scientific settings. We propose Dynamic Causal Network Autoregression (DCNAR), a two-stage neural causal modeling framework that integrates data-driven causal discovery with time-varying causal inference. In the first stage, a neural autoregressive causal discovery model learns a sparse directed causal network from multivariate time series. In the second stage, this learned structure is used as a structural prior for a time-varying neural network autoregression, enabling dynamic estimation of causal influence without requiring pre-specified network structure. We evaluate the scientific validity of DCNAR using behavioral diagnostics that assess causal necessity, temporal stability, and sensitivity to structural change, rather than predictive accuracy alone. Experiments on multi-country panel time-series data demonstrate that learned causal networks yield more stable and behaviorally meaningful dynamic causal inferences than coefficient-based or structure-free alternatives, even when forecasting performance is comparable. These results position DCNAR as a general framework for using AI as a scientific instrument for dynamic causal reasoning under structural uncertainty.", "date": "2026-03-21", "categories": []}
{"id": "2606.05413", "title": "CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting", "abstract": "As urban environments continue to evolve rapidly, accurately modeling the dynamic behaviour of Points of Interest is essential for supporting data-driven urban planning and commercial decision-making. While recent advancements in spatio-temporal graph learning have improved POI forecasting, most methods rely on proximity-based graphs and correlation-driven modeling, which overlook the functional dependencies between POIs and fail to capture the causal effects of urban interventions. In this paper, we introduce a novel research problem -- cold-start POI check-in forecasting, which aims to predict the future check-in pattern of a newly introduced POI, by modeling its temporal evolution and functional interactions with nearby POIs in a structured urban spatial context. To address these challenges, we propose CausalPOI, a spatio-temporal graph-based causal representation learning framework. CausalPOI leverages Spatio-Temporal Functional Interaction Graph to model semantic and spatial relationships between POIs, and constructs structurally aligned treatment and control graphs to simulate factual and counterfactual scenarios. Extensive experiments on real-world SafeGraph datasets demonstrate that CausalPOI significantly outperforms state-of-the-art baselines across the board, validating its effectiveness in spatio-temporal forecasting, semantic interaction modeling, and causal effect estimation, providing a more interpretable and actionable foundation for urban intervention analysis. Source code is available at Github.", "date": "2026-06-03", "categories": []}
{"id": "2606.05636", "title": "StableRCA: Robust Graph-Agnostic Mechanism-Level Root Cause Analysis", "abstract": "Root-Cause Analysis (RCA) seeks to identify the variables responsible for abnormal system behavior in complex domains such as manufacturing, cloud computing, and healthcare. Existing approaches face a critical bottleneck: graph-based causal methods can identify intervention targets but typically require a known or accurately estimated causal graph, while graph-free statistical methods either localize marginal anomalies rather than structural causes, or rely on restrictive assumptions about graph structure or functional form. We propose StableRCA, a local mechanism-level RCA framework that avoids global graph discovery by estimating local Markov boundaries and detecting conditional distribution shifts within them. Leveraging the Independent Causal Mechanism principle, we show that intervention targets can be identified with probability converging exponentially in sample size under faithful Markov boundary recovery and non-degenerate mechanism shifts. Experiments on synthetic benchmarks and five real-world datasets demonstrate that StableRCA is robust to graph misspecification, effective under multiple intervention targets, scalable to large systems, and reliable across diverse application domains. Code is available at: https://anonymous.4open.science/r/StableRCA-E362", "date": "2026-06-04", "categories": []}
{"id": "2606.05689", "title": "Causal Modeling of Selection in Evolution", "abstract": "Understanding potential selection in data is crucial for causal discovery; we argue that \"selection\" in common narratives takes two forms, which we term static and evolutionary selection, respectively. Static selection refers to a one-shot filtering process where observed data consist of a subset of the population of interest, as in survey volunteer bias. Evolutionary selection, in contrast, operates through repeated rounds of differential fitness in reproduction, where observed data constitute the latest generation shaped by a historical trajectory, as in immune adaptation, antibiotic resistance, and social norm emergence. Existing methods largely conflate these two forms and rely on an identical graphical model of selection. We show that this model is valid for static settings but fails to characterize data under evolution, yielding false discovery results. To address this, we introduce a new model that specifically characterizes evolutionary selection, and develop a sound and complete procedure for identifying such models from data across one or multiple environments or generations. Experimental results validate the method's ability to uncover the relevant mechanisms underlying evolution from data.", "date": "2026-06-04", "categories": []}
{"id": "2606.05942", "title": "EML-CD: Causal Mechanism Recovery via EML Symbolic Trees in Structure Learning", "abstract": "Neural network (NN)-based nonlinear causal discovery methods recover DAG structure but leave each causal mechanism as a black box. Waxman et al. argued that extracting causal mechanisms from NN weights is ill-posed. We propose EML-CD, a framework that integrates the EML operator (capable of composing elementary functions from a single binary operator) into causal structure learning, with interpretable mechanism recovery as the primary objective. EML-CD represents each edge mechanism as a gated EML binary tree and automatically discovers closed-form causal equations. Analytical Jacobians can be directly computed from the output equations, enabling quantitative understanding of causal effects. On real data (Sachs protein signaling, d=11), EML-CD achieves SHD=11.2 +/- 0.4 (5-seed mean; baselines are single deterministic runs), on par with PC/GES within seed variance and below CAM, while attaching closed-form equations to each detected edge (precision 0.756, recall 0.365). In a controlled bivariate test with known mechanisms, EML-CD recovers 10 of 11 elementary function families faithfully (held-out shape correlation >= 0.96; only high-frequency sine is partial). On a symbolic synthetic benchmark, EML-CD attains a substantially lower and more stable held-out mechanism f-MSE than a fixed SINDy dictionary (mean 3.67 vs. 7644, the latter inflated by catastrophic extrapolation on one seed), although its structure recovery (SHD 14.0) only matches the dictionary and stays below specialized optimizers; on the Causal Chambers light-tunnel subset, a depth-2 model improves F1 over linear OLS-BIC (0.444 vs. 0.273).", "date": "2026-06-04", "categories": []}
{"id": "2606.05966", "title": "Causal Scaffolding for Physical Reasoning: A Benchmark for Causally-Informed Physical World Understanding in VLMs", "abstract": "Understanding and reasoning about the physical world is the foundation of intelligent behavior, yet state-of-the-art vision-language models (VLMs) still fail at causal physical reasoning, often producing plausible but incorrect answers. To address this gap, we introduce CausalPhys, a benchmark of over 3,000 carefully curated video- and image-based questions spanning four domains: Perception, Anticipation, Intervention, and Goal Orientation. Each question is paired with an expert-annotated causal graph capturing object-attribute-event dependencies, enabling interpretable and fine-grained evaluation of causal understanding. Building on this, we formulate a causal-graph-grounded metric that quantitatively measures how well a model's chain-of-thought reasoning aligns with the correct causal relations, moving beyond answer-only accuracy and enabling systematic diagnosis of VLMs' causal reasoning failures. Using this metric, we conduct a comprehensive analysis of leading VLMs, revealing systematic gaps in capturing causal dependencies and underscoring the need for causality-aware learning. To address these limitations, we further propose Causal Rationale-informed Fine-Tuning (CRFT), which explicitly aligns VLM reasoning with causal structures. Extensive experiments demonstrate that CRFT substantially enhances both reasoning accuracy and interpretability across multiple model backbones. By unifying dataset curation, causal evaluation, and causality-informed learning, CausalPhys establishes a strong foundation for advancing modern VLMs toward causally grounded physical reasoning.", "date": "2026-06-04", "categories": []}
{"id": "2606.06288", "title": "Discrete Causal Representations from Heterogeneous Domains: A Bayesian Approach with Social Survey Applications", "abstract": "Causal representation learning aims to infer the high-level latent causal concepts that give rise to observed low-level measurements. This is particularly relevant for heterogeneous data from different environments or domains since distribution shifts often arise through sparse, localized changes in some of the underlying causal mechanisms, while other parts of the generative process remain unchanged. Whereas identifiability of causal representations has been studied extensively, practical uncertainty-aware methods and real-world use cases remain less explored. In this work, we propose a Bayesian approach to learning causal representations from multi-environment data, focusing on the case of discrete causal concepts and unknown multi-node soft interventions. To this end, we translate causal assumptions and interpretability desiderata into suitable priors and parametric choices within a hierarchical model. We then devise an inference scheme based on sequential Monte Carlo sampling to approximate the resulting multimodal posterior. We showcase our approach through case studies on social survey data, where latent causal concepts correspond to cultural values or political opinions, measurements to survey responses, and environments to different countries or states. Our model infers meaningful high-level concepts and plausible causal relations among them, demonstrating its utility for learning causal representations of complex real-world data.", "date": "2026-06-04", "categories": []}
{"id": "2606.06440", "title": "Causal Atlases from Entropic Inference: Bayesian Networks beyond Optimal DAGs", "abstract": "Data-driven causal relationship identification is pertinent to advancing understanding of complex systems both within and beyond science. Bayesian networks offer a probabilistic method for modelling generic causal relationships via directed acyclic graphs (DAGs). However, typical techniques for constructing Bayesian networks rely on optimization, which can be ill-suited for learning causal relationships because the underlying data may admit multiple chains of causation. More data-faithful representations of causal relationships would provide frameworks for constructing multiple causal maps that are consistent with the variability that is inherent in underlying data. Here, we show that entropy-based inference generates atlases of plausible causal relationships that are consistent with underlying data. On simulated noisy data of 2- and 20-node linear structural equation models, we sample a maximum-entropy ensemble of graphs that allow us to quantify the inherent structural ambiguity in underlying causal relationships. Our method shows that \"optimized\" DAGs can contain causal artifacts are not consistent across equivalently accurate topologies.", "date": "2026-06-04", "categories": []}
{"id": "2209.15448", "title": "Blessing from Human-AI Interaction: Super Reinforcement Learning in Confounded Environments", "abstract": "As AI becomes more prevalent throughout society, effective methods of integrating humans and AI systems that leverage their respective strengths and mitigate risk have become an important priority. In this paper, we introduce the paradigm of super policy learning that takes advantage of Human-AI interaction for data driven sequential decision making. This approach utilizes the observed action, either from AI or humans, as input for achieving a stronger oracle in policy learning for the decision maker (humans or AI). In the decision process with unmeasured confounding, the actions taken by past agents can offer valuable insights into undisclosed information. By including this information for the policy search in a novel and legitimate manner, the proposed super policy learning will yield a super-policy that is guaranteed to outperform both the standard optimal policy and the behavior one (e.g., past agents' actions). We call this stronger oracle a blessing from human-AI interaction. Furthermore, to address the issue of unmeasured confounding in finding super-policies using the batch data, a number of nonparametric and causal identifications are established under the framework of proximal causal inference. Building upon on these novel identification results, we develop several super-policy learning algorithms and systematically study their theoretical properties such as finite-sample regret guarantee. Finally, we illustrate the effectiveness of our proposal through extensive simulations and real-world applications.", "date": "2022-09-29", "categories": []}
{"id": "2604.00915", "title": "Orthogonal Learner for Estimating Heterogeneous Long-Term Treatment Effects", "abstract": "Estimation of heterogeneous long-term treatment effects (HLTEs) is relevant for personalized decision-making in marketing, economics, and medicine, where short-term observational datasets are often combined with long-term observational datasets. However, HLTE estimation is challenging due to limited overlap in treatment assignments or in long-term outcomes for certain subpopulations, which can lead to unstable HLTE estimates with large finite-sample variance. To address this challenge, we introduce the LT-O-learners (Long-Term Orthogonal Learners), a set of novel orthogonal learners for HLTE estimation in the canonical HLTE setting with surrogacy. The key idea of our LT-O-learners is to retarget the loss via custom overlap weights that downweight low-overlap samples. We show that the retargeted loss recovers the true HLTE pointwise and satisfies Neyman-orthogonality. We further prove two key theoretical results: (i) The nuisance error enters the error bound only through higher-order terms, which means our learners are robust to nuisance estimation error. (ii) Under a linear function class, the retargeting effectively controls the asymptotic variance of the HLTE estimator via the overlap weights in low-overlap regimes. We conduct experiments on synthetic and real-world datasets to confirm the theoretical properties of our LT-O-learners, particularly robustness in low-overlap regimes. To our knowledge, ours are the first orthogonal learners for HLTE estimation robust to low overlap in long-term settings.", "date": "2026-04-01", "categories": []}
{"id": "2604.14575", "title": "Generative Augmented Inference", "abstract": "Large language models enable inexpensive AI-generated annotations, but using them reliably for causal inference remains challenging. Naively pooling AI and human data induces bias, while existing methods such as Prediction-Powered Inference (PPI; Angelopoulos et al., 2023a) treat AI outputs as proxies of true labels -- an assumption often violated for generative model outputs in practice. We propose Generative Augmented Inference (GAI), a framework that treats AI outputs as general, potentially high-dimensional informative features for learning human labels rather than as surrogates. GAI flexibly models this relationship using nonparametric methods, enabling consistent estimation and valid inference from combined human and AI data. We establish asymptotic normality and show that, under random labeling, GAI strictly improves asymptotic efficiency over human-data-only estimation whenever AI outputs are informative for true labels. Empirical studies on real-world datasets demonstrate that GAI significantly reduces estimation error and improves confidence interval quality across diverse generative data sources relative to human-only and PPI-based estimation.", "date": "2026-04-16", "categories": []}
{"id": "2605.24358", "title": "Treatment Effect Estimation with Differentiated Networked Effect on Graph Data", "abstract": "Estimating individual treatment effect (ITE) from observational graph data is crucial for decision-making in the fields such as commerce and medicine. This task is challenging due to interference, where individual outcomes can be influenced by the treatments and covariates of their neighbors. Existing methods attempt to model such interference for accurate ITE estimation. However, a critical issue is often overlooked: differentiated networked effect (DNE), an effect caused by local networks consisting of neighbors with varying importance and scales. Capturing DNE is vital; otherwise, we will end up with imprecise ITE estimation due to an erroneous characterization of interference, which can result in misguided decisions. To address this challenge, we propose a novel interference modeling mechanism that incorporates two partial attention mechanisms and a message amplifier. The partial attention mechanisms automatically estimate the importance of different neighbors in contributing to interference, while the message amplifier adjusts the results of the interference modeling mechanism based on the scale of neighbors, all of which enables the model to capture DNE. Experiments on three real-world graphs demonstrate that our methods outperform existing approaches for ITE estimation from graph data, which corroborates the importance of explicitly capturing DNE.", "date": "2026-05-23", "categories": []}
{"id": "2606.04822", "title": "Reconciling Causality and Non-Equilibrium Thermodynamics with Hamiltonian Causal Models", "abstract": "Causal modeling of physical temporal phenomena must handle interventions that act along trajectories, nonstationary induced laws, path-dependent effects, and feedback mediated by dynamics, all challenging in standard causal models. We introduce Hamiltonian Causal Models (HCMs), a trajectory-level framework in which observed variables interact with local environments and interventions act as controls of Hamiltonian mechanisms. HCMs separate immutable equations of motion from intervenable mechanisms and define causal effects as discrepancies between interventional path laws. A key motivation for HCMs is their natural interface with non-equilibrium thermodynamics. Entropy production quantifies the irreversibility of a process and is a central causal observable: it is estimable from data and witnesses causal effects along the system's evolution that are invisible to endpoint and cumulative versions of the standard average treatment effect. As in physics, cause and effect are not primitives of the relation between two random variables but arise from the non-invertibility of the thermodynamic arrow. With this, our paper reconciles the language of statistical causal models and non-stationary thermodynamics, offering new tools to describe causality in a wide range of physical systems.", "date": "2026-06-03", "categories": []}
{"id": "2606.05129", "title": "Preserving Data Privacy in Learning Causal Structure with Fully Homomorphic Encryption", "abstract": "Preserving data privacy is an important topic in structural data management and data mining. However, the issue of privacy leakage in distributed causal structure learning is a persistent challenge, especially in cases where data transmission and computation are required. In this paper, we propose a method based on fully homomorphic encryption (FHE) that performs calculations on ciphertexts, keeping data encrypted in transition and computation. Nevertheless, adopting FHE to causal structure learning is challenging due to the high computation cost and limited support on division as well as logarithm operations in FHE. To tackle this challenge, we propose a series of novel techniques including (i) circuit simplification for better efficiency, (ii) approximation of division and logarithm through Newton-Raphson Reciprocal and Taylor expansion, and (iii) a batching technique with SIMD-acceleration to enhance the whole learning process. Additionally, our method can be easily extended beyond FHE by demonstration of its portability to support differential privacy. Empirical results show that our method achieves high consistency and comparable causal structure with the plaintext version in the datasets tested. Last, our method is efficient and practical to complete learning causal structures in tens of minutes even under the privacy protection of FHE.", "date": "2026-06-03", "categories": []}
{"id": "2506.13107", "title": "Honesty in Causal Forests: When It Helps and When It Hurts", "abstract": "Causal forests estimate how treatment effects vary across individuals, guiding personalized interventions in areas like marketing, operations, and public policy. A standard practice is honest estimation: dividing the data into two samples, one to define subgroups and another to estimate treatment effects within them. This is intended to reduce overfitting and is the default in many software packages. But is it the right choice? We show that honest estimation can reduce the accuracy of estimates of individual treatment effects, especially when effect heterogeneity is substantial and datasets are large enough to detect it. The reason is a bias-variance trade-off: honesty lowers the risk of overfitting but increases the risk of underfitting by limiting the data available to detect and model heterogeneity. Across more than 7,000 benchmark datasets, we find that the cost of using honesty by default can be as high as requiring 27% more data to match the performance of models trained without it. Honesty is best understood as a form of regularization. Whether to adopt it should depend on the goals of the application and its empirical performance, not on reflexive default use.", "date": "2025-06-16", "categories": []}
{"id": "2511.04469", "title": "Towards Causal Market Simulators", "abstract": "Market generators using deep generative models have shown promise for synthetic financial data generation, but existing approaches lack causal reasoning capabilities essential for counterfactual analysis and risk assessment. We propose a Time-series Neural Causal Model VAE (TNCM-VAE) that combines variational autoencoders with structural causal models to generate counterfactual financial time series while preserving both temporal dependencies and causal relationships. Our approach enforces causal constraints through directed acyclic graphs in the decoder architecture and employs the causal Wasserstein distance for training. We validate our method on synthetic autoregressive models inspired by the Ornstein-Uhlenbeck process, demonstrating superior performance in counterfactual probability estimation with L1 distances as low as 0.03-0.10 compared to ground truth. The model enables financial stress testing, scenario analysis, and enhanced backtesting by generating plausible counterfactual market trajectories that respect underlying causal mechanisms.", "date": "2025-11-06", "categories": []}
{"id": "2511.05050", "title": "Estimating Bidirectional Causal Effects with Large Scale Online Kernel Learning", "abstract": "In this study, a scalable online kernel learning framework is proposed for estimating bidirectional causal effects in systems characterized by mutual dependence and heteroskedasticity. Traditional causal inference often focuses on unidirectional effects, overlooking the common bidirectional relationships in real-world phenomena. Building on heteroskedasticity-based identification, the proposed method integrates a quasi-maximum likelihood estimator for simultaneous equation models with large scale online kernel learning. It employs random Fourier feature approximations to flexibly model nonlinear conditional means and variances, while an adaptive online gradient descent algorithm ensures computational efficiency for streaming and high-dimensional data. Results from extensive simulations demonstrate that the proposed method achieves superior accuracy and stability than single equation and polynomial approximation baselines, exhibiting lower bias and root mean squared error across various data-generating processes. These results confirm that the proposed approach effectively captures complex bidirectional causal effects with near-linear computational scaling. By combining econometric identification with modern machine learning techniques, the proposed framework offers a practical, scalable, and theoretically grounded solution for large scale causal inference in natural/social science, policy making, business, and industrial applications.", "date": "2025-11-07", "categories": []}
{"id": "2602.01483", "title": "Causal Preference Elicitation", "abstract": "We propose causal preference elicitation, a Bayesian framework for expert-in-the-loop causal discovery that actively queries local edge relations to concentrate a posterior over directed acyclic graphs (DAGs). From any black-box observational posterior, we model noisy expert judgments with a three-way likelihood over edge existence and direction. Posterior inference uses a flexible particle approximation, and queries are selected by an efficient expected information gain criterion on the expert's categorical response. Experiments on synthetic graphs, protein signaling data, and a human gene perturbation benchmark show faster posterior concentration and improved recovery of directed effects under tight query budgets.", "date": "2026-02-01", "categories": []}
{"id": "2606.01184", "title": "Topological Ignorability for Structural Causal Effects Beyond Means", "abstract": "Many interventions alter the structure of an outcome distribution rather than its mean: they can split a population into disconnected regimes, create loops or holes, generate branches, or reorganize an outcome cloud while leaving the average response nearly unchanged. In such settings, mean-based causal estimands such as the average treatment effect may miss important structural effects.\n  We introduce topological-geometrical causal metrics based on summaries of interventional outcome laws, including density-superlevel Betti summaries, Euler signatures, and persistent-homology summaries. These metrics quantify structural differences between treated and untreated outcome laws beyond averages. We also study the assumptions needed for causal interpretation. We introduce topological ignorability, a topological analogue of conditional ignorability that requires invariance of the chosen structural feature rather than the full counterfactual distribution. When the chosen summary is injective, this condition coincides with weak ignorability; for noninjective summaries, it can identify the structural feature of interest without identifying the full interventional law.\n  We define a covariate-standardized topological-geometrical causal effect and develop practical estimators. We validate the framework in two hidden-confounding benchmarks: a fully synthetic exact benchmark and a real-covariate semi-synthetic benchmark using Wisconsin breast-cancer covariates. In both, weak ignorability fails and balancing observed covariates nearly eliminates standardized mean differences, yet the coordinate-mean average treatment effect remains biased. By contrast, selected finite density-superlevel Betti and Euler contrasts remain stable across oracle, observational, and weighted analyses.", "date": "2026-05-31", "categories": []}
{"id": "2606.03227", "title": "Learning Temporal Causal Structure via Smooth Differentiable Optimization", "abstract": "Causal discovery with instantaneous effects in multivariate time series is challenging, as the instantaneous structure must be acyclic. Prior methods enforce this by either separating instantaneous and lagged estimation into multi-stage pipelines or imposing algebraic acyclicity constraints via complex augmented Lagrangian optimization, both of which incur high computational cost. In this work, we propose a different approach: we learn a differentiable permutation of variables using the Gumbel--Sinkhorn operator and triangularize the instantaneous coefficient matrix of a Structural Vector Autoregressive (SVAR) model in the learned order. This converts acyclicity from a hard constraint into a parameterization and keeps it valid throughout optimization. In doing so, our method enables unified, continuous optimization with gradient-based learning, leading to improved efficiency in time--series causal discovery. Across three real-world benchmarks, our method achieves the best overall performance compared with 12 baselines in both discovery accuracy and efficiency. On the large-scale benchmark, it further demonstrates strong scalability, achieving more than a 6x speedup over competing methods.", "date": "2026-06-02", "categories": []}
{"id": "2606.03251", "title": "Do Real-World Datasets Contain Natural Experiments? An Empirical Study Using Causal Feature Selection", "abstract": "In nature, events that affect some individuals or groups but not others constitute an implicit intervention and are known as natural experiments. For example, the COVID-19 pandemic was an intervention by the coronavirus on the sub-population infected with COVID. We ask, do natural experiments occur in existing real-world datasets? If yes, how should we treat them? To detect natural experiments in data, we use causal discovery to recover the underlying causal graph and perform feature selection based on causal links. If downstream performance improves by treating the data as interventional rather than observational, we argue that this suggests the dataset contains natural experiments. We first validate this hypothesis by simulating datasets with and without natural experiments using synthetic graphs. We then perform a systematic empirical evaluation on a large suite of real-world datasets. Our results indicate that real-world datasets do contain natural experiments and we can take advantage of those natural experiments to improve model performance using causal inference. Our work represents the initial foray into this area, offering a preliminary exploration within a limited scope.", "date": "2026-06-02", "categories": []}
{"id": "2606.03332", "title": "Tailoring Strictly Proper Scoring Rules for Downstream Tasks: An Application to Causal Inference", "abstract": "Probabilistic models are typically trained using task-agnostic objectives like log-loss, which can lead to significant errors in downstream estimation. This disconnect is especially critical in Inverse Probability Weighting (IPW) for causal inference, where propensity score errors near $0$ and $1$ often lead to high bias and variance. We propose a principled framework for deriving task-specific strictly proper scoring rules by matching the local curvature of the downstream error metric. We apply this to the Average Treatment Effect (ATE) estimation, deriving a closed-form loss and its corresponding canonical probability mapping that can be readily integrated with any model like a neural network or a gradient boosting algorithm. Extensive evaluations on causal inference benchmarks demonstrate that our tailored objective consistently outperforms standard likelihood-based and covariate-balancing approaches.", "date": "2026-06-02", "categories": []}
{"id": "2606.03602", "title": "CauTion: Knowing When to Trust LLMs for Ensemble Causal Discovery", "abstract": "Causal discovery from observational data remains challenging due to the fundamental limitations of purely statistical methods, such as statistical distinguishability within equivalence classes and sensitivity to finite sample sizes. While large language models (LLMs) offer a promising source of domain knowledge to complement statistical inference, existing LLM-augmented methods are vulnerable to LLM errors and incur high token costs. Moreover, reliance on a single data-centric algorithm can make results sensitive to algorithm-specific biases. To address these limitations, we propose CauTion, a framework that reliably integrates LLM domain knowledge into an ensemble of statistical causal discovery algorithms through consensus filtering and LLM reliability estimation. CauTion proceeds in three stages. First, an algorithm ensemble utilizes a consensus voting to resolve up to 96% of edges on which algorithms agree, achieving near-perfect accuracy on the filtered consensus edges. Second, a trust-calibrated arbitration mechanism estimates the relative reliability of the LLM and the algorithms via an annotation-free trust calibration procedure, which is then utilized to govern a trust-weighted voting process that restricts LLM arbitration exclusively to edges with unreliable algorithmic evidence. Third, a cycle repair step is applied to guarantee the final causal graph is validly acyclic. Experiments on six datasets demonstrate that CauTion consistently outperforms both data-centric and LLM-augmented baselines, with larger gains on larger graphs and strong robustness to LLM errors. Code is available at https://github.com/OpenCausaLab/CauTion.", "date": "2026-06-02", "categories": []}
{"id": "2606.03719", "title": "Unveiling the Structure of Do-Calculus Reasoning via Derivation Graphs", "abstract": "The do-calculus defines a general system of inference for interventional queries, allowing causal quantities to be transformed through successive applications of its rules. This process induces a rich space of equivalent interventional expressions, but combining and ordering these rules remains challenging. In this work, we introduce derivation graphs, which represent how do-calculus rules are applied and combined, and characterize the full space of observational and interventional probabilities which are equivalent under the do-calculus. The structure of these graphs yields a simple procedure that uses at most four applications of do-calculus rules. Finally, we show how applying identification algorithms to equivalent causal queries produces multiple valid estimands for the same causal quantity, eventually yielding more efficient estimators.", "date": "2026-06-02", "categories": []}
{"id": "2606.03878", "title": "Privacy-Robust Incrementality Measurement for Advertising Systems under Signal Loss", "abstract": "Advertising platforms use randomized lift tests to measure incrementality, but privacy-preserving reporting systems degrade the observed signal through match-rate loss, linkability loss, attribution-window loss, aggregation-threshold suppression, randomized reporting noise, and segment-heterogeneous signal loss. This paper formulates privacy-constrained advertising measurement as a robust causal decision problem under the mentioned signal losses. Given a randomized experiment and an ambiguity set for privacy-induced degradation, the framework projects the observation-compatible fiber of clean/unfiltered experimental worlds onto the incrementality functional and returns certified, rejected, and unresolved decisions. The main result gives a sharp decision frontier. Reports outside the frontier support uniformly valid certification or rejection, whereas reports inside it contain too little information for any method to uniformly distinguish above-threshold incrementality from non-incrementality. Supporting results give finite-sample certification, sample-complexity guarantees, a minimax lower bound showing that signal loss reduces effective information, and a reporting-granularity tradeoff. On 2.0M Criteo Uplift rows and the 64K-row Hillstrom email experiment, clean conversion lift is positive in both datasets, with lifts 0.00112 and 0.00495, respectively. Population certification survives mild degradation in Criteo and severe degradation in Hillstrom, while all considered finite-sample stress settings in both datasets remain unresolved after simultaneous uncertainty and reporting noise are included. Overall, the research contributes a decision-theoretic layer for privacy-aware incrementality measurement whose output is the strongest causal-claim justified by degraded ads signals.", "date": "2026-06-02", "categories": []}
{"id": "2410.14483", "title": "Interventional Processes for Causal Uncertainty Quantification", "abstract": "Reliable uncertainty quantification for causal effects is crucial in high-stakes applications, but remains challenging when the target is an entire function rather than a scalar estimand. In this work, we introduce a GP-based approach for uncertainty quantification of interventional functions. The central idea is to build on recent work representing interventional functions as an inner-product of observational functions in a reproducing kernel Hilbert space (RKHS), by constructing appropriate GP priors for such functions and inferring posteriors from observational data. Our approach yields closed-form posterior moments and tractable training and inference, while avoiding pathologies of previous GP prior constructions for RKHS functions. We further derive a practical procedure for posterior coverage calibration. Across synthetic benchmarks, causal Bayesian optimization tasks, and a large-scale real dataset, our method improves uncertainty quantification while remaining competitive in causal effect estimation.", "date": "2024-10-18", "categories": []}
{"id": "2508.02812", "title": "Evaluating and Learning Robust Bandit Policies Under Uncertain Causal Mechanisms", "abstract": "Causal graphical models can encode large amounts structural knowledge, both from the background knowledge of domain experts and the structural knowledge discovered from randomized experiments or observational data. However, though we may know the general structure of causal relationships, we often do not know the exact causal mechanisms. In this work, we propose a causal multi-armed bandit evaluation and learning algorithm that can reason effectively despite uncertainty over conditional probability distributions. Further, we show how conditional independence testing can be used to choose variables for modeling. We find that the structural equation model (SEM) approach gives more accurate evaluations compared to traditional approaches, particularly as the range of possible causal mechanisms grows. Further, the SEM approach learns low-variance policies, and it learns an optimal policy, assuming the model is sufficiently well-specified. Traditional approaches can converge to local extrema or fail to converge at all.", "date": "2025-08-04", "categories": []}
{"id": "2602.12972", "title": "Jointly Optimizing Debiased CTR and Uplift for Coupons Marketing: A Unified Causal Framework", "abstract": "In online advertising, marketing interventions such as coupons introduce significant confounding bias into Click-Through Rate (CTR) prediction. Observed clicks reflect a mixture of users' intrinsic preferences and the uplift induced by these interventions. This causes conventional models to miscalibrate base CTRs, which distorts downstream ranking and billing decisions. Furthermore, marketing interventions often operate as multi-valued treatments with varying magnitudes, introducing additional complexity to CTR prediction.\n  To address these issues, we propose the \\textbf{Uni}fied \\textbf{M}ulti-\\textbf{V}alued \\textbf{T}reatment Network (UniMVT). Specifically, UniMVT disentangles confounding factors from treatment-sensitive representations, enabling a full-space counterfactual inference module to jointly reconstruct the debiased base CTR and intensity-response curves. To handle the complexity of multi-valued treatments, UniMVT employs an auxiliary intensity estimation task to capture treatment propensities and devise a unit uplift objective that normalizes the intervention effect. This ensures comparable estimation across the continuous coupon-value spectrum. UniMVT simultaneously achieves debiased CTR prediction for accurate system calibration and precise uplift estimation for incentive allocation. Extensive experiments on synthetic and industrial datasets demonstrate UniMVT's superiority in both predictive accuracy and calibration. Furthermore, real-world A/B tests confirm that UniMVT significantly improves business metrics through more effective coupon distribution.", "date": "2026-02-13", "categories": []}
{"id": "2603.12037", "title": "Frequentist Consistency of Prior-Data Fitted Networks for Causal Inference", "abstract": "Foundation models based on prior-data fitted networks (PFNs) have shown strong empirical performance in causal inference by framing the task as an in-context learning problem. However, it is unclear whether PFN-based causal estimators provide uncertainty quantification that is consistent with classical frequentist estimators. In this work, we address this gap by analyzing the frequentist consistency of PFN-based estimators for the average treatment effect (ATE). (1) We show that existing PFNs, when interpreted as Bayesian ATE estimators, can exhibit prior-induced confounding bias: the prior is not asymptotically overwritten by data, which, in turn, prevents frequentist consistency. (2) As a remedy, we suggest employing a calibration procedure based on a one-step posterior correction (OSPC). We show that the OSPC helps to restore frequentist consistency and can yield a semi-parametric Bernstein-von Mises theorem for calibrated PFNs (i.e., both the calibrated PFN-based estimators and the classical semi-parametric efficient estimators converge in distribution with growing data size). (3) Finally, we implement OSPC through tailoring martingale posteriors on top of the PFNs. In this way, we are able to recover functional nuisance posteriors from PFNs, required by the OSPC. In multiple (semi-)synthetic experiments, PFNs calibrated with our martingale posterior OSPC produce ATE uncertainty that (i) asymptotically matches frequentist uncertainty and (ii) is well calibrated in finite samples in comparison to other Bayesian ATE estimators.", "date": "2026-03-12", "categories": []}
{"id": "2605.02640", "title": "Trustworthy AI Suffers from Invariance Conflicts and Causality is The Solution", "abstract": "As artificial intelligence (AI), including machine learning (ML) models and foundation models (FMs), are increasingly deployed in high-stakes domains, ensuring their trustworthiness has become a central challenge. However, the core trustworthy AI objectives, such as fairness, robustness, privacy, and explainability, are hard to achieve simultaneously, especially while preserving utility. This position paper argues that causality is necessary to understand and balance trade-offs in performance and multiple objectives of trustworthy AI. We ground our arguments in re-interpreting trustworthy AI trade-offs as incompatible invariance requirements under different changes to the data-generating process. We then illustrate this argument through case-study analyses from the literature and a stylized synthetic-data simulation, showing that causality provides a unifying framework for understanding how trade-offs in trustworthy AI arise and how they can be softened or resolved through selective invariance. This perspective applies to both classical ML models and large-scale FMs. Finally, we outline open challenges and opportunities for using causality to build both trustworthy and high-performing AI.", "date": "2026-05-04", "categories": []}
{"id": "2605.13430", "title": "Towards a holistic understanding of Selection Bias for Causal Effect Identification", "abstract": "Selection bias is pervasive in observational studies. For example, large scale biobanks data can exhibit ``healthy volunteer bias'' when respondents are healthier and of higher socio-economic status than the population they are meant to represent. Recovering causal effects from such sub-population is an important problem in causal inference, as estimating average treatment effects (ATE) from selected populations can result in a severely biased estimate of the ATE from the whole population. In this paper, we investigate the identifiability of the ATE under selection bias. We provide necessary and sufficient conditions for ATE identifiability, leveraging weak assumptions on probability classes to characterize propensity score and selection probability. Compared to previous works, our results extend existing graphical identifiability criteria and offer a more comprehensive understanding of causal effect identification with strictly weaker conditions in the presence of selection bias.", "date": "2026-05-13", "categories": []}
{"id": "2606.00278", "title": "Evaluating Bivariate Causal Statements Based on Mutual Compatibility", "abstract": "For many real-world systems, causal ground truth is difficult to obtain, making claims about causal effects hard to assess. We develop methods for evaluating collections of $\\binom{n}{2}$ bivariate causal statements over a set of $n$ variables. In the setting of acyclic linear statements, any such collection can be extended to a unique multivariate causal model, but we argue that this induced model is implausible if it imposes substantial additional confounding to explain observed correlations. We introduce a compatibility score that quantifies this notion of plausibility, notably without relying on the faithfulness assumption. Additionally, we define an incompatibility score for purely graphical bivariate causal statements, based on global consistency constraints that are derived from acyclicity and faithfulness assumptions. We give theoretical and empirical evidence that both scores can successfully distinguish correct from incorrect causal statements in generic settings. Moreover, we demonstrate the practical applicability of our methods by analyzing causal claims made by large language models. Our work aims to provide a foundation for assessing the reliability of causal information derived from human experts or artificial intelligence in settings where alternative forms of validation are unavailable.", "date": "2026-05-29", "categories": []}
{"id": "2606.00322", "title": "Perturbative methods for non-parametric instrumental variable", "abstract": "We introduce a perturbative approach for nonparametric instrumental variable (NPIV) estimation. By drawing inspiration from perturbation theory in physics, we extend standard kernel ridge methods with systematic higher perturbation order corrections that significantly improve estimation accuracy. Spectrally, the perturbation introduces mixing between different eigenmodes of the expectation integral operator, which becomes especially useful when the integral equation is ill-defined. One source for such ill-definedness can be the curse of dimensionality. Our method performs across various dimensionality regimes, particularly when the dimensionality parameter $\\beta$ which is defined through the number of samples $n$ and dimension $d$ as $n^\\beta = d$, becomes large. Experimental results show that our first-order perturbative corrections can reduce prediction error by up to 99\\% in high-dimensional ill-defined cases ($\\beta > 0.7$) compared to standard ridge regression approaches. The performance improvement is maintained across a wide range of dimensions, with the advantage becoming more pronounced as dimensionality increases.", "date": "2026-05-29", "categories": []}
{"id": "2606.00674", "title": "The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs", "abstract": "Large Language Models (LLMs) aligned via outcome-based Reinforcement Learning (RL) frequently exhibit a critical failure mode: they achieve high performance on in-distribution benchmarks while demonstrating brittle reasoning capabilities on out-of-distribution (OOD) tasks. We term this phenomenon Reward-Induced Manifold Collapse. We establish a theoretical framework bridging Structural Causal Models (SCM) and the Information Bottleneck (IB) principle to explain this paradox. We define reasoning as a high-complexity causal process and shortcut learning as the exploitation of low-complexity spurious correlations. Under the implicit inductive bias of Stochastic Gradient Descent (SGD), models optimized for outcome rewards are biased toward shortcut solutions whenever the training distribution allows for a ``Markovian Screening'' of the true causal mechanism. We derive a new generalization bound based on Semantic Coverage Measure ($\\eta$) rather than sample size, showing why data scaling on homogeneous distributions may fail to correct reasoning flaws. We also show that Process Reward Models (PRMs) function as Topological Filters, enforcing step-wise mutual information constraints that render the low-complexity shortcut manifold inadmissible. These results provide a mathematical grounding for the role of process supervision beyond simple credit assignment.", "date": "2026-05-30", "categories": []}
{"id": "2606.00754", "title": "Causal Density Functions", "abstract": "We introduce causal density functions: Radon-Nikodym derivatives that compare interventional laws to observational laws and therefore act as local density ratios for causal effects. Whereas many causal-strength measures compare whole distributions after graph surgery, causal density functions provide a pointwise change-of-measure object that can be estimated, calibrated, and used to score directed influence. The basic identity\n  \\[\n  \\mathbb{E}_{\\mathrm{do}}[f(Y)]\n  =\n  \\mathbb{E}_{\\mathrm{obs}}\\!\\left[f(Y)\\rho(X,Y)\\right]\n  \\]\n  makes causal density directly testable: if the estimated density ratio is correct, observational expectations reweighted by $\\rho$ reproduce interventional expectations. We derive practical estimators for do-curves and directed edge scores, relate the construction to Radon-Nikodym/Kan semantics for conditioning and intervention, and evaluate the resulting estimators on synthetic and real perturbation benchmarks.", "date": "2026-05-30", "categories": []}
{"id": "2606.01457", "title": "Transferring Information Across Interventions in Causal Bayesian Optimization", "abstract": "Bayesian optimization is a popular way to optimize expensive systems, where every experiment, simulation, or intervention costs time or money. In its standard form, it treats the variables we control as plain inputs to a black box and cannot tell apart mere correlation from a real cause and effect. Causal Bayesian optimization closes part of this gap by using a known causal graph together with observational data to decide which variables are worth intervening on. Existing methods, however, learn the effect of each possible intervention almost in isolation, even though in a causal system these effects usually share the same underlying mechanisms. We propose graph-coupled causal Bayesian optimization, which ties the different intervention effects together through the uncertainty we have about a small set of shared causal parameters. The result is a causal kernel that lets evidence collected from one intervention improve our estimate of related interventions. For identifiable linear Gaussian causal models, we show that this kernel has low rank, bounded by the number of shared parameters rather than by the size of the intervention menu. This in turn yields an information-gain bound that grows only logarithmically in the optimization horizon, and a regret bound that cleanly separates three sources of error: optimization, causal estimation, and the choice of which intervention sets to consider. We also describe nonlinear and adaptive extensions. Across theory-aligned Gaussian systems, shared-mechanism stress tests, and standard causal optimization benchmarks, the method keeps the benefits of causal Bayesian optimization while transferring information across related interventions, with the clearest gains when direct interventions on the target's parents are unavailable and sparse interventional data must be reused across a large family of candidate interventions.", "date": "2026-05-31", "categories": []}
{"id": "2606.01539", "title": "Scalable Counterfactual Risk Estimation for Rare Events in Longitudinal Data", "abstract": "Estimating the causal effect of time-varying treatments on survival outcomes in large observational studies is computationally demanding, particularly when outcomes are rare. While g-formula-based methods such as the iterative conditional expectation (ICE) estimator provide a principled framework for longitudinal causal inference, they become computationally expensive, especially when bootstrap-based variance estimation is required. In addition, outcome rarity at each time point induces severe class imbalance, leading to instability and convergence issues in logistic regression and related models. To address these challenges, we propose a principled subsampling and reweighting strategy for longitudinal survival data that can be applied to a range of existing causal effect estimators in this setting, including the ICE estimator. The proposed method substantially reduces computational burden while preserving consistency and improving estimation stability in rare-outcome settings. We evaluate the method through simulations and validate it using a large-scale EHR cohort study on social and behavioral determinants of health (SBDH) and suicide risk, demonstrating its effectiveness for modeling rare outcomes in longitudinal data.", "date": "2026-06-01", "categories": []}
{"id": "2606.01789", "title": "Consistency evaluation of benchmarks used for causal discovery", "abstract": "In graphical causal model, causal discovery aims to construct a causal graph based on numerical data and domain knowledge in plain text. However, the evaluation of causal discovery methods remains a challenge in the area as the progress of domain researches often makes benchmark causal graphs contain mis-aligned knowledge. This problem especially affects the evaluation of large language model (LLM) based causal discovery methods as they are sensitive to the new discoveries in the literature. This work is the first to systematically study the quality of benchmark causal graphs. Specifically, we design a pipeline that automatically retrieves relevant research papers from scientific databases, and prompts LLMs to check the consistency between the benchmark causal graphs and domain research papers. We evaluate 11 popular real-world benchmarks, for which our pipeline in total proceeds 38,081 domain papers. Our results show that popular benchmarks vary significantly in their consistency with domain research, with clear implications for causal discovery research.", "date": "2026-06-01", "categories": []}
{"id": "2501.02672", "title": "Re-examining Granger Causality with Causal Bayesian Networks and Reichenbachs Principles", "abstract": "Characterising cause-effect relationships in complex systems is fundamental to understanding their underlying mechanisms. Granger causality (GC) remains a widely used computational tool for identifying causal relationships in time series data. However, like other causal discovery methods, GC has limitations and has been criticised for lacking a rigorous causal foundation. In this work, we present a fix to this criticism by reinterpreting GC through the lenses of Reichenbach's principles and causal Bayesian networks. This reinterpretation was implemented as an algorithm we call causalized Granger causality (c-GC). We demonstrate, both theoretically and graphically, that this reformulation endows GC with a robust causal interpretation under specific assumptions. c-GC yields satisfactory results on synthetic data, offering a more principled framework for causal discovery in observational datasets.", "date": "2025-01-05", "categories": []}
{"id": "2501.12500", "title": "Learning General Causal Structures with Hidden Dynamic Process for Climate Analysis", "abstract": "Understanding climate dynamics requires going beyond correlations in observational data to uncover the underlying causal process. Latent drivers such as atmospheric processes play a central role in temporal dynamics, while direct causal influences also exist among geographically proximate observed variables. Traditional Causal Representation Learning (CRL) typically focuses on latent factors but overlooks such observable-to-observable causal relations, which limits its applicability to climate analysis. In this paper, we introduce a unified framework that jointly uncovers (i) causal relations among observed variables and (ii) latent driving forces together with their interactions. We establish conditions under which both the hidden dynamic process and the causal structure among observed variables are simultaneously identifiable from time-series data, and our guarantees continue to hold in the nonparametric setting through contextual information that recovers latent variables and causal relations. Building on these insights, we propose CaDRe (Causal Discovery and Representation learning), a time-series generative model with structural constraints that integrates CRL and causal discovery. Experiments on synthetic datasets validate our theoretical results. On real-world climate datasets, CaDRe delivers competitive forecasting accuracy and recovers visualized causal graphs aligned with domain expertise, thereby offering interpretable insights into climate systems. Code is available at https://github.com/MinghaoFu/CaDRe.", "date": "2025-01-21", "categories": []}
{"id": "2506.11653", "title": "DISCO: Mitigating Bias in Deep Learning with Conditional Distance Correlation", "abstract": "Dataset bias often leads deep learning models to exploit spurious correlations instead of task-relevant signals. We introduce the Standard Anti-Causal Model (SAM), a unifying causal framework that characterizes bias mechanisms and yields a conditional independence criterion for causal stability. Building on this theory, we propose DISCO$_m$ and sDISCO, efficient and scalable estimators of conditional distance correlation that enable independence regularization in gradient-based models. Across six diverse datasets, our methods consistently outperform or are competitive in existing observed bias mitigation approaches, while requiring fewer hyperparameters and scaling seamlessly to multi-bias scenarios. This work bridges causal theory and practical deep learning, providing both a principled foundation and effective tools for robust prediction. Source Code: https://github.com/yakamoz5/DISCO.", "date": "2025-06-13", "categories": []}
{"id": "2512.00919", "title": "Outcome-Aware Spectral Feature Learning for Instrumental Variable Regression", "abstract": "We address the problem of causal effect estimation in the presence of hidden confounders using nonparametric instrumental variable (IV) regression. An established approach is to use estimators based on learned spectral features, that is, features spanning the top singular subspaces of the operator linking treatments to instruments. While powerful, such features are agnostic to the outcome variable. Consequently, the method can fail when the true causal function is poorly represented by these dominant singular functions. To mitigate, we introduce Augmented Spectral Feature Learning, a framework that makes the feature learning process outcome-aware. Our method learns features by minimizing a novel contrastive loss derived from an augmented operator that incorporates information from the outcome. By learning these task-specific features, our approach remains effective even under spectral misalignment. We provide a theoretical analysis of this framework and validate our approach on challenging benchmarks.", "date": "2025-11-30", "categories": []}
{"id": "2512.23626", "title": "Regret-Based Federated Causal Discovery with Unknown Interventions", "abstract": "Most causal discovery methods recover a completed partially directed acyclic graph representing a Markov equivalence class from observational data. Recent work has extended these methods to federated settings to address data decentralization and privacy constraints, but often under idealized assumptions that all clients share the same causal model. Such assumptions are unrealistic in practice, as client-specific policies or protocols, for example, across hospitals, naturally induce heterogeneous and unknown interventions. In this work, we address federated causal discovery under unknown client-level interventions. We propose I-PERI, a novel federated algorithm that first recovers the CPDAG of the union of client graphs and then orients additional edges by exploiting structural differences induced by interventions across clients. This yields a tighter equivalence class, which we call the $\\mathbf{\\Phi}$-Markov Equivalence Class, represented by the $\\mathbf{\\Phi}$-CPDAG. We provide theoretical guarantees on the convergence of I-PERI, as well as on its privacy-preserving properties, and present empirical evaluations on synthetic data demonstrating the effectiveness of the proposed algorithm.", "date": "2025-12-29", "categories": []}
{"id": "2605.30901", "title": "Density-Guided Robust Counterfactual Explanations on Tabular Data under Model Multiplicity", "abstract": "Counterfactual explanations (CEs) are essential for actionable recourse, yet their reliability is often compromised in low-density regions, where classifiers exhibit high variance. Unlike existing methods that rely on expensive ensemble intersections to define stability, we propose \\textit{DensityFlow}, a generative framework that constructs robust CEs by adhering to the high-confidence data manifold. Specifically, we model the counterfactual generation as continuous-time dynamics parameterized by Neural ODE, guided by a differentiable density score to actively avoid uncertain, low-density areas. This density score is learned via Noise Contrastive Estimation, effectively leveraging a $(K{+}1)$-way discriminator to estimate density ratios. For black-box settings, we introduce a local proxy distillation mechanism that aligns a lightweight surrogate with the target model strictly within the trajectory of CE generation, enabling efficient gradient-based optimization with minimal queries. Experiments demonstrate that \\textit{DensityFlow} achieves superior validity under model multiplicity while significantly reducing query costs compared to ensemble-based baselines. Our implementation is available at https://github.com/G-AILab/DensityFlow.", "date": "2026-05-29", "categories": []}
{"id": "2605.31156", "title": "TabCausal: Pretraining Across Causal Environments for Tabular Causal Discovery", "abstract": "Causal discovery aims to recover directed causal relations from observational and interventional data, providing a basis for mechanistic understanding and reliable decision-making. Causal discovery foundation models (CDFMs) seek to amortize this problem by mapping a dataset directly to a causal graph in a single forward pass, avoiding per-dataset testing, search, or optimization. However, existing CDFMs remain limited, often failing to consistently match strong classical methods, and we find that a key bottleneck is how causal pretraining tasks are constructed. Based on this observation, we propose TabCausal, a data-driven CDFM trained with broad causal pretraining over diverse graph priors, structural mechanisms, noise models, dimensions, sample sizes, and intervention regimes. A dynamic task construction strategy composes these causal environments into varied discovery tasks, enabling more transferable structural learning from observational and mixed-interventional data. On large-scale synthetic benchmarks, TabCausal achieves better macro-averaged performance than a diverse set of causal discovery baselines. To further bridge abstract synthetic generators and realistic causal reasoning scenarios, we introduce a protocol-guided and LLM-audited semantic causal environment benchmark, where domain-grounded SCMs generate interpretable observational and interventional datasets for out-of-distribution analysis. Across both synthetic and semantic environments, TabCausal demonstrates robust structure recovery, especially under interventional evidence, highlighting broad causal pretraining as a key ingredient for transferable amortized causal discovery.", "date": "2026-05-29", "categories": []}
{"id": "2605.31254", "title": "Formalizing and falsifying causal pathways of rare events", "abstract": "Building on recent formalizations of root cause analysis for rare events (``outliers'') in structural equation models, we propose a formal definition of a causal pathway and discuss its testable implications. We identify conditions under which these implications depend only on a causal abstraction defined by the pathway of rare events, rather than on the full causal graph of the underlying system. Accordingly, we introduce an abstraction of causal structure to pathways of rare events that bridges simple verbal causal explanations and detailed causal modeling.", "date": "2026-05-29", "categories": []}
{"id": "2605.31443", "title": "Modeling Covariate Transition for Efficient Estimation of Longitudinal Treatment Effects in Randomized Experiments", "abstract": "We present a regression-adjustment framework designed for the estimation of longitudinal treatment effects in randomized experiments under static regimes. While regression-adjustment methods are useful for variance reduction in randomized experiments by using pre-treatment covariates, they usually focus only on average effects, from which we cannot obtain valuable insights into when the effects appear and how long they continue. To address this issue, we consider intermediate outcomes and evolving post-treatment covariates over time, and we represent such dynamic trajectories using transition kernels. Furthermore, we establish the asymptotic normality and the semiparametric efficiency bound for our estimator, enabling more powerful statistical inference. Simulation studies and empirical analysis using A/B test data from a streaming platform in Japan show the practical advantages of our method.", "date": "2026-05-29", "categories": []}
{"id": "2604.13410", "title": "Estimating Continuous Treatment Effects with Two-Stage Kernel Ridge Regression", "abstract": "We study the problem of estimating the effect function for a continuous treatment, which maps each treatment value to a population-averaged outcome. A central challenge in this setting is confounding: treatment assignment often depends on covariates, creating selection bias that makes direct regression of the response on treatment unreliable. To address this issue, we propose a two-stage kernel ridge regression method. In the first stage, we learn a model for the response as a function of both treatment and covariates; in the second stage, we use this model to construct pseudo-outcomes that correct for distribution shift, and then fit a second model to estimate the treatment effect. Although the response varies with both treatment and covariates, the induced effect function obtained by averaging over covariates is typically much simpler, and our estimator adapts to this structure. Our optimal learning bounds are achieved without estimating the conditional treatment density, thereby bypassing a major bottleneck in existing methods. Furthermore, we introduce a fully data-driven model selection procedure that achieves provable adaptivity to both the unknown degree of overlap and the spectral decay of the underlying kernel.", "date": "2026-04-15", "categories": []}
{"id": "2605.08786", "title": "PRIM: Meta-Learned Bayesian Root Cause Analysis", "abstract": "Root cause analysis (RCA) in complex systems is challenging due to error propagation across multiple variables, the need for structural causal knowledge, and the computational cost of inference at test time. We introduce PRIM (Prior-fitted Root cause Identification with Meta-learning), a causal meta-learning approach that frames RCA as a Bayesian inference task over a synthetic prior of causal models. By marginalising out structural uncertainty, PRIM implicitly identifies changes in the data-generating mechanism between baseline and anomalous periods. In doing so, PRIM infers distributional differences without explicit statistical testing, and implicitly learns causal structure without model fitting at test time. Following the simulation-based meta-learning paradigm of prior-fitted networks, PRIM uses a Model-Averaged Causal Estimation (MACE) transformer neural process that jointly attends over observational and anomalous samples and the causal structure of nodes, enabling zero-shot inference in 17,ms for systems with up to 100 variables. Across synthetic benchmarks and two realistic benchmark datasets, PetShop and CausRCA, PRIM is competitive with methods that are aware of the system's causal graphical structure a priori while outperforming graph-unaware methods on several tasks. Lightweight fine-tuning to specific domains and data dynamics improves performance further.", "date": "2026-05-09", "categories": []}
{"id": "2605.26029", "title": "CausaLab: A Scalable Environment for Interactive Causal Discovery Toward AI Scientists", "abstract": "We introduce CausaLab, a scalable environment for evaluating interactive causal discovery by LLM agents. Unlike prior evaluations, CausaLab evaluates both whether an agent can solve a problem using causal evidence and whether its answer is grounded in a faithful recovered causal mechanism. Each episode places an agent in a synthetic laboratory: it receives prior measurement records, intervenes on a manipulator crystal, and predicts the resonance frequency of a held-out reactor crystal governed by the same mechanism. The hidden data-generating process is a randomly sampled structural causal model (SCM), so success requires recovering both a causal graph and structural equations rather than recalling prior knowledge.\n  Experiments show a persistent gap between prediction and mechanism recovery: in the purely observational 6-node setting, GPT-5.2-high reaches 92% task accuracy but only 0.471 all-edge $F_1$. Mixed observation-intervention strategies improve structural fidelity, while pure intervention remains difficult even for strong agents. We identify premature stopping as a major weakness and show that consistency verification mitigates it. CausaLab therefore separates predictive success from causal understanding and exposes current LLM agents' limits as experimental causal reasoners.", "date": "2026-05-25", "categories": []}
{"id": "2605.26408", "title": "Function-Valued Causal Influence in Nonlinear Time Series", "abstract": "Causal discovery in time series is increasingly performed using nonlinear machine-learning models, yet the resulting causal relationships are almost always summarized by scalar edge scores. We argue that this practice obscures the true object learned by nonlinear autoregressive models: a state-dependent function whose effect varies across regimes, magnitudes, and contexts. We formalize function-valued causal influence for additive, contribution-decomposable architectures and show that scalar causal scores constitute a severe information bottleneck, conflating between-state variation with within-state residual noise. Using Neural Additive Vector Autoregression as a representative architecture, we introduce a practical framework based on Individual Conditional Expectation for estimating causal response functions directly from trained models. Through controlled synthetic experiments, we demonstrate that edges with indistinguishable scalar scores can exhibit qualitatively different functional behaviors, including monotonic, thresholded, saturating, and sign-changing effects. An applied case study on democratic development further shows that function-valued analysis reveals regime-specific and asymmetric causal structure systematically missed by score-centric approaches.", "date": "2026-05-26", "categories": []}
{"id": "2605.27474", "title": "Stop Suppressing the Tail: Causal Inference for Extreme Events", "abstract": "Estimating how an outcome responds to a continuous treatment (the Average Dose-Response Function, or ADRF) is a core causal-inference primitive. However, when outcomes possess heavy tails, standard robust double machine learning (DML) deliberately suppresses these extremes to stabilize the bulk average. In high-stakes settings, such as financial returns or climate losses, this omitted 1-in-1000 extreme event is the actual target quantity. Furthermore, current methods that read the tail from a model's residuals suffer from circular dependence, causing tail shape inferences to shift drastically based solely on whether the core estimator is switched between Huber and Welsch. The research proposes an ADRF estimator that emits a structured tail-shape output alongside the standard point estimate. Its tail diagnostic (PDHTE+JK) evaluates the per-treatment tail shape from the outcome centered by a pilot median, successfully breaking the circular dependence and rendering the diagnostic invariant to the choice of core method. The output encompasses four treatment-conditional quantities: tail shape $\\hat{\\xi}(t)$, deep-tail return levels $\\hat{Q}_{\\alpha}(t)$, conditional shortfalls $\\hat{S}_{\\alpha}(t)$, the recovered mean ADRF, and an explicit refusal mechanism that declines extrapolation when extreme-value modeling is unsupported by the data. Compared to kernel-weighted quantile regression (QR), the proposed estimator reduces deep-tail ($\\alpha=0.001$) return-level MAE by 11% and conditional-shortfall MAE by 25.5% across a heavy-tailed panel. It also achieves a 20-29% MAE reduction in sample-scarce regimes ($n\\le2000$). On freMTPL2 motor-insurance claims, it successfully triggered an explicit extrapolation refusal on the log-claim scale, which neither QR nor loss-only DML can produce.", "date": "2026-05-26", "categories": []}
{"id": "2605.28880", "title": "Towards Continuous-time Causal Foundation Models", "abstract": "Extending discrete-time causal Prior-data Fitted Networks for time series to continuous time invites writing the mechanism as a stochastic differential equation (SDE) -- but if the SDE is integrated \\emph{once per observation gap}, the trajectory law depends on when it is observed, and the prior remains a discrete-time Markov model in SDE clothing.\n  We propose a precise continuity criterion -- trajectory-law invariance to the observation schedule -- together with a three-tier taxonomy (discrete; naive observation-grid integration; fine-grid integration with decoupled observation) and a construction realising the top tier on a random DAG with OU or small-MLP nonlinear drifts, irregular observation schedules, and hard / soft / time-varying interventions.\n  A $2 \\times 2$ encoder $\\times$ integrator ablation, run independently on a linear and a nonlinear prior, finds fine-grid integration beats naive on 8/8 cells (sign-consistency $p < 1/256$) with the gap growing as the eval grid refines; the encoder axis is null with fine integration but time-aware-leading with naive.\n  We release the prior and a preliminary zero-shot protocol on pharmacokinetic and physical-system data.", "date": "2026-05-26", "categories": []}
{"id": "2605.29008", "title": "Causal Intelligence for Constraint-Aware Intervention Design to Induce State Transitions", "abstract": "Driving a system from one state to another through targeted interventions is a fundamental challenge in science, yet most predictive models offer limited mechanistic insight and no principled framework for decision-making. Here we present COAST (Causally Optimal Actions for State Transitions), a causal-intelligence approach for the in-silico design of constrained interventions that induce user-defined state transitions. Given data characterizing source and target states, COAST learns context-specific causal graphs and structural causal models, attributes observed distributional shifts to mechanism-level causal drivers, and introduces a novel constraint-aware multi-objective optimization formulation that balances transition efficacy, intervention complexity, and target-state stability. The approach is modular and domain-agnostic, integrating feature selection, causal discovery, causal modeling, and intervention identification and evaluation through interchangeable components. Across synthetic benchmarks and real biological datasets, COAST recovers key causal drivers and identifies robust single- and multi-target intervention strategies that achieve desired state transitions, accompanied by transparent mechanistic rationales to guide experimental validation.", "date": "2026-05-27", "categories": []}
{"id": "2605.29272", "title": "Causal Label Recovery in Payment Networks", "abstract": "Fraud detection models in payment networks train on chargeback labels that are systematically biased. Every label must survive three sequential gates: authorization (declined transactions generate no labels), issuer reporting (unreported fraud is invisible), and delay (pending chargebacks are missing at training time). Labels that do arrive may be corrupted by first-party misuse or issuer misclassification. A companion paper [arXiv:2605.27557] proved that these four impairments impose a minimax lower bound on detection performance. This paper asks: can that bound be achieved?\n  We formalize the observation pipeline as a sequential missing-data problem with three propensity stages and a corruption layer, and construct the Sequential Triply Robust (STR) estimator. The STR corrects for all four impairments simultaneously and achieves the semiparametric efficiency bound -- no estimator can have lower asymptotic variance. It is sequentially triply robust: at each gate, consistency requires only that either the propensity model or the outcome regression is correctly specified, not both. We provide corruption correction via noise-rate-adjusted pseudo-labels, empirical Bayes shrinkage to stabilize inverse-propensity weights for small issuers, a plug-in variance estimator yielding valid confidence intervals, and a Bernstein concentration inequality for finite-sample guarantees.\n  On the operational side, we derive the optimal training delay -- the maturity window that minimizes the sum of label-quality loss and model staleness -- and prove that the STR permits training on data that is days old rather than months old, decoupling model freshness from the chargeback maturity cycle. The STR provably dominates naive chargeback-based training in mean squared error for any sample size.", "date": "2026-05-28", "categories": []}
{"id": "2605.29464", "title": "Deep Optimal Individualized Treatment Rules for Bivariate Survival Outcomes via Adaptive Prediction-Powered Learning", "abstract": "In randomized trials involving multiple treatments, bivariate survival outcomes present significant analytical challenges for making decisions. This paper addresses the problem of deriving optimal individualized treatment rules to maximize the joint survival probability beyond fixed time points $(t_1, t_2)$ through deep neural networks, while accounting for right censoring. We propose a novel approach that models treatment rules via stochastic policies, coupling marginal accelerated failure time models via link function to capture bivariate dependence. To enhance robustness and effectiveness of decision making, we introduce an adaptive prediction-powered method that leverages auxiliary predictions from machine learning models.", "date": "2026-05-28", "categories": []}
{"id": "2605.29631", "title": "Predicting Causal Effects from Natural Language Queries using Structured Representations", "abstract": "Randomized controlled trials are a cornerstone of medicine and the social sciences as they enable reliable estimates of causal effects. However, they are costly and time-consuming to conduct, motivating interest in predicting causal effects from existing experimental evidence. Recent advances in large language models (LLMs) have demonstrated strong performance on knowledge-intensive tasks, raising the question of whether these models can be used for forecasting causal effect sizes. To investigate this, we introduce Query2Effect, a new large-scale benchmark consisting of more than 72,000 natural language questions aligned with experiment descriptions, created to simulate realistic information-seeking scenarios by varying query specificity along dimensions of implicitness, abstraction, and ambiguity. We then propose a two-step framework that first generates a synthetic structured representation of a query before predicting effect size using a supervised encoder model. Experiments show that finetuning plays a crucial role in improving prediction performance, with absolute error reducing by -27% up to -71% compared to prompted out-of-the-box LLMs, and that our two-step framework is beneficial for out-of-domain generalization, highlighting the benefits of separating semantic interpretation from numerical effect estimation.", "date": "2026-05-28", "categories": []}
{"id": "2605.29788", "title": "Certified Policy Optimisation for Nested Causal Bandits via PAC-Bayes Risk", "abstract": "Critical sequential decisions are rarely single-timescale: a strategic decision causally shapes the context in which every subsequent tactical choice is made; standard bandit and reinforcement-learning theory does not capture this causal coupling between timescales. We formalise the problem class as Nested Contextual Causal Bandits (NCCBs), a hierarchical SCM where each level's action sets the next level's context distribution, and propose Nested Causal Thompson Sampling (NCTS), which draws one mechanism-factorised belief per episode and acts recursively under it. Our main theoretical result is a causal PAC-Bayesian excess-risk bound that certifies any candidate deployment policy from historic data alone, off-policy and anytime, answering the deployment question: can we trust this agent here, and at what risk? Experiments on a hierarchical SCM show that, against a matched RFF-GP joint regression on the same function class, the factorised SCM-mechanism posterior transfers significantly better zero-shot under exogenous distribution shifts, the recursive meta-to-inner commit significantly dominates the joint-commit alternative in distribution, and the certificate significantly contracts as offline data accumulates. Combining these results, we establish progressive certified handover, a safe-deployment method: each timescale flips from a legacy controller to NCTS when gains can be certified, independently of the others.", "date": "2026-05-28", "categories": []}
{"id": "2605.30015", "title": "Test Time Training for Supervised Causal Learning", "abstract": "Supervised Causal Learning (SCL) has shown promise in causal discovery by framing it as a supervised learning problem. However, it suffers from significant out-of-distribution generalization challenges. We reveal three limitations of previous SCL practices: a significant performance gap between synthetic benchmarks and real-world data, fragility to distribution shifts, and failure in compositional generalization, collectively questioning its real-world applicability. To address this, we propose Test-Time Training for Supervised Causal Learning (TTT-SCL), a novel framework that dynamically generates training sets explicitly aligned with any specific test instance. We demonstrate the correlation between TTT-SCL and score-based methods, and design an efficient module for generating training sets based on the classic scoring function. Experiments on synthetic benchmarks, pseudo-real and real-world datasets demonstrate that TTT-SCL significantly outperforms existing SCL and traditional causal discovery methods.", "date": "2026-05-28", "categories": []}
{"id": "2605.30319", "title": "Improved Guarantees for Heterogeneous Treatment-Effect Estimation via Matrix Completion", "abstract": "A central goal of modern causal inference is estimating heterogeneous treatment effects to answer questions like \"how does an intervention affect each unit,\" rather than only on average. We study this problem with panel-data where we observe $n$ units across $m$ times under unknown, non-uniform treatment assignments. The data in this setting is naturally represented as a matrix of all unit--time treatment effects. Estimating heterogeneous treatment effects can then be expressed as obtaining a good estimation of each row's average in this matrix. This allows us to formulate the problem as matrix completion, which can be solved under natural low-rankness assumptions. However, existing matrix-completion guarantees are not powerful enough to get meaningful bounds for the per-row guarantee required for estimating the heterogeneous treatment effect; roughly speaking, they are only useful for estimating average treatment effect bounds, as also illustrated in a recent line of work. We give a simple, computationally efficient estimator that, without knowledge of the propensities and under standard low-rankness and regularity assumptions, achieves a row-wise $\\ell_2$ error of $\\tilde{O}(\\sqrt{\\frac{1}{n} + \\frac{n}{m^2}})$. Technically, our analysis establishes the first sharp row-wise $\\ell_2$-perturbation bound for low-rank approximation, complementing existing spectral-, Frobenius-, and entrywise perturbation theory.", "date": "2026-05-28", "categories": []}
{"id": "2206.15475", "title": "Causal Machine Learning: A Survey and Open Problems", "abstract": "Causal Machine Learning (CausalML) is an umbrella term for machine learning methods that formalize the data-generation process as a structural causal model (SCM). This perspective enables us to reason about the effects of changes to this process (interventions) and what would have happened in hindsight (counterfactuals). We categorize work in CausalML into five groups according to the problems they address: (1) causal supervised learning, (2) causal generative modeling, (3) causal explanations, (4) causal fairness, and (5) causal reinforcement learning. We systematically compare the methods in each category and point out open problems. Further, we review data-modality-specific applications in computer vision, natural language processing, and graph representation learning. Finally, we provide an overview of causal benchmarks and a critical discussion of the state of this nascent field, including recommendations for future work.", "date": "2022-06-30", "categories": []}
{"id": "2503.11477", "title": "Heterogeneous Causal Discovery of Repeated Undesirable Health Outcomes", "abstract": "Understanding the factors that trigger or prevent undesirable health outcomes across patient subpopulations is essential for designing targeted interventions. While randomized controlled trials and expert-led patient interviews are standard methods for identifying these factors, they can be time-consuming or infeasible. Causal discovery offers an alternative to conventional approaches by generating cause-and-effect hypotheses from observational data, yet its practical utility is limited by strong or untestable assumptions. This work presents a novel, end-to-end framework that uniquely integrates an ensemble of causal structure learning (CSL) algorithms with heterogeneous causal effect estimation. By aggregating results across multiple algorithms, the framework identifies robust causal relationships that persist under different modeling assumptions while simultaneously revealing how these effects vary across specific patient contexts. The proposed heterogeneous causal discovery framework improves robustness and provides practitioners with a prioritized set of actionable, clinically interpretable hypotheses. We demonstrate the framework's effectiveness through two large-scale healthcare applications: identifying drivers and inhibitors of repeat emergency department visits among diabetic patients and hospital readmissions among ICU patients, using insurance claims and electronic health record datasets. Our results, across both settings, identify chronic disease management and care coordination as key interventions, while revealing that intervention effectiveness depends on specific patient-level modifiers. We employ a multi-layered validation strategy, including ground-truth recovery via simulations, alignment with clinical literature, validation by expert clinicians, and portability in modern healthcare systems using an external dataset, to demonstrate the framework's practical utility.", "date": "2025-03-14", "categories": []}
{"id": "2505.09861", "title": "LiDDA: Data Driven Attribution at LinkedIn", "abstract": "Data Driven Attribution, which assigns conversion credits to marketing interactions based on causal patterns learned from data, is the foundation of modern marketing intelligence and vital to any marketing business and advertising platform. In this paper, we introduce a unified transformer-based attribution approach that can handle member-level data, aggregate-level data, and integration of external macro factors. We detail the large scale implementation of the approach at LinkedIn, showcasing significant impact. We also share learnings and insights which are broadly applicable to the marketing and ad tech fields.", "date": "2025-05-14", "categories": []}
{"id": "2509.22553", "title": "Linear Causal Representation Learning by Topological Ordering, Pruning, and Disentanglement", "abstract": "Causal representation learning (CRL) has garnered increasing interest from the causal inference and artificial intelligence communities due to its potential to disentangle complex data-generating mechanism into causally interpretable latent features by leveraging the heterogeneity of modern datasets. In this paper, we further contribute to the CRL literature, by focusing on the stylized linear structural causal model over latent features and assuming a linear mixing function that maps latent features to the observed data or measurements. Existing linear CRL methods often rely on stringent assumptions, such as access to single-node interventional data or restrictive distributional constraints on latent features and/or exogenous measurement noise. However, these prerequisites can be easy to violate in practice. In this work, we propose a novel linear CRL algorithm that, unlike existing methods, operates under weaker assumptions on environment heterogeneity and data-generating distributions while still recovering latent causal features up to an equivalence class. We further validate our new algorithm via synthetic experiments and an interpretability analysis of large language models, demonstrating both its superiority over competing methods in finite samples and its potential in integrating causality into understanding artificial intelligence. The source code is available at https://github.com/utulie/code_for_linear_crl_paper_creator.", "date": "2025-09-26", "categories": []}
{"id": "2512.22777", "title": "Adapting, Fast and Slow: On Few-Shot Transportability of Compositions", "abstract": "Generalization across domains requires stable structure that links the source and target distributions. Building on causal transportability theory, we study a sequential prediction setting in which the target predictor can be represented as a circuit composed of causal mechanisms that are learnable from source data. We introduce two classes of transportability. Module transportability captures the atomic case, where the target predictor is given by a mechanism learnable from a single source domain. Circuit transportability generalizes this idea to target predictors obtained by composing several modules learned from source data, enabling zero-shot prediction even when no source mechanism directly predicts the target label. We study these classes of circuits under increasingly relaxed assumptions. First, we provide conditions under which the relevant circuits can be learned from source data alone, given causal knowledge about the source and target domains. We then relax these structural assumptions by allowing limited data from the target domain. In particular, we develop a supervised domain adaptation scheme that learns circuits without requiring explicit causal structure. The resulting few-shot guarantees tie the achievable error to the size of the smallest target circuit composable from modules learned from source data. Finally, we propose a gradient-based relaxation of the symbolic circuit search and evaluate it empirically, showing that it qualitatively tracks the predicted regimes of fast adaptation -- with and without process supervision over intermediate positions -- and slow adaptation when no source mechanism matches.", "date": "2025-12-28", "categories": []}
{"id": "2602.23602", "title": "Moment Matters: Mean and Variance Causal Graph Discovery from Heteroscedastic Observational Data", "abstract": "Heteroscedasticity -- where the variance of a variable changes with other variables -- is pervasive in real data, and elucidating why it arises from the perspective of statistical moments is crucial in scientific knowledge discovery and decision-making. However, standard causal discovery does not reveal which causes act on the mean versus the variance, as it returns a single moment-agnostic graph, limiting interpretability and downstream intervention design. We propose a Bayesian, moment-driven causal discovery framework that infers separate \\textit{mean} and \\textit{variance} causal graphs from observational heteroscedastic data. We first derive the identification results by establishing sufficient conditions under which these two graphs are separately identifiable. Building on this theory, we develop a variational inference method that learns a posterior distribution over both graphs, enabling principled uncertainty quantification of structural features (e.g., edges, paths, and subgraphs). To address the challenges of parameter optimization in heteroscedastic models with two graph structures, we take a curvature-aware optimization approach and develop a prior incorporation technique that leverages domain knowledge on node orderings, improving sample efficiency. Experiments on synthetic, semi-synthetic, and real data show that our approach accurately recovers mean and variance structures and outperforms state-of-the-art baselines.", "date": "2026-02-27", "categories": []}
{"id": "2603.22335", "title": "Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation", "abstract": "Direct Preference Optimization (DPO) guides large language models (LLMs) to generate recommendations aligned with user historical behavior distributions by minimizing preference alignment loss. However, our systematic empirical research and theoretical analysis reveal that DPO tends to amplify spurious correlations caused by environmental confounders during the alignment process, significantly undermining the generalization capability of LLM-based generative recommendation methods in out of distribution (OOD) scenarios. To mitigate this issue, we propose CausalDPO, an extension of DPO that incorporates a causal invariance learning mechanism. This method introduces a backdoor adjustment strategy during the preference alignment phase to eliminate interference from environmental confounders, explicitly models the latent environmental distribution using a soft clustering approach, and enhances robust consistency across diverse environments through invariance constraints. Theoretical analysis demonstrates that CausalDPO can effectively capture users stable preference structures across multiple environments, thereby improving the OOD generalization performance of LLM-based recommendation models. We conduct extensive experiments under four representative distribution shift settings to validate the effectiveness of CausalDPO, achieving an average performance improvement of 17.17% across four evaluation metrics.", "date": "2026-03-21", "categories": []}
{"id": "2605.14284", "title": "Smooth Multi-Policy Causal Effect Estimation in Longitudinal Settings", "abstract": "Comparative evaluation of multiple dynamic treatment policies is essential for healthcare and policy decisions, yet conventional longitudinal causal inference methods estimate each in isolation, preventing information sharing across counterfactuals. We demonstrate that this separate estimation paradigm induces a structurally uncontrolled second-order bias, inflating finite-sample variance even after standard debiasing with longitudinal targeted maximum likelihood estimation(LTMLE). To address this, we propose a policy-aware reparameterization of Iterative Conditional Expectation (ICE) Q-functions that enables joint estimation through shared representations. We implement this approach in the Policy-Encoded Q Network (PEQ-Net), an architecture centered on a shared policy encoder. The encoder is trained using kernel mean embeddings, ensuring that the learned representation space reflects population-level policy dissimilarities. After applying an LTMLE correction step, we prove this design imposes a structural constraint on the second-order remainder, thereby stabilizing finite-sample variance. Experiments on semi-synthetic datasets demonstrate that PEQ-Net consistently outperforms existing ICE-based methods, achieving substantial reductions in root-mean-square error, particularly when evaluating closely related policies.", "date": "2026-05-14", "categories": []}
{"id": "2605.27473", "title": "Calibrated Inference for the Conditional Average Treatment Effect in the Few-Placebo Regime via Gaussian Processes", "abstract": "Estimating how much an intervention helps a given individual the conditional average treatment effect (CATE) is increasingly central to decision-making in medicine, economics, and policy, where an estimate is most useful when accompanied by a calibrated uncertainty interval. We study the few-placebo regime, in which one treatment arm is much smaller than the other, as arises in unequal-allocation trials and small-holdout $A/B$ tests. The standard estimator in this setting is the X-Learner, and a natural way to obtain credible intervals is to make its second stage Bayesian. We show that these intervals under-cover: they contain the true effect less often than their nominal level. We trace this to a structural cause the X-Learner's regression target inherits the bias of a nuisance model fitted to the small arm, so the posterior is centered away from the true effect and we find that the standard remedy, regressing an orthogonal doubly-robust score, is also unreliable here, since the regime's limited overlap leaves the estimator either highly variable or, once stabilized, biased once more. Both consequences reflect a pattern that extends beyond causal inference: a separately estimated variance is attached to a point estimate of a hard-to-learn quantity, and the point estimate's bias is not captured by that variance. We propose GP-CATE, which models each arm's outcome surface with a Gaussian process, so the scarce arm's uncertainty enters the posterior directly rather than as an unmodelled bias. Across synthetic and semi-synthetic benchmarks, GP-CATE attains calibrated coverage where the estimators we compare against including Causal Forest and BART do not, at the cost of intervals that are appropriately wide when the data are uninformative.", "date": "2026-05-26", "categories": []}
{"id": "2605.27477", "title": "Iterative Causal Discovery: Per-Edge Impossibility Certificates, Tier-Aware Oracle Queries, and the $1+K$ Lower Bound", "abstract": "Causal-discovery algorithms return a directed graph, yet provide no principled means of distinguishing edge directions identified by the data from those assigned without an identifying assumption. Under the standard Markov and faithfulness conditions, the observational distribution identifies only a Markov equivalence class; orientations within that class are not determined by the joint distribution and cannot be recovered from additional samples alone, but require either a functional restriction or an intervention. We introduce a protocol for observational causal discovery on continuous data that attaches to each candidate edge a discrete impossibility certificate: a RESOLVED code records the identifiability theorem under which the direction was committed, while an IMPOSSIBLE code records the failure mode together with the specific question a domain expert must answer to resolve it. The bivariate cascade is extended with five gated identifiability tiers LSNM, IGCI, Stein, MDL, and PEIT that abstain when their precondition test rejects. Two oracle primitives, the meta-hub query and the node-children query, jointly establish an upper bound of $1+K$ expert interactions sufficient to recover any DAG, where $K$ denotes the number of non-leaf vertices. Under an ideal-oracle assumption, the bound is met exactly on the asia, sachs, child, and alarm benchmarks.", "date": "2026-05-26", "categories": []}
{"id": "2605.27567", "title": "Why LLMs Fail at Causal Discovery and How Interventional Agents Escape", "abstract": "Causal discovery is a cornerstone of scientific reasoning, yet whether large language models can perform it reliably remains an open question. Recent benchmarks show that even fine-tuned models plateau on simple causal graphs and degrade as complexity grows, but why they fail has not been established. We prove the failure is fundamental: supervised fine-tuning, direct preference optimization, and in-context learning all produce predictors that cannot distinguish between causal graphs generating similar observational data, and any attempt to do so requires the model's internal representations to grow unboundedly, violating the very conditions under which these methods work. We formalize this as a kernel obstruction theorem, establishing that the limitation is intrinsic to the learning paradigm, \\emph{not any particular model or dataset}. We propose Agentic Causal Bayesian Optimization (A-CBO), wherein a frozen language model serves as an interventional oracle answering targeted queries about intervention effects, while an external Bayesian loop concentrates beliefs over candidate graphs in logarithmically many rounds. Because the decision operates outside the space where the obstruction applies, A-CBO provably converges while the underlying model remains unchanged. On Corr2Cause, A-CBO matches fine-tuned baselines without any training. On Extended Corr2Cause, a new benchmark scaling to 24 variables with 18K test samples, A-CBO significantly outperforms both fine-tuning and preference optimization, with the advantage growing", "date": "2026-05-26", "categories": []}
{"id": "2407.15073", "title": "Multi-Agent Causal Discovery Using Large Language Models", "abstract": "Causal discovery aims to identify causal relationships between variables and is a fundamental problem across the sciences. Traditional statistical causal discovery (SCD) methods rely solely on observational data and ignore the contextual information available in metadata, whereas recent LLM-based methods exploit metadata but treat the large language model (LLM) as a single agent, leaving its judgments vulnerable to memorized or biased associations. To address this gap, we introduce MAC (Multi-Agent Causal Discovery Framework), which casts causal discovery as a multi-agent debate coupled with the autonomous selection of an SCD algorithm. MAC combines two complementary modules, bridged by a Meta Fusion mechanism: a Debate-Coding Module (DCM) that grounds an initial graph in data by autonomously selecting and executing the best-suited SCD algorithm, and a Meta-Debate Module (MDM) that refines the graph through an adversarial Affirmative-Negative-Judge debate over the metadata. Across five benchmark datasets and three metrics (F1, SHD, NHD), MAC achieves the best aggregate performance among five statistical and four LLM-based baselines, ranking first on 10 of 15 evaluation points with Gemini-2.0-Flash -- including a perfect reconstruction of the Earthquake graph -- and remains robust across three backbone LLMs.", "date": "2024-07-21", "categories": []}
{"id": "2512.16111", "title": "BUILD with Precision: Bottom-Up Inference of Linear DAGs", "abstract": "Learning the structure of directed acyclic graphs (DAGs) from observational data is a central problem in causal discovery, statistical signal processing, and machine learning. Under a linear Gaussian structural equation model (SEM) with equal noise variances, the problem is identifiable and we show that the ensemble precision matrix of the observations exhibits a distinctive structure that facilitates DAG recovery. Exploiting this property, we propose BUILD (Bottom-Up Inference of Linear DAGs), a deterministic stepwise algorithm that identifies leaf nodes and their parents, then prunes the leaves by removing incident edges to proceed to the next step, exactly reconstructing the DAG from the true precision matrix. In practice, precision matrices must be estimated from finite data, and ill-conditioning may lead to error accumulation across BUILD steps. As a mitigation strategy, we periodically re-estimate the precision matrix (with less variables as leaves are pruned), trading off runtime for enhanced robustness. Reproducible results on challenging synthetic benchmarks demonstrate that BUILD compares favorably to state-of-the-art DAG learning algorithms, while offering an explicit handle on complexity.", "date": "2025-12-18", "categories": []}
{"id": "2602.03517", "title": "Rank-Learner: Orthogonal Ranking of Treatment Effects", "abstract": "Many decision-making problems require ranking individuals by their treatment effects rather than estimating the exact effect magnitudes. Examples include prioritizing patients for preventive care interventions, or ranking customers by the expected incremental impact of an advertisement. Surprisingly, while causal effect estimation has received substantial attention in the literature, the problem of directly learning rankings of treatment effects has largely remained unexplored. In this paper, we introduce Rank-Learner, a novel two-stage learner that directly learns the ranking of treatment effects from observational data. We first show that naive approaches based on precise treatment effect estimation solve a harder problem than necessary for ranking, while our Rank-Learner optimizes a pairwise learning objective that recovers the true treatment effect ordering, without explicit CATE estimation. We further show that our Rank-Learner is Neyman-orthogonal and thus comes with strong theoretical guarantees, including robustness to estimation errors in the nuisance functions. In addition, our Rank-Learner is model-agnostic, and can be instantiated with arbitrary machine learning models (e.g., neural networks). We demonstrate the effectiveness of our method through extensive experiments where Rank-Learner consistently outperforms standard CATE estimators and non-orthogonal ranking methods. Overall, we provide practitioners with a new, orthogonal two-stage learner for ranking individuals by their treatment effects.", "date": "2026-02-03", "categories": []}
{"id": "2604.18751", "title": "Beyond Coefficients: Forecast-Necessity Testing for Interpretable Causal Discovery in Nonlinear Time-Series Models", "abstract": "Nonlinear machine-learning models are increasingly used to discover causal relationships in time-series data, yet the interpretation of their outputs remains poorly understood. In particular, causal scores produced by regularized neural autoregressive models are often treated as analogues of regression coefficients, leading to misleading claims of statistical significance.\n  In this paper, we argue that causal relevance in nonlinear time-series models should be evaluated through forecast necessity rather than coefficient magnitude, and we present a practical evaluation procedure for doing so. We present an interpretable evaluation framework based on systematic edge ablation and forecast comparison, which tests whether a candidate causal relationship is required for accurate prediction. Using Neural Additive Vector Autoregression as a case study model, we apply this framework to a real-world case study of democratic development, modeled as a multivariate time series of panel data - democracy indicators across 139 countries. We show that relationships with similar causal scores can differ dramatically in their predictive necessity due to redundancy, temporal persistence, and regime-specific effects.\n  Our results demonstrate how forecast-necessity testing supports more reliable causal reasoning in applied AI systems and provides practical guidance for interpreting nonlinear time-series models in high-stakes domains.", "date": "2026-04-20", "categories": []}
{"id": "2605.26288", "title": "Beyond Differences: Doubly Robust Meta-Learners for Ratio-Based Treatment Effects", "abstract": "When treatment effects are naturally expressed as ratios -- as in medicine, pricing, and marketing -- the ratio-based CATE $\\tau(x) = E[Y|W=1,X=x] / E[Y|W=0,X=x]$ is the appropriate estimand. Yet existing estimators either impose a log-linear parametric structure or apply generic regression without robustness guarantees for this functional. We introduce the Q-Learner, which decomposes $\\tau(x)$ into a product of two odds ratios, reducing ratio-CATE estimation for binary outcomes to two propensity classification tasks. We further derive doubly robust augmentations for both S/T- and Q-style ratio learners and characterize their distinct robustness properties. In benchmarks on seven RCT datasets, the Q-Learner is the most consistently competitive method in low-conversion regimes, where its propensity-only construction sidesteps the imbalanced regression that hurts outcome-based estimators. On four observational datasets, where propensity must be estimated and confounding cannot be ruled out, the DR learners introduced here decisively come out on top, making them practitioners' natural default for confounded observational data.", "date": "2026-05-25", "categories": []}
{"id": "2605.26413", "title": "Confounder Detection via Treatment Intent: A New Observational Study Design", "abstract": "Understanding the effects of interventions is central to scientific progress, with randomized controlled trials (RCTs) regarded as the gold standard for causal inference in many applied fields. However, RCTs are costly, time-consuming, and often constrained by ethical or practical limitations, motivating the need for causal methods able to draw conclusions from observational data. While such data is collected at ever larger scale, making its use for causal inference is often hindered by the fact that not all variables affecting treatment allocation and the outcome are observed: an issue known as unobserved confounding. In this paper, we introduce a new study design called confounder detection via treatment intent. The idea is to query a human expert who makes treatment decisions, and ask them to compare pairs of units proposed by a principled matching strategy, with the goal of eliciting unobserved variables that explain why treatment decisions differ. We provide a theoretical basis for such a procedure, ascertaining conditions under which such a study design may elicit unobserved confounders. Building on this newly established foundations, we study treatment effects of interventions in the intensive care unit (ICU). First, we show empirical evidence strongly indicating that electronic health records (EHRs) collected in ICUs are subject to unobserved confounding. By using clinical text notes as a proxy for physicians' knowledge and leveraging natural language processing, we provide a proof of concept for our methodology in a semi-synthetic environment with a known ground truth.", "date": "2026-05-26", "categories": []}
{"id": "2605.26679", "title": "Certified Causal Attribution for Real-Time Attack Forensics in 6G Network Slicing", "abstract": "Cross-slice attack attribution in 6G networks requires identifying causal propagation chains through shared infrastructure in under 100 ms. Existing methods struggle to satisfy this strict SLA without sacrificing accuracy, because shared resource contention creates spurious correlations that are indistinguishable from genuine causal links under standard Granger tests. We propose DA-GC, a certified causal attribution framework that integrates resource-conditioned Granger causality with an axiomatically derived Resource Contention Model (RCM) to systematically block resource-mediated confounding. On a 15-slice production-emulation 6G testbed with 1,100 attack scenarios, DA-GC achieves 89.2% attribution accuracy at 87 ms. This represents a 7.9 percentage-point improvement over the strongest baseline at 2.7x lower latency, alongside demonstrated cross-topology generalization and concept-drift resilience. Crucially, DA-GC is backed by a comprehensive formal certification stack. We provide mathematically proven validity certificates for statistical soundness under serially dependent telemetry and piecewise-stationarity. Furthermore, we establish strict security bounds, including an adversarial utilization spoofing breakdown point of $\\delta^* \\approx 0.95$, and define the minimum differential-privacy noise required for a provably private and robust deployment.", "date": "2026-05-26", "categories": []}
{"id": "2605.27022", "title": "ORCA: An End-to-End Interactive Copilot for Optimized Root Cause Analysis", "abstract": "Causal analysis is a crucial task in many domains, including manufacturing, social science, and medicine. However, despite recent progress, the conceptual and methodological complexity of causal methods makes them largely inaccessible to domain experts. This gap prevents experts from leveraging these advances and hinders researchers who lack access to real-world data for validation. To bridge this divide, we introduce ORCA, a copilot for end-to-end causal analysis. ORCA orchestrates agents to understand the user's goals and guide them through the most appropriate causal analysis workflow, from fully automatic to highly user-guided execution. It features causal discovery, causal effect estimation, explainability and Root-Cause-Analysis (RCA). ORCA evaluates and compares performance, generates key metrics and diagrams, and generates insights through structured reports. We highlight its effectiveness across several real-world use-cases.", "date": "2026-05-26", "categories": []}
{"id": "2605.27043", "title": "Causal Representation Learning for Generalisable Recommendation", "abstract": "Predictive models trained on observational data often fail to generalise to the distributions they encounter when deployed, especially when the training data is a product of the system being optimised. Recommender systems are a canonical example: they are trained on interaction logs confounded by the deployed policy, past user behaviour, and platform filtering. As a result, the training distribution differs substantially from the candidate distribution scored at serving time, a gap that makes offline metrics unreliable predictors of online performance. We address the distribution shift problem with a method motivated by causal representation learning (CRL). We propose an information-theoretic disentanglement criterion and prove that its optimum depends only on the causal components of the input. We then derive a tractable variational lower bound that makes the criterion optimisable from finite observational data alone. The scope of our method is narrower than that of much of the CRL literature, in that we target better generalisation under distribution shift, not full identification of all latent causal factors. This narrower target is what makes the method practical, requiring only the existing confounded logs, applying to any standard supervised model, and adding no inference-time cost. Our headline evaluation is an A/B test with millions of users on Spotify, applied to a production ranker for personalised playlist generation. A capacity-matched CRL variant performed on par offline but delivered substantial online gains in listener engagement. Complementary evidence on the public KuaiRand recommendation dataset and a synthetic benchmark with known causal structure shows the same pattern: offline parity with baseline, gains under distribution shift. Across all three settings, adding our causal disentanglement objective yields meaningfully better out-of-distribution generalisation.", "date": "2026-05-26", "categories": []}
{"id": "2605.27163", "title": "The Role of Causal Features in Strategic Classification for Robustness and Alignment", "abstract": "In strategic classification, an institution (e.g., a bank) anticipates adaptation from users who change their features to increase utility in a classification task (e.g., loan repayment). Since a key challenge is the distribution shift induced by users, we turn to causal models, which have been shown to bound the worst-case out-of-distribution (OOD) risk, and establish several new results that link causality and strategic classification. First, we show that causal classification leads to optimal classification error after any sufficiently large adaptation, when the noise is bounded in a certain way. Second, when these assumptions do not hold, we show OOD cross-entropy risk of optimal classifiers decomposes into an OOD bias term and a term arising from not using all observable features, allowing us to understand when causal classifiers have an advantage. Finally, we show that the use of causal features can allow alignment of long-term incentives between institutions and users, contrasting with previous work that highlights social costs of such approaches. We validate our theory empirically on synthetic data, finding that our results predict behavior in practice.", "date": "2026-05-26", "categories": []}
{"id": "2605.27281", "title": "Causal Risk Minimization for High-Dimensional Treatments", "abstract": "Predicting the effect of interventions with many possible variations, e.g., therapeutic content that affects mental health outcomes or an earnings call transcript that drives movement in share price, is useful across several domains. However, classical causal estimators tend to assume that all possible interventions are observed, which is infeasible when interventions vary widely, for instance, in the space of all text strings. We adapt a well-known approach of recasting causal inference as a learning problem, to address high-dimensional treatment spaces. Specifically, under standard assumptions like no unobserved confounding, we show that causal error decomposes into a series of moment-balancing errors of increasing order, and design objectives that directly improve causal estimation. We also show how to project the effect of a high-dimensional treatment onto lower-dimensional treatment attributes, which allows a single model to answer several causal questions without additional attribute-specific training. We empirically evaluate our estimators in settings with high-dimensional continuous, discrete, and text treatments, the last of which used a semi-synthetic dataset of Amazon Reviews. Our experiments demonstrate the benefit of higher-order balance error optimization and competitive performance of projected causal estimates with attribute-specific estimators.", "date": "2026-05-26", "categories": []}
{"id": "2605.10430", "title": "Real vs. Semi-Simulated: Rethinking Evaluation for Treatment Effect Estimation", "abstract": "Estimating heterogeneous treatment effects with machine learning has attracted substantial attention in both academic research and industrial practice. However, the two communities often evaluate models under markedly different conditions. Methodological work typically relies on semi-simulated benchmarks and metrics that require counterfactual outcomes, whereas real-world applications rely on observable metrics based on ranking or test outcomes. Despite the well-known gap between methodological progress and practical deployment, the relationship between these evaluation regimes has not been examined systematically. We conduct a large-scale empirical study of treatment effect evaluation across standard semi-simulated benchmark families and real-world datasets. Our benchmark covers meta-learners paired with multiple base learners, as well as specialized causal machine learning models. We evaluate these methods using observable metrics common in application-oriented literature, alongside counterfactual metrics commonly used in methods papers. Our results reveal two complementary gaps. First, counterfactual metrics do not reliably recover the estimators preferred by observable metrics, even on the same semi-simulated benchmarks. Second, rankings obtained on semi-simulated benchmarks do not transfer to real datasets. We further find that simple meta-learners with strong base models are consistently competitive, in contrast to specialized causal models. Overall, our findings suggest that progress in treatment effect estimation research should not be assessed solely through counterfactual metrics and semi-simulated benchmarks, but it would benefit from incorporating observable metrics and real-data validation.", "date": "2026-05-11", "categories": []}
{"id": "2605.24076", "title": "Causality as the Statistical Conscience of Artificial Intelligence: From Pearl's Ladder to Trustworthy Machines", "abstract": "Modern Artificial Intelligence achieves remarkable predictive power by optimizing statistical risk functionals over vast corpora. Yet a gap separates this from genuine intelligence: the inability to distinguish correlation from causation. This paper argues that causal inference (identifying mechanisms invariant under intervention) is AI's indispensable statistical conscience. Without causal grounding, AI systems are correlation machines: powerful in familiar domains, brittle under distribution shift, and biased in high-stakes settings. Three contributions develop this argument. First, a Statistical Necessity Theorem for Causal Generalization: any algorithm achieving out-of-distribution generalization must encode causal structure, formalizing the distinction between prediction P(Y|X) and intelligence P(Y|do(X)). Second, a unified framework connects Pearl's do-calculus, the Potential Outcomes framework, Double Machine Learning, and Invariant Risk Minimization as a family of Causal Statistical Estimators, each identifying interventional distributions under different assumptions. Third, three AI failure modes (hallucination in large language models, reward hacking in reinforcement learning from human feedback, and degradation under distribution shift) are manifestations of causal blindness, each admitting a principled statistical remedy. Trustworthy AI is, at its core, a problem of causal statistics. The statistical community is not merely equipped to solve it -- it is the only community with the foundational tools to do so rigorously.", "date": "2026-05-22", "categories": []}
{"id": "2605.24808", "title": "Disentangled Double Machine Learning for Accurate Causal Effect Estimation", "abstract": "Confounding bias is a key challenge in causal effect estimation from observational data. Double Machine Learning (DML) addresses this issue by estimating treatment and outcome nuisance functions, constructing treatment and outcome residuals, and estimating causal effects from the residuals. However, DML often produces biased and unstable estimates in highdimensional or finite-sample scenarios. One reason is that DML estimates nuisance functions using all covariates without disentangling distinct latent factors, resulting in unreliable nuisance function estimation. Another is that imprecise nuisance estimation further introduces residual dependence between the treatment residual and the remaining outcome error, undermining the accuracy of causal effect estimates. To address these issues, in this paper, we propose Disentangled Double Machine Learning (DDML), a novel algorithm that integrates two key strategies. First, a causal role disentanglement strategy decomposes covariates into confounders, treatment-specific factors, and outcomespecific factors for enabling reliable nuisance function estimation. And second, a residual dependence orthogonalization strategy mitigates residual dependence caused by nuisance estimation errors for enhancing the precision of causal effect estimates. Experimental results on synthetic, semi-synthetic, and real-world datasets demonstrate that DDML significantly outperforms 13 state-of-the-art baseline algorithms in both MAE and RMSE.", "date": "2026-05-24", "categories": []}
{"id": "2605.24873", "title": "Towards a Universal Causal Reasoner", "abstract": "Despite the importance of causal reasoning, training LLMs to reason causally remains underexplored. Existing data efforts mostly focus on benchmarking LLMs on specific aspects of causality, making them less suitable for training generalizable causal reasoners. To address this, we propose UniCo, a data generation framework that both (1) addresses 18 causal query types across Pearl's Causal Ladder and (2) translates natively symbolic examples into code and natural language forms to simulate real-world use cases where causal terms are not explicitly specified. To ensure data quality, UniCo grounds answers with exact causal inference and filters cases with reasoning shortcuts. Upon supervised finetuning with 66.6K UniCo-generated instances, Qwen3-4B, Qwen3-8B and Olmo-3-7B-Instruct achieve an average of 22.9% improvements across all 18 in-distribution query types, and 8.1% over state-of-the-art causal data generation frameworks on 7 established causal benchmarks outside the training distribution. More importantly, in real-world medical understanding, legal decision, and tabular reasoning, UniCo-trained models consistently display more faithful reasoning traces, outperforming the base models by an average of 20.2% in faithfulness metrics. These suggest that causality-centered training not only strengthens causal reasoning, but also equips LLMs with a causal mindset in general reasoning tasks.", "date": "2026-05-24", "categories": []}
{"id": "2605.25169", "title": "Learning Treatment Effects during Resource Allocation via Priority-Queue Randomization", "abstract": "Public service programs often allocate limited resources under uncertainty about their benefits, creating a need for randomization to support credible evaluation. In practice, however, applicants commonly enter waitlists where resources are prioritized toward individuals judged to have higher need through tiered priority queues, making direct randomization difficult. Motivated by this, we develop an experimental design framework for learning treatment effects while treating those most in need where incoming applicants are randomized into priority queues based on their assessed risk scores. Treatments are then provided across queues in priority order and first-in-first-out within queue as budget becomes available. Our contributions are two-fold. First, we characterize what causal effects are identified under this priority-queue allocation. When arrivals are exogenous, treatments are conditionally randomized, and hence standard estimands are identified; when arrivals are endogenous, queue randomization instead provides an instrument for treatment, identifying local treatment effects induced by the queuing process. Second, we develop optimized queue-assignment designs that trade off statistical efficiency against prioritizing higher-need applicants. We show in the process that, despite dependence in treatment assignments induced by the design, usual iid efficiency bounds remain well-justified design objectives. We illustrate the proposed designs using data from a housing allocation program in a large U.S. county.", "date": "2026-05-24", "categories": []}
{"id": "2605.25211", "title": "Evolving Causal Regulatory Networks (ECR-Net)", "abstract": "Modern machine learning models excel at pattern recognition but remain brittle, often failing to generalize out of distribution (OOD) because they capture spurious correlations rather than the underlying causal data-generating process. Current causal discovery methods, while powerful, typically assume a static graph structure, rendering them unable to model systems that adapt or undergo structural changes across different environments. We introduce ECR-Net, Evolving Causal Regulatory Networks, a novel, bio-inspired framework for adaptive causal mechanism discovery. Our approach models the data-generating process not as a static graph, but as a dynamic system analogous to a Gene Regulatory Network (GRN), composed of localized, recursive functions where variables can activate and inhibit one another. To discover the latent structure of this network, we employ an evolutionary search algorithm that evolves a population of candidate regulatory graphs, optimizing for a fitness function that measures how well the simulated system dynamics reconstruct the observed data. The key innovation of ECR-Net is its ability to model structural adaptation, it explicitly ingests shifts in the data's statistical properties as signals of an environmental shock. In response, the evolutionary search identifies parsimonious modifications to the causal graph topology, such as link inhibitions or activations that explain the new data regime. We posit that ECR-Net represents a new class of adaptive Structural Causal Models capable of discovering how and why a system's fundamental rules change, offering a path toward robust generalization in complex, non-stationary systems.", "date": "2026-05-24", "categories": []}
{"id": "2605.25338", "title": "CausalFlow: Causal Attribution and Counterfactual Repair for LLM Agent Failures", "abstract": "Large language model (LLM) agents frequently fail on multi-step tasks involving reasoning, tool use, and environment interaction. While such failures are typically logged or retried heuristically, they contain structured signals about where execution broke down. We introduce CausalFlow, an interventional framework that converts failed agent traces into minimal counterfactual repairs and reusable supervision. CausalFlow models execution traces as sequential chains of dependent steps and computes Causal Responsibility Scores(CRS) via step-level counterfactual intervention to identify failure-inducing steps. For these steps, we generate minimally edited repairs that flip the final outcome to success, producing validated contrastive pairs of the form (wrong step, corrected step). CausalFlow supports two complementary uses: targeted test-time repair that recovers from failures with minimal behavioral drift, and training-time supervision suitable for offline preference optimization or reward modeling. Across four benchmarks spanning mathematical reasoning, code generation, question answering, and medical browsing, CausalFlow converts failed executions into validated minimal repairs with high minimality and causal-consensus scores, and demonstrates that causal attribution is necessary for reliable improvement across diverse agent tasks, outperforming heuristic refinement in complex retrieval settings while producing more localized repairs throughout. These results demonstrate that interventional analysis over structured execution traces provides a principled and scalable mechanism for transforming agent failures into reliability gains and learning-ready supervision.", "date": "2026-05-25", "categories": []}
{"id": "2605.25811", "title": "Geometry Adaptive Counterfactual Distribution Learning with Diffusion-Guided Smoothing", "abstract": "We study counterfactual distribution learning for high-dimensional outcomes whose counterfactual law may concentrate near lower-dimensional structure. Standard isotropic smoothing treats all ambient directions equally, leading to unfavorable scaling and unstable local inference. We propose two diffusion-guided estimators based on semiparametric debiasing: diffusion-informed smoothing for counterfactual densities and diffusion-informed score smoothing for counterfactual scores. The estimators combine causal nuisance adjustment with geometry-adaptive localization driven by diffusion score information, removing first-order nuisance bias while aligning smoothing with local outcome geometry. We establish asymptotic expansions, risk bounds, and inference procedures for smoothed density and score-based targets, with ambient density inference obtained under additional approximation conditions. Under structural geometry conditions, the leading stochastic error is governed by an effective dimension induced by the diffusion-guided kernel, rather than by the ambient dimension. Semi-synthetic experiments based on CelebA show steeper error decay for geometry-adaptive methods, supporting the proposed effective-dimension theory.", "date": "2026-05-25", "categories": []}
{"id": "2605.25998", "title": "Causal methods for LLM development and evaluation", "abstract": "Large language model (LLM) development is currently driven by large-scale empirical iteration over data mixtures, reward models, routing strategies, and evaluation pipelines. Here, we argue that many central questions in LLM development and evaluation are inherently causal: What is the effect of adding a data domain during pretraining? How do annotator preferences change when LLMs generate text in a different style? Should a prompt be routed to a larger or smaller model given inference cost constraints? In general, causal methods are well-suited to such settings where interventions change outcomes but, surprisingly, are underrepresented in LLM development. Our contribution is threefold: (1) We explain how causal methods can help develop modern LLM development and evaluation: LLM development relies heavily on logged data, which are often subject to confounding and distribution shifts; evaluation uses learned but potentially biased judges; and deployment environments are non-stationary. These conditions make purely predictive approaches fragile and create opportunities for principled identification and estimation methods from causal inference. (2) We further map opportunities for causal methods in the entire LLM development pipeline, including pretraining, alignment, routing, agentic workflows, and evaluation. (3) We discuss new research opportunities around leveraging causal methods for LLM development and evaluation. Overall, we argue that causal methods are potentially underutilized for the LLM development and evaluation pipeline, despite the fact that such methods can ensure a reliable and scientifically grounded design.", "date": "2026-05-25", "categories": []}
{"id": "2502.07295", "title": "Targeted Regularization for Causal Effect Estimation with Exponential Dispersion Family Outcomes", "abstract": "Neural Networks (NNs) for causal effect estimation have shown strong empirical performance, yet endowing them with desirable semiparametric properties -- doubly robustness and fast convergence rates -- remains challenging. A common approach to address this is targeted regularization, which modifies the objective function of NNs. However, existing work on neural causal effect estimation is largely limited to continuous outcomes, restricting its applicability to settings involving binary, count, or other skewed outcomes commonly encountered in practice. We propose a unified targeted regularization framework for the Exponential Dispersion Family (EDF) to address this limitation. Specifically, we first derive the von Mises expansion of the average dose function of canonical functions (ADCF) for discrete treatments and of the sieve-projected ADCF for continuous treatments. Second, we use this expansion to construct a unified targeted regularization, that corrects first-order bias at the distributional level. We integrate this objective into a NN architecture that jointly estimates the outcome model, propensity score model, and fluctuation parameter end-to-end. Experimental results demonstrate the effectiveness of our method.", "date": "2025-02-11", "categories": []}
{"id": "2502.07646", "title": "Causal Additive Models with Unobserved Causal Paths and Backdoor Paths", "abstract": "Causal additive models provide a tractable yet expressive framework for causal discovery in the presence of hidden variables. When unobserved backdoor or causal paths exist between two variables, their causal relationship is often unidentifiable under existing theories. We establish sufficient conditions under which causal directions can be identified in many such cases. These conditions rely on new characterizations of regression sets to determine independence among regression residuals and conditional independencies among observed variables. Building on these results, we introduce a search algorithm that incorporates these innovations and prove its soundness and completeness. Empirical evaluations demonstrate its competitive performance against state-of-the-art methods.", "date": "2025-02-11", "categories": []}
{"id": "2502.13731", "title": "Robust Counterfactual Inference in Markov Decision Processes", "abstract": "This paper addresses a key limitation in existing counterfactual inference methods for Markov Decision Processes (MDPs). Current approaches assume a specific causal model to make counterfactuals identifiable. However, there are usually many causal models that align with the observational and interventional distributions of an MDP, each yielding different counterfactual distributions, so fixing a particular causal model limits the validity (and usefulness) of counterfactual inference. We propose a novel non-parametric approach that computes tight bounds on counterfactual transition probabilities across all compatible causal models. Unlike previous methods that require solving prohibitively large optimisation problems (with variables that grow exponentially in the size of the MDP), our approach provides closed-form expressions for these bounds, making computation highly efficient and scalable for non-trivial MDPs. Once such an interval counterfactual MDP is constructed, our method identifies robust counterfactual policies that optimise the worst-case reward w.r.t. the uncertain interval MDP probabilities. We evaluate our method on various case studies, demonstrating improved robustness over existing methods.", "date": "2025-02-19", "categories": []}
{"id": "2605.23861", "title": "Leveraging Foundation Models for Causal Generative Modeling", "abstract": "Causal generative modeling is essential for developing reliable and transparent AI systems capable of counterfactual reasoning. While existing approaches focus on integrating causal constraints during the training of generative models, they often lack a unified framework to leverage the zero-shot reasoning capabilities of pretrained foundation models. We introduce FM-CGM, a modular framework for end-to-end visual causal reasoning using pretrained foundation models. FM-CGM formalizes the causal pipeline through three core components: a concept extractor, a concept manipulator, and a counterfactual generator. By leveraging a large reasoning model for causal inference and a text-to-image diffusion model for generation, our approach enables zero-shot causal discovery, intervention, and counterfactual generation. We then develop Causal Semantic Guidance (CSG), a cross-attention-based mechanism that ensures semantic interventions propagate to descendant concepts while preserving invariant regions. We empirically show that our approach can identify plausible causal structures and is suitable for faithful counterfactual image generation.", "date": "2026-05-22", "categories": []}
{"id": "2408.13002", "title": "Measuring Variable Importance in Heterogeneous Treatment Effects with Confidence", "abstract": "Causal machine learning holds promise for estimating individual treatment effects from complex data. For successful real-world applications of machine learning methods, it is of paramount importance to obtain reliable insights into which variables drive heterogeneity in the response to treatment. We propose PermuCATE, an algorithm based on the Conditional Permutation Importance (CPI) method, for statistically rigorous global variable importance assessment in the estimation of the Conditional Average Treatment Effect (CATE). Theoretical analysis of the finite sample regime and empirical studies show that PermuCATE has lower variance than the Leave-One-Covariate-Out (LOCO) reference method and provides a reliable measure of variable importance. This property increases statistical power, which is crucial for causal inference in the limited-data regime common to biomedical applications. We empirically demonstrate the benefits of PermuCATE in simulated and real-world health datasets, including settings with up to hundreds of correlated variables.", "date": "2024-08-23", "categories": []}
{"id": "2605.21548", "title": "Local Covariate Selection for Average Causal Effect Estimation without Pretreatment and Causal Sufficiency Assumptions", "abstract": "We study the problem of selecting covariates for unbiased estimation of the total causal effect.Existing approaches typically rely on global causal structure learning over all variables, or on strong assumptions such as causal sufficiency - where observed variables share no latent confounders - or the pretreatment assumption, which limits covariates to those unaffected by the treatment or outcome. These requirements are often unrealistic in practice, and global learning becomes computationally prohibitive in high-dimensional settings.To address these challenges, we propose a novel local learning method for covariate selection in nonparametric causal effect estimation that avoids both the pretreatment and causal sufficiency assumptions. We first characterize a local boundary that contains at least one valid adjustment set whenever one exists for identifying the causal effect, and then develop local identification procedures to efficiently search within this boundary.We prove that the proposed method is sound and complete. Experiments on multiple synthetic datasets and two real-world datasets show that our approach achieves accurate causal effect estimation while substantially improving computational efficiency.", "date": "2026-05-20", "categories": []}
{"id": "2605.21846", "title": "Causal Discovery in Structural VAR Models Under Equal Noise Variance", "abstract": "Causal discovery from multivariate time series is challenging when causal effects may occur both across time and within the same sampling interval. This issue is especially important in applications such as neuroscience, where the sampling rate may be coarse relative to the underlying dynamics and contemporaneous effects need not form an acyclic graph. We study causal discovery in linear Gaussian structural VAR models under an equal noise variance assumption, meaning that the structural noise terms have a common variance. Unlike the DAG-based cross-sectional equal noise variance setting, the time-series setting considered here does not generally yield point identification of a unique causal graph. Instead, multiple structural VAR parameterizations can induce the same stationary observed process law. We introduce a notion of observational equivalence tailored to this setting and show that the corresponding equivalence class is characterized by orthogonal transformations of the structural equations together with a global positive scale. This characterization leads to an equivalence-aware model discrepancy, the observational alignment discrepancy, which compares structural models modulo transformations that preserve the observed law. Building on this theory, we propose ENVAR, a sparsity-based procedure that searches over the induced observational equivalence class for a sparse normalized structural representative. We evaluate the proposed methodology on synthetic structural VAR data and on an fMRI dataset.", "date": "2026-05-21", "categories": []}
{"id": "2605.21928", "title": "CausalGuard: Conformal Inference under Graph Uncertainty", "abstract": "Estimating treatment effects from observational data requires choosing an adjustment set, but valid adjustment depends on an unknown causal graph. Graph misspecification can cause under-coverage, while graph-agnostic conformal wrappers may regain nominal coverage only through large padding. We introduce CausalGuard, a structure-weighted conformal framework that calibrates after aggregating graph-conditional doubly robust pseudo-outcomes. Candidate DAGs are proposed from an LLM-derived edge prior, pruned by conditional-independence tests, and reweighted by Bayesian Information Criterion. A composite nonconformity score then calibrates the posterior-weighted pseudo-outcome. CausalGuard provides distribution-free finite-sample marginal coverage for this aggregated pseudo-outcome; under causal identification, overlap, conditional-mean nuisance stability, and concentration on target-aligned valid adjustment strategies, its conditional mean converges to the true Conditional Average Treatment Effect. Across five benchmarks, CausalGuard attains mean coverage above the nominal 90% level for the directly evaluable target and reduces width when graph-agnostic conformal baselines require large padding. Stress tests show that CausalGuard suppresses invalid collider adjustment and remains stable under misspecified priors when the retained candidate set is data-supported.", "date": "2026-05-21", "categories": []}
{"id": "2605.22075", "title": "Can Breath Biomarkers Causally Influence Blood Glucose? Investigating VOC-Mediated Modulation in Diabetes", "abstract": "Diabetes is a global health burden, and early detection is critical for timely intervention. This study explores a non-invasive, data-driven framework to identify individuals at risk of diabetes using Volatile Organic Compounds (VOCs) and lifestyle variables. We use causal inference techniques to estimate the impact of VOCs such as acetone, isopropanol, isoprene, and ethanol on blood glucose levels. Additionally, we designed a classifier to distinguish diabetics from non-diabetics using non-invasive markers. We created a risk-based ranking system for individuals in the \"gray zone,\" and identified natural clusters in the population using Gaussian Mixture Model. Our results suggest that specific VOCs exhibit a strong causal influence on glucose levels and that machine learning models can reliably classify and stratify individuals at high risk. This integrated causal-explainable analysis can support the development of tool for non-invasive early screening of diabetes.", "date": "2026-05-21", "categories": []}
{"id": "2605.22335", "title": "Learning Causal Orderings for In-Context Tabular Prediction", "abstract": "In-context learning for tabular data sets strong predictive standards in observational settings; it however primarily relies on correlational structure, which becomes unreliable under distribution shift or intervention. While established methods to discover causal structure exist, they are often focused on structure identifiability and decoupled from the predictive architectures that could benefit from them.\n  To bridge these perspectives, we study how to simultaneously infer and enforce causal structure in the form of topological variable orderings into tabular prediction. Unlike standard architectures, our model TabOrder uses causal order-constrained attention, basing predictions only on features that precede a target under a learned causal order. Similar to causal discovery methods, TabOrder learns the optimal variable ordering in an unsupervised manner through a likelihood-based objective. We justify this choice under standard functional model classes and also study how sample missingness, a common challenge in tabular data, interacts with causal direction identification. Empirically, we confirm that TabOrder recovers accurate variable orderings while addressing prediction and imputation tasks, as well as gives insight into real-world biological data under intervention.", "date": "2026-05-21", "categories": []}
{"id": "2605.22462", "title": "From Correlation to Cause: A Five-Stage Methodology for Feature Analysis in Transformer Language Models", "abstract": "We propose a five-stage methodology for causal feature analysis in transformer language models (probe design, feature extraction, causal validation, robustness testing, and deployment integration) and demonstrate it end-to-end on GPT-2 small performing the Indirect Object Identification (IOI) task. Activation patching recovers the canonical IOI circuit (layer-9 head 9 alone gives recovery +1.02). A sparse autoencoder recovers per-name selective features with effect sizes of 30 to 50 activation units. Causal validation finds these features specifically but only partially causal: ablating fifteen of them leaves the model accurate on 98% of prompts. Two NLA-inspired evaluations strengthen this picture: the fifteen selective features explain only 31% of activation variance versus the SAE's 99.7%, and selectivity ratio anticorrelates with causal force (r = -0.56). Robustness testing under three distribution shifts finds that the circuit transfers cleanly but feature ablation effects degrade substantially, exposing a gap between detection robustness and causal robustness. A cost-based deployment evaluation (assumed $50/FN, $0.42/FP, 2% error rate) finds an optimal monitor configuration yielding $8.96 per 1000 queries against a $1000 baseline, a 99.1% saving. Optimal composition strategy varies with cost ratio and base rate. The conjunction of stages produces findings no single stage would.", "date": "2026-05-21", "categories": []}
{"id": "2602.04907", "title": "Causal Discovery from Heteroscedastic Stochastic Dynamical Systems under Imperfect Physical Models", "abstract": "Causal discovery is a data-driven paradigm for analyzing complex systems, while physics-based models, such as ordinary differential equations (ODEs), provide mechanistic structure for real-world dynamical processes. Integrating these paradigms can improve identifiability, stability, and robustness. However, real dynamical systems often exhibit cyclic interactions and nonstationarity, whereas many causal discovery methods rely on acyclicity, stationarity, or equilibrium assumptions. We propose an integrative causal discovery framework for dynamical systems that leverages partial physical knowledge through stochastic differential equations (SDEs). The drift term encodes known ODE dynamics, while the diffusion term captures unknown causal couplings beyond the prescribed physics. We develop a scalable sparsity-inducing maximum quasi-likelihood estimator with a theoretically justified stabilization technique to improve the optimization landscape. Under mild conditions, we establish causal graph recovery guarantees for both stable and unstable SDEs. We also analyze robustness of our causal graph estimate to ODE misspecification and clarify how the introduced stabilization technique balances numerical stability and statistical recoverability. Experiments on linear SDEs and nonlinear benchmarks, including Lotka-Volterra and Lorenz dynamics with acyclic and cyclic structures, show improved graph recovery and robustness over data-driven baselines. We also demonstrate practical utility on real-world epidemic data by reconstructing stochastic SIR dynamics within our causal discovery framework.", "date": "2026-02-03", "categories": []}
{"id": "2605.20396", "title": "Score-Based Causal Discovery of Latent Variable Causal Models", "abstract": "Identifying latent variables and the causal structure involving them is essential across various scientific fields. While many existing works fall under the category of constraint-based methods (with e.g. conditional independence or rank deficiency tests), they may face empirical challenges such as testing-order dependency, error propagation, and choosing an appropriate significance level. These issues can potentially be mitigated by properly designed score-based methods, such as Greedy Equivalence Search (GES) (Chickering, 2002) in the specific setting without latent variables. Yet, formulating score-based methods with latent variables is highly challenging. In this work, we develop score-based methods that are capable of identifying causal structures containing causally-related latent variables with identifiability guarantees. Specifically, we show that a properly formulated scoring function can achieve score equivalence and consistency for structure learning of latent variable causal models. We further provide a characterization of the degrees of freedom for the marginal over the observed variables under multiple structural assumptions considered in the literature, and accordingly develop both exact and continuous score-based methods. This offers a unified view of several existing constraint-based methods with different structural assumptions. Experimental results validate the effectiveness of the proposed methods.", "date": "2026-05-19", "categories": []}
{"id": "2605.20400", "title": "Understanding Deterioration Random Effects for Causal Discovery in Infrastructure Management", "abstract": "Infrastructure deterioration poses significant challenges for asset management, yet existing approaches rely on population-averaged models that overlook equipment-specific heterogeneity. We present a novel framework that combines Bayesian hierarchical hazard modeling with causal discovery to identify operational patterns that drive heterogeneous deterioration rates in pump equipment. Our approach first estimates pump-specific random effects $u_i$ using GPU-accelerated No-U-Turn Sampling (NUTS), achieving 3--5$\\times$ speedup over CPU implementations. We then employ DirectLiNGAM to discover causal relationships between 22 engineered time-series features and deterioration rates, stratified by positive ($u_i > 0$, faster deterioration) versus negative ($u_i \\leq 0$, slower deterioration) random effects. Analyzing 112 pumps with 92,861 observations over 650 days, we uncover striking heterogeneity: the negative group exhibits causal effects 400$\\times$ larger than the positive group, with standard deviation (std) showing a strong positive causal effect ($+1.515$) on deterioration rates in low-risk equipment. We validate linearity assumptions through NonlinearLiNGAM comparison and demonstrate practical scalability through GPU acceleration. Our findings enable targeted maintenance strategies by revealing that different operational regimes require fundamentally distinct management approaches, advancing predictive maintenance from population-averaged to heterogeneity-aware decision making.", "date": "2026-05-19", "categories": []}
{"id": "2605.20767", "title": "The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study", "abstract": "Large language models (LLMs) show potential as simulators of human behavior, offering a scalable way to study responses to interventions. However, because LLMs are trained largely on observational data, interventions in experiments with LLM-simulated synthetic users can induce unintended shifts in latent user attributes, causing user drift where the implicit simulated population differs across treatment conditions, potentially distorting effect estimates. We formalize the confounding or selection bias that can arise due to user drift and show how intervention-dependent shifts can inflate or attenuate observed differences in user responses under intervention. To diagnose confounding, we propose using negative control outcomes--attributes that should remain invariant under intervention--to identify distribution shifts across intervention conditions, providing evidence of user drift. To mitigate drift, we study adjusting the persona specification by eliciting additional confounders, finding that targeted, setting-relevant confounders can substantially reduce bias across survey-style and multi-turn agent evaluations.", "date": "2026-05-20", "categories": []}
{"id": "2605.20782", "title": "Causal Machine Learning Is Not a Panacea: A Roadmap for Observational Causal Inference in Health", "abstract": "Objective: The growing availability of large-scale observational clinical datasets and challenges in conducting randomized controlled trials have spurred enthusiasm in using causal machine learning (ML) for causal inference in observational data. We present a roadmap for applying causal ML to observational data. Materials and methods: We outline the importance of assessing validity assumptions within available data and applying causal ML responsibly for clinical experts using causal ML and ML practitioners with limited clinical expertise. Observations: Despite advances in causal ML, its limitations remain largely under-appreciated across disciplines. This gap in shared knowledge may impact the validity of findings. Discussion: Causal assumptions must be satisfied and modeling choices justified. Otherwise, these approaches risk producing biased or misleading results, with consequences for clinical research and patient care. Conclusion: Causal ML can be a powerful tool for generating causal hypotheses. We provide a template to strengthen the rigor and interpretability of causal analyses.", "date": "2026-05-20", "categories": []}
{"id": "2605.21058", "title": "A Dialogue between Causal and Traditional Representation Learning: Toward Mutual Benefits in a Unified Formulation", "abstract": "Causal representation learning (CRL) and traditional representation learning have largely developed along different trajectories. Traditional representation learning has been driven mainly by applications and empirical objectives, whereas CRL has focused more on theoretical questions, particularly identifiability. This difference in emphasis has created a gap between the two fields in terminology, problem formulation, and evaluation, limiting communication and sometimes leading to disconnected or redundant efforts. In this paper, we argue that these two fields should be brought into dialogue rather than treated as separate paradigms. To this end, we introduce a unified formulation in which the representation learning is characterized by two components: a task component, which specifies what information the learned representation is required to preserve, and a constraint component, which specifies what structure is imposed on the latent space. Under this formulation, the benefits run in both directions. CRL provides theoretical tools for understanding when structured latent constraints are useful or necessary, while traditional representation learning offers practical insights on task design and objective choice that can improve the development of CRL methods. To illustrate this interaction, we experimentally study how different task components affect the behavior of CRL methods under different structured constraints. Results on CausalVerse show that the effectiveness of causal constraints depends strongly on the tasks with which they are paired.", "date": "2026-05-20", "categories": []}
{"id": "2605.21066", "title": "Robust Personalized Recommendation under Hidden Confounding in MNAR", "abstract": "Recommender systems often rely on observational user--item interaction data, which is prone to selection bias due to users' selective interactions with items. Inverse propensity weighting and doubly robust estimators effectively mitigate selection bias under observed confounding, but are unreliable in the presence of hidden confounders. Existing approaches relying on randomized controlled trials (RCTs) or global sensitivity bounds are constrained in practice: RCTs demand costly experimental data, while global sensitivity bounds presume a uniformly bounded effect of unmeasured confounders on propensities through sensitivity analysis, thereby neglecting heterogeneity across user--item interactions. To overcome this limitation, we propose a novel framework, which estimates user--item level sensitivity bounds, thereby substantially relaxing the homogeneity assumption inherent in global sensitivity bounds named Personalized Unobserved-Confounding-aware Interaction Deconfounder (PUID). To ensure both robustness and predictive accuracy, we further develop an adversarial optimization strategy and propose a benchmark-guided variant (BPUID) that incorporates pre-trained models as stabilizing references. Extensive experiments on three real-world datasets demonstrate that our approach significantly outperforms global methods under hidden confounding, without requiring RCT data.", "date": "2026-05-20", "categories": []}
{"id": "2603.07018", "title": "TEA-Time: Transporting Effects Across Time", "abstract": "Treatment effects estimated from a randomized controlled trial are local not only to the study population but also to the time at which the trial was conducted. The literature on generalizing experimental findings to new populations is extensive, yet transporting effects across time has received far less attention, and even defining the target estimand is nonobvious. We formalize the transported average treatment effect under a separable temporal effects assumption, derive two identification strategies: replicated trials and common arm, and develop doubly robust, semiparametrically efficient estimators for each. Applied to a large archive of headline A/B tests, the common arm strategy is substantially more precise but exhibits systematic bias when the temporal factor depends on the gap between intervention and measurement rather than on measurement time alone, while the replicated trials strategy, which allows this dependence, tracks the ground truth more faithfully. Simulation studies investigate when each strategy is reliable and when it silently fails.", "date": "2026-03-07", "categories": []}
{"id": "2605.19135", "title": "Identifiable Multimodal Causal Representation Learning under Partial Latent Sharing", "abstract": "Causal representation learning (CRL) seeks to uncover meaningful latent variables and their corresponding causal structure from high-dimensional observational data. Although its significance, CRL identifiability remains a crucial property, as it ensures the recovery of the mechanisms behind the data generation process, and hence the interpretability and robustness of the representation. Proving identifiability in CRL is intrinsically difficult, and we address in this work an even more challenging setting: multimodality. We consider multimodal observed data with a latent partially shared structure. Each modality is generated, through non linear mixing functions, from a specific subset of causal latent variables. Under flexible assumptions and without imposing any parametric distribution on the latent variables, we establish component-wise identifiability guarantees for the causal latent representation. Our identifiability results, furthermore, apply to the undercomplete scenario where we have, for each modality, more observed than latent variables. To instantiate our theoretical analysis, we introduce a Wasserstein-based module to recover the partially shared latent structure. Due to its differentiability, the latter can be easily integrated into all types of architecture, only requiring minimal changes. Extensive experiments on synthetic and realistic datasets validate the superiority of our approach over SOTA methods.", "date": "2026-05-18", "categories": []}
{"id": "2605.19313", "title": "A Unified Framework for Structure-Aware Clustering and Heterogeneous Causal Graph Learning", "abstract": "In complex multivariate systems, interactions among variables are defined by dependency structures, often encoded as directed acyclic graphs ($\\text{DAGs}$). However, dependency structures can vary across subjects, and ignoring this structural heterogeneity introduces bias and obscures subpopulation-specific dependencies. To address this, we propose Directed Acyclic Graph-based Dependency Clustering via Alternating Direction Method of Multipliers (DAG-DC-ADMM), a unified framework built upon Structural Equation Modeling (SEM) that jointly learns cluster assignments and cluster-specific dependency structures. We encode acyclicity via a smooth constraint and integrate a groupwise truncated Lasso fusion penalty (gTLP) to cluster subjects based on their structural similarity. This yields a nonconvex optimization problem that incorporates sparsity, acyclicity, and structural consensus constraints. We address the nonconvexity by using the augmented Lagrangian method and solve it with an adapted version of the Alternating Direction Method of Multipliers (ADMM) for difference-of-convex programs. For certain graph structures, such as upper triangular adjacency matrices, our algorithm is guaranteed to converge to a Karush-Kuhn-Tucker (KKT) point. Experiments demonstrate that our method recovers cluster-specific causal dependency structures with a high true positive rate and a low false discovery rate. This capability enables the robust discovery of heterogeneous dependencies across subjects where the subpopulation label is unknown.", "date": "2026-05-19", "categories": []}
{"id": "2605.19830", "title": "Set-Valued Policy Learning", "abstract": "Conventional treatment policies map patient covariates to a single recommended intervention in order to maximize expected clinical outcomes. Although a rich body of causal inference methods has been developed to estimate such policies, point-valued recommendations can be highly sensitive to estimation uncertainty, model specification, and finite-sample variability, while typically providing little guidance about how confident one should be in the recommended action. In this work, we propose a set-valued policy learning paradigm for the multiple-treatment setting, in which policies output a set of plausible treatments rather than a single recommendation. This formulation enables intrinsic uncertainty quantification, with the size of the predicted set reflecting the degree of decision ambiguity. We extend the learning-to-defer framework to multiple treatments via a novel \\textit{greatest Lower Bound} method, and introduce \\textit{conformal policy learning}, which bridges the gap between unobserved ground-truth optimal treatments and estimated optimal treatment rules. Drawing on insights from the noisy-label literature, we develop a randomness-injection approach that guarantees marginal coverage without requiring assumptions on underlying black-box optimal treatment rules. Through experiments on synthetic data and a real-world application to In-Vitro Fertilization (IVF), we demonstrate that our methods produce robust and actionable policies that naturally incorporate clinical considerations while effectively balancing performance and reliability.", "date": "2026-05-19", "categories": []}
{"id": "2605.19947", "title": "Exploiting Non-Negativity in DAG Structure Learning", "abstract": "This work addresses the problem of learning directed acyclic graphs (DAGs) from nodal observations generated by a linear structural equation model. DAG learning is a central task in signal processing, machine learning, and causal inference, but it remains challenging because acyclicity is a global combinatorial property. Continuous acyclicity constraints have led to important algorithmic advances by replacing the discrete DAG constraint with smooth equality constraints. However, existing formulations still involve difficult non-convex optimization landscapes and may suffer from degenerate first-order optimality conditions. Here, we restrict attention to DAGs with non-negative edge weights and exploit this additional structure to obtain a simpler characterization of acyclicity. Building on this characterization, we formulate a regularized non-negative DAG learning problem and develop an algorithm based on the method of multipliers. We further analyze the benign optimization landscape induced by non-negativity. In the population regime, we show that the true DAG is the unique global minimizer of the proposed augmented-Lagrangian formulation; moreover, the landscape contains no spurious interior stationary points, and the true DAG is the only acyclic KKT point. Numerical experiments on synthetic and real-world data show that the proposed method improves over state-of-the-art continuous DAG-learning alternatives.", "date": "2026-05-19", "categories": []}
{"id": "2307.09575", "title": "Causal Influences over Social Learning Networks", "abstract": "This paper investigates causal influences between agents linked by a social graph and interacting over time. In particular, the work examines the dynamics of social learning models and distributed decision-making protocols, and derives expressions that reveal the causal relations between pairs of agents and explain the flow of influence over the network. The results turn out to be dependent on the graph topology and the level of information that each agent has about the inference problem they are trying to solve. Using these conclusions, the paper proposes an algorithm to rank the overall influence between agents to discover highly influential agents. It also provides a method to learn the necessary model parameters from raw observational data. The results and the proposed algorithm are illustrated by considering both synthetic data and real social media data.", "date": "2023-07-13", "categories": []}
{"id": "2601.21170", "title": "The Powers of Precision: Structure-Informed Detection in Complex Systems -- From Customer Churn to Seizure Onset", "abstract": "Emergent phenomena -- onset of epileptic seizures, sudden customer churn, or pandemic outbreaks -- often arise from hidden causal interactions in complex systems. We propose a machine learning method for their early detection that addresses a core challenge: unveiling and harnessing a system's latent causal structure despite the data-generating process being unknown and partially observed. The method learns an optimal feature representation from a one-parameter family of estimators -- powers of the empirical covariance or precision matrix -- offering a principled way to tune in to the underlying structure driving the emergence of critical events. A supervised learning module then classifies the learned representation. We prove structural consistency of the family and demonstrate the empirical soundness of our approach on seizure detection and churn prediction, attaining competitive results in both. Beyond prediction, and toward explainability, we ascertain that the optimal covariance power exhibits evidence of good identifiability while capturing structural signatures, thus reconciling predictive performance with interpretable statistical structure.", "date": "2026-01-29", "categories": []}
{"id": "2601.21350", "title": "Factored Causal Representation Learning for Robust Reward Modeling in RLHF", "abstract": "A reliable reward model is essential for aligning large language models with human preferences through reinforcement learning from human feedback. However, standard reward models are susceptible to spurious features that are not causally related to human labels. This can lead to reward hacking, where high predicted reward does not translate into better behavior. In this work, we address this problem from a causal perspective by proposing a factored representation learning framework that decomposes the model's contextual embedding into (1) causal factors that are sufficient for reward prediction and (2) non-causal factors that capture reward-irrelevant attributes such as length or sycophantic bias. The reward head is then constrained to depend only on the causal component. In addition, we introduce an adversarial head trained to predict reward from the non-causal factors, while applying gradient reversal to discourage them from encoding reward-relevant information. Experiments on both mathematical and dialogue tasks demonstrate that our method learns more robust reward models and consistently improves downstream RLHF performance over state-of-the-art baselines. Analyses on length and sycophantic bias further validate the effectiveness of our method in mitigating reward hacking behaviors.", "date": "2026-01-29", "categories": []}
{"id": "2602.02830", "title": "SC3D: Dynamic and Differentiable Causal Discovery for Temporal and Instantaneous Graphs", "abstract": "Discovering causal structures from multivariate time series is a key problem because interactions span across multiple lags and possibly involve instantaneous dependencies. Additionally, the search space of the dynamic graphs is combinatorial in nature. In this study, we propose Stable Causal Dynamic Differentiable Discovery (SC3D), a two-stage differentiable framework that jointly learns lag-specific adjacency matrices and, if present, an instantaneous directed acyclic graph (DAG). In Stage 1, SC3D performs edge preselection through node-wise prediction to obtain masks for lagged and instantaneous edges, whereas Stage 2 refines these masks by optimizing a likelihood with sparsity along with enforcing acyclicity on the instantaneous block. Numerical results across synthetic SVAR systems, nonlinear and chaotic benchmarks, nonstationary dynamics and real-world datasets demonstrate that SC3D achieves improved stability and more accurate recovery of both lagged and instantaneous causal structures compared to existing baselines.", "date": "2026-02-02", "categories": []}
{"id": "2603.06984", "title": "Masking Causality and Conditional Dependence", "abstract": "Many regulatory and analytic problems require that a prohibited variable influence a decision only through a designated allowable channel -- a conditional-independence requirement that arises in path-specific fairness, the handling of classified information, and the regulation of trading on non-public information, among other settings. Such requirements may be enforced either stratum-by-stratum or, more commonly (and more efficiently), through a single averaged constraint on the conditional effect. We study the resulting enforcement problem from two perspectives. From the regulator's side, we formulate causal masking as a linear program and show that averaged-constraint optimization almost surely produces policies that violate the stratum-wise requirement while satisfying the averaged one exactly. The gains from masking grow with confounding and outcome heterogeneity, and detection requires precisely the conditional-independence tests that average constraints aim to avoid. From the optimizer's side, the same construction shows that masked policies recover most of the reward of unconstrained exploitation while being far harder to detect, making them attractive in any setting where the basis of decisions is itself sensitive. Together, these results argue that regulating direct dependence through averaged statistics on observed decisions is structurally limited, and that meaningful enforcement must operate at the level of the decision rule itself.", "date": "2026-03-07", "categories": []}
{"id": "2603.07438", "title": "How Wrong Can Your Counterfactual Be? Quantifying Confounding Bias for Continuous Treatments without a Control Group", "abstract": "Stress testing poses a causal question: how would portfolio credit losses change if the macroeconomy followed an adverse counterfactual path? Yet standard practice remains predictive and might be therefore vulnerable to omitted-variable bias. We propose a partial identification framework for causal stress testing in panel data with a continuous common treatment and no control group. By assuming that the unobserved confounder affects outcome and macro variables additively, we derive a closed-form confounding envelope parameterized by two interpretable sensitivity parameters. We further analyze two practical estimators -- recursive rollout and direct multi-horizon prediction -- derive non-asymptotic error bounds, and characterize when recursive compounding makes direct estimation preferable. For inference, we combine the identification envelope with importance-weighted conformal prediction, yielding finite-sample intervals that separate estimation uncertainty from identification uncertainty under covariate shift. In semi-synthetic experiments built from real U.S. unemployment paths, standard high-accuracy predictive models remain causally biased and substantially under-cover, whereas the proposed framework achieves near-nominal coverage across stress horizons.", "date": "2026-03-08", "categories": []}
{"id": "2605.11365", "title": "Causal Bias Detection in Generative Artificial Intelligence", "abstract": "Automated systems built on artificial intelligence (AI) are increasingly deployed across high-stakes domains, raising critical concerns about fairness and the perpetuation of demographic disparities that exist in the world. In this context, causal inference provides a principled framework for reasoning about fairness, as it links observed disparities to underlying mechanisms and aligns naturally with human intuition and legal notions of discrimination. Prior work on causal fairness primarily focuses on the standard machine learning setting, where a decision-maker constructs a single predictive mechanism $f_{\\widehat Y}$ for an outcome variable $Y$, while inheriting the causal mechanisms of all other covariates from the real world. The generative AI setting, however, is markedly more complex: generative models can sample from arbitrary conditionals over any set of variables, implicitly constructing their own beliefs about all causal mechanisms rather than learning a single predictive function. This fundamental difference requires new developments in causal fairness methodology. We formalize the problem of causal fairness in generative AI and unify it with the standard ML setting under a common theoretical framework. We then derive new causal decomposition results that enable granular quantification of fairness impacts along both (a) different causal pathways and (b) the replacement of real-world mechanisms by the generative model's mechanisms. We establish identification conditions and introduce efficient estimators for causal quantities of interest, and demonstrate the value of our methodology by analyzing race and gender bias in large language models across different datasets.", "date": "2026-05-12", "categories": []}
{"id": "2605.16620", "title": "SCOUT: Cyclic Causal Discovery Under Soft Interventions with Unknown Targets", "abstract": "Learning causal relationships between variables from data is a fundamental research area with many applications across disciplines. Most existing causal discovery algorithms rely on the assumptions that (i) the underlying system is acyclic, (ii) the exogenous noise variables are Gaussian, and (iii) the intervention targets for the data-generating experiments are known. While these assumptions simplify the analysis, they are violated in real-life systems. Most existing methods that address these issues either assume the underlying model is linear or are constrained to operate in limited interventional settings. To that end, we propose SCOUT, a novel causal discovery framework for learning nonlinear cyclic causal relationships from soft interventional data with unknown targets. Our approach maximizes the data log-likelihood to recover the graph structure, using two normalizing-flow architectures: contractive residual flows and neural spline flows. Through experiments on synthetic and real-world data, we show that SCOUT outperforms state-of-the-art methods in both causal graph recovery and unknown target recovery across various interventional and noise settings.", "date": "2026-05-15", "categories": []}
{"id": "2605.16828", "title": "Prediction-Intervention Games and Invariant Sets", "abstract": "We consider the following two-player game: using observational data, the leader chooses a prediction function for a response variable $Y$ from given covariates. The follower then reacts with an intervention on some covariates in the underlying structural causal model to maximize their own objective. The leader knows the intervention targets, but may have limited knowledge of the follower's objective. We call this setup a prediction-intervention game, a special case of a Stackelberg game. Finding an optimal strategy for the leader is generally difficult. To avoid severe performance loss, the leader may base their prediction on the causal parents of $Y$, or more generally on an invariant subset of covariates. We prove, for two common classes of follower objectives, that predictors based on the stable blanket, a specific invariant subset, are always better or as good as those based on the causal parents. We further upper bound the leader's post-intervention risk by a worst-case risk over allowed interventions and strengthen existing distribution generalization results to analyze this bound: we give sufficient conditions under which stable-blanket predictors are worst-case optimal, and show by examples that these conditions cannot in general be dropped. Finally, we discuss practical strategies for settings with known and unknown graph, and test them on simulated and real-world data.", "date": "2026-05-16", "categories": []}
{"id": "2605.16927", "title": "From Static Risk to Dynamic Trajectories: Toward World-Model-Inspired Clinical Prediction", "abstract": "Clinical decision-making is a feedback system where risk estimates influence treatment, which in turn changes disease trajectories, and both shape clinicians' measurement practices. Static prediction often fails clinically: models trained on observational care logs conflate disease biology with clinician behavior, particularly under treatment confounder feedback and irregular or informative observation. This Review focuses on intervention-aware disease trajectory modeling in clinical AI--methods estimating patient-specific longitudinal disease evolution and assessing trajectory changes under alternative treatments. We organize the field around six linked components: three decision tasks (factual forecasting, counterfactual estimation, policy evaluation) and three data-generating mechanisms (disease evolution, treatment assignment, observation process) that determine identifiability. We present the first unified framework bridging forecasting, counterfactual trajectories, and policy evaluation across discrete/continuous time, explicitly addressing treatment assignment, time-varying confounding, and observation bias. We synthesize key method families (multistate/joint models, temporal point-process, deep sequence architectures, longitudinal causal inference), map them to relevant components, and align evaluation with claim strength via overlap diagnostics, uncertainty quantification, off-policy robustness, and target-trial validation. This synthesis advances benchmark prediction to decision-grade clinical evidence, enabling treatment-sensitive individualized futures, pre-deployment policy stress-testing, and safer closed-loop learning health systems that adapt/abstain when evidence is insufficient.", "date": "2026-05-16", "categories": []}
{"id": "2605.16989", "title": "Decision-Aware Proximal Bridge Learning for Optimal Treatment Selection", "abstract": "Individualized treatment selection with continuous actions requires accurate causal response estimation in decision-relevant regions, rather than uniformly over the entire action space. Estimating a global causal response surface and then choosing the treatment that maximizes it can therefore be suboptimal, since standard estimation objectives allocate modeling effort according to the observed treatment distribution rather than the regions that determine the optimal decision. While decision-aware approaches have been studied in unconfounded settings, this problem remains underexplored in proximal causal inference, where proxy variables and bridge functions enable identification under suitable assumptions even in the presence of hidden confounding. Despite recent progress, proximal methods have primarily focused on treatment-effect and potential-outcome estimation rather than treatment selection and optimal decision-making. To bridge this gap, we introduce a policy-targeted weighted bridge loss that emphasizes decision-relevant treatment regions while retaining global stabilization. We prove a regret bound showing that the proposed weighted bridge loss controls treatment-selection regret through a weighted ill-posedness constant. We instantiate the framework in decision-aware variants of several proximal bridge solvers, yielding practical algorithms that alternate between weighted bridge estimation, response-surface projection, policy update, and weight refinement. Empirically, we find that decision-aware weighting reduces regret across several bridge solvers, suggesting improved treatment selection in proximal settings.", "date": "2026-05-16", "categories": []}
{"id": "2605.17465", "title": "TriOpt: A Scalable Algorithm for Linear Causal Discovery", "abstract": "Learning causal relations from observational data is challenging because the graph search space grows super-exponentially with the number of variables. Ordering-based methods reduce this space by first identifying the topological ordering, whereas continuous optimization methods explore most likely regions of the space by casting DAG learning as a differentiable objective with an acyclicity constraint. Despite their conceptual appeal, both paradigms face significant scalability limitations in high-dimensional settings, restricting their practical applicability. In this work, we introduce a new formulation for linear causal discovery that tightly integrates these two paradigms to achieve substantial gains in scalability without sacrificing accuracy. Our approach, TriOpt, decomposes the problem into two efficient stages. First, it recovers the topological ordering by exploiting the Sherman-Morrison rank-1 downdate together with the additive structure of linear kernels, enabling fast and scalable ordering estimation. Second, given this ordering, we reformulate structure learning as a convex continuous optimization problem that entirely avoids the need for enforcing costly acyclicity constraints. We theoretically show that, under the true ordering, TriOpt exactly recovers the underlying linear DAG. Empirically, across synthetic, semi-synthetic, and real-world datasets, TriOpt achieves orders-of-magnitude speedups over state-of-the-art linear causal discovery methods in high-dimensional regimes, while maintaining comparable or superior accuracy.", "date": "2026-05-17", "categories": []}
{"id": "2605.17528", "title": "CasualSynth: Generating Structurally Sound Synthetic Data", "abstract": "Large Language Models (LLMs) generate realistic synthetic data but offer no guarantee that their outputs respect the causal mechanisms governing the target domain. We introduce CausalSynth, a framework that decouples causal structure generation from semantic realization, yielding synthetic data that is both causally valid and linguistically rich. The framework operates in three phases. First, a Structural Causal Model (SCM) - a tuple of structural equations defined over a directed acyclic graph (DAG) generates causal skeletons, i.e., variable assignments that satisfy the Global Markov Property of the governing DAG, via ancestral sampling. Second, an LLM acts as a constrained \\emph{realizer}, a conditional translator that maps each skeleton to a high-dimensional observation such as a clinical note or a transaction log. Third, an Iterative Consistency Verification module detects structural violations through deterministic extraction and feeds targeted corrections back to the LLM, forming a closed-loop refinement process. We identify the Semantic Backdoor problem the systematic tendency of LLMs to override imposed causal facts with pre-training priors -- and prove that our iterative mechanism reduces the resulting selection bias relative to standard rejection sampling. On three causal benchmarks (ASIA, ALARM, and MIMIC-Struct), CausalSynth preserved conditional independencies with false-positive rates near the nominal $\\alpha=0.05$ level and achieved realizability rates above 96% with 70B-parameter LLM backbones. The framework additionally supports principled interventional and counterfactual generation through noise retention and graph mutilation.", "date": "2026-05-17", "categories": []}
{"id": "2605.18327", "title": "Causely: A Causal Intelligence Layer for Enterprise AI A Benchmark Study on SRE and Reliability Workflows", "abstract": "AI agents deployed into SRE workflows currently derive their understanding of environment state from raw observability telemetry at query time, paying a semantic-interpretation tax in tokens, latency, and inferential reliability. We propose Causely, a causal intelligence layer that maintains a structured representation of environment topology, attribute dependencies, and causal relationships that are anchroed to a ontological representation of the managed environment. Causely transforms raw telemetry into a live, queryable model providing the semantic and causal foundation AI agents require to diagnose, evaluate impact, and act safely in production. We evaluate this value proposition through a benchmark study conducted in a controlled setting with injected faults in a 24-microservice OpenTelemetry demo application. Our experiments compare four agent configurations (Claude Code, OpenAI Codex, HolmesGPT with Sonnet and Gemini backends). Experiments are run with and without access to Causely under two scenarios: an active incident and a healthy baseline. On the active-fault scenario, causal grounding reduces mean time-to-diagnosis by 63\\%, mean token consumption by 60\\%, and mean tool-call count by 78\\%, compressing the investigation footprint by 4.8$\\times$ and lowering direct API cost per run by 57\\%; root-cause-diagnosis accuracy rises from 75\\% to 100\\%.", "date": "2026-05-18", "categories": []}
{"id": "2605.18459", "title": "Adaptive Experimentation for Censored Survival Outcomes", "abstract": "Adaptive experimentation enables efficient estimation of causal effects, but existing methods are not designed for survival data with censoring, where event times are only partially observed (e.g., overall survival in cancer trials but with dropout). In this paper, we develop a novel framework for adaptive experimentation to estimate causal effects under right censoring. For this, we derive the semiparametric efficiency bound for the average survival effect curve as a function of the treatment allocation policy and thereby obtain a closed-form efficiency-optimal allocation policy. The policy generalizes classical Neyman allocation to survival settings by prioritizing patient strata where both event and censoring dynamics induce high uncertainty. Building on this, we propose the Adaptive Survival Estimator (ASE), an adaptive framework that learns the allocation policy and estimates the average survival effect curve sequentially. Our framework has three main benefits: (i) it accommodates arbitrary machine learning models for nuisance estimation; (ii) it is guided by a closed-form efficiency-optimal allocation policy; and (iii) it admits strong theoretical guarantees, including asymptotic normality via a martingale central limit theorem. We demonstrate our framework across various numerical experiments to show consistent efficiency gains over uniform randomization and censoring-agnostic baselines.", "date": "2026-05-18", "categories": []}
{"id": "2605.15353", "title": "PACER: Acyclic Causal Discovery from Large-Scale Interventional Data", "abstract": "Inferring the structure of directed acyclic graphs (DAGs) from data is a central challenge in causal discovery, particularly in modern high-dimensional settings where large-scale interventional data are increasingly available. While interventional data can improve identifiability, existing methods remain limited by soft acyclicity constraints, leading to optimization over invalid cyclic graphs, numerical instability, and reduced scalability. We introduce PACER (Perturbation-driven Acyclic Causal Edge Recovery), a scalable framework for causal discovery that guarantees acyclicity by construction. PACER parameterizes a distribution over DAGs through a joint model of variable permutations and edge probabilities, enabling direct optimization over valid causal structures without surrogate penalties. The framework supports a unified likelihood-based treatment of observational and interventional data, flexible conditional density models, and the incorporation of structural prior knowledge. For linear-Gaussian mechanisms, we derive closed-form expressions for the expected interventional log-likelihood and its gradients, yielding substantial computational gains. Empirically, PACER matches or exceeds state-of-the-art methods on protein signaling and large-scale genetic perturbation benchmarks, while scaling efficiently to networks with thousands of variables and achieving up to two orders of magnitude speedups over penalty-based differentiable approaches. These results demonstrate that exact and scalable causal discovery from high-dimensional perturbation data is achievable through principled search space design.", "date": "2026-05-14", "categories": []}
{"id": "2605.15967", "title": "Deterministic Event-Graph Substrates as World Models for Counterfactual Reasoning", "abstract": "We study event-graph substrates: a class of world models that represent agent state as an append-only log of typed RDF triples and answer counterfactual queries by forking the log under a structured intervention vocabulary. Substrates are inspectable at the triple level, support exact counterfactuals, and transfer across domains without learned components. We formalize the class, prove a duality between explanatory and counterfactual queries that reduces both to the same causal-ancestor traversal, and evaluate a 1,400-line CLEVRER-DSL interpreter atop a domain-agnostic substrate runtime at full CLEVRER validation scale (n=75,618). The substrate exceeds the NS-DR symbolic oracle on all four per-question categories (by 9.89, 20.26, 17.65, and 0.80 percentage points), and exceeds the parametric ALOE baseline on descriptive and explanatory while lagging on predictive and counterfactual. We also introduce twin-EventLog, a 500-specification Park-canonical Smallville counterfactual benchmark on which the substrate exceeds Llama-3.1-8B with full context by 18.80 points joint accuracy.", "date": "2026-05-15", "categories": []}
{"id": "2509.20846", "title": "Causal Time Series Generation via Diffusion Models", "abstract": "Time series generation (TSG) synthesizes realistic sequences and has achieved remarkable success. Among TSG, conditional models generate sequences given observed covariates, however, such models learn observational correlations without considering unobserved confounding. In this work, we propose a causal perspective on conditional TSG and introduce causal time series generation as a new TSG task family, formalized within Pearl's causal ladder, extending beyond observational generation to include interventional and counterfactual settings. To instantiate these tasks, we develop CaTSG, a unified diffusion-based framework with backdoor-adjusted guidance that causally steers sampling toward desired interventions and individual counterfactuals while preserving observational fidelity. Specifically, our method derives causal score functions via backdoor adjustment and the abduction-action-prediction procedure, thus enabling principled support for all three levels of TSG. Extensive experiments on both synthetic and real-world datasets show that CaTSG achieves superior fidelity and also supporting interventional and counterfactual generation that existing baselines cannot handle. Overall, we propose the causal TSG family and instantiate it with CaTSG, providing an initial proof-of-concept and opening a promising direction toward more reliable simulation under interventions and counterfactual generation.", "date": "2025-09-25", "categories": []}
{"id": "2510.13583", "title": "On the Identifiability of Causal Graphs with the Invariance Principle", "abstract": "Causal discovery from i.i.d. observational data is known to be generally ill-posed. We demonstrate that if we have access to the distribution {induced} by a structural causal model, and additional data from (in the best case) \\textit{only two} environments that sufficiently differ in the noise statistics, the unique causal graph is identifiable. Notably, this is the first result in the literature that guarantees the entire causal graph recovery with a constant number of environments and arbitrary nonlinear mechanisms. Our only constraint is the Gaussianity of the noise terms; however, we propose potential ways to relax this requirement. Of interest on its own, we expand on the well-known duality between independent component analysis (ICA) and causal discovery; recent advancements have shown that nonlinear ICA can be solved from multiple environments, at least as many as the number of sources: we show that the same can be achieved for causal discovery while having access to much less auxiliary information.", "date": "2025-10-15", "categories": []}
{"id": "2602.09969", "title": "Causal Multi-Task Demand Learning", "abstract": "We study a canonical multi-task demand-learning problem motivated by retail pricing, where a firm seeks to estimate heterogeneous linear price-response functions across multiple decision contexts. Each context is described by rich covariates but exhibits limited price variation, motivating transfer learning across tasks. A central challenge in leveraging cross-task transfer is endogeneity: prices may be arbitrarily correlated with unobserved task-level demand determinants across tasks.\n  We propose a new meta-learning framework that identifies the conditional mean of task-specific causal demand parameters given a subset of task-specific observables despite such confounding, assuming that each task contains at least two distinct locally exogenous price points. This subset is carefully designed to include all of the prices to address cross-task confounding, while masking two demand outcomes that provide randomized supervision to address identifiability issues arising from the inclusion of all prices. We show that this information design is maximally uniformly valid, in that any refinement of the conditioning set that reveals withheld-outcome information is not guaranteed to identify the conditional mean causal target. We validate our method on real and synthetic data, demonstrating improved recovery of demand responses relative to standard transfer-learning baselines.", "date": "2026-02-10", "categories": []}
{"id": "2602.20571", "title": "CausalReasoningBenchmark: A Real-World Benchmark for Disentangled Evaluation of Causal Identification and Estimation", "abstract": "Many benchmarks for automated causal inference evaluate a system's performance based on a single numerical output, such as an Average Treatment Effect (ATE). This approach conflates two distinct steps in causal analysis: identification - formulating a valid research design under stated assumptions - and estimation - implementing that design numerically on finite data. We introduce CausalReasoningBenchmark, a benchmark of 173 queries across 132 real-world datasets, curated from 79 peer-reviewed research papers and three widely-used causal-inference textbooks. For each query a system must produce (i) a structured identification specification that names the strategy, the treatment, outcome, and control variables, and all design-specific elements, and (ii) a point estimate with a standard error. By scoring these two components separately, our benchmark enables granular diagnosis: it distinguishes failures in causal reasoning from errors in numerical execution. Baseline results with a state of the art LLM show that, while the model correctly identifies the high-level strategy in 79% of cases, full identification-specification correctness drops to only 34%, revealing that the bottleneck lies in the nuanced details of research design rather than in computation. CausalReasoningBenchmark is publicly available on Hugging Face and is designed to foster the development of more robust automated causal-inference systems.", "date": "2026-02-24", "categories": []}
{"id": "2605.15133", "title": "Causal Foundation Models with Continuous Treatments", "abstract": "Causal inference, estimating causal effects from observational data, is a fundamental tool in many disciplines. Of particular importance across a variety of domains is the continuous treatment setting, where the variable of intervention has a continuous range. This setting is far less explored and represents a substantial shift from the binary treatment setting, with models needing to represent effects across a continuum of treatment values. In this paper, we present the first causal foundation model for the continuous treatment setting. Our model meta-learns the ability to predict causal effects across a wide variety of unseen tasks without additional training or fine-tuning. First, we design a novel prior over data-generating processes with continuous treatment variables in order to generate a rich causal training corpus. We then train a transformer to reconstruct individual treatment-response curves given only observational data, leveraging in-context learning to amortize expensive Bayesian posterior inference. Our model achieves state-of-the-art performance on individual treatment-response curve reconstruction tasks compared to causal models which are trained specifically for those tasks.", "date": "2026-05-14", "categories": []}
{"id": "2410.14375", "title": "Causal Fine-Tuning under Latent Confounded Shift", "abstract": "Adapting to latent confounded shift remains a core challenge in modern AI. This setting is driven by hidden variables that induce spurious correlations between inputs and outputs during training, leading models to rely on non-causal shortcuts. For example, a model may learn to treat metadata (e.g., data source like \"Amazon\") as a proxy for positive sentiment, causing failure when the source becomes predominantly negative during deployment. To address this latent confounded shift, we introduce Causal Fine-Tuning(CFT). Using a structural causal model as an inductive bias, we derive sufficient identification conditions that motivate a fine-tuning objective for decomposing representations into high-level stable and low-level shift-sensitive components. Instantiating this framework in BERT, we show that learning such causal/spurious representations and adjusting them accordingly yield a more robust predictor. Experiments on spurious correlation injection attacks in text demonstrate that our method outperforms black-box domain generalization baselines, highlighting the benefits of explicitly modeling causal structure.", "date": "2024-10-18", "categories": []}
{"id": "2604.26070", "title": "Observable Neural ODEs for Identifiable Causal Forecasting in Continuous Time", "abstract": "Causal inference in continuous-time sequential decision problems is challenged by hidden confounders. We show that, in latent state-space models with time-varying interventions, observability of the latent dynamics from observed data is necessary for identifying dynamic treatment effects, linking control-theoretic observability to causal identifiability, even when hidden confounders affect both treatments and outcomes. We derive a continuous-time adjustment formula expressing potential outcome distributions under treatment trajectories via the measurement model, latent dynamics, and the filtering distribution over latent states given observed histories. We propose Observable Neural ODEs (ObsNODEs), Neural ODE models in observable normal form for causal forecasting. ObsNODEs learn continuous-time dynamics with states reconstructible from observations, enabling outcome prediction under alternative treatment paths. Experiments on synthetic cancer data, semi-synthetic data based on MIMIC-IV, and real-world sepsis data show strong performance over recent sequence models.", "date": "2026-04-28", "categories": []}
{"id": "2605.12835", "title": "PROMETHEUS: Automating Deep Causal Research Integrating Text, Data and Models", "abstract": "Large language models can extract local causal claims from text, but those claims become more useful when organized as persistent, navigable world models rather than as flat summaries. We introduce PROMETHEUS, a framework that turns retrieved literature, filings, reviews, reports, agent traces, source data, code, simulations, and scientific models into causal atlases: sheaf-like families of local causal predictive-state models over an explicit cover of a research substrate. Each local region contains causal episodes, structured claim tables, predictive tests, support statistics, and provenance; restriction maps compare overlapping regions; gluing diagnostics expose agreement, drift, contradiction, and underdetermination. The resulting Topos World Model is not a single universal graph. It is a research instrument for navigating what a corpus says, where it says it, how strongly it is supported, and where local claims fail to assemble into a coherent global view. Three literature-atlas case studies -- ocean-temperature impacts on marine populations, GLP-1 weight-loss evidence, and resveratrol/red-wine health-benefit claims -- illustrate deep causal research from text with explicit locality, evidence, persistent state, and gluing tension. Four grounded-counterfactual case studies -- a Nature Climate Change microplastics forcing paper, an Indus Valley hydrology paper with VIC-derived figure data and model code, the canonical Sachs protein-signaling study with single-cell perturbation data, and a Nature singing-mouse study with MAPseq projection matrices -- show a stronger mode: when a paper ships source data, simulation outputs, or code, PROMETHEUS can evaluate a counterfactual against that scientific substrate and then rebuild the sheaf world model around the", "date": "2026-05-13", "categories": []}
{"id": "2605.12924", "title": "IV-ICL: Bounding Causal Effects with Instrumental Variables via In-Context Learning", "abstract": "The instrumental-variables (IV) setting is standard for partial identification of causal effects when unobserved confounding makes point identification impossible. Existing approaches face methodological bottlenecks: closed-form bound estimands are required -- e.g., Balke-Pearl equations in binary IV -- and even when available, designing accurate estimators requires manual effort tailored to each estimand. While direct Bayesian inference of the causal effects, instead of the bounds, circumvents these challenges, it is often computationally intensive and suffers from high prior sensitivity or under-dispersed posteriors. As a remedy, we introduce IV-ICL, an amortized Bayesian in-context learning method that learns the marginal posterior distribution of the causal effects directly and derives bounds as its quantiles. Unlike standard variational inference that optimizes exclusive KL divergence, amortized Bayesian inference minimizes the expected inclusive KL, a mass-covering objective. We empirically observe that optimizing inclusive KL can recover the entire identified set across diverse data-generating processes, while exclusive-KL (e.g. with variational inference) of the same Bayesian formulation collapses onto a single mode and fails to cover the identified set. We evaluate IV-ICL on synthetic and semi-synthetic IV benchmarks and show it produces intervals that are more reliably valid and more informative compared to efficient semi-parametric, Bayesian, and plug-in baselines, at 20-500x lower inference time. Beyond methodology, we propose a procedure to convert randomized controlled trials into IV benchmarks with provably preserved ground-truth causal effects that enables a more realistic evaluation of partial-identification methods.", "date": "2026-05-13", "categories": []}
{"id": "2605.13589", "title": "Causal Learning with the Invariance Principle", "abstract": "Causal discovery, the problem of inferring the direction of causality, is generally ill-posed. We use the language of structural causal models (SCM) to show that assuming that the causal relations are acyclic and invariant across multiple environments (e.g., the way minimum wage affects employment rate is stable across different geographical regions), \\textit{only} two auxiliary environments are sufficient to infer the causal graph for arbitrary nonlinear mechanisms. Moreover, we demonstrate that this implies identifiability of the SCM functional mechanisms: as a corollary, we show that \\textit{two} auxiliary environments are sufficient to guarantee correct counterfactual inference. We empirically support our theoretical results on synthetic data.", "date": "2026-05-13", "categories": []}
{"id": "2501.06857", "title": "A Counterfactual Cause in Situation Calculus", "abstract": "Recently, Batusov and Soutchanski proposed a notion of actual achievement cause in the situation calculus, amongst others, they can determine the cause of quantified effects in a given action history. While intuitively appealing, this notion of cause is not defined in a counterfactual perspective. In this paper, we propose a notion of cause based on counterfactual analysis. In the context of action history, we show that our notion of cause generalizes naturally to a notion of achievement cause. We analyze the relationship between our notion of the achievement cause and the achievement cause by Batusov and Soutchanski. Finally, we relate our account of cause to Halpern and Pearl's account of actual causality. Particularly, we note some nuances in applying a counterfactual viewpoint to disjunctive goals, a common thorn in definitions of actual causes.", "date": "2025-01-12", "categories": []}
{"id": "2505.13770", "title": "Ice Cream Doesn't Cause Drowning: Benchmarking LLMs Against Statistical Pitfalls in Causal Inference", "abstract": "Reliable causal inference is essential for making decisions in high-stakes areas like medicine, economics, and public policy. However, it remains unclear whether large language models (LLMs) can handle rigorous and trustworthy statistical causal inference. Current benchmarks usually involve simplified tasks. For example, these tasks might only ask LLMs to identify semantic causal relationships or draw conclusions directly from raw data. As a result, models may overlook important statistical pitfalls, such as Simpson's paradox or selection bias. This oversight limits the applicability of LLMs in the real world. To address these limitations, we propose CausalPitfalls, a comprehensive benchmark designed to rigorously evaluate the capability of LLMs in overcoming common causal inference pitfalls. Our benchmark features structured challenges across multiple difficulty levels, each paired with grading rubrics. This approach allows us to quantitatively measure both causal reasoning capabilities and the reliability of LLMs' responses. We evaluate models using two protocols: (1) direct prompting, which assesses intrinsic causal reasoning, and (2) code-assisted prompting, where models generate executable code for explicit statistical analysis. Additionally, we validate the effectiveness of this judge by comparing its scoring with assessments from human experts. Our results reveal significant limitations in current LLMs when performing statistical causal inference. The CausalPitfalls benchmark provides essential guidance and quantitative metrics to advance the development of trustworthy causal reasoning systems.", "date": "2025-05-19", "categories": []}
{"id": "2506.02084", "title": "Adversarial Causal Tuning for Realistic Time-series Generation", "abstract": "We address the problem of generating simulated, yet realistic, time-series data from a causal model with the same observational and interventional distributions as a given real dataset (probabilistic causal digital twin). While non-causal models (e.g., GANs) also strive to simulate realistic data, causal models are fundamentally more powerful, able to simulate the effect of interventions (what-if scenarios), optimize decisions, perform root-cause analysis, and counterfactual causal reasoning. We introduce the Adversarial Causal Tuning (ACT) methodology, which outputs the optimal causal model that fits the data, along with a quantification of the goodness-of-fit. The returned causal model can then be employed to simulate new data or to perform other causal reasoning tasks. ACT adopts ideas from Generative Adversarial Network training and AutoML to search for optimal causal pipelines and discriminators that detect deviations between the distributions of real and simulated data. It also adapts a permutation testing procedure from established causal tuning methods to penalize models for complexity. Through extensive experiments on real, semi-synthetic, and synthetic datasets, we show that (a) employing multiple optimized discriminators is paramount for selecting the optimal causal models and quantifying goodness-of-fit, (b) ACT selects the optimal causal model in synthetic datasets while avoiding overfitting, generating data indistinguishable from the true data distribution (c) all state-of-the-art generative and causal simulation methods, exhibit room for improvement in reproducing real data distributions; generating realistic temporal data is still an open research challenge.", "date": "2025-06-02", "categories": []}
{"id": "2507.03622", "title": "Localising Dropout Variance in Twin Networks", "abstract": "Accurate individual treatment-effect estimation demands not only reliable point predictions but also uncertainty measures that help practitioners \\emph{locate} the source of model failure. We introduce a layer-wise variance decomposition for deep twin-network models: by toggling Monte Carlo Dropout independently in the shared encoder and the outcome heads, we split total predictive variance into an \\emph{encoder component} ($\\sigma_{\\mathrm{enc}}^2$) and a \\emph{head component} ($\\sigma_{\\mathrm{head}}^2$), with $\\sigma_{\\mathrm{enc}}^2 + \\sigma_{\\mathrm{head}}^2 \\approx \\sigma_{\\mathrm{tot}}^2$ by the law of total variance. Across three synthetic covariate-shift regimes, the encoder component dominates under distributional shift ($\\rho_{\\mathrm{enc}}=0.53$) while the head component becomes informative only once encoder uncertainty is controlled. On a real-world twins cohort with induced multivariate shift, only $\\sigma_{\\mathrm{enc}}^2$ spikes on out-of-distribution samples and becomes the primary error predictor ($\\rho_{\\mathrm{enc}}\\!\\approx\\!0.89$), while $\\sigma_{\\mathrm{head}}^2$ remains flat. The decomposition adds negligible cost over standard MC Dropout and provides a practical diagnostic for deciding whether to collect more diverse covariates or more outcome data.", "date": "2025-07-04", "categories": []}
{"id": "2510.16127", "title": "Learning density ratios in causal inference using Bregman-Riesz regression", "abstract": "The ratio of two probability density functions is a fundamental quantity that appears in many areas of statistics and machine learning, including causal inference, reinforcement learning, covariate shift, outlier detection, independence testing, importance sampling, and diffusion modeling. Naively estimating the numerator and denominator densities separately using, e.g., kernel density estimators, can lead to unstable performance and suffer from the curse of dimensionality as the number of covariates increases. For this reason, several methods have been developed for estimating the density ratio directly based on (a) Bregman divergences or (b) recasting the density ratio as the odds in a probabilistic classification model that predicts whether an observation is sampled from the numerator or denominator distribution. Additionally, the density ratio can be viewed as the Riesz representer of a continuous linear map, making it amenable to estimation via (c) minimization of the so-called Riesz loss, which was developed to learn the Riesz representer in the Riesz regression procedure in causal inference. In this paper we show that all three of these methods can be unified in a common framework, which we call Bregman--Riesz regression. We further show how data augmentation techniques can be used to apply density ratio learning methods to causal problems, where the numerator distribution typically represents an unobserved intervention. We show through simulations how the choice of Bregman divergence and data augmentation strategy can affect the performance of the resulting density ratio learner. A Python package is provided for researchers to apply Bregman--Riesz regression in practice using gradient boosting, neural networks, and kernel methods.", "date": "2025-10-17", "categories": []}
{"id": "2510.18071", "title": "Arbitrated Indirect Treatment Comparisons", "abstract": "Matching-adjusted indirect comparison (MAIC) has been increasingly employed in health technology assessments (HTA). By reweighting subjects from a trial with individual participant data (IPD) to match the covariate summary statistics of another trial with only aggregate data (AgD), MAIC facilitates the estimation of a treatment effect defined with respect to the AgD trial population. This manuscript introduces a new class of methods, termed arbitrated indirect treatment comparisons, designed to address the ``MAIC paradox'' -- a phenomenon highlighted by Jiang et al.~(2025). The MAIC paradox arises when different sponsors, analyzing the same data, reach conflicting conclusions regarding which treatment is more effective. The underlying issue is that each sponsor implicitly targets a different population. To resolve this inconsistency, the proposed methods focus on estimating treatment effects in a common target population, specifically chosen to be the overlap population.", "date": "2025-10-20", "categories": []}
{"id": "2602.04611", "title": "Targeted Synthetic Control Method", "abstract": "The synthetic control method (SCM) estimates causal effects in panel data with a single-treated unit by constructing a counterfactual outcome as a weighted combination of untreated control units that matches the pre-treatment trajectory. In this paper, we introduce the targeted synthetic control (TSC) method, a new two-stage estimator that directly estimates the counterfactual outcome. Specifically, our TSC method (1) yields a targeted debiasing estimator, in the sense that the targeted updating refines the initial weights to produce more stable weights; and (2) ensures that the final counterfactual estimation is a convex combination of observed control outcomes to enable direct interpretation of the synthetic control weights. TSC is flexible and can be instantiated with arbitrary machine learning models. Methodologically, TSC starts from an initial set of synthetic-control weights via a one-dimensional targeted update through the weight-tilting submodel, which calibrates the weights to reduce bias of weights estimation arising from pre-treatment fit. Furthermore, TSC avoids key shortcomings of existing methods (e.g., the augmented SCM), which can produce unbounded counterfactual estimates. Across extensive synthetic and real-world experiments, TSC consistently improves estimation accuracy over state-of-the-art SCM baselines.", "date": "2026-02-04", "categories": []}
{"id": "2602.13004", "title": "Towards Uncertainty-Aware Federated Granger Causal Learning", "abstract": "Granger causality recovers directed interactions from time-series data, but in many distributed systems, the data are vertically partitioned across clients, with each client observing only the variables of its own subsystem. Federated Granger causality (FedGC) recovers cross-client interactions without sharing raw data. Existing FedGC methods, however, return deterministic point estimates with no calibrated measure of uncertainty, leaving operators without a principled basis for identifying reliable cross-client interactions. We address this limitation by characterizing how uncertainty propagates through the FedGC framework. We derive closed-form covariance recursions for the cross-covariances induced by the coupled client-server feedback loop, and establish spectral-radius-based convergence conditions yielding closed-form expressions for the steady-state variances at both the client and server. Under mild stability conditions, we prove that the steady-state uncertainty depends only on client data statistics (aleatoric) and is independent of the priors placed on the model parameters (epistemic). Building on this asymptotic characterization, we construct a post-training hypothesis testing procedure that separates genuine cross-client interactions from spurious edges. Experiments on synthetic and real-world datasets show that the predicted uncertainty propagation matches the theory across multiple operating regimes, while consistently outperforming the state-of-the-art federated causal structure learning baselines.", "date": "2026-02-13", "categories": []}
{"id": "2605.11362", "title": "Causal Fairness for Survival Analysis", "abstract": "In the data-driven era, large-scale datasets are routinely collected and analyzed using machine learning (ML) and artificial intelligence (AI) to inform decisions in high-stakes domains such as healthcare, employment, and criminal justice, raising concerns about the fairness behavior of these systems. Existing works in fair ML cover tasks such as bias detection, fair prediction, and fair decision-making, but largely focus on static settings. At the same time, fairness in temporal contexts, particularly survival/time-to-event (TTE) analysis, remains relatively underexplored, with current approaches to fair survival analysis adopting statistical fairness definitions, which, even with unlimited data, cannot disentangle the causal mechanisms that generate disparities. To address this gap, we develop a causal framework for fairness in TTE analysis, enabling the decomposition of disparities in survival into contributions from direct, indirect, and spurious pathways. This provides a human-understandable explanation of why disparities arise and how they evolve over time. Our non-parametric approach proceeds in four steps: (1) formalizing the necessary assumptions about censoring and lack of confounding using a graphical model; (2) recovering the conditional survival function given covariates; (3) applying the Causal Reduction Theorem to reframe the problem in a form amenable to causal pathway decomposition; (4) estimating the effects efficiently. Finally, our approach is used to analyze the temporal evolution of racial disparities in outcome after admission to an intensive care unit (ICU).", "date": "2026-05-12", "categories": []}
{"id": "2605.11373", "title": "Causal Algorithmic Recourse: Foundations and Methods", "abstract": "The trustworthiness of AI decision-making systems is increasingly important. A key feature of such systems is the ability to provide recommendations for how an individual may reverse a negative decision, a problem known as algorithmic recourse. Existing approaches treat recourse outcomes as counterfactuals of a fixed unit, ignoring that real-world recourse involves repeated decisions on the same individual under possibly different latent conditions. We develop a causal framework that models recourse as a process over pre- and post-intervention outcomes, allowing for partial stability and resampling of latent variables. We introduce post-recourse stability conditions that enable reasoning about recourse from observational data alone, and develop a copula-based algorithm for inferring the effects of recourse under these conditions. For settings where paired observations of the same individual before and after intervention are available (called recourse data), we develop methods for inferring copula parameters and performing goodness-of-fit testing. When the copula model is rejected, we provide a distribution-free algorithm for learning recourse effects directly from recourse data. We demonstrate the value of the proposed methods on real and semi-synthetic datasets.", "date": "2026-05-12", "categories": []}
{"id": "2605.12051", "title": "Learning plug-in surrogate endpoints for randomized experiments", "abstract": "Surrogate endpoints are used in place of long-term outcomes in randomized experiments when observing the real outcome for a large enough cohort is prohibitively expensive or impractical. A short-term surrogate is good if the result of an experiment using the surrogate is predictive of the result of a hypothetical study using the real outcome. Much attention has been paid to formalizing this property in causal terms, but most criteria are unidentifiable and cannot be turned into practical algorithms for learning surrogate endpoints from data. To address this, we study plug-in composite surrogates, functions of post-treatment variables that may be substituted directly for the primary outcome in a randomized experiment. We propose two methods for learning plug-in surrogates that maximize effect predictiveness, and characterize the possibility of finding endpoints that yield unbiased effect estimates in representative scenarios. Finally, in both synthetic experiments with known effects and in data from a real-world experiment, we find that our method, based on directly modeling the surrogate effect, returns plug-in endpoints more predictive of the primary effect than established methods.", "date": "2026-05-12", "categories": []}
{"id": "2506.05967", "title": "Preference Learning for AI Alignment: a Causal Perspective", "abstract": "Reward modelling from preference data is a crucial step in aligning large language models (LLMs) with human values, requiring robust generalisation to novel prompt-response pairs. In this work, we propose to frame this problem in a causal paradigm, providing the rich toolbox of causality to identify the persistent challenges, such as causal misidentification, preference heterogeneity, and confounding due to user-specific factors. Inheriting from the literature of causal inference, we identify key assumptions necessary for reliable generalisation and contrast them with common data collection practices. We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness. Finally, we outline desiderata for future research and practices, advocating targeted interventions to address inherent limitations of observational data.", "date": "2025-06-06", "categories": []}
{"id": "2506.16234", "title": "Sequential Causal Discovery with Noisy Language Model Priors", "abstract": "Causal discovery from observational data typically assumes access to complete data and availability of perfect domain experts. In practice, data often arrive in batches, are subject to sampling bias, and expert knowledge is scarce. Language Models (LMs) offer a surrogate for expert knowledge but suffer from hallucinations, inconsistencies, and bias. We present a hybrid framework that bridges these gaps by adaptively integrating sequential batch data with LM-derived noisy, expert knowledge while accounting for both data-induced and LM-induced biases. We propose a representation shift from Directed Acyclic Graph (DAG) to Partial Ancestral Graph (PAG), that accommodates ambiguities within a coherent framework, allowing grounding the global LM knowledge in local observational data. To guide LM interactions, we use a sequential optimization scheme that adaptively queries the most informative edges. Across varied datasets and LMs, we outperform prior work in structural accuracy and extend to parameter estimation, showing robustness to LM noise.", "date": "2025-06-19", "categories": []}
{"id": "2507.03310", "title": "Causal Discovery for Irregularly Time Series with Consistency Guarantees", "abstract": "This paper studies causal discovery in irregularly sampled time series-a key challenge in risk-sensitive domains like finance, healthcare, and climate science, where missing data and inconsistent sampling frequencies distort causal mechanisms. The main challenge comes from the interdependence between missing data imputation and causal structure recovery: errors in imputation and structure learning can reinforce each other, leading to an inaccurate causal graph. Existing methods either impute first and then discover, or jointly optimize both via neural representation learning, but lack explicit mechanisms to ensure mutual consistency of imputation and structure learning. We address this challenge with ReTimeCausal, an EM-based framework that alternates between imputation and structure learning, which encourages structural consistency throughout the optimization process. Our framework provides theoretical consistency guarantees for structure recovery and extends classical results to settings with irregular sampling and high missingness. ReTimeCausal combines kernel-based sparse regression and structural constraints in an alternating process that updates the completed data and the causal graph in turn. Experiments on synthetic and real-world datasets show that ReTimeCausal is more effective than existing methods under challenging irregular sampling and missing data.", "date": "2025-07-04", "categories": []}
{"id": "2509.18484", "title": "Estimating Heterogeneous Causal Effect on Networks via Orthogonal Learning", "abstract": "Estimating causal effects on networks is challenging because treatments may affect both treated units and their neighbors, while network homophily induces dependence and confounding. These challenges are amplified when causal effects are heterogeneous across units and edges. We propose a two-stage orthogonal learning framework for estimating heterogeneous direct and spillover effects on networks. The first stage uses graph neural networks to estimate nuisance components that capture complex dependence on covariates and network structure. The second stage residualizes these nuisance components and estimates causal effects through an interpretable attention-based interference model, yielding edge-level spillover estimates as well as node- and population-level summaries. Neyman orthogonalization and cross-fitting reduce sensitivity to first-stage estimation error, so nuisance errors enter only at higher order. We further develop a bootstrap-based uncertainty quantification procedure for the estimated spillover matrix, enabling pointwise and simultaneous inference for heterogeneous edge- and node-level effects. Experiments show that our method improves heterogeneous effect estimation while supporting interpretable downstream analyses such as influential-neighbor detection and spillover-sign recovery.", "date": "2025-09-23", "categories": []}
{"id": "2509.22531", "title": "Debiased Front-Door Learners for Heterogeneous Effects", "abstract": "In observational settings where treatment and outcome share unmeasured confounders but an observed mediator remains unconfounded, the front-door (FD) adjustment identifies causal effects through the mediator. We study the heterogeneous treatment effect (HTE) under FD identification and introduce two debiased learners: FD-DR-Learner and FD-R-Learner. Under explicit sample-splitting, bounded-overlap, moment, and stage-learning assumptions, we show that FD-DR satisfies a product-error bound and FD-R satisfies a stage-error decomposition; these results yield conditional quasi-oracle corollaries when the relevant nuisance remainders are no larger than the target or stage oracle terms. We provide error analyses establishing this debiasedness and demonstrate robust empirical performance in synthetic studies and a real-world case study of primary seat-belt laws using Fatality Analysis Reporting System (FARS) dataset. Together, these results indicate that the proposed learners can deliver reliable and sample-efficient HTE estimates in FD scenarios when the stated assumptions are credible. The implementation is available at https://github.com/yonghanjung/FD-CATE.", "date": "2025-09-26", "categories": []}
{"id": "2510.13397", "title": "Assessing the robustness of heterogeneous treatment effects in survival analysis under informative censoring", "abstract": "Dropout is common in clinical studies, with up to half of patients leaving early due to side effects or other reasons. When dropout is informative (i.e., dependent on survival time), it introduces censoring bias, because of which treatment effect estimates are also biased. In this paper, we propose an assumption-lean framework to assess the robustness of conditional average treatment effect (CATE) estimates in survival analysis when facing censoring bias. Unlike existing works that rely on strong assumptions, such as non-informative censoring, to obtain point estimation, we use partial identification to derive informative bounds on the CATE. Thereby, our framework helps to identify patient subgroups where treatment is effective despite informative censoring. We further propose a novel model-agnostic meta-learner, called SurvB-learner, to estimate the bounds that can be used in combination with arbitrary machine-learning models, and that has favorable theoretical properties such as double-robustness and quasi-oracle efficiency. We finally demonstrate the effectiveness of our meta-learner across various experiments using both simulated and real-world data.", "date": "2025-10-15", "categories": []}
{"id": "2512.00175", "title": "Comparing Two Proxy Methods for Causal Identification", "abstract": "Identifying causal effects in the presence of unmeasured variables is a fundamental challenge in causal inference, for which proxy variable methods have emerged as a powerful solution. We contrast two major approaches in this framework: (1) bridge equation methods, which leverage solutions to integral equations to recover causal targets, and (2) array decomposition methods, which recover latent factors used to identify counterfactual quantities via eigendecomposition tasks. We compare the model restrictions underlying these two approaches and provide insight into implications of the underlying assumptions, clarifying the scope of applicability for each method.", "date": "2025-11-28", "categories": []}
{"id": "2601.23026", "title": "Root Cause Analysis of Measurement and Mechanistic Anomalies", "abstract": "Root cause analysis of anomalies aims to identify how and why a sample deviates from the normal process. Existing methods primarily focus on telling which features are responsible, ignoring that anomalies can arise through two fundamentally different processes: measurement errors, where the sample is generated normally but one or more values is recorded incorrectly, and mechanism shifts, where the causal process that generated the sample was changed. While measurement errors can often be safely corrected, mechanistic anomalies require careful consideration. In this paper, we formally define a causal model that explicitly captures both types by treating outliers as latent interventions on latent (\"true\") and observed (\"measured\") variables and show under which conditions the distinction is possible. Based on this model, we develop an efficient inference procedure for localizing root causes and distinguishing anomaly types. Experiments on synthetic and real-world data show that our method provides state-of-the-art and highly robust performance in both root cause localization and classification of anomaly types.", "date": "2026-01-30", "categories": []}
{"id": "2603.02678", "title": "Causal Discovery Should Embrace the Wisdom of the Crowd", "abstract": "This paper argues for recognizing an emerging paradigm of causal learning by wisdom of the crowd. Recent developments in government, industry, and research point to the rise of decentralized and crowd-based approaches within causal modeling, where causal knowledge distributed across many contributors can be systematically elicited and integrated with causal learning workflows. In this paradigm, causal learning becomes a distributed decision-making problem: each participant contributes partial and potentially noisy knowledge, while collective contributions help construct a global causal structure. This direction is enabled by advances in crowdsourcing platforms, expert knowledge elicitation, aggregation techniques, and large language model (LLM)-augmented information acquisition. Its promise is increasingly visible in early research and emerging real-world practices. Building on this momentum, we outline a framework for crowd-based causal learning spanning elicitation, modeling, aggregation, and optimization. We further discuss the opportunities and challenges introduced by this paradigm and call for interdisciplinary collaboration across causal learning, collective intelligence, human-AI interaction, and decision science.", "date": "2026-03-03", "categories": []}
{"id": "2604.23904", "title": "Generative Synthetic Data for Causal Inference: Pitfalls, Remedies, and Opportunities", "abstract": "Synthetic tabular data are often evaluated by distributional similarity, privacy distance, or train-on-synthetic-test-on-real predictive performance, but these criteria do not ensure validity for causal inference. We show that fully generative tabular synthesizers, including GAN- and LLM-based models, can preserve predictive utility while distorting average treatment effect (ATE) estimates. The failure is structural: ATE preservation requires both a realistic covariate law and an accurate treatment-effect contrast, whereas prediction loss penalizes treatment-effect error only through an overlap-weighted term. We formalize this mismatch through sensitivity and loss-decomposition results, and identify an analogous decomposition in block-level next-token prediction under log loss. Motivated by the tabular causal analysis, we propose a hybrid synthetic-data framework that generates covariates while modeling treatment and outcome mechanisms separately, allowing causal-purpose treatment assignment such as randomized synthetic assignment. We evaluate this framework in three settings: ATE preservation under fully generative versus hybrid synthesis, targeted augmentation for practical positivity problems, and synthetic simulation engines for comparing OR, IPW, AIPW, and TMLE before real-data analysis. Across synthetic and ACTG experiments, hybrid synthesis improves causal fidelity relative to fully generative baselines; LLM-based hybrid synthesis is often more faithful than CTGAN for ATE preservation and finite-sample estimator benchmarking.", "date": "2026-04-26", "categories": []}
{"id": "2605.05743", "title": "Fourier Feature Methods for Nonlinear Causal Discovery: FFML Scoring, TRFF Scoring, and FFCI Testing in Mixed Data", "abstract": "Gaussian process (GP) marginal likelihood scores and kernel conditional independence tests are theoretically appealing for nonlinear causal discovery but computationally prohibitive at scale. We present three complementary RFF-based methods forming a practical toolkit for score-based, constraint-based, and hybrid causal discovery.\n  The Fourier Feature Marginal Likelihood (FFML) score approximates the exact GP marginal likelihood by replacing the $n x n$ kernel Gram matrix with a finite-dimensional feature representation, reducing cost to $O(nm^2 + m^3)$ while retaining the probabilistic interpretation and automatic complexity penalty of the exact score. FFML extends to mixed (continuous and discrete) parent sets via a product-kernel construction, with a Kronecker path for small discrete parent sets and a Hadamard-product path otherwise.\n  The Tetrad Random Fourier Feature (TRFF) score is a complementary BIC-style alternative using penalized Student-t regression with random Fourier features. TRFF offers robustness to heavy-tailed noise and faster runtime than FFML. Empirically, TRFF and FFML exhibit a complementary precision-recall profile: TRFF achieves higher precision while FFML achieves better recall and lower SHD overall.\n  The Fourier Feature Conditional Independence (FFCI) test is a fast nonparametric CI test for mixed data, using ridge residualization in feature space and a Frobenius-norm cross-covariance statistic approximated as a weighted sum of chi-squared variables. Empirically, BOSS+FFML achieves the lowest SHD on nonlinear data, while BOSS+TRFF offers the highest precision. When run through PC-Max, FFCI and RCIT exhibit complementary precision-recall profiles: RCIT is more precise while FFCI achieves better recall and substantially lower SHD, at approximately twice the runtime.", "date": "2026-05-07", "categories": []}
{"id": "2605.08197", "title": "ReplaySCM: A Benchmark for Executable Causal Mechanism Induction from Interventions", "abstract": "Most causal benchmarks for language models score local answers or graph structure. We introduce ReplaySCM, a 1,300 item benchmark for executable causal mechanism induction from finite interventional evidence. Each item contains binary worlds generated by a latent fully observed acyclic Boolean structural causal model (SCM). A system must output a mechanism map in a restricted Boolean DSL; the submission is parsed, checked for legality and acyclicity, and replayed on training and held-out intervention worlds. Scoring uses replay behavior rather than formula strings, so syntactically different mechanisms receive credit when they behave correctly. ReplaySCM varies the structural information disclosed to the model through Ordered, Block-order, Hidden-order, and Hidden-roots settings, and includes Alternative-SCM tasks that supply a valid reference SCM and ask for a semantically distinct alternative that fits the training worlds, together with a separating intervention and witness. Frontier LLMs infer parts of the functional-parent structure, but held-out replay drops sharply when order or root structure is hidden. We also evaluate a matched support-audit ladder: Original, Extra Worlds, and Counterexample Audit (CEx), that raises mean local predecessor-pattern coverage from 0.8949 to 0.9815 to 1.0; under the audited searches, no discovered semantic alternative remains consistent with the training worlds. The Ordered/Hidden-order gap persists under this stronger evidence. ReplaySCM complements answer-level causal reasoning and graph-discovery benchmarks by evaluating executable replay generalization from finite interventional evidence, without claiming unique identification of the latent SCM.", "date": "2026-05-05", "categories": []}
{"id": "2605.08485", "title": "Sinkhorn Treatment Effects: A Causal Optimal Transport Measure", "abstract": "We introduce the Sinkhorn treatment effect, an entropic optimal transport measure of divergence between counterfactual distributions. Unlike classical quantities such as the average treatment effect, this measure captures differences across entire distributions. We analyze this divergence as a statistical functional and show it can be written as a smooth transformation of counterfactual mean embeddings with an appropriate kernel. This characterization allows us to establish first-order pathwise differentiability in general, and second-order pathwise differentiability under the null hypothesis of equal counterfactual distributions. Leveraging this smoothness, we construct debiased estimators and use them to obtain asymptotically valid tests for distributional treatment effects with a fixed entropic regularization parameter. Because the power of the test depends on this unknown parameter, we further propose an aggregated test that combines evidence across a grid of regularization choices. Experiments on simulated and image data demonstrate the practical advantages of our estimator and testing procedure.", "date": "2026-05-08", "categories": []}
{"id": "2605.08999", "title": "Non-Parametric Rehearsal Learning via Conditional Mean Embeddings", "abstract": "In machine learning, a critical class of decision-related problems concerns preventing predicted undesirable outcomes, referred to as the \\textit{avoiding undesired future} (AUF) problem. To address this, the \\textit{rehearsal learning} framework has been proposed to model influence relations for effective decisions. However, existing rehearsal methods rely on restrictive parametric assumptions such as linear systems or additive noise, limiting their practical applicability. In this paper, we propose the first non-parametric rehearsal learning approach for AUF without assuming specific functional forms of data generation processes. Specifically, we use kernel machinery to reformulate the AUF objective into a unified representation that disentangles desirability modeling from action-induced distributional changes. To handle the discontinuity of desirability indicator, we present a smooth Probit surrogate and provide an approximation error bound. Meanwhile, we capture the action-induced changes via conditional mean embeddings, and develop a kernel ridge regression based nested estimator for AUF objective with consistency guarantees. Such a formulation naturally accommodates nonlinear systems and non-additive noise, and empirical results on synthetic and real-data-derived semi-synthetic benchmarks demonstrate the effectiveness and flexibility of our approach.", "date": "2026-05-09", "categories": []}
{"id": "2605.09079", "title": "CauSim: Scaling Causal Reasoning with Increasingly Complex Causal Simulators", "abstract": "Despite surpassing human performance across mathematics, coding, and other knowledge-intensive tasks, large language models (LLMs) continue to struggle with causal reasoning. A core obstacle is the target data itself: causal systems are complex and often expressed in non-executable forms, while ground-truth answers to causal queries are inherently scarce. We introduce CauSim, a framework that turns causal reasoning from a scarce-label problem into a scalable supervised one. CauSim constructs increasingly complex causal simulators: executable structural causal models (SCMs), incrementally built by LLMs, that scale to globally complex systems while maintaining verifiable answers to causal queries. CauSim operates across representations by formalizing non-executable causal knowledge into code, enabling data augmentation, and translating executable SCMs into natural language, enabling supervision in previously difficult-to-supervise representations. We structure our research into two parts: (1) how to construct increasingly complex causal simulators, and (2) a systematic study of what CauSim enables, demonstrating generalization across representations, consistent gains from curriculum scaling and data volume, LLM self-improvement through self-generated simulators, and data augmentation via formalization of existing domain knowledge.", "date": "2026-05-09", "categories": []}
{"id": "2605.09168", "title": "CIVeX: Causal Intervention Verification for Language Agents", "abstract": "A valid tool call is not necessarily a valid intervention. Tool-using language agents are guarded by schema validators, policy filters, provenance checks, state predictors, and self-verification, yet such safeguards do not certify that a state-changing action has an identifiable causal effect. In confounded workflows, the action that looks optimal in observational logs can reduce utility when executed.\n  We introduce CIVeX, a causal intervention verifier that maps proposed actions to structural causal queries over a committed action-state graph, checks identifiability, and returns one of four auditable verdicts: EXECUTE, REJECT, EXPERIMENT, or ABSTAIN. Execution requires an assumption-scoped causal certificate carrying graph commitments, an identification argument, a one-sided lower confidence bound (LCB), provenance, and risk limits.\n  On Causal-ToolBench (1,890 instances, 7 seeds), CIVeX yields zero observed false executions across moderate and adversarial confounding. Under adversarial confounding it reaches 84.9% accuracy and 81.1% of oracle utility (+2.23 vs +2.76) and is the only non-oracle method whose constrained utility under a zero-false-execution constraint exceeds the AlwaysAbstain floor. On IHDP and ZOZO Open Bandit (real production logs with uniform-random ground truth), CIVeX matches Oracle correct-execution within 0.1pp and cuts per-execute false-execution by >=50x over naive baselines. A chain-of-thought LLM verifier (Claude Opus, Sonnet) cuts false-execution by an order of magnitude over a terse baseline, yet under adversarial confounding Opus's utility falls to 74% of CIVeX's. Intervention identifiability, not action validity, is the missing primitive for reliable tool use.", "date": "2026-05-09", "categories": []}
{"id": "2605.09422", "title": "Perception Without Engagement: Dissecting the Causal Discovery Deficit in LMMs", "abstract": "Although Large Multimodal Models (LMMs) have achieved strong performance on general video understanding, their susceptibility to textual prior shortcuts during causal discovery has been recognized as a critical deficit. The underlying mechanisms of this phenomenon remain incompletely understood, as existing benchmarks only measure response accuracy without revealing the sources and extent of the deficit. We introduce ProCauEval, a perturbation-based evaluation protocol that shifts from outcome assessment to mechanism diagnosis, probing causal discovery through five controlled configurations that systematically manipulate visual and textual modalities to decompose their respective contributions to model behavior and dissect the failure modes. Evaluating 17 mainstream LMMs, we find that models faithfully perceive video content yet systematically underexploit it during causal reasoning. We further observe that stronger post-training amplifies rather than mitigates textual prior reliance, and that higher baseline performance correlates with greater fragility under perturbation. To address these, we propose Anti-Distillation Policy Optimization (ADPO), a reinforcement learning framework built on negative teacher alignment, which augments GRPO by explicitly pushing the policy away from a prior-only counterfactual teacher induced by visual corruption. Specifically, ADPO maximizes the divergence between the policy distributions conditioned on the original and visually corrupted inputs, thereby forcing the model to ground its reasoning in visual evidence rather than textual shortcuts. Extensive experiments show that ADPO improves visual engagement without sacrificing fundamental comprehension, thus offering a preliminary step toward reliable causal discovery.", "date": "2026-05-10", "categories": []}
{"id": "2605.09514", "title": "Doubly Robust Proxy Causal Learning with Neural Mean Embeddings", "abstract": "Unobserved confounding prevents standard covariate adjustment from identifying causal response functions in observational studies. Proxy causal learning addresses this problem through bridge equations involving treatment- and outcome-inducing proxies, avoiding direct recovery of the latent confounder. Existing doubly robust proxy estimators combine outcome and treatment bridges, but typically rely on fixed kernels, sieves, or low-dimensional semiparametric models; existing neural proxy methods are more flexible, but are largely single-bridge estimators. We develop a neural doubly robust framework for proxy causal learning with continuous and structured treatments. Our method introduces a neural mean-embedding estimator for the treatment bridge, combines it with a neural outcome bridge, and estimates the doubly robust correction through a final regression stage. The framework covers population, heterogeneous, and conditional dose-response functions, yielding full response-curve estimators rather than binary-treatment effects. The algorithms use two stages for each bridge and history-aware updates of the final linear layers to stabilize stochastic multi-stage training. We prove consistency of the algorithms showing that the doubly robust error is controlled by the final averaging and regression errors together with the smaller of the outcome- and treatment-side weak-norm bridge errors. Across synthetic and image-valued benchmarks, the proposed estimators outperform existing baselines and single-bridge neural estimators, showing the benefit of combining learned outcome and treatment bridges in a doubly robust construction. Our implementation is available at https://github.com/BariscanBozkurt/DRPCL-Neural-Mean-Embedding.", "date": "2026-05-10", "categories": []}
{"id": "2605.09663", "title": "Causal Parametric Drift Simulation: A Digital Twin Framework for Classifier Robustness Evaluation", "abstract": "Machine learning classifiers in dynamic environments face concept drift -- changes in the data-generating process that degrade performance. Conventional evaluation via static test sets or noise perturbations fails to preserve causal dependencies in tabular data, often producing causally invalid assessments. Post-hoc tools like SHAP and LIME offer correlational insights that may not reflect the causal mechanisms driving model failure.\n  We propose a framework that complements existing drift detection by leveraging Structural Causal Models as \"Digital Twins\" of data-generating processes, enabling precise causal interventions while preserving structural dependencies. Our technique, Causal Parametric Drift Simulation, stress-tests classifiers to identify vulnerabilities before deployment. Experiments on the Open Sourcing Mental Illness (OSMH) dataset demonstrate that this approach exposes latent vulnerabilities invisible to standard statistical monitors.", "date": "2026-05-10", "categories": []}
{"id": "2605.09870", "title": "Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions", "abstract": "We propose SVAR-FM (Structural VAR with Flow Matching), a framework for time series causal discovery that treats a physics-based simulator as a mechanical realization of Pearl's do operator. Clamping a variable inside the simulator physically severs confounding paths, producing interventional data by construction. Conditional Flow Matching then learns the nonlinear interventional conditionals. Theoretically, we prove that the full structural VAR becomes identifiable under a coverage condition on the simulator-clampable variables, and derive an end-to-end error bound that decomposes into Monte Carlo, simulator fidelity, and Flow Matching terms. A sign-flip corollary predicts that when simulator accuracy falls below a threshold, the estimated causal effect reverses sign. Empirically, a benchmark across four scientific domains confirms that SVAR-FM recovers the correct causal sign where observational methods produce sign-reversed estimates due to confounding. A case study in ultrafast laser physics verifies the sign-flip prediction by physically varying the accuracy level of a first-principles quantum solver: the low-accuracy setting reverses the causal sign, while the high-accuracy setting recovers the correct direction (R-squared = 0.983, zero bias).", "date": "2026-05-11", "categories": []}
{"id": "2605.10163", "title": "Coarsening Linear Non-Gaussian Causal Models with Cycles", "abstract": "Recent work on causal abstraction, in particular graphical approaches focusing on causal structure between clusters of variables, aims to summarize a high-dimensional causal structure in terms of a low-dimensional one. Existing methods for learning such summaries from data assume that both the high- and low-dimensional structures are acyclic, which is helpful for causal effect identification and reasoning but excludes many high-dimensional models and thus limits applicability. We show that in the linear non-Gaussian (LiNG) setting, the high-dimensional acyclicity assumption can be relaxed while still allowing recovery of a low-dimensional causal directed acyclic graph (DAG). We further connect identifiability of this low-dimensional DAG to existing results: LiNG models with cycles are observationally identifiable only up to an equivalence class whose members differ by reversals of directed cycles; our low-dimensional DAG, which is invariant across all members of a given equivalence class, thus forms a natural representative of the class. While existing approaches for learning this observational equivalence class over high-dimensional variables have exponential time complexity, our low-dimensional summary is learned in worst-case cubic time and comes with explicit bounds on the sample complexity. We provide open source code and experiments on synthetic data to corroborate our theoretical results.", "date": "2026-05-11", "categories": []}
{"id": "2605.10206", "title": "Extended Wasserstein-GAN Approach to Causal Distribution Learning: Density-Free Estimation and Minimax Optimality", "abstract": "Distributional causal inference requires estimating not only average treatment effects but also interventional outcome distributions, including quantiles, tail risks, and policy-dependent uncertainty. As a method for distributional causal inference, generative adversarial network (GAN)-based counterfactual methods are flexible tools for this task. However, these methods have several limitations. First, the objectives of certain techniques do not coincide with the statistical risk of the identifiable causal target, and therefore provide limited theoretical guarantees regarding estimable counterfactual distributions or optimality. Second, they tend to rely on unstable density-based methods, such as density ratio estimation. In this paper, we propose GANICE (GAN for Interventional Conditional Estimation) with several advantages: it (i) clarifies the conditional interventional distribution for each treatment--covariate state as the causal estimation target; (ii) estimates the conditional distribution such that its averaged Wasserstein risk is minimized; (iii) establishes minimax optimality. GANICE achieves these advantages through the introduction of the extended Wasserstein distance, the incorporation of a cellwise critic in its dual, and an optimality proof based on Besov space theory. Our experiments demonstrate that GANICE consistently outperforms existing methods.", "date": "2026-05-11", "categories": []}
{"id": "2605.10533", "title": "ConfoundingSHAP: Quantifying confounding strength in causal inference", "abstract": "In causal inference, confounders are variables that influence both treatment decisions and outcomes. However, unlike as in randomized clinical trials, the treatment assignment mechanism in observational studies is not known, and it is thus unclear which covariates act as confounders. Here, we aim to generate insight for causal inference and answer: which of the observed covariates act as confounders? We introduce ConfoundingSHAP, a Shapley-based method for attributing confounding strength to individual covariates. Our contributions are twofold. First, we propose a Shapley game targeted to infer the confounding strength of the covariates. Our resulting Shapley values differ from the standard applications of SHAP explanations on causal targets, such as understanding treatment effect heterogeneity, which are ill-suited for our task. Second, as our task requires evaluating the value function over many adjustment sets, we provide a scalable TabPFN-based estimation that avoids exhaustive refitting. We demonstrate the practical value across various datasets, where ConfoundingSHAP provides informative explanations of which observed covariates drive confounding and thereby helps to provide more insight for causal inference in practice.", "date": "2026-05-11", "categories": []}
{"id": "2605.10590", "title": "Amortizing Causal Sensitivity Analysis via Prior Data-Fitted Networks", "abstract": "Causal sensitivity analysis aims to provide bounds for causal effect estimates in the presence of unobserved confounding. However, existing methods for causal sensitivity analysis are per-instance procedures, meaning that changes to the dataset, causal query, sensitivity level, or treatment require new computation. Here, we instead present an in-context learning approach. Specifically, we propose an amortized approach to causal sensitivity analysis based on prior-data fitted networks. A key challenge is that the sensitivity bounds are not directly available when sampling training data. To address this, we develop a general prior-data construction that is applicable across the class of generalized treatment sensitivity models. Our construction involves a Lagrangian scalarization of the objective to generate training labels for the bounds through a tradeoff between causal effect min/max-imization and sensitivity model violation, which avoids model-specific analytical derivations. We further show that, under standard convexity and linearity conditions, our objective recovers the full Pareto frontier of solutions. Empirically, we demonstrate our amortized approach across various datasets, causal queries, and sensitivity levels, where our approach achieves a test-time computation that is orders of magnitude faster than per-instance methods. To the best of our knowledge, ours is the first foundation model for in-context learning for causal sensitivity analysis.", "date": "2026-05-11", "categories": []}
{"id": "2605.10651", "title": "A Recursive Decomposition Framework for Causal Structure Learning in the Presence of Latent Variables", "abstract": "Constraint-based causal discovery is widely used for learning causal structures, but heavy reliance on conditional independence (CI) testing makes it computationally expensive in high-dimensional settings. To mitigate this limitation, many divide-and-conquer frameworks have been proposed, but most assume causal sufficiency, i.e., no latent variables. In this paper, we show that divide-and-conquer strategies can be theoretically generalized beyond causal sufficiency to settings with latent variables. Specifically, we propose a recursive decomposition framework, termed DiCoLa, that enables divide-and-conquer causal discovery in the presence of latent variables. It recursively decomposes the global learning task into smaller subproblems and integrates their solutions through a principled reconstruction step to recover the global structure. We theoretically establish the soundness and completeness of the proposed framework. Extensive experiments on synthetic data demonstrate that our approach significantly improves computational efficiency across a range of causal discovery algorithms, while experiments on a real-world dataset further illustrate its practical effectiveness.", "date": "2026-05-11", "categories": []}
{"id": "2506.14399", "title": "Factored Classifier-Free Guidance", "abstract": "Counterfactual generation aims to simulate realistic hypothetical outcomes under causal interventions. Diffusion models have emerged as a powerful tool for this task, combining DDIM inversion with conditional generation and classifier-free guidance (CFG). In this work, we identify a key limitation of CFG for counterfactual generation: it prescribes a global guidance scale for all attributes, leading to significant spurious changes in inferred counterfactuals. To mitigate this, we propose Factored Classifier-Free Guidance (FCFG), a flexible and model-agnostic guidance technique that enables attribute-wise control following a causal graph. FCFG complements recent advances in classifier-free guidance and can be seamlessly extended to advanced guidance schemes such as CFG++ and APG. Our experiments demonstrate that FCFG significantly improves the axiomatic soundness of inferred counterfactuals across both natural and medical image datasets, mitigating spurious amplification effects, and enhancing counterfactual reversibility.", "date": "2025-06-17", "categories": []}
{"id": "2601.07247", "title": "Multi-environment Invariance Learning with Missing Data", "abstract": "Learning models that can handle distribution shifts is a key challenge in domain generalization. Invariance learning, an approach that focuses on identifying features invariant across environments, improves model generalization by capturing stable relationships, which may represent causal effects when the data distribution is encoded within a structural equation model (SEM) and satisfies modularity conditions. This has led to a growing body of work that builds on invariance learning, leveraging the inherent heterogeneity across environments to develop methods that provide causal explanations while enhancing robust prediction. However, in many practical scenarios, obtaining complete outcome data from each environment is challenging due to the high cost or complexity of data collection. This limitation in available data hinders the development of models that fully leverage environmental heterogeneity, making it crucial to address missing outcomes to improve both causal insights and robust prediction. In this work, we derive an estimator from the invariance objective under missing outcomes. We establish non-asymptotic guarantees on variable selection property and $\\ell_2$ error convergence rates, which are influenced by the proportion of missing data and the quality of imputation models across environments. We evaluate the performance of the new estimator through extensive simulations and demonstrate its application using the UCI Bike Sharing dataset to predict the count of bike rentals. The results show that despite relying on a biased imputation model, the estimator is efficient and achieves lower prediction error, provided the bias is within a reasonable range.", "date": "2026-01-12", "categories": []}
{"id": "2602.09034", "title": "Latent-Space Causal Discovery from Indirect Neuroimaging Observations", "abstract": "Neuroimaging does not observe causal variables directly: hemodynamics and volume conduction distort signals so that statistical dependence need not reflect latent neural influence. Before estimating graphs, one must specify under what assumptions delayed directed structure can be studied from such indirect observations. We formalize a conditional setting - recoverable inversion under modality physics together with nonstationary latent dynamics - and derive an inversion-error propagation bound under explicit assumptions. Building on this framing, we propose INCAMA (INdirect CAusal MAmba): physics-aware inversion coupled with a delay-aware Mamba encoder that uses mechanism shifts as informative variation for directed graph scoring. We use controlled simulations for quantitative validation and HCP motor-task fMRI as a zero-shot external transfer check based on anatomical and task-network consistency. Across TVB simulations, INCAMA improves directed-structure recovery by 2-3x in F1 over observation-space and two-stage baselines, and on HCP motor-task fMRI it produces sparse directed estimates concentrated in canonical visuo-motor pathways.", "date": "2026-01-30", "categories": []}
{"id": "2603.00041", "title": "Econometric vs. Causal Structure-Learning for Time-Series Policy Decisions: Evidence from the UK COVID-19 Policies", "abstract": "Causal machine learning (ML) recovers graphical structures that inform us about potential cause-and-effect relationships. Most progress has focused on cross-sectional data with no explicit time order, whereas recovering causal structures from time series data remains the subject of ongoing research in causal ML. In addition to traditional causal ML, this study assesses econometric methods that some argue can recover causal structures from time series data. The use of these methods can be explained by the significant attention the field of econometrics has given to causality, and specifically to time series, over the years. This presents the possibility of comparing the causal discovery performance between econometric and traditional causal ML algorithms. We seek to understand if there are lessons to be incorporated into causal ML from econometrics, and provide code to translate the results of these econometric methods to the most widely used Bayesian Network R library, bnlearn. We investigate the benefits and challenges that these algorithms present in supporting policy decision-making, using the real-world case of COVID-19 in the UK as an example. Four econometric methods are evaluated in terms of graphical structure, model dimensionality, and their ability to recover causal effects, and these results are compared with those of eleven causal ML algorithms. Amongst our main results, we see that econometric methods provide clear rules for temporal structures, whereas causal-ML algorithms offer broader discovery by exploring a larger space of graph structures that tends to lead to denser graphs that capture more identifiable causal relationships.", "date": "2026-02-09", "categories": []}
{"id": "2605.04081", "title": "Time series causal discovery with variable lags", "abstract": "Causal Bayesian Networks (CBNs) are a powerful tool for reasoning under uncertainty about complex real-world problems. Such problems evolve over time, responding to external shocks as they occur. To support decision-making, CBNs require a cause-and-effect map of the variables under consideration, known as the network's structure. Learning the graphical structure of a causal model from data remains challenging; learning it from time-series data is even harder because dependencies may arise at different time lags. Existing time-series causal discovery methods often assume a fixed lag window and do not explicitly optimise edge-specific lags. We propose a Tabu-based structure learning algorithm that searches for a time-ordered directed structure (i.e., where every edge respects time) while allowing edge-specific lags up to a specified maximum lag. The approach uses a decomposable BIC-based score with node-specific effective sample sizes and an explicit lag-length penalty encouraging parsimonious delay assignments while preserving efficient local score updates. We provide theoretical guarantees of validity and local optimality, and we also describe a parallel implementation for improved scalability. In simulations, the method recovered graph structure competitively and estimated lags accurately when true adjacencies were recovered. On a real-world UK COVID-19 policy dataset, the learnt structure was dominated by short delays while retaining a substantial minority of longer-lag dependencies, consistent with delayed behavioural and epidemiological effects.", "date": "2026-04-15", "categories": []}
{"id": "2605.06941", "title": "Causal-Aware Foundation-Model for Bilevel Optimization in Discrete Choice Settings", "abstract": "We introduce a causal aware foundation-model framework for real time optimal decision making in discrete choice environments. We propose a constrained triple-head price optimization (C3PO) network to solve a bilevel decision problem in which a service provider selects an optimal assortment while heterogeneous users make personalized acceptance or rejection choices optimizing their own personalized preferences. C3PO integrates imitation learning of prices, multi-task learning of revenue responses, and in context learning of price elasticity to generate pricing recommendations while adhering to business constraints. During inference, frontier model prompting retrieves an enhanced elasticity prior for new products from behavioral economics literature, improving pricing effectiveness. We demonstrate strong in context learning performance using simulated, synthetic, and real-world datasets. C3PO is trained on simulated data generated from multiple classical discrete choice models in economics. The model is trained on data comprising simulated customer segments and counterfactual action and outcome pairs and evaluated on randomly generated choice environments with no access to the underlying preference structure. The trained model consistently improves the pricing KPIs, with gains increasing as customer price sensitivity increases. We also deploy the tuned foundation model for optimal pricing in real-world applications such as healthcare, tender pricing, airline ancillary pricing, and other domains, achieving substantial gains across multiple products, markets, and divisions.", "date": "2026-05-07", "categories": []}
{"id": "2605.06979", "title": "PLOT: Progressive Localization via Optimal Transport in Neural Causal Abstraction", "abstract": "Causal abstraction offers a principled framework for mechanistic interpretability, aligning a high-level causal model with the low-level computation realized by a neural network through counterfactual intervention analysis. Existing methods such as distributed alignment search (DAS) learn expressive subspace interventions, but the relevant neural site is unknown a priori, so finding a handle requires a computationally burdensome search over candidate sites. We introduce PLOT (Progressive Localization via Optimal Transport), a transport-based framework that localizes causal variables from the output effect geometry of abstract and neural interventions. PLOT fits an optimal transport coupling between abstract variables and candidate neural sites, yielding a global soft correspondence that can be calibrated into intervention handles. In simple settings, a single coupling over individual neurons suffices. In larger models, PLOT is applied progressively, moving from coarse sites such as tokens, timesteps, or layers to finer supports such as coordinate groups or PCA spans, and optionally guiding DAS based on the localized signal. Across experiments of increasing complexity, transport-only PLOT handles are exceedingly fast and competitive on accuracy, while PLOT-guided DAS reaches DAS-level accuracy at a fraction of full DAS runtime, providing an efficient localization engine for causal abstraction research at scale.", "date": "2026-05-07", "categories": []}
{"id": "2605.06993", "title": "Optimal Experiments for Partial Causal Effect Identification", "abstract": "Causal queries are often only partially identifiable from observational data, and experiments that could tighten the resulting bounds are typically costly. We study the problem of selecting, prior to observing experimental outcomes, a cost-constrained subset of experiments that maximally tightens bounds on a target query. We formalize this as the max-potency problem, where epistemic potency measures the worst-case reduction in bound width guaranteed by an experiment, and show that this problem is NP-hard via a reduction from 0-1 knapsack. Building on the polynomial-programming framework of Duarte et al. (2023), we give a general procedure for evaluating epistemic potency in discrete settings. To control the super-exponential search space, we introduce two graphical pruning criteria that depend only on the causal graph and the query: a novel path-interception rule that exploits district structure to certify zero potency in linear time, and an identifiability check based on the ID algorithm. On Erdos-Renyi random graphs and 11 bnlearn benchmark networks, the two criteria together prune 50-88% of candidate experiments on average without solving a single polynomial program. For the general subset search, we show that ID-pruned experiments are combinatorially inert, yielding a super-exponential reduction in the number of subsets evaluated. We close with an end-to-end demonstration on observational NHANES data, selecting optimal experiments for estimating the effect of physical activity on diabetes.", "date": "2026-05-07", "categories": []}
{"id": "2605.07029", "title": "BGM-IV: an AI-powered Bayesian generative modeling approach for instrumental variable analysis", "abstract": "Instrumental-variable (IV) regression enables causal estimation under endogeneity, but modern IV problems often involve nonlinear structural effects and high-dimensional covariates. Existing nonlinear IV methods directly learn the causal relation in observed feature space or rely on learned representations within two-stage or moment-based procedures, which can struggle when the causal information is embedded in a high-dimensional representation. We propose BGM-IV, a latent Bayesian generative modeling approach that reframes nonlinear IV regression as posterior inference in a causally structured latent space. BGM-IV infers latent components that separately capture shared confounding structure, outcome-specific variation, treatment-specific variation, and covariate-only nuisance information. To account for endogeneity, BGM-IV replaces the confounded outcome likelihood with an IV-integrated pseudo-likelihood that averages over instrument-induced treatment values within the latent model. Across various benchmark datasets, BGM-IV remains competitive in the classical low-dimensional regime and performs best in high-dimensional covariate regimes. Together, these results show that structured latent generative modeling provides a principled and effective strategy to nonlinear IV estimation with rich covariates. The code of BGM-IV is available at https://github.com/liuq-lab/BGM-IV.", "date": "2026-05-07", "categories": []}
{"id": "2605.07057", "title": "Integrating Causal DAGs in Deep RL: Activating Minimal Markovian States with Multi-Order Exposure", "abstract": "Online reinforcement learning (RL) relies on the Markov property for guaranteed performance, but real-world applications often lack well-defined states given raw observed variables. While causal RL has attracted growing interest, existing work typically assumes Markovian states are provided and focuses on using causality to accelerate learning, leaving a fundamental gap: \\emph{given a longitudinal causal graph over observed variables, how does one construct MDP states that provably satisfy the Markov property?} We address this by providing a procedure that constructs a provably minimal state representation. In deep RL, we observe that the minimal representation alone empirically fails to improve performance, indicating that neural networks cannot directly exploit Markovian minimality. To address this, we propose \\textbf{MOSE} (Multi-Order State Exposure), which feeds multi-order historical state constructions into the same $Q$-function. MOSE consistently outperforms both the minimal state construction and single-window policies on common benchmarks and synthetic datasets. Including the minimal representation alongside MOSE can further improve performance. Our results establish a core principle for causal deep RL: minimal sufficiency is not enough, and \\emph{controlled redundancy} is necessary to unlock the benefit of causal state information.", "date": "2026-05-08", "categories": []}
{"id": "2605.07065", "title": "Causal EpiNets: Precision-corrected Bounds on Individual Treatment Effects using Epistemic Neural Networks", "abstract": "Individual treatment effects are not point-identified from data. The Probability of Necessity and Sufficiency (PNS) circumvents this limitation by characterizing individual-level causality through intersection bounds derived from combined experimental and observational data. In finite samples, however, standard plug-in estimators systematically fail: they violate structural probability constraints and suffer from extremum bias induced by max-min operators, yielding spuriously narrow intervals. We propose a neural framework for finite-sample PNS estimation that resolves both pathologies. We introduce an anchored neural architecture that guarantees structural constraint satisfaction by construction. To correct extremum bias, we employ precision-corrected intersection-bound inference, leveraging Epistemic Neural Networks for scalable, high-dimensional uncertainty quantification. Empirical evaluations confirm that this approach maintains nominal coverage and exact constraint validity in high-dimensional regimes where standard estimators systematically undercover.", "date": "2026-05-08", "categories": []}
{"id": "2605.07199", "title": "Three-in-One World Model: Energy-Based Consistency, Prediction, and Counterfactual Inference for Marketing Intervention", "abstract": "Marketing decisions reflect the interaction of latent consumer heterogeneity, time-varying internal states, and explicit interventions, a structure that current prediction- and language-oriented models do not capture in a unified manner. We propose a Three-in-One world-model architecture in which a Deep Boltzmann Machine (DBM) learns a frozen belief representation from demographics, time, and lagged actions and outcomes, with lightweight task-specific adapters attached on top. The same belief supports three tasks within a single framework: (i) energy-based consistency evaluation through the DBM's free energy, (ii) outcome prediction through adapters, and (iii) counterfactual inference by holding the belief fixed and varying only the action input given to the adapter. Using a controlled simulation in which the latent price sensitivity, promotion responsiveness, and base preference of each consumer are known, we show that the adapters match a strong MLP baseline on visit- and purchase-AUC while recovering heterogeneous treatment effects substantially better than S-, T-, X-, and DR-learner meta-learners and a Causal Forest baseline built on the same raw features, with the largest gap on a confounded price-promotion intervention. Complementing this, free-energy clamps systematically penalize counterfactual purchase trajectories that lack prior promotional exposure, and the penalty itself depends on the latent base preference in the expected direction. These results indicate that DBM beliefs disentangle latent traits in a form that survives counterfactual queries, providing an integrated world-model substrate for marketing intervention.", "date": "2026-05-08", "categories": []}
{"id": "2605.07204", "title": "Arrow: A Foundation Model for Causal Discovery", "abstract": "We introduce Arrow, a foundation model for zero-shot causal discovery on observational tabular data. Arrow factorizes a directed acyclic graph into an undirected skeleton and a topological order, guaranteeing acyclicity by construction. Given a new dataset, it uses a transformer-based architecture to contextualize variables within and across observations, then predicts skeleton edge probabilities and node order scores that together define a graph. Arrow is trained in a supervised fashion on synthetic datasets with ground-truth graphs, using an end-to-end differentiable directed edge composite likelihood induced by the skeleton-order factorization. The training distribution spans diverse graph families, functional forms, noise models, and dataset shapes. Across in- and out-of-distribution synthetic, semi-synthetic, and real datasets, Arrow matches or outperforms existing causal discovery methods at substantially lower inference cost than competitive alternatives. Our results demonstrate that large-scale pretraining on diverse synthetic data can yield zero-shot causal discovery models that are fast, accurate, and reusable on new datasets.", "date": "2026-05-08", "categories": []}
{"id": "2605.07267", "title": "PerCaM-Health: Personalized Dynamic Causal Graphs for Healthcare Reasoning", "abstract": "Personalized healthcare decisions require reasoning about how physiological and behavioral variables influence an individual patient over time. Existing temporal causal discovery methods are poorly matched to this setting: cohort-level models provide stable but non-personalized structures, while per-patient discovery is unreliable because individual trajectories are short, noisy, irregular, and non-stationary. This creates a fundamental gap between population-level causal modeling and the patient-specific, time-varying mechanisms needed for intervention reasoning. We introduce PerCaM-Health, a framework for learning personalized dynamic causal graphs from longitudinal health data. The framework learns a knowledge-guided population temporal graph, then conservatively adapts and evolves it using patient-specific temporal evidence and rolling-window updates, producing interpretable and auditable graph sequences. By coupling these graphs with temporal structural equations, the framework enables patient-level counterfactual queries, such as estimating short-horizon outcome changes under hypothetical behavioral interventions. Experiments on a semi-synthetic dynamic health benchmark show that PerCaM-Health improves graph recovery, dynamic edge tracking, and intervention direction accuracy compared to cohort-level, per-patient, and non-personalized temporal baselines. These results demonstrate that jointly modeling personalization and temporal evolution yields more reliable causal structure and intervention reasoning.", "date": "2026-05-08", "categories": []}
{"id": "2605.07280", "title": "Mask2Cause: Causal Discovery via Adjacency Constrained Causal Attention", "abstract": "Leveraging deep learning for causal discovery in time series remains challenging because existing neural methods predominantly rely on component-wise architectures that fail to capture shared system dynamics or employ decoupled post-hoc graph extraction that risks overfitting to spurious correlations. We propose $\\textbf{Mask2Cause}$, an end-to-end framework that recovers the underlying causal graph directly during the forecasting forward pass. Our approach introduces an Inverted Variable Embedding and an Adjacency-Constrained Masked Attention mechanism, trained with homoscedastic or heteroscedastic objectives to capture causal influences in both mean and variance. Empirical results on diverse benchmarks, from synthetic chaotic dynamics to realistic biological simulations, demonstrate state-of-the-art causal discovery with significantly reduced parameter complexity compared to standard baselines. We further show that inferred causal structures can be used to reduce parameter count of forecasting models by more than 70% on average while maintaining predictive accuracy.", "date": "2026-05-08", "categories": []}
{"id": "2605.07467", "title": "Physical Simulators as Do-Operators: Causal Discovery under Latent Confounders for AI-for-Science", "abstract": "Existing interventional causal discovery methods -- IGSP, DCDI, ENCO -- assume causal sufficiency (no latent confounders) and rely on virtual interventions in synthetic simulators. In AI-for-Science settings such as molecular design and materials science, latent confounders are ubiquitous and real interventions (e.g., physics-based simulations) require hours to days per data point. We propose CFM-SD (Causal Flow Matching with Simulation Data), which uses first-principles physical simulators as do-operators in Pearl's interventional calculus to simultaneously handle latent confounders and real interventional data. Theoretically, $d$-variable causal structure is identifiable with $O(d)$ single-variable interventions -- the minimum under physical realizability constraints. In Intrinsic Evaluation on synthetic data ($\\gamma=0.2$--$0.8$), CFM-SD achieves average F1$=0.800$ vs. F1$=0.127$--$0.562$ for all baselines. In Extrinsic Evaluation on real scientific data, CFM-SD achieves 57--58\\% bias reduction in molecular toxicity prediction and battery electrolyte optimization, demonstrating practical value beyond synthetic benchmarks.", "date": "2026-05-08", "categories": []}
{"id": "2605.07600", "title": "Mathematical Reasoning via Intervention-Based Time-Series Causal Discovery Using LLMs as Concept Mastery Simulators", "abstract": "Recent methods for improving LLM mathematical reasoning, whether through MCTS-based test-time search or causal graph-guided knowledge injection, cannot identify which concepts causally contribute to a correct answer, as the observed association may be spurious, driven by confounders such as problem difficulty.\n  We propose CIKA (Causal Intervention for Knowledge Activation), a framework that uses the LLM itself as an interventional simulator: a prompt sets the concept state to ``mastered'' and the correctness change estimates the causal effect. We formalize this quantity as an Interventional Capability Probe (ICP), which diagnoses whether the LLM can use a given concept -- distinct from merely possessing knowledge. Because the intervention exogenously sets the concept state independently of problem difficulty, ICP separates confounding that observational methods cannot.\n  On 67 screened problems, the ICP of the top-ranked concept (+0.219) is significantly larger than that of the negative control (+0.039; paired $t$-test, $p < 10^{-6}$, Cohen's $d = 0.86$), confirming that the probe discriminates causally relevant concepts from irrelevant ones. Analysis of 601 Omni-MATH problems further shows that solved problems have 6.1$\\times$ higher ATE than unsolved ones (0.338 vs. 0.055), confirming that ICP is predictive of problem-solving success. With a 7B-parameter LLM whose weights are entirely frozen, CIKA achieves 69.7\\% on the contamination-free Omni-MATH-Rule benchmark and 64.0\\% overall, compared to 60.5\\% for o1-mini, and 97.2\\% on GSM8K, 46--50\\% on AIME 2024--2026, and 46.2\\% on MathArena. The Causal Knowledge Activation component contributes 33.8\\% of correct answers on problems where the base model alone fails, demonstrating that the LLM already possessed but had not activated the requisite knowledge.", "date": "2026-05-08", "categories": []}
{"id": "2605.07665", "title": "Debiased Counterfactual Generation via Flow Matching from Observations", "abstract": "Estimating counterfactual distributions under interventions is central to treatment risk assessment and counterfactual generation tasks. Existing approaches model the counterfactual distribution as a standalone generative target, without exploiting its relationship to the observational data. In this work, we show that under standard assumptions, observational and counterfactual outcome distributions are tightly linked: they have identical support and tail behavior, remain statistically close under weak confounding, and share any features of high-dimensional outcomes which are invariant to confounders. These properties motivate learning counterfactual distributions not from scratch, but via a deconfounding flow from the observational distribution. We formulate this problem via flow-matching and derive a semiparametrically efficient estimator based on a novel efficient influence function correction. We subsequently extend our estimator to target minimal-energy flows in high-dimensions, which we show can be especially simple targets between observational and counterfactual distributions. In experiments, deconfounding flows outperform existing debiased counterfactual distribution estimators, while also mitigating known failure modes of flow-based methods.", "date": "2026-05-08", "categories": []}
{"id": "2605.07993", "title": "Bayesian Sensitivity of Causal Inference Estimators under Evidence-Based Priors", "abstract": "Causal inference, especially in observational studies, relies on untestable assumptions about the true data-generating process. Sensitivity analysis helps us determine how robust our conclusions are when we alter these underlying assumptions. Existing frameworks for sensitivity analysis are concerned with worst-case changes in assumptions. In this work, we argue that using such pessimistic criteria can often become uninformative or lead to conclusions contradicting our prior knowledge about the world. To demonstrate this claim, we generalize the recent s-value framework (Gupta & Rothenh\\\"ausler, 2023) to estimate the sensitivity of three different common assumptions in causal inference. Empirically, we find that, indeed, worst-case conclusions about sensitivity can rely on unrealistic changes in the data-generating process. To overcome this, we extend the s-value framework with a new sensitivity analysis criterion: Bayesian Sensitivity Value (BSV), which computes the expected sensitivity of an estimate to assumption violations under priors constructed from real-world evidence. We use Monte Carlo approximations to estimate this quantity and illustrate its applicability in an observational study on the effect of diabetes treatments on weight loss.", "date": "2026-05-08", "categories": []}
{"id": "2605.08034", "title": "Semiparametric Efficient Test for Interpretable Distributional Treatment Effects", "abstract": "Distributional treatment effects can be invisible to means: a treatment may preserve average outcomes while changing tails, modes, dispersion, or rare-event probabilities. Kernel tests can detect discrepancies between interventional outcome laws, but global tests do not reveal where the laws differ. We propose DR-ME, to our knowledge the first semiparametrically efficient finite-location test for interpretable distributional treatment effects. DR-ME evaluates an interventional kernel witness at learned outcome locations, returning causal-discrepancy coordinates rather than only a global rejection. From observational data, we derive orthogonal doubly robust kernel features whose centered oracle form is the canonical gradient of this finite witness. For fixed locations, we characterize the local testing limit: DR-ME is chi-square calibrated under the null, has noncentral chi-square local power, and uses the covariance whitening that optimizes local signal-to-noise for discrepancies visible through the selected coordinates. This efficient local-power geometry yields a principled location-learning criterion, with sample splitting preserving post-selection validity. Experiments show near-nominal type-I error, competitive power against global doubly robust kernel tests, and interpretable learned locations that localize distributional effects in a semi-synthetic medical-imaging study.", "date": "2026-05-08", "categories": []}
{"id": "2409.00417", "title": "Learning linear acyclic causal model including Gaussian noise using ancestral relationships", "abstract": "This paper discusses algorithms for learning causal DAGs. The PC algorithm makes no assumptions other than the faithfulness to the causal model and can identify only up to the Markov equivalence class. LiNGAM assumes linearity and continuous non-Gaussian disturbances for the causal model, and the causal DAG defining LiNGAM is shown to be fully identifiable. The PC-LiNGAM, a hybrid of the PC algorithm and LiNGAM, can identify up to the distribution-equivalence pattern of a linear causal model, even in the presence of Gaussian disturbances. However, in the worst case, the PC-LiNGAM has factorial time complexity for the number of variables. In this paper, we propose an algorithm for learning the distribution-equivalence patterns of a linear causal model with a lower time complexity than PC-LiNGAM, using the causal ancestor finding algorithm in Maeda and Shimizu, which is generalized to account for Gaussian disturbances.", "date": "2024-08-31", "categories": []}
{"id": "2509.04112", "title": "Synthetic Counterfactual Labels for Efficient Conformal Counterfactual Inference", "abstract": "This work addresses the problem of constructing reliable prediction intervals for individual counterfactual outcomes. Existing conformal counterfactual inference (CCI) methods provide marginal coverage guarantees but often produce overly conservative intervals, particularly under treatment imbalance when counterfactual samples are scarce. We introduce synthetic data-powered CCI (SP-CCI), a new framework that augments the calibration set with synthetic counterfactual labels generated by a pre-trained counterfactual model. To ensure validity, SP-CCI incorporates synthetic samples into a conformal calibration procedure based on risk-controlling prediction sets (RCPS) with a debiasing step informed by prediction-powered inference (PPI). We prove that SP-CCI achieves tighter prediction intervals while preserving marginal coverage, with theoretical guarantees under both exact and approximate importance weighting. Empirical results on different datasets confirm that SP-CCI consistently reduces interval width compared to standard CCI across all settings.", "date": "2025-09-04", "categories": []}
{"id": "2510.16811", "title": "Graph Learning Is Suboptimal in Causal Bandits", "abstract": "We study regret minimization in causal bandits under causal sufficiency where the underlying causal structure is not known to the agent. Previous work has focused on identifying the reward's parents and then applying classic bandit methods to them, or jointly learning the parents while minimizing regret. We investigate whether such strategies are optimal. Somewhat counterintuitively, our results show that learning the parent set is suboptimal. We do so by proving that there exist instances where regret minimization and parent identification are fundamentally conflicting objectives. We further analyze both the known and unknown parent set size regimes, establish novel regret lower bounds that capture the combinatorial structure of the action space. Building on these insights, we propose nearly optimal algorithms that bypass graph and parent recovery, demonstrating that parent identification is indeed unnecessary for regret minimization. Experiments confirm that there exists a large performance gap between our method and existing baselines in various environments.", "date": "2025-10-19", "categories": []}
{"id": "2601.16715", "title": "Dynamic Expert-Guided Model Averaging for Causal Discovery", "abstract": "Would-be practitioners of causal discovery face a dizzying array of algorithms without a clear best choice. This abundance of competitive methods makes ensembling a natural strategy for practical applications. At the same time, real-world use cases frequently violate the assumptions on which common causal discovery algorithms are based, forcing reliance on expert knowledge. Inspired by recent work on dynamically requested expert knowledge and large language models (LLMs) as experts, we present a flexible model averaging method that integrates selective expert querying to ensemble a diverse set of causal discovery algorithms. Crucially, we distinguish between edge existence and orientation, enabling the method to leverage the complementary strengths of data-driven discovery and expert input. We further consider the realistic setting of limited access to an imperfect expert, using disagreement among algorithms to query the expert in cases of greater uncertainty. Experiments demonstrate that our method consistently outperforms strong baselines on both clean and noisy data. Code and data are available at https://anonymous.4open.science/r/expert-cd-ensemble-3282/.", "date": "2026-01-23", "categories": []}
{"id": "2601.21092", "title": "MapPFN: Learning Causal Perturbation Maps in Context", "abstract": "Planning effective interventions in biological systems requires treatment-effect models that adapt to unseen biological contexts by identifying their specific underlying mechanisms. Yet single-cell perturbation datasets span only a handful of biological contexts, and existing methods cannot leverage new interventional evidence at inference time to adapt beyond their training data. To meta-learn a perturbation effect estimator, we present MapPFN, a prior-data fitted network (PFN) pre-trained on a synthetic biological prior with causal interventions, decoupling pre-training from limited wet-lab data. Unlike existing methods, MapPFN uses in-context learning to map a sequence of experiments to a post-perturbation distribution, enabling a single pre-trained model to adapt to new datasets and arbitrary gene sets at inference time. Zero-shot, MapPFN identifies differentially expressed genes on par with models trained on real single-cell data, and fine-tuning further improves predictions across biological contexts. Our code, model and data are available at https://marvinsxtr.github.io/MapPFN.", "date": "2026-01-28", "categories": []}
{"id": "2604.27307", "title": "A Novel Computational Framework for Causal Inference: Tree-Based Discretization with ILP-Based Matching", "abstract": "Causal inference is essential for data-driven decision-making, as it aims to uncover causal relationships from observational data. However, identifying causality remains challenging due to the potential for confounding and the distinction between correlation and causation. While recent advances in causal machine learning and matching algorithms have improved estimation accuracy, these methods often face trade-offs between interpretability and computational efficiency. This paper proposes a novel approach that combines a tree-based discretization technique, tailored for causal inference, with an integer linear programming-based matching algorithm. The discretization ensures approximately linear relationships for control datasets within strata, enabling effective matching, while the optimization framework optimizes for global balance. The resulting algorithm yields computational efficiency and less biased ATT estimates compared to state-of-the-art algorithms. Empirical evaluations demonstrate the proposed method's practical advantages over existing techniques in causal inference scenarios.", "date": "2026-04-30", "categories": []}
{"id": "2605.01669", "title": "PRCD-MAP: Learning How Much to Trust Imperfect Priors in Causal Discovery", "abstract": "External priors of unknown reliability create a brittle trade-off in causal discovery: blind trust amplifies errors, blind rejection wastes signal. Real priors are also heterogeneously reliable -- physical laws are trustworthy, LLM-suggested edges are speculative -- yet existing methods either ignore priors or impose them through globally uniform trust. We propose PRCD-MAP, a soft prior-consumption layer that assigns per-edge trust to an imperfect prior and uses it to modulate a prior-aware $\\ell_1$ and prior-weighted $\\ell_2$ regularizer in a MAP objective. Trust is calibrated by empirical Bayes on a Laplace-approximated marginal likelihood and propagated along the prior graph by an MLP, so data-confirmed neighborhoods boost trust and contradictions suppress it. PRCD-MAP enjoys a population-level safety guarantee: it is $\\varepsilon$-safe in expectation over the prior-generation distribution, with $\\varepsilon\\leq C\\cdot\\mathrm{acc}(1{-}\\mathrm{acc})\\cdot d^2/T$ at the parametric $T^{-1}$ rate and vanishing at the prior-quality endpoints. When the prior is uninformative, learned trust provably collapses to its floor and the method recovers a no-prior baseline. Empirically, on real CausalTime data PRCD-MAP exploits informative LLM priors (LLM-prior gain $+0.067/+0.089$ AUROC on AQI/Medical over a no-prior PRCD-MAP backbone; combined backbone+prior lead $+0.123/+0.043$ over PCMCI+), auto-attenuates on the anonymous-variable Traffic stress test, and retains a lead at $d{=}300$; against BayesDAG, the closest soft-Bayesian baseline, PRCD-MAP wins on every CausalTime dataset under a matched $W_0$-only protocol. A four-way ablation isolates each component: EB calibration and MLP trust propagation jointly carry the plurality of the gain, with positive sign on every dataset. Extensions to nonlinear (NAM) and cross-sectional settings show the calibrated-trust principle is setting-agnostic.", "date": "2026-05-03", "categories": []}
{"id": "2605.05438", "title": "On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning", "abstract": "Standard fine-tuning of transformer models on causal reasoning tasks leads to catastrophic model collapse, where models learn trivial solutions such as always predicting \"Yes\" or \"No\" regardless of input structure. We demonstrate that fine-tuning Gemma 270M on transitivity and d-separation tasks without semantic loss results in 100% collapse rate, with models achieving misleadingly high accuracy (73.9%) while learning no causal reasoning. We propose a semantic loss function with graph-based logical constraints and dynamic lambda scheduling that prevents this collapse. Our approach achieves 70.4% accuracy on transitivity tasks and 68.6% on d-separation tasks with stable, context-dependent predictions, representing a 42.7% improvement over collapsed baselines. Adversarial evaluation on 1,000 structural reasoning samples shows semantic models achieve 67-70% accuracy while collapsed models fail catastrophically at 43-71%. We validate our findings through comprehensive benchmarking on 200,000+ evaluation samples across five model variants, demonstrating that semantic loss is essential and not optional, for stable causal reasoning in transformers.", "date": "2026-05-06", "categories": []}
{"id": "2605.05524", "title": "MOSAIC: Module Discovery via Sparse Additive Identifiable Causal Learning for Scientific Time Series", "abstract": "Causal representation learning (CRL) seeks to recover latent variables with identifiability guarantees, typically up to permutation and component-wise reparameterization under appropriate assumptions. However, identifiability does not imply interpretability: latent semantics are typically assigned post hoc by alignment with known ground-truth factors. This limitation is particularly acute in scientific time series, where underlying mechanisms are unknown and discovering interpretable structure is a primary goal. In contrast, scientific observations (such as residue-pair distances, climate indices, or process sensors) are inherently semantic, as they correspond to named physical quantities. This raises a key question: can the interpretability of observations be transferred to the identifiable latent space? We propose MOSAIC (Module discovery via Sparse Additive Identifiable Causal learning), a sparse temporal VAE that integrates temporal CRL identifiability with support recovery over observed variables. MOSAIC identifies latent variables via regime-conditioned temporal variation, and recovers for each latent a sparse set of associated observations through an additive decoder, yielding module-level interpretability. We show that ANOVA main-effect supports are identifiable under general smooth mixing functions, and provide finite-sample recovery guarantees for a tractable sparse-additive variant. Empirically, MOSAIC recovers domain-consistent variable groups across RNA molecular dynamics, solar wind, ENSO climate, the Tennessee Eastman process, and a synthetic tokamak benchmark, enabling interpretable discovery of latent mechanisms in scientific time series.", "date": "2026-05-06", "categories": []}
{"id": "2605.05568", "title": "Relaxed Sparsest-Permutation Formulation for Causal Discovery at Scale", "abstract": "Despite the growing availability of large datasets, causal structure learning remains computationally prohibitive at scale. We revisit sparsest-permutation learning for linear structural equation models and show that exact Cholesky factorization is unnecessary for structure recovery. This observation motivates a support-level relaxation that searches for sparse triangular factors over a precision-support screening graph. The relaxed formulation can be efficiently evaluated via masked zero-fill incomplete Cholesky factorization, enabling scalable comparison of candidate orderings. At the population level, we establish soundness for Markov equivalence class (MEC) recovery under no-cancellation and sparsest Markov representation assumptions, as well as robustness to ordering misspecification. Motivated by these guarantees, we introduce SCOPE, a sparse-Cholesky pipeline that provides a scalable implementation of the relaxed formulation. Experiments on synthetic and real datasets demonstrate that SCOPE matches the MEC recovery accuracy of substantially slower baselines, while achieving significantly reduced runtime and scaling to 10k variables.", "date": "2026-05-07", "categories": []}
{"id": "2605.05706", "title": "Resolving the bias-precision paradox with stochastic causal representation learning for personalized medicine", "abstract": "Estimating individualized treatment effects from longitudinal observational data is central to data-driven medicine, yet existing methods face a fundamental limitation: reducing confounding bias often suppresses clinically informative heterogeneity, degrading patient-specific predictions. Here, we identify this tension as a bias-precision paradox in causal representation learning and introduce sampling-based maximum mean discrepancy (sMMD), a stochastic alignment strategy that replaces global adversarial balancing with subset-level matching. We instantiate this approach in a framework for counterfactual outcome prediction with attribution-grounded interpretability. Across two large-scale ICU cohorts (n = 27,783), our framework improves accuracy under distribution shift, reducing error by up to 11.5% and substantially increasing recall in high-risk tasks. Mechanistic analyses show that sMMD selectively preserves clinically decisive variables. In human-AI evaluation, our method outperforms clinicians-in-training and large language models, and improves clinician accuracy by 14.7% while reducing decision time, enabling interpretable, real-time clinical decision support.", "date": "2026-05-07", "categories": []}
{"id": "2605.05882", "title": "Tuning Derivatives for Causal Fairness in Machine Learning", "abstract": "Artificial-intelligence systems are becoming ubiquitous in society, yet their predictions typically inherit biases with respect to protected attributes such as race, gender, or age. Classical fairness notions, most notably Statistical Parity (SP), demand that predictions be independent of the protected attributes, but are overly restrictive when these attributes influence mediating variables that are considered business necessities. Recent causal formulations relax SP by distinguishing allowed from not-allowed causal paths and by complementing SP with Predictive Parity (PP), requiring the predictor to replicate the legitimate influence of business-necessities. Existing path-based definitions are mainly practical when applied to categorical attributes. This paper introduces a new framework for fairness in structural causal models that is tailored to continuous protected attributes. We formalize SP and PP through path-specific partial derivatives, establish conditions under which these criteria coincide with prior causal definitions, and characterize when a fair predictor, one that satisfies SP along not-allowed paths while achieving PP along allowed paths, exists. Building on this theory, we propose a fair tuning algorithm that either constructs such a predictor or, when not possible, allows for a trade-off between SP and PP. We present experiments on simulated and real data to evaluate our proposal, compare it with previously proposed methods, and show that it performs better when PP is considered.", "date": "2026-05-07", "categories": []}
{"id": "2605.05890", "title": "RepFlow: Representation Enhanced Flow Matching for Causal Effect Estimation", "abstract": "Estimating causal effects from observational data has become increasingly critical in diverse fields including healthcare, economics, and social policy. The fundamental challenge in causal inference arises from the missing counterfactuals and the selection bias. Existing methods are largely limited to point estimates and lack the capacity for distribution modeling. In this work, we propose RepFlow, a novel framework that formulates causal effect estimation as a joint optimization problem integrating representation learning with Conditional Flow Matching (CFM).\n  RepFlow mitigates selection bias by minimizing the entropically regularized Wasserstein distance between treated and control representations.\n  To enhance numerical stability, we further introduce an $L_2$ normalization constraint on latent representations.\n  This balanced representation enables the flow model to accurately capture the distribution of potential outcomes. Extensive experiments across a wide range of benchmarks demonstrate that RepFlow consistently outperforms existing methods in both point and distributional causal effect estimation.", "date": "2026-05-07", "categories": []}
{"id": "2605.05993", "title": "TabCF: Distributional Control Function Estimation with Tabular Foundation Models", "abstract": "Instrumental variable (IV) and control function (CF) methods are powerful tools for causal effect estimation in the presence of unmeasured confounding, yet most existing approaches target only mean effects and/or demand substantial fitting and tuning effort. In this paper, we introduce a simple method, TabCF, for control function regression using tabular foundation models, which enables accurate, fast, identification-transparent, and tuning-light causal estimation of distributional quantities, such as interventional means and quantiles; we also propose a copula-based approximation for multivariate outcomes. TabCF performs favorably against representative methods across a broad range of small- to medium-sized synthetic and real data scenarios. The central message is two-fold: for practitioners, it highlights that TabCF is an effective tool for distributional causal inference; for researchers, it suggests that the proposed approach could be considered a strong baseline for future method development. Code is available at https://github.com/GepingChen/TabCF.", "date": "2026-05-07", "categories": []}
{"id": "2605.06059", "title": "Correcting heterogeneous diagnostic bias when developing clinical prediction models using causal hidden Markov models", "abstract": "In routine care, individuals identified a priori as high-risk are usually tested for conditions more frequently. Protected attributes, such as sex or ethnicity may also determine testing frequency. Such heterogeneous detection rates across a population induce label error. This causes systematic model error for specific groups and biases performance metrics during validation.\n  This paper proposes a method to correct for such bias in prediction models due to differential diagnostic delay. We use a causal inference framework to define our target estimand: an individual's diagnosis probability in a counterfactual scenario where their diagnosis rate matches that of a reference group. We model the longitudinal process as a hidden Markov model, in which confirmatory test results are emissions from a latent progressive disease stage. We validate our approach in simulated data and apply it to a case study of chronic kidney disease prediction using electronic health records.\n  In simulations, our method reduces prediction bias and improves calibration-in-the-large, correcting the Observed:Expected ratio in the underdiagnosed group from 1.34 (standard deviation: 0.09) in a model developed without any correction for underdiagnosis bias to 1.02 (0.09). Violations of assumptions in the simulation affected the estimation of model parameters, but the proposed approach nonetheless remained better calibrated than the standard model. In the clinical case study, we identify diabetes as the main driver of observability, with an odds ratio of 10.36 (95% confidence interval, 9.80 - 11.02) in 6-month urine albumin-creatinine ratio testing rate. Using our approach to predict the counterfactual diagnostic rate in patients without diabetes, we improved the Observed:Expected ratio of a developed clinical prediction model from 1.55 (1.51 - 1.59) to 1.01 (0.98 - 1.04).", "date": "2026-05-07", "categories": []}
{"id": "2605.06066", "title": "Causal Reinforcement Learning for Complex Card Games: A Magic The Gathering Benchmark", "abstract": "Causal reinforcement learning (RL) lacks benchmarks for complex systems that combine sequential decision making, hidden information, large masked action spaces, and explicit causal structure. We introduce MTG-Causal-RL, a Gymnasium benchmark built on Magic: The Gathering with a 3,077-dimensional partial observation, a 478-action masked discrete action space, five competitive Standard archetypes, three reward schemes, and a hand-specified Structural Causal Model (SCM) over strategic variables. Every episode exposes causal variables, SCM-predicted intervention effects, and per-factor credit traces, making causal credit assignment, leave-one-out cross-archetype transfer, and policy auditability first-class metrics. We adapt a panel of reference baselines: random, heuristic, masked PPO, a causal-world-model PPO variant, and an architecture-matched scalar control. We propose Causal Graph-Factored Advantage PPO (CGFA-PPO) as a reference causal agent that uses SCM parents of win probability as factor-aligned critic targets with an intervention-calibration loss. All comparisons use paired seeds, paired-bootstrap confidence intervals, and Holm-Bonferroni correction within pre-registered families. Masked PPO and CGFA-PPO reach competitive in-distribution win rates and exceed the random baseline; per-factor calibration trajectories and leave-one-out transfer gaps expose diagnostic structure that scalar win rate alone cannot. We release the benchmark, reference-baseline results, and full evaluation protocol openly. By coupling a strategically rich, partially observed domain with an explicit causal interface and statistical protocol, MTG-Causal-RL gives causal-RL, world-model, and LLM-agent research a shared testbed for questions current benchmarks cannot pose together: causal credit assignment under masked action spaces, structural transfer across archetypes, and SCM-grounded policy auditability.", "date": "2026-05-07", "categories": []}
{"id": "2605.06288", "title": "A Topological Sorting Criterion for Random Causal Directed Acyclic Graphs", "abstract": "Random directed acyclic graphs (DAGs) based on imposing an order on Erd\\H{o}s-R\\'enyi and scale free random graphs are widely used for evaluating causal discovery algorithms. We show that in such DAGs, the set of nodes reachable via open paths, termed relatives, increases monotonically along the causal order. We assess the prevalence of this pattern numerically, and demonstrate that it can be exploited for causal order recovery via sorting by the estimated number of relatives. We note that many simulations in the literature feature settings where this yields an excellent proxy for the causal order, and show that a strict increase of relatives along the causal order leads to a singular Markov equivalence class. We propose sampling time-series DAGs as a possible alternative and discuss implications for causal discovery algorithms and their evaluation on synthetic data.", "date": "2026-05-07", "categories": []}
{"id": "2605.06385", "title": "Data-Driven Covariate Selection for Nonparametric and Cycle-Agnostic Causal Effect Estimation", "abstract": "Estimating causal effects from observational data requires identifying valid adjustment sets. This task is especially challenging in realistic settings where latent confounding and feedback loops are present. Existing approaches typically assume acyclicity or rely on global causal structure learning, limiting applicability and computational efficiency. In this work, we study a local, data-driven method for covariate selection based on conditional independence information. While this method is known to be sound and complete in acyclic causal models, its validity in the presence of cycles has remained unclear. Our main contribution is to show that these guarantees extend to cyclic causal models. In particular, our result relies on the invariance of conditional independence assertions under $\\sigma$-acyclification. These findings establish a unified, cycle-agnostic perspective on covariate selection and causal effect estimation, showing that the method applies across cyclic and acyclic settings without modification. Empirically, we validate this on extensive synthetic data, showing reliable performance in cyclic causal models.", "date": "2026-05-07", "categories": []}
{"id": "2605.06386", "title": "Covariate Balancing and Riesz Regression Should Be Guided by the Neyman Orthogonal Score in Debiased Machine Learning", "abstract": "This position paper argues that, in debiased machine learning, balancing functions should be derived from the Neyman orthogonal score, not chosen only as functions of covariates. Covariate balancing is effective when the regression error entering the score can be represented by functions of covariates alone, and it is the natural finite-dimensional approximation for targets such as ATT counterfactual means. For ATE estimation under treatment effect heterogeneity, however, the score error generally contains treatment-specific components because the outcome regression is a function of the full regressor $X=(D,Z)$. In that case, balancing common functions of $Z$ can leave the treatment-specific component unbalanced. We therefore advocate regressor balancing, implemented by Riesz regression with basis functions of $X$, as the general balancing principle for DML. The position is not that covariate balancing is invalid, but that covariate balancing should be understood as the special case that is appropriate when the score-relevant regression error is a function of covariates alone.", "date": "2026-05-07", "categories": []}
{"id": "2605.06608", "title": "DARTS: Targeting Prognostic Covariates in Budget-Constrained Sequential Experiments", "abstract": "Randomized controlled trials typically assume that prognostic covariates are known and available at no cost. In practice, obtaining high-dimensional pretreatment data is costly, forcing a trade-off between covariate-adaptive precision and a measurement budget. We introduce Dynamic Adaptive Rerandomization via Thompson Sampling (DARTS), which treats covariate acquisition as a sequential optimization problem embedded within a design-based causal inference task. A budgeted combinatorial Thompson sampler learns which covariates are most prognostic across successive batches; selected covariates then drive rerandomization and regression adjustment to reduce batch-level average treatment effect variance. Our primary theoretical contribution is a decoupling result: adaptive covariate selection based on past batches preserves batch-level randomization validity, and the cumulative inverse-variance weighted estimator achieves at least nominal asymptotic coverage. We further derive a Bayes risk bound for the acquisition layer that matches the minimax lower bound up to logarithmic factors. Empirically, DARTS systematically concentrates the budget on informative features, significantly closing the efficiency gap to oracle designs while maintaining strict inferential validity.", "date": "2026-05-07", "categories": []}
{"id": "2209.15111", "title": "Quantifying Harm", "abstract": "In earlier work we defined a qualitative notion of harm: either harm is caused, or it is not. For practical applications, we often need to quantify harm; for example, we may want to choose the least harmful of a set of possible interventions. In this work, which is an expanded version of an earlier conference paper, we develop a quantitative notion of harm. We first present a quantitative definition of harm in a deterministic context involving a single individual, then we consider the issues involved in dealing with uncertainty regarding the context and going from a notion of harm for a single individual to a notion of \"societal harm\", which involves aggregating the harm to individuals. We show that the \"obvious\" way of doing this (just taking the expected harm for an individual and then summing the expected harm over all individuals) can lead to counterintuitive or inappropriate answers, and discuss alternatives, drawing on work from the decision-theory literature. Finally, we connect our work to a recent debate over harm within the context of precision medicine.", "date": "2022-09-29", "categories": []}
{"id": "2407.14861", "title": "Improving Bias Correction Standards by Quantifying its Effects on Treatment Outcomes", "abstract": "With the growing access to administrative health databases, retrospective studies have become crucial evidence for medical treatments. Yet, non-randomized studies frequently face selection biases, requiring mitigation strategies. Propensity score matching (PSM) addresses these biases by selecting comparable populations, allowing for analysis without further methodological constraints. However, PSM has several drawbacks. Different matching methods can produce significantly different Average Treatment Effects (ATE) for the same task, even when meeting all validation criteria. To prevent cherry-picking the best method, public authorities must involve field experts and engage in extensive discussions with researchers.\n  To address this issue, we introduce a novel metric, A2A, to reduce the number of valid matches. A2A constructs artificial matching tasks that mirror the original ones but with known outcomes, assessing each matching method's performance comprehensively from propensity estimation to ATE estimation. When combined with Standardized Mean Difference, A2A enhances the precision of model selection, resulting in a reduction of up to 50% in ATE estimation errors across synthetic tasks and up to 90% in predicted ATE variability across both synthetic and real-world datasets. To our knowledge, A2A is the first metric capable of evaluating outcome correction accuracy using covariates not involved in selection.\n  Computing A2A requires solving hundreds of PSMs, we therefore automate all manual steps of the PSM pipeline. We integrate PSM methods from Python and R, our automated pipeline, a new metric, and reproducible experiments into popmatch, our new Python package, to enhance reproducibility and accessibility to bias correction methods.", "date": "2024-07-20", "categories": []}
{"id": "2605.02475", "title": "Shadow-Loom: Causal Reasoning over Graphical World Models of Narratives", "abstract": "Stories hold a reader's attention because they have causes, secrets, and consequences. Shadow-Loom is an experimental open-source framework that turns a narrative into a versioned graphical world model and lets two engines act on it: a causal physics grounded in Pearl's ladder of causation and a recently proposed counterfactual calculus over Ancestral Multi-World Networks; and a narrative physics that scores the same graph against four structural reader-states -- mystery, dramatic irony, suspense, and surprise -- in the tradition of Sternberg's curiosity/suspense/surprise triad, with suspense formalised in the structural-affect line of work on story comprehension and computational suspense. Large language models are used only at the boundary: extraction, rendering, and audit; identification, intervention, and counterfactual reasoning are carried out in typed code over the graph. The system is offered as a research artefact rather than as a benchmarked NLP model; code, fixtures, and pipeline are released open source.", "date": "2026-05-04", "categories": []}
{"id": "2605.03278", "title": "Copula-Based Endogeneity Correction for Doubly Robust Estimation of Treatment Effect", "abstract": "Doubly Robust (DR) estimation of treatment effect relies on an untestable assumption that is the absence of unobserved confounding. This assumption is par- ticularly problematic in the context of healthcare research, where variables like pre- scription refill rates serve as proxies for unobserved behaviors such as medication adherence. These proxy variables are often endogenous, exhibiting correlation with the regression error term due to unmeasured confounding or measurement error. We propose a copula-corrected doubly robust estimator that addresses endogeneity in both the treatment and outcome models without requiring instrumental variables. Gaussian copulas model the joint distribution of endogenous covariates and the error term, enabling consistent estimation while preserving the doubly robust property that requires correct specification of either the treatment or outcome model, not both. Monte Carlo simulations demonstrate that naive DR estimation exhibits substantial bias under endogeneity, whereas our corrected estimator recovers unbiased treatment effects across different data-generating processes. We apply our method to examine the effect of nutritional counseling on blood pressure using the National Health and Nutrition Examination Survey (NHANES) data. Naive DR estimation suggests counseling is associated with increased blood pressure. After copula correction, this effect becomes statistically insignificant, consistent with literature showing modest effects of nutri- Counseling in reducing blood pressure. Our methodology provides researchers with a practical tool for obtaining treatment effects in the presence of endogeneity.", "date": "2026-05-05", "categories": []}
{"id": "2605.04191", "title": "Heterogeneous Ordinal Structure Learning with Bayesian Nonparametric Complexity Discovery", "abstract": "Public attitudes toward artificial intelligence are heterogeneous, ordinally measured, and poorly captured by any single dependency graph. Existing ordinal structure learners assume a shared directed acyclic graph (DAG) across all respondents; recent heterogeneous ordinal graphical-model approaches focus on subgroup discovery rather than confirmatory cluster-specific DAG estimation; and latent profile analyses discard dependency structure entirely. We introduce a heterogeneous ordinal structure-learning framework combining monotone Gaussian score embedding, Bayesian nonparametric (BNP) complexity discovery via a truncated stick-breaking prior, and confirmatory fixed-K estimation with cluster-specific sparse DAG learning. The key methodological insight is a discovery-to-confirmation workflow: the nonparametric stage calibrates plausible archetype complexity, while inner-validated confirmatory refitting yields stable, interpretable structural estimates. On the 2024 Pew American Trends Panel AI attitudes survey, Wave 152 (W152) survey, (N = 4,788, 8 ordinal items), the confirmatory K*=5 model reduces holdout transformed-score mean squared error (MSE) by 25.8% over a single-graph baseline and by 4.6% over mixture-only clustering. A controlled tiered semi-synthetic benchmark calibrated to W152 structure validates recovery across difficulty regimes and transparently reveals failure modes under stress conditions.", "date": "2026-05-05", "categories": []}
{"id": "2605.04313", "title": "NoisyCausal: A Benchmark for Evaluating Causal Reasoning Under Structured Noise", "abstract": "Causal reasoning in natural language requires identifying relevant variables, understanding their interactions, and reasoning about effects and interventions, often under noisy or ambiguous conditions. While large language models (LLMs) exhibit strong general reasoning abilities, they struggle to disentangle correlation from causation, particularly when observations are partially incorrect or irrelevant information is present. In this work, we introduce NoisyCausal, a new benchmark designed to evaluate causal reasoning under structured noise. Each instance is generated from a ground-truth causal graph and contextualized with a natural language scenario by injecting controllable forms of noise, such as irrelevant distractors, value perturbations, confounding, and partial observability. Moreover, we propose a modular reasoning framework that combines LLMs with explicit causal structure to address these challenges. Our method prompts the LLM to extract variables, construct a causal graph from context, and then reformulates the reasoning task as a structured prompt grounded in this graph. Rather than relying on statistical patterns alone, the LLM is guided by symbolic structure, enabling more interpretable and robust inference. Experimental results show that our method significantly outperforms standard prompting and reasoning baselines on NoisyCausal. Furthermore, it generalizes well to external benchmarks such as Cladder without task-specific tuning. Our findings highlight the importance of combining causal abstractions with language-driven reasoning to achieve faithful and robust causal understanding in LLMs.", "date": "2026-05-05", "categories": []}
{"id": "2605.04381", "title": "Causal discovery under mean independence and linearity", "abstract": "Causal discovery methods such as LiNGAM identify causal structure from observational data by assuming mutually independent disturbances. This assumption is fragile: shared volatility, common scale effects, or other forms of dependence can cause the methods to recover the wrong causal order, even with infinite data. We introduce the Linear Mean-Independent Acyclic Model (LiMIAM), which replaces full independence with weaker one-sided mean-independence restrictions on the disturbances. Under finite-order consequences of these restrictions, source nodes are generically identifiable, and hence a compatible causal order can be recovered recursively. Our proof is constructive and leads to DirectLiMIAM, a sequential residual-based algorithm for causal discovery under dependent noise. In simulations with mean-independent but dependent disturbances, DirectLiMIAM outperforms LiNGAM methods. A large-scale empirical application to the oil market highlights the implausibility of the independence assumption and the ability of DirectLiMIAM to recover a realistic causal ordering, from policy to production and from prices to inflation.", "date": "2026-05-06", "categories": []}
{"id": "2605.04413", "title": "Counterfactual identifiability beyond global monotonicity: non-monotone triangular structural causal models", "abstract": "Structural causal models provide a unified semantics for interventions and counterfactuals, but most identifiability results rely on restrictive assumptions like global monotonicity, which are often violated in embodied interaction, where the same exogenous perturbation can induce opposite responses under different contact contexts. We ask what structure still suffices once global monotonicity is dropped. We introduce non-monotone triangular structural causal models (NM-TM-SCM), which retain triangular recursion but replace global monotonicity with mechanism-wise invertibility and context-independent inverse transport. We prove that these conditions are equivalent to exogenous isomorphism and imply complete counterfactual identifiability, and we give a counterexample showing that local invertibility alone is insufficient. We instantiate the theory in CausalInverter, with triangular invertible layers, orientation gates, and transport-stability regularization. On synthetic non-monotonic mechanisms, the structural bias yields systematic counterfactual gains as non-monotonicity increases. On MuJoCo Door, our model achieves perfect event-level counterfactual recovery, lowers continuous angle error relative to a Transformer baseline, and delivers substantially more stable recovery than Transformer and conditional-flow predictors. On MuJoCo Push, where non-monotonicity is weaker, the same low-data predictors remain competitive or better, consistent with a bias-variance boundary. These results identify a broader identifiable regime between globally monotone triangular models and unconstrained black-box world models.", "date": "2026-05-06", "categories": []}
{"id": "2605.04838", "title": "PAIR-CI: Calibrated Conditional Independence Testing for Causal Discovery with Incomplete Data", "abstract": "The standard constraint-based paradigm for causal discovery with incomplete data -- impute first, test second -- is frequently miscalibrated: any consistent conditional independence (CI) test rejects a true null with probability approaching 1 when imputation error induces spurious conditional dependence. We introduce PAIR-CI, a nonparametric CI test that restores calibration by integrating multiple imputation directly into the inferential procedure via a paired permutation design. PAIR-CI compares cross-validated models that include and exclude the candidate variable while receiving the same imputed conditioning set, forcing imputation error to cancel in their loss difference rather than contaminate the test statistic. A provably consistent variance estimator jointly accounts for uncertainty arising from cross-validation and multiple imputation -- to our knowledge, the first formal unification of these two inferential frameworks. In simulations, existing imputation-based CI tests exhibit false positive rates of 28--45% when data are missing not at random (MNAR), whereas PAIR-CI averages below the nominal 5% level across data-generating processes and missingness mechanisms. These gains are largest in nonlinear settings and grow with causal graph size: when integrated into the PC algorithm, PAIR-CI reduces structural Hamming distance by 8% on 10-variable nonlinear graphs, 15% on 30-variable equivalents, and up to 44% on the 56-variable HAILFINDER network, with stable performance in all settings.", "date": "2026-05-06", "categories": []}
{"id": "2605.04930", "title": "When Does Gene Regulatory Network Inference Break? A Controlled Diagnostic Study of Causal and Correlational Methods on Single-Cell Data", "abstract": "Despite theoretical advantages, causal methods for Gene Regulatory Network (GRN) inference from single-cell RNA-seq data consistently fail to match or outperform correlation-based baselines in many realistic benchmarks, a persistent puzzle which casts doubt on the value of causality for this task. We argue that existing benchmarks are insufficiently controlled to answer this question because they evaluate on real or semi-real data where multiple pathologies co-occur, confounding failure modes, and obscuring the specific conditions under which different inference methods excel or fail. To address this gap, we introduce a controlled diagnostic framework that isolates seven biologically motivated pathologies (dropout, latent confounders, cell-type mixing, feedback loops, network density, sample size, and pseudotime drift) and measure how six representative methods spanning three inference paradigms degrade as each pathology intensifies. Across 6,120 controlled experiments, we find that causal methods genuinely dominate in clean and structurally favorable regimes, but specific pathologies (notably dropout and latent confounders) selectively neutralize their advantages. We further introduce an error-type decomposition that reveals methods with similar aggregate accuracy commit qualitatively different errors. To probe whether single-pathology effects persist when multiple stressors co-occur, we perform an interaction sweep over the three most impactful pathologies and find that their joint effects are sub-additive, while also exposing density-conditional cross-overs invisible to single-dial analysis. Our findings offer a nuanced understanding of when and why different methods succeed or fail for GRN inference, providing actionable insights for method development and practical guidance for practitioners.", "date": "2026-05-06", "categories": []}
{"id": "2605.04955", "title": "Order-based Rehearsal Learning", "abstract": "When a machine learning (ML) model forecasts an undesired event, one often seeks a decision to avoid it, known as the avoiding undesired future (AUF) problem. Many rehearsal learning methods have been proposed for AUF, but they rely on an underlying graph structure; learning such a graph from observational data is challenging and can incur substantial estimation error. In this work, we demonstrate that the order structure can be sufficient for AUF decision-making, and propose the first order-based rehearsal learning method. Although an order is less informative than a graph, it can be sufficient to identify the influence of decisions from observational data, suggesting that learning the entire graph is not always necessary. To learn the order, we develop an information-theoretic method that imposes no restrictions on the form of structural functions or the type of noise distributions. For AUF decision-making, we construct an order-based sampler to approximate the influence of decisions and, combined with a surrogate objective for maximizing the post-decision success probability, reduce the AUF task to a differentiable optimization problem. Experiments show that our order learning method outperforms existing methods, and that our AUF approach not only surpasses methods relying on learned graphs or learned orders, but also matches or even exceeds oracle baselines that are given the true graph.", "date": "2026-05-06", "categories": []}
{"id": "2605.05029", "title": "The Predictive-Causal Gap: An Impossibility Theorem and Large-Scale Neural Evidence", "abstract": "We report a systematic failure mode in predictive representation learning. Across 2695 neural network configurations trained to predict linear-Gaussian dynamics, the optimal encoder tracks the environment rather than the system it is meant to model. The mean causal fidelity -- the fraction of encoder sensitivity allocated to system degrees of freedom -- is 0.49, and only 2.5% of configurations exceed 0.70. The failure intensifies with dimension: at N=100, the optimal encoder becomes causally blind (fidelity ~10^{-8}) while achieving 92% lower prediction error than the causal representation. We prove this is not an optimization artifact but a structural property of the predictive objective: when environment modes are slower or less noisy than system modes, every minimizer of the population risk encodes the former. The set of dynamics exhibiting this predictive-causal gap is open and of positive measure in parameter space. In a nonlinear Duffing-GRU sweep, unconstrained predictors learn environment-dominant representations in 55% of tasks (95% CI 41--68%) versus 24% under operational grounding (p=2.3e-3); the median out-of-distribution MSE inflation under environment shift is 1.82x versus 1.00x. Operational grounding -- restricting the loss to system observables -- partially suppresses the gap, but causal fidelity is never recovered without an explicit system-environment boundary. The results identify the predictive-causal gap as a structural limit of learning, with implications for self-supervised representation learning, world models, and the scaling paradigm.", "date": "2026-05-06", "categories": []}
{"id": "2605.05125", "title": "Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation", "abstract": "Target trial emulation (TTE) enables causal questions to be studied with observational data when randomized controlled trials (RCTs) are infeasible. Yet treatment-effect methods often address causal estimation, missingness, and temporal structure separately, limiting their robustness in electronic health records (EHRs), where time-varying confounding and missing-not-at-random (MNAR) biomarkers can reach 50%--80%. We propose a two-stage pipeline for treatment effect estimation from incomplete longitudinal EHRs. First, CausalFlow-T, a directed acyclic graph (DAG)-constrained normalizing flow with long short-term memory (LSTM)-encoded patient history, performs exact invertible counterfactual inference, avoiding approximation errors from variational inference and separating confounding through explicit causal structure. Ablations on four synthetic and one semi-synthetic benchmark with known counterfactuals show that DAG constraints and exact inference address distinct failure modes: neither compensates for the other. Second, because CausalFlow-T requires completed inputs, we introduce an LLM-driven evolutionary imputer that proposes executable imputation operators rather than individual entries, and evaluate it with three large language model (LLM) backends, including two open-source models. Across 30%--80% MNAR missingness, this imputer achieves the best pooled rank over biomarker and causal metrics, leading in point-wise accuracy and temporal extrapolation while preserving average treatment effect (ATE) recovery as statistical baselines degrade. On Swiss primary-care EHRs from adults with type 2 diabetes initiating a GLP-1 receptor agonist or SGLT-2 inhibitor, the pipeline estimates a per-protocol weight-loss difference of -0.98 kg [95% CI -1.01, -0.96] favoring GLP-1 receptor agonists, consistent with randomized evidence and obtained from realistically incomplete real-world EHRs.", "date": "2026-05-06", "categories": []}
{"id": "2605.02962", "title": "ISAAC: Auditing Causal Reasoning in Deep Models for Drug-Target Interaction", "abstract": "Deep learning models for drug--target interaction (DTI) prediction often achieve strong benchmark performance without necessarily relying on mechanistically meaningful molecular features, a limitation that standard accuracy-based evaluation cannot detect. We introduce ISAAC (Intervention-based Structural Auditing Approach for Causal Reasoning), a post-hoc framework that evaluates prior-relative structural sensitivity by probing frozen models through matched mechanistic and spurious input-level interventions, independently of predictive accuracy. Applied to three sequence-based DTI architectures on the Davis benchmark, ISAAC reveals approximately 25\\% relative differences in reasoning scores across models with comparable AUROC (within around 3\\%), stable across training and intervention seeds and two distinct perturbation operators. These discrepancies, undetectable under conventional accuracy metrics, motivate the use of post-hoc structural auditing as a complement to standard performance evaluation in scientific machine learning for molecular modeling.", "date": "2026-05-03", "categories": []}
{"id": "2605.03045", "title": "TCD-Arena: Assessing Robustness of Time Series Causal Discovery Methods Against Assumption Violations", "abstract": "Causal Discovery (CD) is a powerful framework for scientific inquiry. Yet, its practical adoption is hindered by a reliance on strong, often unverifiable assumptions and a lack of robust performance assessment. To address these limitations and advance empirical CD evaluation, we present TCD-Arena, a modularized, highly customizable, and extendable testing kit to assess the robustness of time series CD algorithms against stepwise more severe assumption violations. For demonstration, we conduct an extensive empirical study comprising around 30 million individual CD attempts and reveal nuanced robustness profiles for 33 distinct assumption violations. Further, we investigate CD ensembles and find that they have the potential to improve general robustness, which has implications for real-world applications. With this, we strive to ultimately facilitate the development of CD methods that are reliable for a diverse range of synthetic and potentially real-world data conditions.", "date": "2026-05-04", "categories": []}
{"id": "2605.03267", "title": "Partial Effective Information Decomposition for Synergistic Causality", "abstract": "Causality is a central topic in scientific inquiry, yet for complex systems, the identification and analysis of synergistic causation remain a challenging and fundamental problem. In the context of causal relations among multivariate variables, a decomposition framework grounded in interventionist causation is still lacking. To address this gap, this paper proposes Partial Effective Information Decomposition (PEID), a framework that decomposes the influence of multiple source variables on a target variable under maximum-entropy interventions into unique and synergistic information, thereby providing a unified and computable characterization of synergistic causal relations. Theoretically, in the three-variable case, the proposed framework is compatible with the major axioms of Partial Information Decomposition (PID). Empirically, under maximum-entropy interventions, correlations among input variables are removed, causing redundancy to vanish and thereby enabling PEID to compute synergistic relations. Furthermore, based on this framework, it is possible to define causal graphs containing hyperedges as well as downward causation, thus offering a unified toolkit for analyzing cross-scale and multivariate causal mechanisms in complex systems. Finally, applying the framework to a machine-learning-based air quality forecasting task on KnowAir-V2, we demonstrate that PEID can extract interpretable inter-station causal structures from a learned dynamical model. These results suggest that PEID provides a general interventionist information-theoretic tool for analyzing multivariate and synergistic causal mechanisms in complex systems.", "date": "2026-05-05", "categories": []}
{"id": "2605.03268", "title": "Partially Observed Structural Causal Models", "abstract": "Here we introduce Partially Observed Structural Causal Models (POSCMs) that formalize causal systems where latent contexts co-determine both the interaction structure and downstream mechanisms on observed variables. POSCMs provide an extension of structural causal models (SCMs), as a self-contained causal modeling framework for endogenous graphs, allowing for an intervention hierarchy spanning node- and edge-level context and endogenous variable interventions. To enable surgical edge interventions, we adopt a Kolmogorov-Arnold-Sprecher edge-functional decomposition, an existence theorem for representing each node mechanism as a sum of univariate functions of its parents, yielding an explicit parametrization of dyadic functional contributions. We provide an identifiability theory that clarifies which intervention families would suffice to disentangle structure formation from mechanisms. We empirically validate these predictions in a biophysically detailed virtual human retina simulator, constructing intervention protocols that (i) reproduce the non-identifiability predicted when context is latent and no context-level interventions are available, (ii) exhibit structure-mechanism confounding under latent edges when only node interventions are observed, and (iii) recover synaptic input-output relationships via targeted node interventions, consistent with our positive kernel identifiability result. Our work generalizes SCMs in a way that allows it to work in a world closer to the one we live in.", "date": "2026-05-05", "categories": []}
{"id": "2510.10245", "title": "Kernel Treatment Effects with Adaptively Collected Data", "abstract": "Adaptive experiments improve efficiency by adjusting treatment assignments based on past outcomes, but this adaptivity breaks the i.i.d.\\ assumptions that underpin classical asymptotics. At the same time, many questions of interest are distributional, extending beyond average effects. Kernel treatment effects (KTE) provide a flexible framework by representing interventional outcome distributions in an RKHS and comparing them via kernel distances. We present the first kernel-based framework for distributional inference under adaptive data collection. Our method combines doubly robust RKHS scores with a witness function learned on one fold, and performs inference on a second fold using a projected, sequentially normalized scalar statistic with valid type-I error. Experiments show that the resulting procedure is well calibrated and effective for both mean shifts and higher-moment differences, outperforming adaptive baselines limited to scalar effects.", "date": "2025-10-11", "categories": []}
{"id": "2511.08303", "title": "Semi-Supervised Treatment Effect Estimation with Unlabeled Covariates for Prediction-Powered Causal Inference", "abstract": "This study investigates treatment effect estimation in the semi-supervised setting, also can be interpreted as prediction-powered inference. In our setting, we can use not only the standard triple of covariates, treatment indicator, and outcome, but also unlabeled auxiliary covariates. For this problem, we develop efficiency bounds and efficient estimators whose asymptotic variance aligns with the efficiency bound. In the analysis, we introduce two different data-generating processes: the one-sample setting and the two-sample setting. The one-sample setting considers the case where we can observe treatment indicators and outcomes for a part of the dataset, which is also called the censoring setting. In contrast, the two-sample setting considers two independent datasets with labeled and unlabeled data, which is also called the case-control setting or the stratified setting. In both settings, we find that by incorporating auxiliary covariates, we can lower the efficiency bound and obtain an estimator with an asymptotic variance smaller than that without such auxiliary covariates. We frame our framework as prediction-powered causal inference.", "date": "2025-11-11", "categories": []}
{"id": "2603.11907", "title": "Causal Representation Learning with Optimal Compression under Complex Treatments", "abstract": "Estimating Individual Treatment Effects (ITE) in multi-treatment scenarios faces two critical challenges: the Hyperparameter Selection Dilemma for balancing weights and the Curse of Dimensionality in computational scalability. This paper derives a novel multi-treatment generalization bound and proposes a theoretical estimator for the optimal balancing weight $\\alpha$, eliminating expensive heuristic tuning. We investigate three balancing strategies: Pairwise, One-vs-All (OVA), and Treatment Aggregation. While OVA achieves superior precision in low-dimensional settings, our proposed Treatment Aggregation ensures both accuracy and O(1) scalability as the treatment space expands. Furthermore, we extend our framework to a generative architecture, Multi-Treatment CausalEGM, which preserves the Wasserstein geodesic structure of the treatment manifold. Experiments on semi-synthetic and image datasets demonstrate that our approach significantly outperforms traditional models in estimation accuracy and efficiency, particularly in large-scale intervention scenarios.", "date": "2026-03-12", "categories": []}
{"id": "2605.01311", "title": "The Partial Testimony of Logs: Evaluation of Language Model Generation under Confounded Model Choice", "abstract": "Offline evaluation of language models from usage logs is biased when model choice is confounded: the same user-side factors that influence which model is used can also influence how its output is judged, so raw comparisons of logged scores mix self-selected populations rather than estimating a common quantity of interest. A small randomized experiment can break this bias by overriding model choice, but in practice such experiments are scarce and costly. We study a three-source design that combines a large confounded observational log (OBS) for scale, a small randomized experiment (EXP) for unconfounded scoring, and an offline simulator (SIM) that replays candidate models on cached contexts. Our main result is an identification theorem showing that the randomized experiment and the simulator are together enough to recover causal model values; the observational log enters only afterward, to reduce estimation error rather than to make the causal comparison valid. Six estimator families are evaluated in a controlled semi-synthetic validation and in two real-task cached benchmarks for summarization and coding. No family dominates every regime; relative performance depends on the amount of unbiased EXP supervision and on how closely the target reward aligns with OBS-derived structure.", "date": "2026-05-02", "categories": []}
{"id": "2605.01579", "title": "Minimum Specification Perturbation: Robustness as Distance-to-Falsification in Causal Inference", "abstract": "Empirical causal claims depend on many analyst decisions, from selecting covariates to choosing estimators. Existing robustness tools summarize how results vary across these choices, but, to the best of our knowledge, do not answer: \\textbf{How many analyst decisions must change to reach a specification, which is a set of choices, whose confidence interval (CI) contains zero?} We introduce \\emph{Minimum Specification Perturbation (MSP)}, the smallest number of changes. MSP is small under the null, grows with effect strength and captures distance-to-falsification information that dispersion-based summaries cannot report; when making decisions under weak effects, an MSP-based rule yields lower false-positive rates than dispersion-based rules. We show that Fragility Index and MSP measure orthogonal vulnerabilities: fragility to influential observations need not imply fragility to specification choices. On the LaLonde benchmark, MSP = 1 implies that one decision change makes the CI contain zero. We further provide exact permutation calibration under randomization and characterize computation, showing tractable cases under additive structure and NP-hardness in general.", "date": "2026-05-02", "categories": []}
{"id": "2605.01756", "title": "The (Marginal) Value of a Search Ad: An Online Causal Framework for Repeated Second-price Auctions", "abstract": "Existing auto-bidding algorithms in digital advertising often treat the value of an ad opportunity as the revenue obtained when an ad is shown and/or clicked, and bid accordingly. This can lead to wasteful spending because the true value is the marginal gain from paid exposure: even without winning a sponsored slot, an advertiser may still earn revenue via an organic search result (e.g., on Google or Amazon). Motivated by recent work, we model ad value as a treatment effect--the outcome difference between winning and losing the auction--and study online learning for bidding in second-price (Vickrey) auctions under this causal perspective. We develop algorithms that attain rate-optimal regret under several feedback models. A key ingredient exploits the information revealed by the second-price payment rule, which strictly improves regret relative to analogous learning problems in first-price auctions.", "date": "2026-05-03", "categories": []}
{"id": "2605.01765", "title": "Distributional Causal Mediation via Conditional Generative Modeling", "abstract": "Mediation analysis has traditionally focused on outcome-level summary contrasts, such as mean effects, which may obscure substantial distributional changes induced by complex and nonlinear causal mechanisms. We propose Distributional Causal Mediation Analysis (DCMA), a generative learning framework for identifying and estimating treatment effects on entire outcome distributions transmitted through multiple mediators. DCMA learns conditional generative models for the mediators and the outcome, recovering the relevant conditional distributions from observational data. Leveraging the identification formulas, it reconstructs interventional outcome distributions via Monte Carlo forward simulation by noise resampling, enabling the capture of both classical summary effects and rich distributional contrasts such as energy distance and the Wasserstein distance. Analytical error bounds are derived to decompose how estimation errors in the learned conditional models propagate to the reconstructed interventional outcome distributions. The empirical effectiveness of DCMA is demonstrated through numerical experiments and real-world data applications.", "date": "2026-05-03", "categories": []}
{"id": "2605.01856", "title": "Stable Blanket with Hidden Variables and Cycles", "abstract": "Stabilized regression aims to identify a set of predictors whose conditional relationship with a response variable remains invariant across different environments. Existing graphical characterizations of the stable blanket are mainly developed for structural causal models (SCMs) without hidden variables or causal cycles. However, latent variables and feedback relationships naturally arise in many applications, and they can change both the Markov blanket and the set of predictors that remain stable under interventions. This paper studies stable blankets in graphical causal models with hidden variables, causal cycles, and both features simultaneously. For models with hidden variables, we use acyclic directed mixed graphs (ADMGs) and $m$-separation to characterize the Markov blanket and to construct intervention-stable predictor sets. We introduce the notion of an intervened sub-district and use it to describe how interventions may affect districts connected to the response. For models with cycles, we work with directed graphs (DGs) and directed mixed graphs (DMGs) together with $\\sigma$-separation, treating strongly connected components (SCCs) as the basic graphical units. We then combine these ideas to analyze models with both hidden variables and cycles. The main results give graphical characterizations of Markov blankets, stable frontiers, and stable blankets in these generalized settings. In particular, we identify conditions under which the response is conditionally independent of intervention variables given a suitable predictor set, and we describe when such sets are minimal or unique. These results extend the graphical interpretation of stabilized regression beyond acyclic fully observed models.", "date": "2026-05-03", "categories": []}
{"id": "2605.02177", "title": "The Causal Description Gap: Information-Theoretic Separations Across Pearl's Hierarchy", "abstract": "Pearl's causal hierarchy shows that observational, interventional, and counterfactual queries are qualitatively distinct. We ask a quantitative version of this question: how many additional bits are needed to specify higher-rung causal answers once lower-rung answers are known? We formalize this via query-class description length, the Kolmogorov complexity of the answer oracle induced by an SCM for a class of queries. Our main construction gives binary acyclic SCMs whose observational distribution has constant description length, while the single-variable interventional answer oracle has description length $\\Theta(n^2)$. A degree-sensitive upper bound shows that finite-gate-schema SCMs of indegree $d$ have observational-interventional gap at most $O(nd \\log(en/d) + n \\log n)$, making the quadratic construction order-optimal in the dense regime and a rooted-tree construction order-optimal for bounded indegree. The quadratic separation persists under $\\varepsilon$-accurate total-variation descriptions for every fixed $\\varepsilon < 1/4$. At the next rung, the full hard-do interventional oracle can still leave a $\\Theta(n)$ counterfactual description gap. A general ambiguity-to-bits theorem and Shannon analogue show that these gaps equal the logarithm of residual higher-rung ambiguity up to lower-order terms.", "date": "2026-05-04", "categories": []}
{"id": "2605.02318", "title": "Can Causal Discovery Algorithms Help in Generating Legal Arguments?", "abstract": "In 2011, Judea Pearl received the Turing Award, considered the Nobel Prize in Computing, for fundamental contributions to artificial intelligence through the development of a calculus for probabilistic and causal reasoning. It includes pioneering the development of causal discovery algorithms. These computer algorithms can analyze large multivariate datasets and automatically discover the causal relationships among the constituent variables. They have been widely used in many critical fields such as medicine and economics to support decisions. However, to our knowledge, they have not been leveraged in law. This paper attempts to alleviate this gap by investigating whether causal discovery algorithms can be leveraged for automated generation of legal arguments. To that end, a novel legal dataset is prepared by identifying 17 legal concepts, such as physical assault and property dispute. A curated collection of 150 homicide cases are annotated with these concepts, e.g., a case is annotated with physical assault only if a physical assault had been reported in that case. Subsequently, a selected set of widely-used causal discovery algorithms is applied to the annotated dataset to discover the causal relationships between the legal concepts. Additionally, the degrees of belief associated with the discovered relationships are quantified in mathematical probabilities. It is shown that some of the causal relationships help generate viable legal arguments, e.g., if one could establish that a physical assault has not taken place during a homicide, it should be a sufficient condition (with probability 1) to establish that the homicide has not been committed due to a property-related dispute. Thus, this paper shows that causal discovery algorithms can be helpful in generating legal arguments, opening up avenues for promising future endeavors.", "date": "2026-05-04", "categories": []}
{"id": "2605.02454", "title": "Causal Software Engineering: A Vision and Roadmap", "abstract": "Software engineering increasingly involves making high-stakes decisions under uncertainty, using signals from code, field data, and socio-technical processes. Recent AI-driven support (e.g., anomaly detection, predictive analytics, AIOps, as well as LLM-based agents) has amplified engineers' ability to detect patterns and synthesize content and recommendations, but many critical questions are interventional or counterfactual: What is the expected impact of changing a load-balancing strategy? Would an outage have been avoided under a different release plan? Correlational models answer \"what tends to co-occur\"; they struggle to answer \"what would happen if we act.\" We propose Causal Software Engineering (CSE) as a future paradigm in which causal models and causal reasoning systematically inform activities across the software lifecycle, augmenting existing practices with explicit assumptions, uncertainty-aware effect estimates, and counterfactual diagnosis. We outline (i) a causal-first workflow view spanning development and operations, (ii) a staged roadmap for tools and organizational adoption, and (iii) an evaluation and benchmark agenda for measuring progress.", "date": "2026-05-04", "categories": []}
{"id": "2605.02681", "title": "The Design and Composition of Structural Causal Decision Processes", "abstract": "We present two new classes of causal models of decision-making agents. Our approach is motivated by the needs of modeling the economics of computing systems. These systems are composed of subsystems and can exhibit endogenous limits on cognitive resources and value discounting. Structural Causal Decision Models (SCDMs) expand on Structural Causal Influence Models. Like SCIMs, they explicitly represent the causal relationships between model variables and the payoffs of agent decisions. Additionally, agent decisions can be constrained by their causal antecedents, and SCDMs can have open root variables for which no probability distribution or structural equation is given. We show that SCDMs have a well-defined and computationally useful property of composability. Building on SCDMs, we then define a Structural Causal Decision Process (SCDP) as a recurring SCDM with a discount variable. SCDPs benefit from the useful composition properties of SCDMs. Moreover, SCDPs are strictly more expressive than POMDPs because they do not assume rational belief formation. Indeed, an SCDP can endogenously model the memory-formation process, and is thus useful for modeling resource rational agents in dynamic settings. SCDPs are also capable of modeling variable discounting, a tool used widely in social scientific modeling. We pose that SCDPs are a useful framework for policy simulation for the digital economy, mechanism design for information systems, and digital twin modeling of cyberinfrastructure.", "date": "2026-05-04", "categories": []}
{"id": "2503.14459", "title": "Doubly robust identification of treatment effects from multiple environments", "abstract": "Practical and ethical constraints often require the use of observational data for causal inference, particularly in medicine and social sciences. Yet, observational datasets are prone to confounding, potentially compromising the validity of causal conclusions. While it is possible to correct for biases if the underlying causal graph is known, this is rarely a feasible ask in practical scenarios. A common strategy is to adjust for all available covariates, yet this approach can yield biased treatment effect estimates, especially when post-treatment or unobserved variables are present. We propose RAMEN, an algorithm that produces unbiased treatment effect estimates by leveraging the heterogeneity of multiple data sources without the need to know or learn the underlying causal graph. Notably, RAMEN achieves doubly robust identification: it can identify the treatment effect whenever the causal parents of the treatment or those of the outcome are observed, and the node whose parents are observed satisfies an invariance assumption. Empirical evaluations on synthetic and real-world datasets show that our approach outperforms existing methods.", "date": "2025-03-18", "categories": []}
{"id": "2510.04378", "title": "Score-based Greedy Search for Structure Identification of Partially Observed Linear Causal Models", "abstract": "Identifying the structure of a partially observed causal system is essential to various scientific fields. Recent advances have focused on constraint-based causal discovery to solve this problem, and yet in practice these methods often face challenges related to multiple testing and error propagation. These issues could be mitigated by a score-based method and thus it has raised great attention whether there exists a score-based greedy search method that can handle the partially observed scenario. In this work, we propose the first score-based greedy search method for the identification of structure involving latent variables with identifiability guarantees. Specifically, we propose Generalized N Factor Model and establish the global consistency:\n  the true structure including latent variables can be identified up to the Markov equivalence class by using score. We then design\n  Latent variable Greedy Equivalence Search (LGES), a greedy search algorithm for this class of model with well-defined operators,\n  which search very efficiently over the graph space to find the optimal structure. Our experiments on both synthetic and real-life data validate the effectiveness of our method (code will be publicly available).", "date": "2025-10-05", "categories": []}
{"id": "2605.00398", "title": "M-CaStLe: Uncovering Local Causal Structures in Multivariate Space-Time Gridded Data", "abstract": "Causal graph discovery for space-time systems is challenging in high-dimensional gridded data, which often has many more grid cells than temporal observations per cell. The Causal Space-Time Stencil Learning (CaStLe) meta-algorithm was developed to address that niche under space-time locality and stationarity assumptions, but it is currently limited to univariate analyses. In this work, we present M-CaStLe. M-CaStLe generalizes the local embedding and parent-identification phases of CaStLe to jointly model local within-variable and cross-variable space-time causal structures in gridded data. Like CaStLe, by constraining candidate parents to a constant-size space-time neighborhood and pooling spatial replicates, M-CaStLe increases effective sample size to make discovery tractable in high-dimensional settings. We further decompose the resulting multivariate stencil graph into reaction and spatial graphs to aid interpretation in complex settings.\n  We study M-CaStLe in four settings: a multivariate space-time vector autoregression benchmark with known ground truth, an advective-diffusive-reaction partial differential equation verification problem with derived physical reference structure, an atmospheric chemistry case study in a low-temporal-sample regime, and an El Ni\\~{n}o Southern Oscillation study on reanalysis data, identifying phase-dependent ocean--atmosphere coupling. Across these settings, M-CaStLe more accurately recovers multivariate causal structure in controlled settings and identifies important physical dynamics in real-world case studies. Overall, M-CaStLe advances causal discovery for multivariate space-time systems while retaining interpretability at the grid level.", "date": "2026-05-01", "categories": []}
{"id": "2602.07915", "title": "CausalCompass: Evaluating the Robustness of Time-Series Causal Discovery in Misspecified Scenarios", "abstract": "Causal discovery from time series is a fundamental task in machine learning. However, its widespread adoption is hindered by a reliance on untestable causal assumptions and by the lack of robustness-oriented evaluation in existing benchmarks. To address these challenges, we propose CausalCompass, a flexible and extensible benchmark framework designed to assess the robustness of time-series causal discovery (TSCD) methods under violations of modeling assumptions. To demonstrate the practical utility of CausalCompass, we conduct extensive benchmarking of representative TSCD algorithms across eight assumption-violation scenarios. Our experimental results indicate that no single method consistently attains optimal performance across all settings. Nevertheless, the methods exhibiting superior overall performance across diverse scenarios are almost invariably deep learning-based approaches. We further provide hyperparameter sensitivity analyses to deepen the understanding of these findings. We additionally conduct ablation experiments to explain the strong performance of deep learning-based methods under assumption violations. We also find, somewhat surprisingly, that NTS-NOTEARS relies heavily on standardized preprocessing in practice, performing poorly in the vanilla setting but exhibiting strong performance after standardization. Finally, our work aims to provide a comprehensive and systematic evaluation of TSCD methods under assumption violations, thereby facilitating their broader adoption in real-world applications. The user-friendly implementation, documentation and datasets are available at https://anonymous.4open.science/r/CausalCompass-anonymous-5B4F/.", "date": "2026-02-08", "categories": []}
{"id": "2604.27394", "title": "Bayesian X-Learner: Calibrated Posterior Inference for Heterogeneous Treatment Effects under Heavy-Tailed Outcomes", "abstract": "Conditional Average Treatment Effect (CATE) estimation in practice demands three properties simultaneously: heterogeneous effects $\\tau(x)$, calibrated uncertainty over them, and robustness to the heavy tails that contaminate real outcome data. Meta-learners (K\\\"unzel et al., 2019) give (i); causal forests and BART give (i)-(ii) with Gaussian-tail assumptions; no widely used tool gives all three. We present Bayesian X-Learner, an X-Learner built on cross-fitted doubly robust pseudo-outcomes (Kennedy, 2020) with a full MCMC posterior over $\\tau(x)$ via a Welsch redescending pseudo-likelihood. On Hill's IHDP benchmark the default configuration attains mean $\\sqrt{\\varepsilon_{\\mathrm{PEHE}}} = 0.56$ on 5 replications (lowest mean; differences from S-/T-/X-learners, full-config Causal BART, and a causal forest baseline are not significant at $\\alpha=0.05$, and rank ordering is unstable at 10 replications -- IHDP comparisons are competitive rather than dominant). On contaminated \"whale\" DGPs with up to 20-25% tail density, a one-flag extension (contamination_severity) that selects a Huber-$\\delta$ nuisance loss per Huber's minimax-$\\delta$ relation recovers RMSE $\\approx 0.13$ with tight credible intervals (single-cross-fit 30-seed coverage 83% [Wilson 66%, 93%] at 20% density; modular-Bayes pooling with Bayesian-bootstrap nuisance draws restores nominal 95% coverage).", "date": "2026-04-30", "categories": []}
{"id": "2509.25236", "title": "Networks of Causal Abstractions: A Sheaf-theoretic Framework", "abstract": "A core challenge in causal artificial intelligence is the principled coordination of multiple, imperfect, and subjective causal perspectives arising from distributed agents with limited and heterogeneous access to the environment. This problem has received little formal treatment, as the existing framework assumes a single shared global causal model. This work introduces the causal abstraction network (CAN), a general sheaf-theoretic framework for representing, learning, and reasoning across collections of mixture of causal models (MCMs) - a class that unifies several existing models of context-dependent causal mechanisms. Sheaf theory provides a natural foundation for this task, offering a rigorous framework to coherently align distributed causal knowledge without requiring explicit causal graphs, functional mechanisms, interventional data, or jointly sampled observations.\n  At the theoretical level, we provide a categorical formulation of MCMs and characterize key properties of CANs, including consistency and smoothness. Under consistency, we establish necessary and sufficient conditions: (i) for the existence of global sections, linked to spectral properties of an associated connection Laplacian; and (ii) for the convergence of causal knowledge diffusion over the CAN to the space of global sections. At the methodological level, we exploit the compositionality of causal abstractions to decompose the learning of consistent CANs into local problems on network edges, extending our prior work on Gaussian variables to Gaussian mixtures via the proposed MIXTURE-CALSEP algorithm. We validate the framework on synthetic data and through a financial application involving a multi-agent trading system, demonstrating CAN recovery, CAN-based portfolio optimization, and counterfactual reasoning.", "date": "2025-09-25", "categories": []}
{"id": "2510.06735", "title": "Incorporating Expert Knowledge into Bayesian Causal Discovery of Mixtures of Directed Acyclic Graphs", "abstract": "Bayesian causal discovery benefits from prior information elicited from domain experts, and in heterogeneous domains any prior knowledge would be badly needed. However, so far prior elicitation approaches have assumed a single causal graph and hence are not suited to heterogeneous domains. We propose a causal elicitation strategy for heterogeneous settings, based on Bayesian experimental design (BED) principles, and a variational mixture structure learning (VaMSL) method -- extending the earlier differentiable Bayesian structure learning (DiBS) method -- to iteratively infer mixtures of causal Bayesian networks (CBNs). We construct an informative graph prior incorporating elicited expert feedback in the inference of mixtures of CBNs. Our proposed method successfully produces a set of alternative causal models (mixture components or clusters), and achieves an improved structure learning performance on heterogeneous synthetic data when informed by a simulated expert. Finally, we demonstrate that our approach is capable of capturing complex distributions in a breast cancer database.", "date": "2025-10-08", "categories": []}
{"id": "2603.09145", "title": "Causally Sufficient and Necessary Feature Expansion for Class-Incremental Learning", "abstract": "Current expansion-based methods for Class Incremental Learning (CIL) effectively mitigate catastrophic forgetting by freezing old features. However, such task-specific features learned from the new task may collide with the old features. From a causal perspective, spurious feature correlations are the main cause of this collision, manifesting in two scopes: (i) guided by empirical risk minimization (ERM), intra-task spurious correlations cause task-specific features to rely on shortcut features. These non-robust features are vulnerable to interference, inevitably drifting into the feature space of other tasks; (ii) inter-task spurious correlations induce semantic confusion between visually similar classes across tasks. To address this, we propose a Probability of Necessity and Sufficiency (PNS)-based regularization method to guide feature expansion in CIL. Specifically, we first extend the definition of PNS to expansion-based CIL, termed CPNS, which quantifies both the causal completeness of intra-task representations and the separability of inter-task representations. We then introduce a dual-scope counterfactual generator based on twin networks to ensure the measurement of CPNS, which simultaneously generates: (i) intra-task counterfactual features to minimize intra-task PNS risk and ensure causal completeness of task-specific features, and (ii) inter-task interfering features to minimize inter-task PNS risk, ensuring the separability of inter-task representations. Theoretical analyses confirm its reliability. The regularization is a plug-and-play method for expansion-based CIL to mitigate feature collision. Extensive experiments demonstrate the effectiveness of the proposed method.", "date": "2026-03-10", "categories": []}
{"id": "2604.26169", "title": "Budget-Constrained Causal Bandits: Bridging Uplift Modeling and Sequential Decision-Making", "abstract": "Treatment allocation under budget constraints is a central challenge in digital advertising: advertisers must decide which users to show ads to while spending a limited budget wisely. The standard approach follows a two-stage offline pipeline - first collect historical data to estimate heterogeneous treatment effects (HTE), then solve a constrained optimization to allocate the budget. This works well with abundant data, but fails in cold-start settings such as new campaigns, new markets, or new customer segments where little historical data exists. We propose Budget-Constrained Causal Bandits (BCCB), an online framework that learns which users respond to ads while simultaneously spending the budget, making treatment decisions one user at a time. BCCB unifies three components into a single sequential process: learning individual-level ad effectiveness, exploring users whose response is uncertain, and pacing the budget over time. We evaluated on the Criteo Uplift dataset, a large-scale advertising dataset from a real randomized controlled trial. Our key finding is a data-efficiency crossover: offline methods require approximately 10,000 historical observations to produce reliable results, while BCCB operates effectively from the very first user. Furthermore, BCCB exhibits 3-5x lower performance variance between runs, making it more practical for real campaign planning. Among purely online methods, BCCB consistently outperforms standard Thompson Sampling, budgeted Thompson Sampling, and greedy HTE estimation across all budget levels tested.", "date": "2026-04-28", "categories": []}
{"id": "2604.26919", "title": "Causal Learning with Neural Assemblies", "abstract": "Can Neural Assemblies -- groups of neurons that fire together and strengthen through co-activation -- learn the direction of causal influence between variables? While established as a computationally general substrate for classification, parsing, and planning, neural assemblies have not yet been shown to internalize causal directionality. We demonstrate that the inherent operations of neural assemblies -- projection, local plasticity control, and sparse winner selection -- are sufficient for directional learning. We introduce DIRECT (DIRectional Edge Coupling/Training), a mechanism that co-activates source and target assemblies under an adaptive gain schedule to internalize directed relations. Unlike backpropagation-based methods, DIRECT relies solely on local plasticity, making the resulting causal claims auditable at the mechanism level. Our findings are verified through a dual-readout validation strategy: (i) synaptic-strength asymmetry, measuring the emergent weight gap between forward and reverse links, and (ii) functional propagation overlap, quantifying the reliability of directional signal flow. Across multiple domains, the framework achieves perfect structural recovery under a supervised, known-structure setting. These results establish neural assemblies as an auditable bridge between biologically plausible dynamics and formal causal models, offering an \"explainable by design\" framework where causal claims are traceable to specific neural winners and synaptic asymmetries.", "date": "2026-04-29", "categories": []}
{"id": "2602.01433", "title": "DCD: Decomposition-based Causal Discovery from Autocorrelated and Non-Stationary Temporal Data", "abstract": "Multivariate time series in domains such as finance, climate science, and healthcare often exhibit long-term trends, seasonal patterns, and short-term fluctuations, complicating causal inference under non-stationarity and autocorrelation. Existing causal discovery methods typically operate on raw observations, making them vulnerable to spurious edges and misattributed temporal dependencies. We introduce a decomposition-based causal discovery framework that separates each time series into trend, seasonal, and residual components and performs component-specific causal analysis. Trend components are assessed using stationarity tests, seasonal components using kernel-based dependence measures, and residual components using constraint-based causal discovery. The resulting component-level graphs are integrated into a unified multi-scale causal structure. This approach isolates long- and short-range causal effects, reduces spurious associations, and improves interpretability. Across extensive synthetic benchmarks and real-world climate data, our framework more accurately recovers ground-truth causal structure than state-of-the-art baselines, particularly under strong non-stationarity and temporal autocorrelation.", "date": "2026-02-01", "categories": []}
{"id": "2604.25295", "title": "Optimization-Free Topological Sort for Causal Discovery via the Schur Complement of Score Jacobians", "abstract": "Continuous causal discovery typically couples representation learning with structural optimization via non-convex acyclicity penalties, which subjects solvers to local optima and restricts scalability in high-dimensional regimes. We propose a decoupled paradigm that shifts the causal discovery bottleneck from non-convex optimization to statistical score estimation. We introduce the Score-Schur Topological Sort (SSTS), an algorithm that extracts topological order directly from unconstrained generative models, bypassing constrained structure optimization. We establish that the causal hierarchy leaves a geometric signature within the score function: iterative graph marginalization is mathematically equivalent to computing the Schur complement of the Score-Jacobian Information Matrix (SJIM) under linear conditions. This translates the acyclicity constraint into an algebraic procedure with a dominant cost of O(d^3) operations. For non-linear systems, we formulate the expectation gap of Schur marginalization and introduce Block-SSTS to compress extraction depth, bounding structural error. Empirically, SSTS allows causal structural analysis on non-linear graphs up to d=1000. At this scale, our framework indicates that once the non-convex optimization bottleneck is mathematically bypassed, the structural fidelity of continuous causal discovery is bounded by the finite-sample estimation variance of the global score geometry. By reducing graph extraction to matrix operations, this work reframes scalable causal discovery from a constrained optimization problem to a statistical estimation challenge.", "date": "2026-04-28", "categories": []}
{"id": "2604.25832", "title": "TrialCalibre: A Fully Automated Causal Engine for RCT Benchmarking and Observational Trial Calibration", "abstract": "Real-world evidence (RWE) studies that emulate target trials increasingly inform regulatory and clinical decisions, yet residual, hard-to-quantify biases still limit their credibility. The recently proposed BenchExCal framework addresses this challenge via a two-stage Benchmark, Expand, Calibrate process, which first compares an observational emulation against an existing randomized controlled trial (RCT), then uses observed divergence to calibrate a second emulation for a new indication causal effect estimation. While methodologically powerful, BenchExCal is resource intensive and difficult to scale. We introduce TrialCalibre, a conceptualized multiagent system designed to automate and scale the BenchExCal workflow. Our framework features specialized agents such as the Orchestrator, Protocol Design, Data Synthesis, Clinical Validation, and Quantitative Calibration Agents that coordi-nate the the overall process. TrialCalibre incorpo-rates agent learning (e.g., RLHF) and knowledge blackboards to support adaptive, auditable, and transparent causal effect estimation.", "date": "2026-04-28", "categories": []}
{"id": "2502.04274", "title": "Orthogonal Representation Learning for Estimating Causal Quantities", "abstract": "End-to-end representation learning has become a powerful tool for estimating causal quantities from high-dimensional observational data, but its efficiency remained unclear. Here, we face a central tension: End-to-end representation learning methods often work well in practice but lack asymptotic optimality in the form of the quasi-oracle efficiency. In contrast, two-stage Neyman-orthogonal learners provide such a theoretical optimality property but do not explicitly benefit from the strengths of representation learning. In this work, we step back and ask two research questions: (1) When do representations strengthen existing Neyman-orthogonal learners? and (2) Can a balancing constraint - a commonly proposed technique in the representation learning literature - provide improvements to Neyman-orthogonality? We address these two questions through our theoretical and empirical analysis, where we introduce a unifying framework that connects representation learning with Neyman-orthogonal learners (namely, OR-learners). In particular, we show that, under the low-dimensional manifold hypothesis, the OR-learners can strictly improve the estimation error of the standard Neyman-orthogonal learners. At the same time, we find that the balancing constraint requires an additional inductive bias and cannot generally compensate for the lack of Neyman-orthogonality of the end-to-end approaches. Building on these insights, we offer guidelines for how users can effectively combine representation learning with the classical Neyman-orthogonal learners to achieve both practical performance and theoretical guarantees.", "date": "2025-02-06", "categories": []}
{"id": "2510.13087", "title": "DeepCausalMMM: A Deep Learning Framework for Marketing Mix Modeling with Causal Structure Learning", "abstract": "Marketing Mix Modeling (MMM) estimates the impact of marketing activities on business outcomes such as sales or revenue. Traditional MMM approaches rely on linear regression or Bayesian hierarchical models that assume channel independence and struggle to capture temporal dynamics and non-linear saturation. DeepCausalMMM addresses these limitations by combining deep learning, causal inference, and marketing science. It uses Gated Recurrent Units (GRUs) to learn temporal patterns (adstock, lag) while learning statistical dependencies between channels through Directed Acyclic Graph (DAG) structure with upper triangular constraints. It implements Hill equation saturation curves for diminishing returns and budget optimization. Key features: (1) data-driven hyperparameters learned from data with defaults, (2) linear mean scaling of the dependent variable, (3) configurable attribution priors with dynamic loss scaling, (4) multi-region modeling with shared and region-specific parameters, (5) robust methods including Huber loss, (6) response curve analysis.", "date": "2025-10-15", "categories": []}
{"id": "2510.15479", "title": "Adversary-Free Counterfactual Prediction via Information-Regularized Representations", "abstract": "We study counterfactual prediction under assignment bias and propose a mathematically grounded, information-theoretic approach that removes treatment-covariate dependence without adversarial training. Starting from a bound that links the counterfactual-factual risk gap to mutual information, we learn a stochastic representation Z that is predictive of outcomes while minimizing I(Z; T). We derive a tractable variational objective that upper-bounds the information term and couples it with a supervised decoder, yielding a stable, provably motivated training criterion. The framework extends naturally to dynamic settings by applying the information penalty to sequential representations at each decision time. We evaluate the method on controlled numerical simulations and a real-world clinical dataset, comparing against recent state-of-the-art balancing, reweighting, and adversarial baselines. Across metrics of likelihood, counterfactual error, and policy evaluation, our approach performs favorably while avoiding the training instabilities and tuning burden of adversarial schemes.", "date": "2025-10-17", "categories": []}
{"id": "2604.23107", "title": "MOCA: A Transformer-based Modular Causal Inference Framework with One-way Cross-attention and Cutting Feedback", "abstract": "Causal effect estimation from observational data requires careful adjustment for confounding. Classical estimators such as inverse probability weighting and augmented inverse probability weighting are effective under favorable model specification, but may become unstable when treatment assignment and outcome mechanisms are complex, non-linear, and high-dimensional. Machine learning and representation learning approaches improve flexibility, yet joint training can allow outcome-related information to influence treatment-side representations, which is undesirable from a causal perspective. We propose MOCA (Modular One-way Causal Attention), a transformer-based framework that separates treatment and outcome modeling through a modular design, and performs confounder adjustment using a one-way attention mechanism. A cutting-feedback strategy, implemented via gradient detachment, prevents the outcome loss from updating the treatment module. This design preserves directional information flow while retaining the representational power of transformer architectures for causal inference. Across multiple simulated scenarios, including linear, nonlinear, heavy-tailed, hidden confounding, and high-dimensional settings, MOCA shows competitive or improved performance relative to IPW, AIPW, X-learner, TARNet, and DragonNet. We further illustrate the method on the Infant Health and Development Program dataset and the Dehejia-Wahba dataset as real-world benchmarks. These results suggest that modular attention with one-way information flow provides a promising and interpretable direction for causal inference with modern deep learning models.", "date": "2026-04-25", "categories": []}
{"id": "2604.23465", "title": "Machine learning models for estimating counterfactuals in a single-arm inflammatory bowel disease study", "abstract": "Single-arm trials accelerate study timelines by reducing the number of patients that must be recruited for a concurrent control group. However, these designs require an alternative comparator to estimate treatment effects. One approach is to construct a virtual control arm using a machine learning (ML) model trained on external control data to predict the counterfactual outcomes of the treatment arm. Our aim in this study was to leverage virtual controls by developing and evaluating ML-based counterfactual outcome models trained on IFX-treated patients to predict 1-year steroid-free clinical remission (SFCR ) and a composite of C-reactive protein remission plus steroid-free clinical remission (CRP-SFCR) for ADA-treated pediatric Crohn's disease patients, and to compare the resulting IFX-versus-ADA treatment effect estimates with those obtained using propensity score matching to external controls. Five ML models were used to train counterfactual models on the observed IFX cohort data. The resulting models were used to predict the counterfactual outcomes for the ADA arm patients. LGBM yields the best OR closest to the propensity score matched reference, and all 95% CI results align with the conclusion from the reference study that no statistical difference in the primary and secondary outcomes has been observed between the patients treated with ADA or IFX. Our study supports virtual controls as a viable and effective substitute for expensive, lengthy or unethical patient recruitment in an inflammatory bowel disease (IBD) trial. The developed gradient boosted prediction model can be used as a pretrained model to generate IFX counterfactual predictions in future studies, pending external validation and assessment of transportability.", "date": "2026-04-25", "categories": []}
{"id": "2604.23633", "title": "Causal Discovery as Dialectical Aggregation: A Quantitative Argumentation Framework", "abstract": "Constraint-based causal discovery is brittle in finite-sample regimes because erroneous conditional-independence (CI) decisions can cascade into substantial structural errors. We propose Quantitative Argumentation for Causal Discovery (QACD), a semantics-driven framework that represents CI outcomes as graded, defeasible arguments rather than irreversible constraints. QACD maps statistical test outcomes to argument strengths and aggregates conflicting evidence through connectivity-mediated witness propagation, producing a fixed-point acceptability labeling over candidate adjacencies. Experiments on standard benchmark Bayesian networks suggest that QACD improves structural coherence and interventional reliability in several noisy or inconsistent CI regimes, while remaining competitive with classical constraint-based, hybrid, and prior argumentation-based baselines.", "date": "2026-04-26", "categories": []}
{"id": "2604.23800", "title": "Causal Representation Learning from General Environments under Nonparametric Mixing", "abstract": "Causal representation learning aims to recover the latent causal variables and their causal relations, typically represented by directed acyclic graphs (DAGs), from low-level observations such as image pixels. A prevailing line of research exploits multiple environments, which assume how data distributions change, including single-node interventions, coupled interventions, or hard interventions, or parametric constraints on the mixing function or the latent causal model, such as linearity. Despite the novelty and elegance of the results, they are often violated in real problems. Accordingly, we formalize a set of desiderata for causal representation learning that applies to a broader class of environments, referred to as general environments. Interestingly, we show that one can fully recover the latent DAG and identify the latent variables up to minor indeterminacies under a nonparametric mixing function and nonlinear latent causal models, such as additive (Gaussian) noise models or heteroscedastic noise models, by properly leveraging sufficient change conditions on the causal mechanisms up to third-order derivatives. These represent, to our knowledge, the first results to fully recover the latent DAG from general environments under nonparametric mixing. Notably, our results match or improve upon many existing works, but require less restrictive assumptions about changing environments.", "date": "2026-04-26", "categories": []}
{"id": "2604.24062", "title": "Grounding Before Generalizing: How AI Differs from Humans in Causal Transfer", "abstract": "Extracting abstract causal structures and applying them to novel situations is a hallmark of human intelligence. While Large Language Models (LLMs) and Vision Language Models (VLMs) have shown strong performance on a wide range of reasoning tasks, their capacity for interactive causal learning -- inducing latent structures through sequential exploration and transferring them across contexts -- remains uncharacterized. Human learners accomplish such transfer after minimal exposure, whereas classical Reinforcement Learning (RL) agents fail catastrophically. Whether state-of-the-art Artificial Intelligence (AI) models possess human-like mechanisms for abstract causal structure transfer is an open question. Using the OpenLock paradigm requiring sequential discovery of Common Cause (CC) and Common Effect (CE) structures, here we show that models exhibit fundamentally delayed or absent transfer: even successful models require initial environmental-specific mapping -- what we term environmental grounding -- before efficiency gains emerge, whereas humans leverage prior structural knowledge from the very first solution attempt. In the text-only condition, models matched or exceeded human discovery efficiency. In contrast, visual information -- in both the image-only and text-and-image conditions -- overall degraded rather than enhanced performance, revealing a broad reliance on symbolic processing rather than integrated multimodal reasoning. Models further exhibited systematic CC/CE asymmetries absent in humans, suggesting heuristic biases rather than direction-neutral causal abstraction. These findings reveal that large-scale statistical learning does not produce the decontextualized causal schemas underpinning human analogical reasoning, establishing grounding-dependent transfer as a fundamental limitation of current LLMs and VLMs.", "date": "2026-04-27", "categories": []}
{"id": "2604.24547", "title": "Dialysis Risk Prediction and Treatment Effect Estimation for AKI patients using Longitudinal Electronic Health Records", "abstract": "Progression to dialysis or end-stage renal disease is a rare but clinically important outcome. Clinicians need evidence on how medication exposures influence downstream risk. We constructed a fixed-window EHR cohort (90-day observation, 730-day prediction; N=81401; dialysis/ESRD prevalence: 1.1%) and modeled sequences of diagnoses, procedures, and medications with kidney laboratory trends (creatinine, BUN, eGFR). A transformer-based causal multi-head model was trained to estimate drug- and ingredient-level average treatment effects (ATEs) using counterfactual exposure removal and insertion under a full medication history setup. On test set, predictive performance reached an AUC of 0.694 and PR-AUC of 0.094. At the selected decision threshold (0.883), the model achieved an F1 score of 0.201 with a Brier score of 0.018. Post-hoc causal analyses of lab changes (eGFR, creatinine, BUN) using IPTW, AIPW, naive, and covariate-adjusted OLS methods assessed clinical directionality. Results showed partial protective-direction support for ACE/ARB exposures and worsening-direction signals for loop diuretics.", "date": "2026-04-27", "categories": []}
{"id": "2604.22337", "title": "TabSCM: A practical Framework for Generating Realistic Tabular Data", "abstract": "Most tabular-data generators match marginal statistics yet ignore causal structure, leading downstream models to learn spurious or unfair patterns. We present TabSCM, a mixed-type generator that preserves those causal dependencies. Starting from a Completed Partially Directed Acyclic Graph (CPDAG) found by any causal structure discovery algorithm, TabSCM (i) orients edges to a DAG, (ii) fits root-node marginals with KDE or categorical frequencies, and (iii) learns topologically ordered structural assignments. Such assignments are achieved using conditional diffusion models for continuous variables as child nodes and gradient-boosted trees for categorical ones. Ancestral sampling yields semantically valid records and enables exact counterfactual queries. On seven public datasets, encompassing healthcare, finance, housing, environment, TabSCM matches or surpasses state-of-the-art GAN, diffusion, and LLM baselines in statistical fidelity, downstream utility, and privacy risk, while also cutting rule-violation rates and providing causally meaningful and robust conditional interventions. Because generation is decomposed into explicit equations, it runs up to 583$\\times$ faster than diffusion-only models and exposes interpretable knobs for fairness auditing and policy simulation, making TabSCM a practical choice for realism, explainability, and causal soundness.", "date": "2026-04-24", "categories": []}
{"id": "2604.22416", "title": "From Local to Cluster: A Unified Framework for Causal Discovery with Latent Variables", "abstract": "Latent variables pose a fundamental challenge to causal discovery and inference. Conventional local methods focus on direct neighbors but fail to provide macro level insights. Cluster level methods enable macro causal reasoning but either assume clusters are known a priori or require causal sufficiency. Moreover, directly applying single variable causal discovery methods to cluster level problems violates causal sufficiency and leads to incorrect results. To overcome these limitations, this paper proposes L2C (Local to Cluster Causal Abstraction), a unified framework that bridges local structure learning and cluster level causal discovery. Unlike prior work that requires a complete manual assignment of micro variables to clusters, L2C discovers the partition automatically from local causal patterns. Our solution leverages a cluster reduction theorem to reduce any cluster to at most three nodes without loss of causal information, applies local causal discovery to identify direct causes, effects, and V structures in the presence of latent variables, and performs macro level causal inference via cluster level calculus on the learned cluster graph. L2C does not assume causal sufficiency, as latent variables are handled through local discovery. Theoretical analysis shows that L2C ensures soundness, atomic completeness, and computational efficiency. Extensive experiments on synthetic and real world data demonstrate that L2C accurately recovers ground truth clusters and achieves superior macro causal effect identification compared to existing baselines.", "date": "2026-04-24", "categories": []}
{"id": "2506.05590", "title": "Nonlinear Causal Discovery through a Sequential Edge Orientation Approach", "abstract": "Recent advances have established the identifiability of a directed acyclic graph (DAG) under additive noise models (ANMs), spurring the development of various causal discovery methods. However, most existing methods make restrictive model assumptions, rely heavily on general independence tests, or require substantial computational time. To address these limitations, we propose a sequential procedure to orient undirected edges in a completed partial DAG (CPDAG), representing an equivalence class of DAGs, by leveraging the pairwise additive noise model (PANM) to identify their causal directions. We prove that this procedure can recover the true causal DAG assuming a restricted ANM. Building on this result, we develop a novel constraint-based algorithm for learning causal DAGs under nonlinear ANMs. Given an estimated CPDAG, we develop a ranking procedure that sorts undirected edges by their adherence to the PANM, which defines an evaluation order of the edges. To determine the edge direction, we devise a statistical test that compares the log-likelihood values, evaluated with respect to the competing directions, of a sub-graph comprising just the candidate nodes and their identified parents in the partial DAG. We further establish the structural learning consistency of our algorithm in the large-sample limit. Extensive experiments on synthetic and real-world datasets demonstrate that our method is computationally efficient, robust to model misspecification, and consistently outperforms many existing nonlinear DAG learning methods.", "date": "2025-06-05", "categories": []}
{"id": "2603.27406", "title": "On the Relationship between Bayesian Networks and Probabilistic Structural Causal Models", "abstract": "In this paper, the relationship between probabilistic graphical models, in particular Bayesian networks, and causal diagrams, also called structural causal models, is studied. Structural causal models are deterministic models, based on structural equations or functions, that can be provided with uncertainty by adding independent, unobserved random variables to the models, equipped with probability distributions. One question that arises is whether a Bayesian network that has obtained from expert knowledge or learnt from data can be mapped to a probabilistic structural causal model, and whether or not this has consequences for the network structure and probability distribution. We show that linear algebra and linear programming offer key methods for the transformation, and examine properties for the existence and uniqueness of solutions based on dimensions of the probabilistic structural model. Finally, we examine in what way the semantics of the models is affected by this transformation.\n  Keywords: Causality, probabilistic structural causal models, Bayesian networks, linear algebra, experimental software.", "date": "2026-03-28", "categories": []}
{"id": "2407.01621", "title": "Deciphering interventional dynamical causality from non-intervention complex systems", "abstract": "Detecting and quantifying causality is a focal topic in the fields of science, engineering, and interdisciplinary studies. However, causal studies on non-intervention systems attract much attention but remain extremely challenging. Delay-embedding technique provides a promising approach. In this study, we propose a framework named Interventional Dynamical Causality (IntDC) in contrast to the traditional Constructive Dynamical Causality (ConDC). ConDC, including Granger causality, transfer entropy and convergence of cross-mapping, measures the causality by constructing a dynamical model without considering interventions. A computational criterion, Interventional Embedding Entropy (IEE), is proposed to measure causal strengths in an interventional manner. IEE is an intervened causal information flow but in the delay-embedding space. Further, the IEE theoretically and numerically enables the deciphering of IntDC solely from observational (non-interventional) time-series data, without requiring any knowledge of dynamical models or real interventions in the considered system. In particular, IEE can be applied to rank causal effects according to their importance and construct causal networks from data. We conducted numerical experiments to demonstrate that IEE can find causal edges accurately, eliminate effects of confounding, and quantify causal strength robustly over traditional indices. We also applied IEE to real-world tasks. IEE performed as an accurate and robust tool for causal analyses solely from the observational data. The IntDC framework and IEE algorithm provide an efficient approach to the study of causality from time series in diverse non-intervention complex systems.", "date": "2024-06-29", "categories": []}
{"id": "2603.01168", "title": "SphUnc: Hyperspherical Uncertainty Decomposition and Causal Identification via Information Geometry", "abstract": "Reliable decision-making in complex multi-agent systems requires calibrated predictions and interpretable uncertainty. We introduce SphUnc, a unified framework combining hyperspherical representation learning with structural causal modeling. The model maps features to unit hypersphere latents using von Mises-Fisher distributions, decomposing uncertainty into epistemic and aleatoric components through information-geometric fusion. A structural causal model on spherical latents enables directed influence identification and interventional reasoning via sample-based simulation. Empirical evaluations on social and affective benchmarks demonstrate improved accuracy, better calibration, and interpretable causal signals, establishing a geometric-causal foundation for uncertainty-aware reasoning in multi-agent settings with higher-order interactions.", "date": "2026-03-01", "categories": []}
{"id": "2604.20039", "title": "Separable Pathways for Causal Reasoning: How Architectural Scaffolding Enables Hypothesis-Space Restructuring in LLM Agents", "abstract": "Causal discovery through experimentation and intervention is fundamental to robust problem solving. It requires not just updating beliefs within a fixed framework but revising the hypothesis space itself, a capacity current AI agents lack when evidence demands representations they have not previously constructed. We extend the blicket detector paradigm from developmental science to test this capacity in AI agents equipped with architectural scaffolding that targets hypothesis-space restructuring. Our compositional architecture has two discrete components: context graphs, which structure exploration as typed state machines, and dynamic behaviors, which monitor for evidence that the current hypothesis space is inadequate and expand it at runtime. Across 1,085 experimental trials, these components make orthogonal contributions: context graphs drive reasoning quality within the post-switch hypothesis space, accounting for 94\\% of the accuracy gain, while dynamic behaviors drive reasoning eligibility by detecting regime changes and preventing premature commitment to outdated hypotheses.", "date": "2026-04-21", "categories": []}
{"id": "2604.20259", "title": "Causal-Transformer with Adaptive Mutation-Locking for Early Prediction of Acute Kidney Injury", "abstract": "Accurate early prediction of Acute Kidney Injury (AKI) is critical for timely clinical intervention. However, existing deep learning models struggle with irregularly sampled data and suffer from the opaque \"black-box\" nature of sequential architectures, strictly limiting clinical trust. To address these challenges, we propose CT-Former, integrating continuous-time modeling with a Causal-Transformer. To handle data irregularity without biased artificial imputation, our framework utilizes a continuous-time state evolution mechanism to naturally track patient temporal trajectories. To resolve the black-box problem, our Causal-Attention module abandons uninterpretable hidden state aggregation. Instead, it generates a directed structural causal matrix to identify and trace the exact historical onset of severe physiological shocks. By establishing clear causal pathways between historical anomalies and current risk predictions, CT-Former provides native clinical interpretability. Training follows a decoupled two-stage protocol to optimize the causal-fusion process independently. Extensive experiments on the MIMIC-IV cohort (N=18,419) demonstrate that CT-Former significantly outperforms state-of-the-art baselines. The results confirm that our explicitly transparent architecture offers an accurate and trustworthy tool for clinical decision-making.", "date": "2026-04-22", "categories": []}
{"id": "2604.20306", "title": "Dual Causal Inference: Integrating Backdoor Adjustment and Instrumental Variable Learning for Medical VQA", "abstract": "Medical Visual Question Answering (MedVQA) aims to generate clinically reliable answers conditioned on complex medical images and questions. However, existing methods often overfit to superficial cross-modal correlations, neglecting the intrinsic biases embedded in multimodal medical data. Consequently, models become vulnerable to cross-modal confounding effects, severely hindering their ability to provide trustworthy diagnostic reasoning. To address this limitation, we propose a novel Dual Causal Inference (DCI) framework for MedVQA. To the best of our knowledge, DCI is the first unified architecture that integrates Backdoor Adjustment (BDA) and Instrumental Variable (IV) learning to jointly tackle both observable and unobserved confounders. Specifically, we formulate a Structural Causal Model (SCM) where observable cross-modal biases (e.g., frequent visual and textual co-occurrences) are mitigated via BDA, while unobserved confounders are compensated using an IV learned from a shared latent space. To guarantee the validity of the IV, we design mutual information constraints that maximize its dependence on the fused multimodal representations while minimizing its associations with the unobserved confounders and target answers. Through this dual mechanism, DCI extracts deconfounded representations that capture genuine causal relationships. Extensive experiments on four benchmark datasets, SLAKE, SLAKE-CP, VQA-RAD, and PathVQA, demonstrate that our method consistently outperforms existing approaches, particularly in out-of-distribution (OOD) generalization. Furthermore, qualitative analyses confirm that DCI significantly enhances the interpretability and robustness of cross-modal reasoning by explicitly disentangling true causal effects from spurious cross-modal shortcuts.", "date": "2026-04-22", "categories": []}
{"id": "2604.20516", "title": "Efficient Symbolic Computations for Identifying Causal Effects", "abstract": "Determining identifiability of causal effects from observational data under latent confounding is a central challenge in causal inference. For linear structural causal models, identifiability of causal effects is decidable through symbolic computation. However, standard approaches based on Gr\\\"obner bases become computationally infeasible beyond small settings due to their doubly exponential complexity. In this work, we study how to practically use symbolic computation for deciding rational identifiability. In particular, we present an efficient algorithm that provably finds the lowest degree identifying formulas. For a causal effect of interest, if there exists an identification formula of a prespecified maximal degree, our algorithm returns such a formula in quasi-polynomial time.", "date": "2026-04-22", "categories": []}
{"id": "2604.20777", "title": "Efficient Multi-Cohort Inference for Long-Term Effects and Lifetime Value in A/B Testing with User Learning", "abstract": "In streaming platforms churn is extremely costly, yet A/B tests are typically evaluated using outcomes observed within a limited experimental horizon. Even when both short- and predicted long-term engagement metrics are considered, they may fail to capture how a treatment affects users' retention. Consequently, an intervention may appear beneficial in the short term and neutral in the long term while still generating lower total value than the control due to users churn.\n  To address this limitation, we introduce a method that estimates long-term treatment effects (LTE) and residual lifetime value change ($\\Delta ERLV$) in short multi-cohort A/B tests under user learning. To estimate time-varying treatment effects efficiently, we introduce an inverse-variance weighted estimator that combines multiple cohorts estimates, reducing variance relative to standard approaches in the literature. The estimated treatment trajectory is then modeled as a parametric decay to recover both the asymptotic treatment effect and the cumulative value generated over time.\n  Our framework enables simultaneous evaluation of steady-state impact and residual user value within a single experiment. Empirical results show improved precision in estimating LTE and $\\Delta ERLV$ and identify scenarios in which relying on either short-term or long-term metrics alone would lead to incorrect product decisions.", "date": "2026-04-22", "categories": []}
{"id": "2502.10605", "title": "Batch-Adaptive Causal Annotations", "abstract": "Estimating the causal effects of interventions is crucial to policy and decision-making, yet outcome data are often missing or subject to non-standard measurement error. While ground-truth outcomes can sometimes be obtained through costly data annotation or follow-up, budget constraints typically allow only a fraction of the dataset to be labeled. We address this challenge by optimizing which data points should be sampled for outcome information in order to improve efficiency in average treatment effect estimation with missing outcomes. We derive a closed-form solution for the optimal batch sampling probability by minimizing the asymptotic variance of a doubly robust estimator for causal inference with missing outcomes. Motivated by our street outreach partners, we extend the framework to costly annotations of unstructured data, such as text or images in healthcare and social services. Across simulated and real-world datasets, including one of outreach interventions in homelessness services, our approach achieves substantially lower mean-squared error and recovers the AIPW estimate with fewer labels than existing baselines. In practice, we show that our method can match confidence intervals obtained with 361 random samples using only 90 optimized samples - saving 75% of the labeling budget.", "date": "2025-02-14", "categories": []}
{"id": "2502.16156", "title": "A Review of Causal Decision Making", "abstract": "To make effective decisions, it is important to have a thorough understanding of the causal relationships among actions, environments, and outcomes. This review aims to surface three crucial aspects of decision-making through a causal lens: 1) the discovery of causal relationships through causal structure learning, 2) understanding the impacts of these relationships through causal effect learning, and 3) applying the knowledge gained from the first two aspects to support decision making via causal policy learning. Moreover, we identify challenges that hinder the broader utilization of causal decision-making and discuss recent advances in overcoming these challenges. Finally, we provide future research directions to address these challenges and to further enhance the implementation of causal decision-making in practice, with real-world applications illustrated based on the proposed causal decision-making. We aim to offer a comprehensive methodology and practical implementation framework by consolidating various methods in this area into a Python-based collection. URL: https://causaldm.github.io/Causal-Decision-Making.", "date": "2025-02-22", "categories": []}
{"id": "2601.17647", "title": "Knowledge-Guided Time-Varying Causal Inference for Arctic Sea Ice Dynamics", "abstract": "Quantifying the causal relationship between sea ice thickness and sea surface height (SSH) is essential for understanding the mechanisms driving polar climate change and global sea-level rise. Conventional deep learning models often struggle with treatment effect estimation in climate settings due to time-varying confounding and the lack of physical constraints. To address these challenges, we propose the Knowledge-Guided Causal Model Variational Autoencoder (KGCM-VAE) to quantify the effect of SSH on sea ice thickness. The framework leverages established physical relationships between SSH and surface velocity to generate physically grounded, time-varying continuous treatments, where each treatment value can change at every time step within a sequence. The model also incorporates Maximum Mean Discrepancy (MMD) to balance treated and control distributions in the latent space, mitigating observed confounding bias. Using synthetic data, we evaluated the model's ability to predict sea ice thickness responses under hypothetical SSH forcing scenarios, demonstrating that KGCM-VAE achieves superior PEHE compared to state-of-the-art baselines. Ablation studies further confirm that MMD consistently enhances treatment effect estimation over the base model. Additionally, we conducted a real-world case study to examine the sensitivity of physical parameters to specific treatments and to compare these findings with an existing modeling study.", "date": "2026-01-25", "categories": []}
{"id": "2604.16763", "title": "LLM-Extracted Covariates for Clinical Causal Inference: Rethinking Integration Strategies", "abstract": "Causal inference from electronic health records (EHR) is fundamentally limited by unmeasured confounding: critical clinical states such as frailty, goals of care, and mental status are documented in free-text notes but absent from structured data. Large language models can extract these latent confounders as interpretable, structured covariates, yet how to effectively integrate them into causal estimation pipelines has not been systematically studied. Using the MIMIC-IV database with 21,859 sepsis patients, we compare seven covariate-integration strategies for estimating the effect of early vasopressor initiation on 28-day mortality, spanning tabular-only baselines, traditional NLP representations, and three LLM-augmented approaches. A central finding is that not all integration strategies are equally effective: directly augmenting the propensity score model with LLM covariates achieves the best performance, while dual-caliper matching on text-derived categorical distances restricts the donor pool and degrades estimation. In semi-synthetic experiments with known ground-truth effects, LLM-augmented propensity scores reduce estimation bias from 0.0143 to 0.0003 relative to tabular-only methods, and this advantage persists under substantial simulated extraction error. On real data, incorporating LLM-extracted covariates reduces the estimated treatment effect from 0.055 to 0.027, directionally consistent with the CLOVERS randomized trial, and a doubly robust estimator yielding 0.019 confirms the robustness of this finding. Our results offer practical guidance on when and how text-derived covariates improve causal estimation in critical care.", "date": "2026-04-18", "categories": []}
{"id": "2604.19066", "title": "Age-Dependent Heterogeneity in the Association Between Physical Activity and Mental Distress: A Causal Machine Learning Analysis of 3.2 Million U.S. Adults", "abstract": "Physical activity (PA) is widely recognized as protective against mental distress, yet whether this benefit varies systematically across population subgroups remains poorly understood. Using pooled data from ten consecutive annual waves of the U.S. Behavioral Risk Factor Surveillance System (2015-2024; n = 3,242,218), we investigate heterogeneity in the association between leisure-time PA and frequent mental distress (FMD, >=14 days/month) across age groups. Survey-weighted logistic regression reveals a striking age gradient: the adjusted odds ratio for PA ranges from 0.89 among young adults (18-24) to 0.50 among adults aged 55-64, with the protective association strengthening monotonically with age. Temporal analysis across all ten years shows that the young-adult PA effect has been eroding over the past decade, with the 18-24 OR reaching 1.01 (null) in both 2018 and 2024 -- paralleling the deepening youth mental health crisis. Causal Forest via Double Machine Learning independently identifies age as the dominant driver of treatment effect heterogeneity (feature importance = 0.39, 2.5x the next predictor). E-value sensitivity analysis, propensity score overlap checks, placebo tests, and imputation comparisons confirm the robustness of the findings. These results suggest that the well-documented exercise--mental health link may not generalize to the youngest adult population, whose distress appears increasingly driven by stressors that PA alone cannot mitigate.", "date": "2026-04-21", "categories": []}
{"id": "2507.20993", "title": "Annotation-Assisted Learning of Treatment Policies From Multimodal Electronic Health Records", "abstract": "We study how to learn treatment policies from multimodal electronic health records (EHRs) that consist of tabular data and clinical text. These policies can help physicians make better treatment decisions and allocate healthcare resources more efficiently. Causal policy learning methods prioritize patients with the largest expected treatment benefit. Yet, existing estimators are designed for tabular covariates under causal assumptions that may be hard to justify in the multimodal setting. A pragmatic alternative is to apply causal estimators directly to multimodal representations, but this can produce biased treatment effect estimates when the representations do not preserve the relevant confounding information. As a result, predictive models of baseline risk are commonly used in practice to guide treatment decisions, although they are not designed to identify which patients benefit most from treatment. We propose AACE (Annotation-Assisted Coarsened Effects), an annotation-assisted approach to causal policy learning for multimodal EHRs. The method uses expert-provided annotations during training to support confounding adjustment, and then predicts treatment benefit from only multimodal representations at inference. We show that the proposed method achieves strong empirical performance across synthetic, semi-synthetic, and real-world EHR datasets, outperforming risk-based and representation-based causal baselines, and offering practical insights for applying causal machine learning in clinical practice.", "date": "2025-07-28", "categories": []}
{"id": "2509.22297", "title": "Large Language Models as Nondeterministic Causal Models", "abstract": "Recent work by Chatzi et al. and Ravfogel et al. has developed, for the first time, a method for generating counterfactuals of probabilistic Large Language Models. Such counterfactuals tell us what would - or might - have been the output of an LLM if some factual prompt ${\\bf x}$ had been ${\\bf x}^*$ instead. The ability to generate such counterfactuals is an important necessary step towards explaining, evaluating, and eventually improving, the behavior of LLMs. I argue, however, that the existing method rests on an ambiguous interpretation of LLMs: it does not interpret LLMs literally, for the method involves the assumption that one can change the implementation of an LLM's sampling process without changing the LLM itself, nor does it interpret LLMs as intended, for the method involves explicitly representing a nondeterministic LLM as a deterministic causal model. I here present a much simpler method for generating counterfactuals that is based on an LLM's intended interpretation by representing it as a nondeterministic causal model instead. The advantage of my simpler method is that it is directly applicable to any black-box LLM without modification, as it is agnostic to any implementation details. The advantage of the existing method, on the other hand, is that it directly implements the generation of a specific type of counterfactuals that is useful for certain purposes, but not for others. I clarify how both methods relate by offering a theoretical foundation for reasoning about counterfactuals in LLMs based on their intended semantics, thereby laying the groundwork for novel application-specific methods for generating counterfactuals.", "date": "2025-09-26", "categories": []}
{"id": "2511.22003", "title": "A Sensitivity Approach to Causal Inference Under Limited Overlap", "abstract": "Limited overlap between treated and control groups is a key challenge in observational analysis. Standard approaches like trimming importance weights can reduce variance but introduce a fundamental bias. We propose a sensitivity framework for contextualizing findings under limited overlap, where we assess how irregular the outcome function has to be in order for the main finding to be invalidated. Our approach is based on worst-case confidence bounds on the bias introduced by standard trimming practices, under explicit assumptions necessary to extrapolate counterfactual estimates from regions of overlap to those without. Empirically, we demonstrate how our sensitivity framework protects against spurious findings by quantifying uncertainty in regions with limited overlap.", "date": "2025-11-27", "categories": []}
{"id": "2604.14691", "title": "CAMO: An Agentic Framework for Automated Causal Discovery from Micro Behaviors to Macro Emergence in LLM Agent Simulations", "abstract": "LLM-empowered agent simulations are increasingly used to study social emergence, yet the micro-to-macro causal mechanisms behind macro outcomes often remain unclear. This is challenging because emergence arises from intertwined agent interactions and meso-level feedback and nonlinearity, making generative mechanisms hard to disentangle. To this end, we introduce \\textbf{\\textsc{CAMO}}, an automated \\textbf{Ca}usal discovery framework from \\textbf{M}icr\\textbf{o} behaviors to \\textbf{M}acr\\textbf{o} Emergence in LLM agent simulations. \\textsc{CAMO} converts mechanistic hypotheses into computable factors grounded in simulation records and learns a compact causal representation centered on an emergent target $Y$. \\textsc{CAMO} outputs a computable Markov boundary and a minimal upstream explanatory subgraph, yielding interpretable causal chains and actionable intervention levers. It also uses simulator-internal counterfactual probing to orient ambiguous edges and revise hypotheses when evidence contradicts the current view. Experiments across four emergent settings demonstrate the promise of \\textsc{CAMO}.", "date": "2026-04-16", "categories": []}
{"id": "2604.17568", "title": "Diverse Dictionary Learning", "abstract": "Given only observational data $X = g(Z)$, where both the latent variables $Z$ and the generating process $g$ are unknown, recovering $Z$ is ill-posed without additional assumptions. Existing methods often assume linearity or rely on auxiliary supervision and functional constraints. However, such assumptions are rarely verifiable in practice, and most theoretical guarantees break down under even mild violations, leaving uncertainty about how to reliably understand the hidden world. To make identifiability actionable in the real-world scenarios, we take a complementary view: in the general settings where full identifiability is unattainable, what can still be recovered with guarantees, and what biases could be universally adopted? We introduce the problem of diverse dictionary learning to formalize this view. Specifically, we show that intersections, complements, and symmetric differences of latent variables linked to arbitrary observations, along with the latent-to-observed dependency structure, are still identifiable up to appropriate indeterminacies even without strong assumptions. These set-theoretic results can be composed using set algebra to construct structured and essential views of the hidden world, such as genus-differentia definitions. When sufficient structural diversity is present, they further imply full identifiability of all latent variables. Notably, all identifiability benefits follow from a simple inductive bias during estimation that can be readily integrated into most models. We validate the theory and demonstrate the benefits of the bias on both synthetic and real-world data.", "date": "2026-04-19", "categories": []}
{"id": "2604.17910", "title": "Physics-Informed Causal MDPs for Sequential Constraint Repair in Engineering Simulation Pipelines", "abstract": "Off-policy learning in constrained MDPs with large binary state spaces faces a fundamental tension: causal identification of transition dynamics requires structural assumptions, while sample-efficient policy learning requires state-space compression. We introduce PI-CMDP, a framework for CMDPs whose constraint dependencies form a layered DAG under a Lifecycle Ordering Assumption (LOA). We propose an Identify-Compress-Estimate pipeline: (i) Identify: LOA enables backdoor identification of causal edge weights for cross-layer pairs, with formal partial-identification bounds when LOA is violated; (ii) Compress: a Markov abstraction compresses state cardinality from 2^(WL) to (W+1)^L under layer-priority regularity and exchangeability; and (iii) Estimate: a physics-guided doubly-robust estimator remains unbiased and reduces the variance constant when the physics prior outperforms a learned model. We instantiate PI-CMDP on constraint repair in engineering simulation pipelines. On the TPS benchmark (4,206 episodes), PI-CMDP achieves 76.2% repair success rate with only 300 training episodes versus 70.8% for the strongest baseline (+5.4 pp), narrowing to +2.8 pp (83.4% vs. 80.6%) in the full-data regime, while substantially reducing cascade failure rates. All improvements are consistent across 5 independent seeds (paired t-test p < 0.02).", "date": "2026-04-20", "categories": []}
{"id": "2604.18460", "title": "Learning Invariant Modality Representation for Robust Multimodal Learning from a Causal Inference Perspective", "abstract": "Multimodal affective computing aims to predict humans' sentiment, emotion, intention, and opinion using language, acoustic, and visual modalities. However, current models often learn spurious correlations that harm generalization under distribution shifts or noisy modalities. To address this, we propose a causal modality-invariant representation (CmIR) learning framework for robust multimodal learning. At its core, we introduce a theoretically grounded disentanglement method that separates each modality into `causal invariant representation' and `environment-specific spurious representation' from a causal inference perspective. CmIR ensures that the learned invariant representations retain stable predictive relationships with labels across different environments while preserving sufficient information from the raw inputs via invariance constraint, mutual information constraint, and reconstruction constraint. Experiments across multiple multimodal benchmarks demonstrate that CmIR achieves state-of-the-art performance. CmIR particularly excels on out-of-distribution data and noisy data, confirming its robustness and generalizability.", "date": "2026-04-20", "categories": []}
{"id": "2604.18469", "title": "A Generalized Synthetic Control Method for Baseline Estimation in Demand Response Services", "abstract": "Baseline estimation is critical to Demand Response (DR) settlement in electricity markets, yet existing machine learning methods remain limited in predictive performance, while methodologies from causal inference and counterfactual prediction are still underutilized in this domain. We introduce a Generalized Synthetic Control Method that builds on the classical Synthetic Control Method (SCM) from econometrics. While SCM provides a powerful framework for counterfactual estimation, classical SCM remains a static estimator: it fits the treated unit as a combination of contemporaneous donor units and therefore ignores predictable temporal structure in the residual error. We develop a generalized SCM framework that transforms baseline estimation into a dynamic counterfactual prediction problem by augmenting the donor representation with exogenous features, lagged treated load, and selected lagged donor signals. This enriched representation allows the estimator to capture autoregressive dependence, delayed donor-response patterns, and error-correction effects beyond the scope of standard SCM. The framework further accommodates nonlinear predictors when linear weighting is inadequate, with the greatest benefit arising in limited-data settings. Experiments on the Ausgrid smart-meter dataset show consistent improvements over classical SCM and strong benchmark methods, with the dominant performance gains driven by dynamic augmentation.", "date": "2026-04-20", "categories": []}
{"id": "2409.01794", "title": "Estimating Joint Interventional Distributions from Marginal Interventional Data", "abstract": "In this paper we show how to exploit interventional data to acquire the joint conditional distribution of all the variables using the Maximum Entropy principle. To this end, we extend the Causal Maximum Entropy method to make use of interventional data in addition to observational data. Using Lagrange duality, we prove that the solution to the Causal Maximum Entropy problem with interventional constraints lies in the exponential family, as in the Maximum Entropy solution. Our method allows us to perform two tasks of interest when marginal interventional distributions are provided for any subset of the variables. First, we show how to perform causal feature selection from a mixture of observational and single-variable interventional data, and, second, how to infer joint interventional distributions. For the former task, we show on synthetically generated data, that our proposed method outperforms the state-of-the-art method on merging datasets, and yields comparable results to the KCI-test which requires access to joint observations of all variables.", "date": "2024-09-03", "categories": []}
{"id": "2603.24448", "title": "Integrating Causal Machine Learning into Clinical Decision Support Systems: Insights from Literature and Practice", "abstract": "Current clinical decision support systems (CDSSs) typically base their predictions on correlation, not causation. In recent years, causal machine learning (ML) has emerged as a promising way to improve decision-making with CDSSs by offering interpretable, treatment-specific reasoning. However, existing research often emphasizes model development rather than designing clinician-facing interfaces. To address this gap, we investigated how CDSSs based on causal ML should be designed to effectively support collaborative clinical decision-making. Using a design science research methodology, we conducted a structured literature review and interviewed experienced physicians. From these, we derived eight empirically grounded design requirements, developed seven design principles, and proposed nine practical design features. Our results establish guidance for designing CDSSs that deliver causal insights, integrate seamlessly into clinical workflows, and support trust, usability, and human-AI collaboration. We also reveal tensions around automation, responsibility, and regulation, highlighting the need for an adaptive certification process for ML-based medical products.", "date": "2026-03-25", "categories": []}
{"id": "2604.03920", "title": "From Plausible to Causal: Counterfactual Semantics for Policy Evaluation in Simulated Online Communities", "abstract": "LLM-based social simulations can generate believable community interactions, enabling ``policy wind tunnels'' where governance interventions are tested before deployment. But believability is not causality. Claims like ``intervention $A$ reduces escalation'' require causal semantics that current simulation work typically does not specify. We propose adopting the causal counterfactual framework, distinguishing \\textit{necessary causation} (would the outcome have occurred without the intervention?) from \\textit{sufficient causation} (does the intervention reliably produce the outcome?). This distinction maps onto different stakeholder needs: moderators diagnosing incidents require evidence about necessity, while platform designers choosing policies require evidence about sufficiency. We formalize this mapping, show how simulation design can support estimation under explicit assumptions, and argue that the resulting quantities should be interpreted as simulator-conditional causal estimates whose policy relevance depends on simulator fidelity. Establishing this framework now is essential: it helps define what adequate fidelity means and moves the field from simulations that look realistic toward simulations that can support policy changes.", "date": "2026-04-05", "categories": []}
{"id": "2505.16051", "title": "Flow-based Generative Modeling of Potential Outcomes and Counterfactuals", "abstract": "Predicting potential and counterfactual outcomes from observational data is central to individualized decision-making, particularly in clinical settings where treatment choices must be tailored to each patient rather than guided solely by population averages. We propose PO-Flow, a continuous normalizing flow (CNF) framework for causal inference that jointly models potential outcome distributions and factual-conditioned counterfactual outcomes. Trained via flow matching, PO-Flow provides a unified approach to individualized potential outcome prediction, conditional average treatment effect estimation, and counterfactual prediction. By encoding an observed factual outcome and decoding under an alternative treatment, PO-Flow provides an encode-decode mechanism for factual-conditioned counterfactual prediction. In addition, PO-Flow supports likelihood-based evaluation of potential outcomes, enabling uncertainty-aware assessment of predictions. A supporting recovery guarantee is established under certain assumptions, and empirical results on benchmark datasets demonstrate strong performance across a range of causal inference tasks within the potential outcomes framework.", "date": "2025-05-21", "categories": []}
{"id": "2512.08230", "title": "Empowerment Gain and Causal Model Construction: Children and adults are sensitive to controllability and variability in their causal interventions", "abstract": "Learning about the causal structure of the world is a fundamental problem for human cognition. Causal models and especially causal learning have proved to be difficult for large pretrained models using standard techniques of deep learning. In contrast, cognitive scientists have applied advances in our formal understanding of causation in computer science, particularly within the Causal Bayes Net formalism, to understand human causal learning. In the very different tradition of reinforcement learning, researchers have described an intrinsic reward signal called \"empowerment\" which maximizes mutual information between actions and their outcomes. \"Empowerment\" may be an important bridge between classical Bayesian causal learning and reinforcement learning and may help to characterize causal learning in humans and enable it in machines. If an agent learns an accurate causal world model, they will necessarily increase their empowerment, and increasing empowerment will lead to a more accurate causal world model. Empowerment may also explain distinctive features of childrens causal learning, as well as providing a more tractable computational account of how that learning is possible. In an empirical study, we systematically test how children and adults use cues to empowerment to infer causal relations, and design effective causal interventions.", "date": "2025-12-09", "categories": []}
{"id": "2604.13218", "title": "Identifiability of Potentially Degenerate Gaussian Mixture Models With Piecewise Affine Mixing", "abstract": "Causal representation learning (CRL) aims to identify the underlying latent variables from high-dimensional observations, even when variables are dependent with each other. We study this problem for latent variables that follow a potentially degenerate Gaussian mixture distribution and that are only observed through the transformation via a piecewise affine mixing function. We provide a series of progressively stronger identifiability results for this challenging setting in which the probability density functions are ill-defined because of the potential degeneracy. For identifiability up to permutation and scaling, we leverage a sparsity regularization on the learned representation. Based on our theoretical results, we propose a two-stage method to estimate the latent variables by enforcing sparsity and Gaussianity in the learned representations. Experiments on synthetic and image data highlight our method's effectiveness in recovering the ground-truth latent variables.", "date": "2026-04-14", "categories": []}
{"id": "2604.12137", "title": "Observing the unobserved confounding through its effects: toward randomized trial-like estimates from real-world survival data", "abstract": "Background: Randomized controlled trials (RCTs) are costly, time-consuming, and often infeasible, while treatment-effect estimation from observational data is limited by unobserved confounding.\n  Methods: We developed a three-step framework to address unobserved confounding in observational survival data. First, we infer a latent prognostic factor (U) from restricted mean survival time (RMST) discrepancies between patients with similar observed factors, the same treatment, and divergent outcomes, leveraging the idea that the aggregate effect of unmeasured factors can be inferred even if individual factors cannot. Second, we balance U with observed baseline covariates using prognostic matching, entropy balancing, or inverse probability of treatment weighting. Third, we apply multivariable survival analysis to estimate hazard ratios (HRs). We evaluated the framework in three observational cohorts with RCT benchmarks, two RCT cohorts, and six multicenter observational cohorts.\n  Results: In three observational cohorts (nine comparisons), balancing U improved agreement with trial HRs in all cases; in the strongest settings, it reduced absolute log-HR error by approximately ten-fold versus using observed covariates alone (mean reduction 0.344; p=0.001). In two RCT cohorts, U was balanced across arms (most SMDs <0.1) and adjustment had minimal impact on log-HRs (mean absolute change 0.08). Across six multicenter cohorts, balancing U within centers reduced cross-center dispersion in chemotherapy log-HR estimates (mean reduction 0.147; p=0.016); when populations were directly balanced across centers to account for case-mix differences, cross-center survival differences were narrowed in 75%-100% of comparisons.\n  Conclusions: Inferring and balancing a latent prognostic signal may reduce unobserved confounding and improve treatment-effect estimation from real-world data.", "date": "2026-04-13", "categories": []}
{"id": "2604.12992", "title": "Causal Diffusion Models for Counterfactual Outcome Distributions in Longitudinal Data", "abstract": "Predicting counterfactual outcomes in longitudinal data, where sequential treatment decisions heavily depend on evolving patient states, is critical yet notoriously challenging due to complex time-dependent confounding and inadequate uncertainty quantification in existing methods. We introduce the Causal Diffusion Model (CDM), the first denoising diffusion probabilistic approach explicitly designed to generate full probabilistic distributions of counterfactual outcomes under sequential interventions. CDM employs a novel residual denoising architecture with relational self-attention, capturing intricate temporal dependencies and multimodal outcome trajectories without requiring explicit adjustments (e.g., inverse-probability weighting or adversarial balancing) for confounding. In rigorous evaluation on a pharmacokinetic-pharmacodynamic tumor-growth simulator widely adopted in prior work, CDM consistently outperforms state-of-the-art longitudinal causal inference methods, achieving a 15-30% relative improvement in distributional accuracy (1-Wasserstein distance) while maintaining competitive or superior point-estimate accuracy (RMSE) under high-confounding regimes. By unifying uncertainty quantification and robust counterfactual prediction in complex, sequentially confounded settings, without tailored deconfounding, CDM offers a flexible, high-impact tool for decision support in medicine, policy evaluation, and other longitudinal domains.", "date": "2026-04-14", "categories": []}
{"id": "2404.18905", "title": "Detecting critical treatment effect bias in small subgroups", "abstract": "Randomized trials are considered the gold standard for making informed decisions in medicine, yet they often lack generalizability to the patient populations in clinical practice. Observational studies, on the other hand, cover a broader patient population but are prone to various biases. Thus, before using an observational study for decision-making, it is crucial to benchmark its treatment effect estimates against those derived from a randomized trial. We propose a novel strategy to benchmark observational studies beyond the average treatment effect. First, we design a statistical test for the null hypothesis that the treatment effects estimated from the two studies, conditioned on a set of relevant features, differ up to some tolerance. We then estimate an asymptotically valid lower bound on the maximum bias strength for any subgroup in the observational study. Finally, we validate our benchmarking strategy in a real-world setting and show that it leads to conclusions that align with established medical knowledge.", "date": "2024-04-29", "categories": []}
{"id": "2412.07469", "title": "Score-matching-based Structure Learning for Temporal Data on Networks", "abstract": "Causal discovery is a crucial initial step in establishing causality from empirical data and background knowledge. Numerous algorithms have been developed for this purpose. Among them, the score-matching method has demonstrated superior performance across various evaluation metrics, particularly for the commonly encountered Additive Nonlinear Causal Models. However, current score-matching-based algorithms are primarily designed to analyze independent and identically distributed (i.i.d.) data. More importantly, they suffer from high computational complexity due to the pruning step required for handling dense Directed Acyclic Graphs (DAGs). To enhance the scalability of score matching, we have developed a new parent-finding subroutine for leaf nodes in DAGs, significantly accelerating the most time-consuming part of the process: the pruning step. This improvement results in an efficiency-lifted score matching algorithm, termed Parent Identification-based Causal structure learning for both i.i.d. and temporal data on networKs, or PICK. The new score-matching algorithm extends the scope of existing algorithms and can handle static and temporal data on networks with weak network interference. Our proposed algorithm can efficiently cope with increasingly complex datasets that exhibit spatial and temporal dependencies, commonly encountered in academia and industry. The proposed algorithm can accelerate score-matching-based methods while maintaining high accuracy in real-world applications.", "date": "2024-12-10", "categories": []}
{"id": "2502.11008", "title": "CounterBench: Evaluating and Improving Counterfactual Reasoning in Large Language Models", "abstract": "Counterfactual reasoning is widely recognized as one of the most challenging and intricate aspects of causality in artificial intelligence. In this paper, we evaluate the performance of large language models (LLMs) in counterfactual reasoning. In contrast to previous studies that primarily focus on commonsense causal reasoning, where LLMs often rely on prior knowledge for inference, we specifically assess their ability to perform counterfactual inference using a set of formal rules. To support this evaluation, we introduce a new benchmark dataset, CounterBench, comprising 1K counterfactual reasoning questions. The dataset is designed with varying levels of difficulty, diverse causal graph structures, distinct types of counterfactual questions, and multiple nonsensical name variants. Our experiments demonstrate that counterfactual reasoning poses a significant challenge for LLMs, with most models performing at levels comparable to random guessing. To enhance LLM's counterfactual reasoning ability, we propose a novel reasoning paradigm, CoIn, which guides LLMs through iterative reasoning and backtracking to systematically explore counterfactual solutions. Experimental results show that our method significantly improves LLM performance on counterfactual reasoning tasks and consistently enhances performance across different LLMs.Our dataset is available at https://huggingface.co/datasets/CounterBench/CounterBench.", "date": "2025-02-16", "categories": []}
{"id": "2510.02050", "title": "Multidata Causal Discovery for Statistical Hurricane Intensity Forecasting", "abstract": "Improving statistical forecasts of tropical cyclone (TC) intensity is limited by complex nonlinear interactions and difficulty in identifying relevant predictors. Conventional methods prioritize correlation or fit, often overlooking confounding variables and limiting generalizability to unseen TCs. To address this, we leverage a multidata causal discovery framework with a replicated dataset based on Statistical Hurricane Intensity Prediction Scheme (SHIPS) using ERA5 meteorological reanalysis. We conduct experiments to identify and select predictors causally linked to TC intensity changes. We then train multiple linear regression models to compare causal feature selection with correlation, random forest feature importance, and no feature selection, across five forecast lead times from 1 to 5 days (24 to 120 hours). Causal feature selection consistently outperforms on unseen test cases, especially for lead times shorter than 3 days. Top causal features include vertical shear, mid-tropospheric potential vorticity and surface moisture conditions, which are physically significant yet often underutilized in TC intensity predictions. We build an extended predictor set (SHIPS+) by adding selected features to the standard SHIPS predictors. SHIPS+ yields increased short-term predictive skill at lead times of 24, 48, and 72 hours. Adding nonlinearity using a multilayer perceptron further extends skill to longer lead times, despite our framework being purely regional and not requiring global forecast data. Operational SHIPS tests confirm that three of the six added causally discovered predictors improve forecast skill, with the largest gains at longer lead times. Our results demonstrate that causal discovery improves TC intensity prediction and pave the way toward more empirical forecasts.", "date": "2025-10-02", "categories": []}
{"id": "2604.10412", "title": "Orthogonal machine learning for conditional odds and risk ratios", "abstract": "Conditional effects are commonly used measures for understanding how treatment effects vary across different groups, and are often used to target treatments/interventions to groups who benefit most. In this work we review existing methods and propose novel ones, focusing on the odds ratio (OR) and the risk ratio (RR). While estimation of the conditional average treatment effect (ATE) has been widely studied, estimators for the OR and RR lag behind, and cutting edge estimators such as those based on doubly robust transformations or orthogonal risk functions have not been generalized to these parameters. We propose such a generalization here, focusing on the DR-learner and the R-learner. We derive orthogonal risk functions for the OR and RR and show that the associated pseudo-outcomes satisfy second-order conditional-mean remainder properties analogous to the ATE case. We also evaluate estimators for the conditional ATE, OR, and RR in a comprehensive nonparametric Monte Carlo simulation study to compare them with common alternatives under hundreds of different data-generating distributions. Our numerical studies provide empirical guidance for choosing an estimator. For instance, they show that while parametric models are useful in very simple settings, the proposed nonparametric estimators significantly reduce bias and mean squared error in the more complex settings expected in the real world. We illustrate the methods in the analysis of physical activity and sleep trouble in U.S. adults using data from the National Health and Nutrition Examination Survey (NHANES). The results demonstrate that our estimators uncover substantial treatment effect heterogeneity that is obscured by traditional regression approaches and lead to improved treatment decision rules, highlighting the importance of data-adaptive methods for advancing precision health research.", "date": "2026-04-12", "categories": []}
{"id": "2604.08935", "title": "A novel hybrid approach for positive-valued DAG learning", "abstract": "Causal discovery from observational data remains a fundamental challenge in machine learning and statistics, particularly when variables represent inherently positive quantities such as gene expression levels, asset prices, company revenues, or population counts, which often follow multiplicative rather than additive dynamics. We propose the Hybrid Moment-Ratio Scoring (H-MRS) algorithm, a novel method for learning directed acyclic graphs (DAGs) from positive-valued data by combining moment-based scoring with log-scale regression. The key idea is that for positive-valued variables, the moment ratio $\\frac{\\mathbb{E}[X_j^2]}{\\mathbb{E}[(\\mathbb{E}[X_j \\mid S])^2]}$ provides an effective criterion for causal ordering, where $S$ denotes candidate parent sets. H-MRS integrates log-scale Ridge regression for moment-ratio estimation with a greedy ordering procedure based on raw-scale moment ratios, followed by Elastic Net-based parent selection to recover the final DAG structure. Experiments on synthetic log-linear data demonstrate competitive precision and recall. The proposed method is computationally efficient and naturally respects positivity constraints, making it suitable for applications in genomics and economics. These results suggest that combining log-scale modeling with raw-scale moment ratios provides a practical framework for causal discovery in positive-valued domains.", "date": "2026-04-10", "categories": []}
{"id": "2604.09135", "title": "Identifying Causal Effects Using a Single Proxy Variable", "abstract": "Unobserved confounding is a key challenge when estimating causal effects from a treatment on an outcome in scientific applications. In this work, we assume that we observe a single, potentially multi-dimensional proxy variable of the unobserved confounder and that we know the mechanism that generates the proxy from the confounder. Under a completeness assumption on this mechanism, which we call Single Proxy Identifiability of Causal Effects or simply SPICE, we prove that causal effects are identifiable. We extend the proxy-based causal identifiability results by Kuroki and Pearl (2014); Pearl (2010) to higher dimensions, more flexible functional relationships and a broader class of distributions. Further, we develop a neural network based estimation framework, SPICE-Net, to estimate causal effects, which is applicable to both discrete and continuous treatments.", "date": "2026-04-10", "categories": []}
{"id": "2604.09271", "title": "The causal relation between off-street parking and electric vehicle adoption in Scotland", "abstract": "The transition to electric mobility hinges on maximising aggregate adoption while also facilitating equitable access. This study examines whether the 'charging divide' between households with and without off-street parking reflects a genuine infrastructure constraint or a by-product of socio-economic disparity. Moving beyond conventional predictive models, we apply a probabilistic causal framework to a nationally representative dataset of Scottish households, enabling estimation of policy interventions while explicitly neutralising the confounding effect of other causal factors. The results reveal a structural hierarchy in the EV adoption process. Private off-street parking functions as a conversion catalyst: enabling access to home-charging increases the probability of EV ownership from 3.3% to 5.6% (a 70% relative, 2.3 percentage point absolute increase). However, this effect primarily accelerates households already economically positioned to purchase an EV rather than recruiting new entrants. By contrast, household income operates as the fundamental affordability ceiling. A causal contrast between lower- and higher-income strata, shows a reduction in market non-participation by 23.1 percentage points, identifying financial capacity as the principal gatekeeper to entering the EV transition funnel. Crucially, the analysis demonstrates that standard observational models overstate the isolated effect of off-street parking infrastructure. The apparent effect emerges from selection bias: higher-income households are disproportionately likely to possess both private parking and the means to purchase EVs. These findings support a dual-track policy strategy: lowering the affordability ceiling for non-participants through financial instruments, while addressing EV home-charging access for the 'latent intent' cohort in high-density urban contexts.", "date": "2026-04-10", "categories": []}
{"id": "2407.19426", "title": "Causal Discovery in Linear Models with Unobserved Variables and Measurement Error", "abstract": "The presence of unobserved common causes and measurement error poses two major obstacles to causal structure learning, since ignoring either source of complexity can induce spurious causal relations among variables of interest. We study causal structure learning in linear systems where both challenges may occur simultaneously. We introduce a causal model called LV-SEM-ME, which contains four types of variables: directly observed variables, variables that are not directly observed but are measured with error, the corresponding measurements, and variables that are neither observed nor measured. Under a separability condition-namely, identifiability of the mixing matrix associated with the exogenous noise terms of the observed variables-together with certain faithfulness assumptions, we characterize the extent of identifiability and the corresponding observational equivalence classes. We provide graphical characterizations of these equivalence classes and develop recovery algorithms that enumerate all models in the equivalence class of the ground truth. We also establish, via a four-node union model that subsumes instrumental variable, front-door, and negative-control-outcome settings, a form of identification robustness: the target effect remains identifiable in the broader LV-SEM-ME model even when the assumptions underlying the specialized identification formulas for the corresponding submodels need not all hold simultaneously.", "date": "2024-07-28", "categories": []}
{"id": "2604.07712", "title": "CausalVAE as a Plug-in for World Models: Towards Reliable Counterfactual Dynamics", "abstract": "In this work, CausalVAE is introduced as a plug-in structural module for latent world models and is attached to diverse encoder-transition backbones. Across the reported benchmarks, competitive factual prediction is preserved and intervention-aware counterfactual retrieval is improved after the plug-in is added, suggesting stronger robustness under distribution shift and interventions. The largest gains are observed on the Physics benchmark: when averaged over 8 paired baselines, CF-H@1 is improved by +102.5%. In a representative GNN-NLL setting on Physics, CF-H@1 is increased from 11.0 to 41.0 (+272.7%). Through causal analysis, learned structural dependencies are shown to recover meaningful first-order physical interaction trends, supporting the interpretability of the learned latent causal structure.", "date": "2026-04-09", "categories": []}
{"id": "2505.02781", "title": "Local Markov Equivalence for PC-style Local Causal Discovery and Identification of Controlled Direct Effects", "abstract": "Identifying controlled direct effects (CDEs) is crucial across numerous scientific domains. While existing methods can identify these effects from causal directed acyclic graphs (DAGs), the true DAG is often unknown in practice. Essential graphs, which represent a Markov equivalence class of DAGs characterized by the same set of conditional independencies, provide a more practical and realistic alternative, and the PC algorithm is one of the most widely used method to learn them using conditional independence tests. However, learning the full essential graph is computationally intensive and relies on strong, untestable assumptions. In this work, we adapt the PC algorithm to recover only the portion of the graph needed for identifying CDEs. In particular, we introduce the local essential graph (LEG), a graph structure defined relative to a target variable, and present LocPC, an algorithm that learns the LEG using solely local conditional independence tests. Building on this, we develop LocPC-CDE, which extracts precisely the portion of the LEG that is both necessary and sufficient for identifying a CDE. Compared to global methods, our algorithms require less conditional independence tests and operate under weaker assumptions while maintaining theoretical guarantees. We illustrate the effectiveness of our approach on synthetic and real data.", "date": "2025-05-05", "categories": []}
{"id": "2603.11090", "title": "Interventional Time Series Priors for Causal Foundation Models", "abstract": "Prior-data fitted networks (PFNs) have emerged as powerful foundation models for tabular causal inference, yet their extension to time series remains limited by the absence of synthetic data generators that provide interventional targets. Existing time series benchmarks generate observational data with ground-truth causal graphs but lack the interventional data required for training causal foundation models. To address this, we propose \\textbf{CausalTimePrior}, a principled framework for generating synthetic temporal structural causal models (TSCMs) with paired observational and interventional time series. Our prior supports configurable causal graph structures, nonlinear autoregressive mechanisms, regime-switching dynamics, and multiple intervention types (hard, soft, time-varying). We demonstrate that PFNs trained on CausalTimePrior can perform in-context causal effect estimation on held-out TSCMs, establishing a pathway toward foundation models for time series causal inference.", "date": "2026-03-11", "categories": []}
{"id": "2602.10370", "title": "Causal Effect Estimation with Learned Instrument Representations", "abstract": "Instrumental variable (IV) methods mitigate bias from unobserved confounding in observational causal inference but rely on the availability of a valid instrument, which can often be difficult or infeasible to identify in practice. In this paper, we propose a representation learning approach that constructs instrumental representations from observed covariates, which enable IV-based estimation even in the absence of an explicit instrument. Our model (ZNet) achieves this through an architecture that mirrors the structural causal model of IVs; it decomposes the ambient feature space into confounding and instrumental components, and is trained by enforcing empirical moment conditions corresponding to the defining properties of valid instruments (i.e., relevance, exclusion restriction, and instrumental unconfoundedness). Importantly, ZNet is compatible with a wide range of downstream two-stage IV estimators of causal effects. Our experiments demonstrate that ZNet can (i) recover ground-truth instruments when they already exist in the ambient feature space and (ii) construct latent instruments in the embedding space when no explicit IVs are available. Our work suggests when ZNet can be used as a module for causal inference in general observational settings.", "date": "2026-02-10", "categories": []}
{"id": "2604.02656", "title": "Transfer Learning for Meta-analysis Under Covariate Shift", "abstract": "Randomized controlled trials often do not represent the populations where decisions are made, and covariate shift across studies can invalidate standard IPD meta-analysis and transport estimators. We propose a placebo-anchored transport framework that treats source-trial outcomes as abundant proxy signals and target-trial placebo outcomes as scarce, high-fidelity gold labels to calibrate baseline risk. A low-complexity (sparse) correction anchors proxy outcome models to the target population, and the anchored models are embedded in a cross-fitted doubly robust learner, yielding a Neyman-orthogonal, target-site doubly robust estimator for patient-level heterogeneous treatment effects when target treated outcomes are available. We distinguish two regimes: in connected targets (with a treated arm), the method yields target-identified effect estimates; in disconnected targets (placebo-only), it reduces to a principled screen--then--transport procedure under explicit working-model transport assumptions. Experiments on synthetic data and a semi-synthetic IHDP benchmark evaluate pointwise CATE accuracy, ATE error, ranking quality for targeting, decision-theoretic policy regret, and calibration. Across connected settings, the proposed method is best or near-best and improves substantially over proxy-only, target-only, and transport baselines at small target sample sizes; in disconnected settings, it retains strong ranking performance for targeting while pointwise accuracy depends on the strength of the working transport condition.", "date": "2026-04-03", "categories": []}
{"id": "2604.05136", "title": "Non-monotonic causal discovery with Kolmogorov-Arnold Fuzzy Cognitive Maps", "abstract": "Fuzzy Cognitive Maps constitute a neuro-symbolic paradigm for modeling complex dynamic systems, widely adopted for their inherent interpretability and recurrent inference capabilities. However, the standard FCM formulation, characterized by scalar synaptic weights and monotonic activation functions, is fundamentally constrained in modeling non-monotonic causal dependencies, thereby limiting its efficacy in systems governed by saturation effects or periodic dynamics. To overcome this topological restriction, this research proposes the Kolmogorov-Arnold Fuzzy Cognitive Map (KA-FCM), a novel architecture that redefines the causal transmission mechanism. Drawing upon the Kolmogorov-Arnold representation theorem, static scalar weights are replaced with learnable, univariate B-spline functions located on the model edges. This fundamental modification shifts the non-linearity from the nodes' aggregation phase directly to the causal influence phase. This modification allows for the modeling of arbitrary, non-monotonic causal relationships without increasing the graph density or introducing hidden layers. The proposed architecture is validated against both baselines (standard FCM trained with Particle Swarm Optimization) and universal black-box approximators (Multi-Layer Perceptron) across three distinct domains: non-monotonic inference (Yerkes-Dodson law), symbolic regression, and chaotic time-series forecasting. Experimental results demonstrate that KA-FCMs significantly outperform conventional architectures and achieve competitive accuracy relative to MLPs, while preserving graph- based interpretability and enabling the explicit extraction of mathematical laws from the learned edges.", "date": "2026-04-06", "categories": []}
{"id": "2604.05829", "title": "Bivariate Causal Discovery Using Rate-Distortion MDL: An Information Dimension Approach", "abstract": "Approaches to bivariate causal discovery based on the minimum description length (MDL) principle approximate the (uncomputable) Kolmogorov complexity of the models in each causal direction, selecting the one with the lower total complexity. The premise is that nature's mechanisms are simpler in their true causal order. Inherently, the description length (complexity) in each direction includes the description of the cause variable and that of the causal mechanism. In this work, we argue that current state-of-the-art MDL-based methods do not correctly address the problem of estimating the description length of the cause variable, effectively leaving the decision to the description length of the causal mechanism. Based on rate-distortion theory, we propose a new way to measure the description length of the cause, corresponding to the minimum rate required to achieve a distortion level representative of the underlying distribution. This distortion level is deduced using rules from histogram-based density estimation, while the rate is computed using the related concept of information dimension, based on an asymptotic approximation. Combining it with a traditional approach for the causal mechanism, we introduce a new bivariate causal discovery method, termed rate-distortion MDL (RDMDL). We show experimentally that RDMDL achieves competitive performance on the T\\\"ubingen dataset. All the code and experiments are publicly available at github.com/tiagobrogueira/Causal-Discovery-In-Exchangeable-Data.", "date": "2026-04-07", "categories": []}
{"id": "2604.06123", "title": "A Large-Scale Empirical Comparison of Meta-Learners and Causal Forests for Heterogeneous Treatment Effect Estimation in Marketing Uplift Modeling", "abstract": "Estimating Conditional Average Treatment Effects (CATE) at the individual level is central to precision marketing, yet systematic benchmarking of uplift modeling methods at industrial scale remains limited. We present UpliftBench, an empirical evaluation of four CATE estimators: S-Learner, T-Learner, X-Learner (all with LightGBM base learners), and Causal Forest (EconML), applied to the Criteo Uplift v2.1 dataset comprising 13.98 million customer records. The near-random treatment assignment (propensity AUC = 0.509) provides strong internal validity for causal estimation. Evaluated via Qini coefficient and cumulative gain curves, the S-Learner achieves the highest Qini score of 0.376, with the top 20% of customers ranked by predicted CATE capturing 77.7% of all incremental conversions, a 3.9x improvement over random targeting. SHAP analysis identifies f8 as the dominant heterogeneous treatment effect (HTE) driver among the 12 anonymized covariates. Causal Forest uncertainty quantification reveals that 1.9% of customers are confident persuadables (lower 95% CI > 0) and 0.1% are confident sleeping dogs (upper 95% CI < 0). Our results provide practitioners with evidence-based guidance on method selection for large-scale uplift modeling pipelines.", "date": "2026-04-07", "categories": []}
{"id": "2405.03083", "title": "Causal K-Means Clustering", "abstract": "Causal effects are often characterized with population summaries. These might provide an incomplete picture when there are heterogeneous treatment effects across subgroups. Since the subgroup structure is typically unknown, it is more challenging to identify and evaluate subgroup effects than population effects. We propose a new solution to this problem: \\emph{Causal k-Means Clustering}, which leverages the k-means clustering algorithm to uncover the unknown subgroup structure. Our problem differs significantly from the conventional clustering setup since the variables to be clustered are unknown counterfactual functions. We present a plug-in estimator which is simple and readily implementable using off-the-shelf algorithms, and study its rate of convergence. We also develop a new bias-corrected estimator based on nonparametric efficiency theory and double machine learning, and show that this estimator achieves fast root-n rates and asymptotic normality in large nonparametric models. Our proposed methods are especially useful for modern outcome-wide studies with multiple treatment levels. Further, our framework is extensible to clustering with generic pseudo-outcomes, such as partially observed outcomes or otherwise unknown functions. Finally, we explore finite sample properties via simulation, and illustrate the proposed methods using a study of mobile-supported self-management for chronic low back pain.", "date": "2024-05-05", "categories": []}
{"id": "2410.18918", "title": "MissNODAG: Differentiable Cyclic Causal Graph Learning from Incomplete Data", "abstract": "Causal discovery in real-world systems, such as biological networks, is often complicated by feedback loops and incomplete data. Standard algorithms, which assume acyclic structures or fully observed data, struggle with these challenges. To address this gap, we propose MissNODAG, a differentiable framework for learning both the underlying cyclic causal graph and the missingness mechanism from partially observed data, including data missing not at random. Our framework integrates an additive noise model with an expectation-maximization procedure, alternating between imputing missing values and optimizing the observed data likelihood, to uncover both the cyclic structures and the missingness mechanism. We establish consistency guarantees under exact maximization of the score function in the large sample setting. Finally, we demonstrate the effectiveness of MissNODAG through synthetic experiments and an application to real-world gene perturbation data.", "date": "2024-10-24", "categories": []}
{"id": "2502.02020", "title": "Causal Bandit Over Unknown Graphs: Upper Confidence Bounds With Backdoor Adjustment", "abstract": "The causal bandit problem seeks to identify, through sequential experimentation, an intervention that maximizes the expected reward in a causal system modeled by a directed acyclic graph (DAG). Existing methods typically assume that the causal graph is known or impose restrictive structural assumptions. In this paper, we study causal bandit problems when the causal graph is unknown. We first consider Gaussian DAG models without latent confounders. By combining observational and experimental data collected sequentially during the bandit process, we identify candidate backdoor adjustment sets for each intervention arm. These sets enable estimation of causal effects and construction of upper confidence bounds that integrate information from both data sources. Based on these estimates, we propose a new algorithm, termed backdoor-adjustment upper confidence bound (BA-UCB), for sequential intervention selection. We establish finite-time upper bounds on the cumulative regret of BA-UCB, showing improved rates and substantially relaxed dependency on the number of intervention arms compared to standard multi-armed bandit methods. We further extend the methodology and theoretical guarantees to settings with latent confounders, where the observed variables are modeled by an acyclic directed mixed graph. Simulation studies demonstrate that BA-UCB achieves substantially lower cumulative regret and favorable computational efficiency relative to existing approaches.", "date": "2025-02-04", "categories": []}
{"id": "2505.11211", "title": "Bayesian Hierarchical Invariant Prediction", "abstract": "We propose Bayesian Hierarchical Invariant Prediction (BHIP) reframing Invariant Causal Prediction (ICP) through the lens of Hierarchical Bayes. We leverage the hierarchical structure to explicitly test invariance of causal mechanisms under heterogeneous data, resulting in improved computational scalability for a larger number of predictors compared to ICP. Moreover, given its Bayesian nature BHIP enables the use of prior information. We evaluate BHIP on both synthetic and real-world datasets, demonstrating its potential as an alternative inference method to ICP and related methods.", "date": "2025-05-16", "categories": []}
{"id": "2507.13920", "title": "Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem", "abstract": "Most neural models of causality assume static causal graphs, failing to capture the dynamic and sparse nature of physical interactions where causal relationships emerge and dissolve over time. We introduce the Causal Process Framework and its neural implementation, Causal Process Models (CPMs), for learning sparse, time-varying causal graphs from visual observations. Unlike traditional approaches that maintain dense connectivity, our model explicitly constructs causal edges only when objects actively interact, dramatically improving both interpretability and computational efficiency. We achieve this by casting dynamic interaction-graph construction for world modeling as a multi-agent reinforcement learning problem, where specialized agents sequentially decide which objects are causally connected at each timestep. Our key innovation is a structured representation that factorizes object and force vectors along three learned dimensions (mutability, causal relevance, and control relevance), enabling the automatic discovery of semantically meaningful encodings. We demonstrate that a CPM significantly outperforms dense graph baselines on physical prediction tasks, particularly for longer horizons and varying object counts.", "date": "2025-07-18", "categories": []}
{"id": "2509.00472", "title": "Partially Functional Dynamic Backdoor Diffusion-based Causal Model", "abstract": "Causal inference in spatio-temporal settings is critically hindered by unmeasured confounders with complex spatio-temporal dynamics and the prevalence of multi-resolution data. While diffusion models present a promising avenue for estimating structural causal models, existing approaches are limited by assumptions of causal sufficiency or static confounding, failing to capture the region-specific, temporally dependent nature of real-world latent variables or to directly handle functional variables. We bridge this gap by introducing the Partially Functional Dynamic Backdoor Diffusion-based Causal Model (PFD-BDCM), a unified generative framework designed to simultaneously tackle causal inference with dynamic confounding and functional data. Our approach formalizes a novel structural causal model that captures spatio-temporal dependencies in latent confounders through conditional autoregressive processes, represents functional variables via basis expansion coefficients treated as standard graph nodes, and integrates valid backdoor adjustment into a diffusion-based generative process. We provide theoretical guarantees on the preservation of causal effects under basis expansion and derive error bounds for counterfactual estimates. Experiments on synthetic data and a real-world air pollution case study demonstrate that PFD-BDCM outperforms existing methods across observational, interventional, and counterfactual queries. This work provides a rigorous and practical tool for robust causal inference in complex spatio-temporal systems characterized by non-stationarity and multi-resolution data.", "date": "2025-08-30", "categories": []}
{"id": "2509.02892", "title": "Improving Generative Methods for Causal Evaluation via Simulation-Based Inference", "abstract": "Generating synthetic datasets that accurately reflect real-world observational data is critical for evaluating causal estimators, but it remains a challenging task. Existing generative methods offer a solution by producing synthetic datasets anchored in the observed data (source data) while allowing variation in key parameters such as the treatment effect and amount of confounding bias. However, it is often unclear which generative methods to use and which values of parameters to choose when generating synthetic datasets. Moreover, existing methods typically require users to provide fixed point estimates of such parameters. This denies users the ability to express uncertainty over both generative methods and parameter values and removes the potential for posterior inference, potentially leading to unreliable estimator comparisons. We introduce simulation-based inference for causal evaluation (SBICE), a framework that treats the generative method and its corresponding generative parameters as uncertain and infers their posterior distribution given a source dataset. Leveraging techniques in simulation-based inference, SBICE identifies suitable generative methods and infers distributions over its parameter configurations to produce synthetic datasets closely aligned with the source data distribution. Empirical results demonstrate that SBICE improves the reliability of estimator evaluations by generating realistic datasets whose causal estimates closely match the estimates of the source data, making it a robust and uncertainty-aware approach to selecting causal estimators.", "date": "2025-09-02", "categories": []}
{"id": "2509.12981", "title": "Causal Discovery via Quantile Partial Effect", "abstract": "Quantile Partial Effect (QPE) is a statistic associated with conditional quantile regression, measuring the effect of covariates at different levels. Our theory demonstrates that when the QPE of cause on effect is assumed to lie in a finite linear span, cause and effect are identifiable from their observational distribution. This generalizes previous identifiability results based on Functional Causal Models (FCMs) with additive, heteroscedastic noise, etc. Meanwhile, since QPE resides entirely at the observational level, this parametric assumption does not require considering mechanisms, noise, or even the Markov assumption, but rather directly utilizes the asymmetry of shape characteristics in the observational distribution. By performing basis function tests on the estimated QPE, causal directions can be distinguished, which is empirically shown to be effective in experiments on a large number of bivariate causal discovery datasets. For multivariate causal discovery, leveraging the close connection between QPE and score functions, we find that Fisher Information is sufficient as a statistical measure to determine causal order when assumptions are made about the second moment of QPE. We validate the feasibility of using Fisher Information to identify causal order on multiple synthetic and real-world multivariate causal discovery datasets.", "date": "2025-09-16", "categories": []}
{"id": "2512.11919", "title": "A fine-grained look at causal effects in causal spaces", "abstract": "The notion of causal effect is fundamental across many scientific disciplines. Traditionally, quantitative researchers have studied causal effects at the level of variables; for example, how a certain drug dose (W) causally affects a patient's blood pressure (Y). However, in many modern data domains, the raw variables-such as pixels in an image or tokens in a language model-do not have the semantic structure needed to formulate meaningful causal questions. In this paper, we offer a more fine-grained perspective by studying causal effects at the level of events, drawing inspiration from probability theory, where core notions such as independence are first given for events and sigma-algebras, before random variables enter the picture. Within the measure-theoretic framework of causal spaces, a recently introduced axiomatisation of causality, we first introduce several binary definitions that determine whether a causal effect is present, as well as proving some properties of them linking causal effect to (in)dependence under an intervention measure. Further, we provide quantifying measures that capture the strength and nature of causal effects on events, and show that we can recover the common measures of treatment effect as special cases.", "date": "2025-12-11", "categories": []}
{"id": "2602.07943", "title": "IV Co-Scientist: Multi-Agent LLM Framework for Causal Instrumental Variable Discovery", "abstract": "In the presence of confounding between an endogenous variable and the outcome, instrumental variables (IVs) are used to isolate the causal effect of the endogenous variable. Identifying valid instruments requires interdisciplinary knowledge, creativity, and contextual understanding, making it a non-trivial task. In this paper, we investigate whether large language models (LLMs) can aid in this task. We perform a two-stage evaluation framework. First, we test whether LLMs can recover well-established instruments from the literature, assessing their ability to replicate standard reasoning. Second, we evaluate whether LLMs can identify and avoid instruments that have been empirically or theoretically discredited. Building on these results, we introduce IV Co-Scientist, a multi-agent system that proposes, critiques, and refines IVs for a given treatment-outcome pair. We also introduce a statistical test to contextualize consistency in the absence of ground truth. Our results show the potential of LLMs to discover valid instrumental variables from a large observational database.", "date": "2026-02-08", "categories": []}
{"id": "2603.19186", "title": "Improving RCT-Based CATE Estimation Under Covariate Mismatch via Calibrated Alignment", "abstract": "Randomized controlled trials (RCTs) are the gold standard for estimating heterogeneous treatment effects, yet they are often underpowered for detecting effect heterogeneity. Large observational studies (OS) can supplement RCTs for conditional average treatment effect (CATE) estimation, but a key barrier is covariate mismatch: the two sources measure different, only partially overlapping, covariates. We propose CALM (Calibrated ALignment under covariate Mismatch), which bypasses imputation by learning embeddings that map each source's features into a common representation space. OS outcome models are transferred to the RCT embedding space and calibrated using trial data, preserving causal identification from randomization. Finite-sample risk bounds decompose into alignment error, outcome-model complexity, and calibration complexity terms, identifying when embedding alignment outperforms imputation. Under the calibration-based linear variant, the framework provides protection against negative transfer; the neural variant can be vulnerable under severe distributional shift. Under sparse linear models, the embedding approach strictly generalizes imputation. Simulations across 51 settings confirm that (i) calibration-based methods are equivalent for linear CATEs, and (ii) the neural embedding variant wins all 22 nonlinear-regime settings with large margins.", "date": "2026-03-19", "categories": []}
{"id": "2603.22620", "title": "Causal Discovery in Action: Learning Chain-Reaction Mechanisms from Interventions", "abstract": "Causal discovery is challenging in general dynamical systems because, without strong structural assumptions, the underlying causal graph may not be identifiable even from interventional data. However, many real-world systems exhibit directional, cascade-like structure, in which components activate sequentially and upstream failures suppress downstream effects. We study causal discovery in such chain-reaction systems and show that the causal structure is uniquely identifiable from blocking interventions that prevent individual components from activating. We propose a minimal estimator with finite-sample guarantees, achieving exponential error decay and logarithmic sample complexity. Experiments on synthetic models and diverse chain-reaction environments demonstrate reliable recovery from a few interventions, while observational heuristics fail in regimes with delayed or overlapping causal effects.", "date": "2026-03-23", "categories": []}
{"id": "2604.03391", "title": "SDVDiag: Using Context-Aware Causality Mining for the Diagnosis of Connected Vehicle Functions", "abstract": "Real-world implementations of connected vehicle functions are spreading steadily, yet operating these functions reliably remains challenging due to their distributed nature and the complexity of the underlying cloud, edge, and networking infrastructure. Quick diagnosis of problems and understanding the error chains that lead to failures is essential for reducing downtime. However, diagnosing these systems is still largely performed manually, as automated analysis techniques are predominantly data-driven and struggle with hidden relationships and the integration of context information. This paper addresses this gap by introducing a multimodal approach that integrates human feedback and system-specific information into the causal analysis process. Reinforcement Learning from Human Feedback is employed to continuously train a causality mining model while incorporating expert knowledge. Additional modules leverage distributed tracing data to prune false-positive causal links and enable the injection of domain-specific relationships to further refine the causal graph.Evaluation is performed using an automated valet parking application operated in a connected vehicle test field. Results demonstrate a significant increase in precision from 14\\% to 100\\% for the detection of causal edges and improved system interpretability compared to purely data-driven approaches, highlighting the potential for system operators in the connected vehicle domain.", "date": "2026-04-03", "categories": []}
{"id": "2604.03772", "title": "Debiased Machine Learning for Conformal Prediction of Counterfactual Outcomes Under Runtime Confounding", "abstract": "Data-driven decision making frequently relies on predicting counterfactual outcomes. In practice, researchers commonly train counterfactual prediction models on a source dataset to inform decisions on a possibly separate target population. Conformal prediction has arisen as a popular method for producing assumption-lean prediction intervals for counterfactual outcomes that would arise under different treatment decisions in the target population of interest. However, existing methods require that every confounding factor of the treatment-outcome relationship used for training on the source data is additionally measured in the target population, risking miscoverage if important confounders are unmeasured in the target population. In this paper, we introduce a computationally efficient debiased machine learning framework that allows for valid prediction intervals when only a subset of confounders is measured in the target population, a common challenge referred to as runtime confounding. Grounded in semiparametric efficiency theory, we show the resulting prediction intervals achieve desired coverage rates with faster convergence compared to standard methods. Through numerous synthetic and semi-synthetic experiments, we demonstrate the utility of our proposed method.", "date": "2026-04-04", "categories": []}
{"id": "2604.04274", "title": "InferenceEvolve: Towards Automated Causal Effect Estimators through Self-Evolving AI", "abstract": "Causal inference is central to scientific discovery, yet choosing appropriate methods remains challenging because of the complexity of both statistical methodology and real-world data. Inspired by the success of artificial intelligence in accelerating scientific discovery, we introduce InferenceEvolve, an evolutionary framework that uses large language models to discover and iteratively refine causal methods. Across widely used benchmarks, InferenceEvolve yields estimators that consistently outperform established baselines: against 58 human submissions in a recent community competition, our best evolved estimator lay on the Pareto frontier across two evaluation metrics. We also developed robust proxy objectives for settings without semi-synthetic outcomes, with competitive results. Analysis of the evolutionary trajectories shows that agents progressively discover sophisticated strategies tailored to unrevealed data-generating mechanisms. These findings suggest that language-model-guided evolution can optimize structured scientific programs such as causal inference, even when outcomes are only partially observed.", "date": "2026-04-05", "categories": []}
{"id": "2604.04290", "title": "DAGAF: A directed acyclic generative adversarial framework for joint structure learning and tabular data synthesis", "abstract": "Understanding the causal relationships between data variables can provide crucial insights into the construction of tabular datasets. Most existing causality learning methods typically focus on applying a single identifiable causal model, such as the Additive Noise Model (ANM) or the Linear non-Gaussian Acyclic Model (LiNGAM), to discover the dependencies exhibited in observational data. We improve on this approach by introducing a novel dual-step framework capable of performing both causal structure learning and tabular data synthesis under multiple causal model assumptions. Our approach uses Directed Acyclic Graphs (DAG) to represent causal relationships among data variables. By applying various functional causal models including ANM, LiNGAM and the Post-Nonlinear model (PNL), we implicitly learn the contents of DAG to simulate the generative process of observational data, effectively replicating the real data distribution. This is supported by a theoretical analysis to explain the multiple loss terms comprising the objective function of the framework. Experimental results demonstrate that DAGAF outperforms many existing methods in structure learning, achieving significantly lower Structural Hamming Distance (SHD) scores across both real-world and benchmark datasets (Sachs: 47%, Child: 11%, Hailfinder: 5%, Pathfinder: 7% improvement compared to state-of-the-art), while being able to produce diverse, high-quality samples.", "date": "2026-04-05", "categories": []}
{"id": "2402.01207", "title": "Efficient Causal Graph Discovery Using Large Language Models", "abstract": "We propose a novel framework that leverages LLMs for full causal graph discovery. While previous LLM-based methods have used a pairwise query approach, this requires a quadratic number of queries which quickly becomes impractical for larger causal graphs. In contrast, the proposed framework uses a breadth-first search (BFS) approach which allows it to use only a linear number of queries. We also show that the proposed method can easily incorporate observational data when available, to improve performance. In addition to being more time and data-efficient, the proposed framework achieves state-of-the-art results on real-world causal graphs of varying sizes. The results demonstrate the effectiveness and efficiency of the proposed method in discovering causal relationships, showcasing its potential for broad applicability in causal graph discovery tasks across different domains.", "date": "2024-02-02", "categories": []}
{"id": "2410.06128", "title": "Amortized Inference of Causal Models via Conditional Fixed-Point Iterations", "abstract": "Structural Causal Models (SCMs) offer a principled framework to reason about interventions and support out-of-distribution generalization, which are key goals in scientific discovery. However, the task of learning SCMs from observed data poses formidable challenges, and often requires training a separate model for each dataset. In this work, we propose an amortized inference framework that trains a single model to predict the causal mechanisms of SCMs conditioned on their observational data and causal graph. We first use a transformer-based architecture for amortized learning of dataset embeddings, and then extend the Fixed-Point Approach (FiP) to infer the causal mechanisms conditionally on their dataset embeddings. As a byproduct, our method can generate observational and interventional data from novel SCMs at inference time, without updating parameters. Empirical results show that our amortized procedure performs on par with baselines trained specifically for each dataset on both in and out-of-distribution problems, and also outperforms them in scarce data regimes.", "date": "2024-10-08", "categories": []}
{"id": "2604.02472", "title": "VALOR: Value-Aware Revenue Uplift Modeling with Treatment-Gated Representation for B2B Sales", "abstract": "B2B sales organizations must identify \"persuadable\" accounts within zero-inflated revenue distributions to optimize expensive human resource allocation. Standard uplift frameworks struggle with treatment signal collapse in high-dimensional spaces and a misalignment between regression calibration and the ranking of high-value \"whales.\" We introduce VALOR (Value Aware Learning of Optimized (B2B) Revenue), a unified framework featuring a Treatment-Gated Sparse-Revenue Network that uses bilinear interaction to prevent causal signal collapse. The framework is optimized via a novel Cost-Sensitive Focal-ZILN objective that combines a focal mechanism for distributional robustness with a value-weighted ranking loss that scales penalties based on financial magnitude. To provide interpretability for high-touch sales programs, we further derive Robust ZILN-GBDT, a tree based variant utilizing a custom splitting criterion for uplift heterogeneity. Extensive evaluations confirm VALOR's dominance, achieving a 20% improvement in rankability over state-of-the-art methods on public benchmarks and delivering a validated 2.7x increase in incremental revenue per account in a rigorous 4-month production A/B test.", "date": "2026-04-02", "categories": []}
{"id": "2604.02488", "title": "Causal-Audit: A Framework for Risk Assessment of Assumption Violations in Time-Series Causal Discovery", "abstract": "Time-series causal discovery methods rely on assumptions such as stationarity, regular sampling, and bounded temporal dependence. When these assumptions are violated, structure learning can produce confident but misleading causal graphs without warning. We introduce Causal-Audit, a framework that formalizes assumption validation as calibrated risk assessment. The framework computes effect-size diagnostics across five assumption families (stationarity, irregularity, persistence, nonlinearity, and confounding proxies), aggregates them into four calibrated risk scores with uncertainty intervals, and applies an abstention-aware decision policy that recommends methods (e.g., PCMCI+, VAR-based Granger causality) only when evidence supports reliable inference. The semi-automatic diagnostic stage can also be used independently for structured assumption auditing in individual studies. Evaluation on a synthetic atlas of 500 data-generating processes (DGPs) spanning 10 violation families demonstrates well-calibrated risk scores (AUROC > 0.95), a 62% false positive reduction among recommended datasets, and 78% abstention on severe-violation cases. On 21 external evaluations from TimeGraph (18 categories) and CausalTime (3 domains), recommend-or-abstain decisions are consistent with benchmark specifications in all cases. An open-source implementation of our framework is available.", "date": "2026-04-02", "categories": []}
{"id": "2507.04754", "title": "Intervening to Learn and Compose Causally Disentangled Representations", "abstract": "In designing generative models, it is commonly believed that in order to learn useful latent structure, we face a fundamental tension between expressivity and structure. In this paper we challenge this view by proposing a new approach to training arbitrarily expressive generative models that simultaneously learn causally disentangled concepts. This is accomplished by adding a simple context module to an arbitrarily complex black-box model, which learns to process concept information by implicitly inverting linear representations from the model's encoder. Inspired by the notion of intervention in a causal model, our module selectively modifies its architecture during training, allowing it to learn a compact joint model over different contexts. We show how adding this module leads to causally disentangled representations that can be composed for out-of-distribution generation on both real and simulated data. The resulting models can be trained end-to-end or fine-tuned from pre-trained models. To further validate our proposed approach, we prove a new identifiability result that extends existing work on identifying structured representations.", "date": "2025-07-07", "categories": []}
{"id": "2510.15555", "title": "Doubly Robust Estimation of Causal Effects in Strategic Equilibrium Systems", "abstract": "We introduce the Strategic Doubly Robust (SDR) estimator, a novel framework that integrates strategic equilibrium modeling with doubly robust estimation for causal inference in strategic environments. SDR addresses endogenous treatment assignment arising from strategic agent behavior, maintaining double robustness while incorporating strategic considerations. Theoretical analysis confirms SDR's consistency and asymptotic normality under strategic unconfoundedness. Empirical evaluations demonstrate SDR's superior performance over baseline methods, achieving 7.6\\%-29.3\\% bias reduction across varying strategic strengths and maintaining robust scalability with agent populations. The framework provides a principled approach for reliable causal inference when agents respond strategically to interventions.", "date": "2025-10-17", "categories": []}
{"id": "2601.10531", "title": "Coarsening Causal DAG Models", "abstract": "Directed acyclic graphical (DAG) models are a powerful tool for representing causal relationships among jointly distributed random variables, especially concerning data from across different experimental settings. However, it is not always practical or desirable to estimate a causal model at the granularity of given features in a particular dataset. There is a growing body of research on causal abstraction to address such problems. We contribute to this line of research by (i) providing novel graphical identifiability results for practically-relevant interventional settings, (ii) proposing an efficient, provably consistent algorithm for directly learning abstract causal graphs from interventional data with unknown intervention targets, and (iii) uncovering theoretical insights about the lattice structure of the underlying search space, with connections to the field of causal discovery more generally. As proof of concept, we apply our algorithm on synthetic and real datasets with known ground truths, including measurements from a controlled physical system with interacting light intensity and polarization.", "date": "2026-01-15", "categories": []}
{"id": "2604.01325", "title": "The Digital Twin Counterfactual Framework: A Validation Architecture for Simulated Potential Outcomes", "abstract": "The fundamental problem of causal inference - that the counterfactual outcome for any individual is never observed - has shaped the entire methodology of the field. Every existing approach substitutes assumptions for missing data: ignorability, parallel trends, exclusion restrictions. None produces the counterfactual itself. This paper proposes the Digital Twin Counterfactual Framework (DTCF): rather than estimating the counterfactual statistically, we simulate it using a digital twin and subject the simulation to a hierarchical validation regime. We formalize the digital twin simulator as a stochastic mapping within the potential outcomes framework and introduce a hierarchy of twin fidelity assumptions - from marginal fidelity through joint fidelity to structural fidelity - each unlocking a progressively richer class of estimands. The central contribution is threefold. First, a five-level validation architecture converts the unfalsifiable claim that the simulator produces correct counterfactuals into falsifiable tests against observable data. Second, a formal decomposition separates causal quantities into those that are marginally validated (ATE, CATE, QTE - testable through observable-arm comparison) and those that are copula-dependent (the ITE distribution, probability of benefit/harm, variance of treatment effects - permanently reliant on the unobservable within-individual dependence structure). Third, bounding, sensitivity, and uncertainty quantification tools make the copula dependence explicit. The DTCF does not resolve the fundamental problem of causal inference. What it provides is a framework in which marginal causal claims become increasingly testable, joint causal claims become explicitly assumption-indexed, and the gap between the two is formally characterized.", "date": "2026-04-01", "categories": []}
{"id": "2604.01406", "title": "Causal Optimal Coupling for Gaussian Input-Output Distributional Data", "abstract": "We study the problem of identifying an optimal coupling between input-output distributional data generated by a causal dynamical system. The coupling is required to satisfy prescribed marginal distributions and a causality constraint reflecting the temporal structure of the system. We formulate this problem as a Schr\"odinger Bridge, which seeks the coupling closest - in Kullback-Leibler divergence - to a given prior while enforcing both marginal and causality constraints. For the case of Gaussian marginals and general time-dependent quadratic cost functions, we derive a fully tractable characterization of the Sinkhorn iterations that converges to the optimal solution. Beyond its theoretical contribution, the proposed framework provides a principled foundation for applying causal optimal transport methods to system identification from distributional data.", "date": "2026-04-01", "categories": []}
{"id": "2604.01946", "title": "PAC-Bayesian Reward-Certified Outcome Weighted Learning", "abstract": "Estimating optimal individualized treatment rules (ITRs) via outcome weighted learning (OWL) often relies on observed rewards that are noisy or optimistic proxies for the true latent utility. Ignoring this reward uncertainty leads to the selection of policies with inflated apparent performance, yet existing OWL frameworks lack the finite-sample guarantees required to systematically embed such uncertainty into the learning objective. To address this issue, we propose PAC-Bayesian Reward-Certified Outcome Weighted Learning (PROWL). Given a one-sided uncertainty certificate, PROWL constructs a conservative reward and a strictly policy-dependent lower bound on the true expected value. Theoretically, we prove an exact certified reduction that transforms robust policy learning into a unified, split-free cost-sensitive classification task. This formulation enables the derivation of a nonasymptotic PAC-Bayes lower bound for randomized ITRs, where we establish that the optimal posterior maximizing this bound is exactly characterized by a general Bayes update. To overcome the learning-rate selection problem inherent in generalized Bayesian inference, we introduce a fully automated, bounds-based calibration procedure, coupled with a Fisher-consistent certified hinge surrogate for efficient optimization. Our experiments demonstrate that PROWL achieves improvements in estimating robust, high-value treatment regimes under severe reward uncertainty compared to standard methods for ITR estimation.", "date": "2026-04-02", "categories": []}
{"id": "2604.02250", "title": "Smoothing the Landscape: Causal Structure Learning via Diffusion Denoising Objectives", "abstract": "Understanding causal dependencies in observational data is critical for informing decision-making. These relationships are often modeled as Bayesian Networks (BNs) and Directed Acyclic Graphs (DAGs). Existing methods, such as NOTEARS and DAG-GNN, often face issues with scalability and stability in high-dimensional data, especially when there is a feature-sample imbalance. Here, we show that the denoising score matching objective of diffusion models could smooth the gradients for faster, more stable convergence. We also propose an adaptive k-hop acyclicity constraint that improves runtime over existing solutions that require matrix inversion. We name this framework Denoising Diffusion Causal Discovery (DDCD). Unlike generative diffusion models, DDCD utilizes the reverse denoising process to infer a parameterized causal structure rather than to generate data. We demonstrate the competitive performance of DDCDs on synthetic benchmarking data. We also show that our methods are practically useful by conducting qualitative analyses on two real-world examples. Code is available at this url: https://github.com/haozhu233/ddcd.", "date": "2026-04-02", "categories": []}
{"id": "2406.15904", "title": "Learning When the Concept Shifts: Confounding, Invariance, and Dimension Reduction", "abstract": "Practitioners often face the challenge of deploying prediction models in new environments with shifted distributions of covariates and responses. With observational data, such shifts are often driven by unobserved confounding, and can in fact alter the concept of which model is best. This paper studies distribution shifts in the domain adaptation problem with unobserved confounding. We postulate a linear structural causal model to account for endogeneity and unobserved confounding, and we leverage exogenous invariant covariate representations to cure concept shifts and improve target prediction. We propose a data-driven representation learning method that optimizes for a lower-dimensional linear subspace and a prediction model confined to that subspace. This method operates on a non-convex objective -- that interpolates between predictability and stability -- constrained to the Stiefel manifold, using an analog of projected gradient descent. We analyze the optimization landscape and prove that, provided sufficient regularization, nearly all local optima align with an invariant linear subspace resilient to distribution shifts. This method achieves a nearly ideal gap between target and source risk. We validate the method and theory with real-world data sets to illustrate the tradeoffs between predictability and stability.", "date": "2024-06-22", "categories": []}
{"id": "2408.07575", "title": "A General Framework on Conditions for Constraint-based Causal Learning", "abstract": "Most constraint-based causal learning algorithms provably return the correct causal graph under certain correctness conditions, such as faithfulness. By representing any constraint-based causal learning algorithm using the notion of a property, we provide a general framework to obtain and study correctness conditions for these algorithms. From the framework, we provide exact correctness conditions for the PC algorithm, which are then related to the correctness conditions of some other existing causal discovery algorithms. The framework also suggests a paradigm for designing causal learning algorithms which allows for the correctness conditions of algorithms to be controlled for before designing the actual algorithm, and has the following implications. We show that the sparsest Markov representation condition is the weakest correctness condition for algorithms that output ancestral graphs or directed acyclic graphs satisfying any existing notions of minimality. We also reason that Pearl-minimality is necessary for meaningful causal learning but not sufficient to relax the faithfulness condition and, as such, has to be strengthened, such as by including background knowledge, for causal learning beyond faithfulness.", "date": "2024-08-14", "categories": []}
{"id": "2509.17180", "title": "Regularizing Extrapolation in Causal Inference", "abstract": "Many common estimators in machine learning and causal inference are linear smoothers, where the prediction is a weighted average of the training outcomes. Some estimators, such as ordinary least squares and kernel ridge regression, allow for arbitrarily negative weights, which improve feature imbalance but often at the cost of increased dependence on parametric modeling assumptions and higher variance. By contrast, estimators like importance weighting and random forests (sometimes implicitly) restrict weights to be non-negative, reducing dependence on parametric modeling and variance at the cost of worse imbalance. In this paper, we propose a unified framework that directly penalizes the level of extrapolation, replacing the current practice of a hard non-negativity constraint with a soft constraint and corresponding hyperparameter. We derive a worst-case extrapolation error bound and introduce a novel \"bias-bias-variance\" tradeoff, encompassing biases due to feature imbalance, model misspecification, and estimator variance; this tradeoff is especially pronounced in high dimensions, particularly when positivity is poor. We then develop an optimization procedure that regularizes this bound while minimizing imbalance and outline how to use this approach as a sensitivity analysis for dependence on parametric modeling assumptions. We demonstrate the effectiveness of our approach through synthetic experiments and a real-world application, involving the generalization of randomized controlled trial estimates to a target population of interest.", "date": "2025-09-21", "categories": []}
{"id": "2601.20589", "title": "Exact Graph Learning via Integer Programming", "abstract": "Learning the dependence structure among variables in complex systems is a central problem across medical, natural, and social sciences. These structures can be naturally represented by graphs, and the task of inferring such graphs from data is known as graph learning or causal discovery. Existing approaches typically rely on restrictive assumptions about the data-generating process, employ greedy oracle algorithms, or solve approximate formulations of the graph learning problem. Therefore, they are either sensitive to violations of central assumptions or fail to guarantee globally optimal solutions. We address these limitations by introducing a nonparametric graph learning framework based on conditional independence testing and integer programming. We reformulate the graph learning problem as a mixed-integer program and prove that solving this integer-programming problem provides a globally optimal solution to the original graph learning problem. Our method leverages efficient encodings of graphical separation criteria, enabling the exact recovery of larger graphs than was previously feasible. We provide an open-source R package 'glip' which supports learning (acyclic) directed (mixed) graphs and chain graphs. We demonstrate that our approach is often faster than existing exact graph learning procedures and achieves state-of-the-art performance on simulated and benchmark data across all aforementioned classes of graphs.", "date": "2026-01-28", "categories": []}
{"id": "2604.00811", "title": "Deconfounding Scores and Representation Learning for Causal Effect Estimation with Weak Overlap", "abstract": "Overlap, also known as positivity, is a key condition for causal treatment effect estimation. Many popular estimators suffer from high variance and become brittle when features differ strongly across treatment groups. This is especially challenging in high dimensions: the curse of dimensionality can make overlap implausible. To address this, we propose a class of feature representations called deconfounding scores, which preserve both identification and the target of estimation; the classical propensity and prognostic scores are two special cases. We characterize the problem of finding a representation with better overlap as minimizing an overlap divergence under a deconfounding score constraint. We then derive closed-form expressions for a class of deconfounding scores under a broad family of generalized linear models with Gaussian features and show that prognostic scores are overlap-optimal within this class. We conduct extensive experiments to assess this behavior empirically.", "date": "2026-04-01", "categories": []}
{"id": "2507.11780", "title": "Inference on Optimal Policy Values and Other Irregular Functionals via Softmax Smoothing", "abstract": "Constructing confidence intervals for the value of an (unknown) optimal treatment policy is a fundamental problem in causal inference. Insight into the optimal policy value can guide the development of reward-maximizing, individualized treatment regimes. However, because the functional that defines the optimal value is non-differentiable, standard semi-parametric approaches for performing inference fail to be directly applicable. Many existing works circumvent non-differentiability by making the unrealistic assumption of zero probability of treatment non-response, i.e. that every unit responds (either positively or negatively) to an assigned treatment. Further, works that don't circumvent this restriction rely on refitting nuisance models a number of times proportional to the sample size. In this paper, we construct and analyze a simple, softmax smoothing-based estimator for the value of an optimal treatment policy. Our estimator applies in both static and dynamic treatment regimes, only requires fitting a constant number of nuisance models, and is statistically efficient when there is zero probability of non-response to treatment. Also, while our estimator does not require making semi-parametric restrictions, it can exploit them when they exist. We further show how our softmax smoothing approach can be used to estimate general parameters that are specified as a maximum of scores involving nuisance components, and look at conditional Balke and Pearl bounds and $L^1$ calibration error as salient examples.", "date": "2025-07-15", "categories": []}
{"id": "2510.14582", "title": "Local Causal Discovery for Statistically Efficient Causal Inference", "abstract": "Causal discovery methods can identify valid adjustment sets for causal effect estimation for a pair of target variables, even when the underlying causal graph is unknown. Global causal discovery methods focus on learning the whole causal graph and therefore enable the recovery of optimal adjustment sets, i.e., sets with the lowest asymptotic variance, but they quickly become computationally prohibitive as the number of variables grows. Local causal discovery methods offer a more scalable alternative by focusing on the local neighborhood of the target variables, but are restricted to statistically suboptimal adjustment sets. In this work, we propose Local Optimal Adjustments Discovery (LOAD), a sound and complete causal discovery approach that combines the computational efficiency of local methods with the statistical optimality of global methods. First, LOAD identifies the causal relation between the targets and tests if the causal effect is identifiable by using only local information. If it is identifiable, it finds the possible descendants of the treatment and infers the optimal adjustment set as the parents of the outcome in a modified forbidden projection. Otherwise, it returns the locally valid parent adjustment sets. In our experiments on synthetic and realistic data LOAD outperforms global methods in scalability, while providing more accurate effect estimation than local methods.", "date": "2025-10-16", "categories": []}
{"id": "2603.24436", "title": "Enes Causal Discovery", "abstract": "Enes The proposed architecture is a mixture of experts, which allows for the model entities, such as the causal relationships, to be further parameterized. More specifically, an attempt is made to exploit a neural net as implementing neurons poses a great challenge for this dataset. To explain, a simple and fast Pearson coefficient linear model usually achieves good scores. An aggressive baseline that requires a really good model to overcome that is. Moreover, there are major limitations when it comes to causal discovery of observational data. Unlike the sachs one did not use interventions but only prior knowledge; the most prohibiting limitation is that of the data which is addressed. Thereafter, the method and the model are described and after that the results are presented.", "date": "2026-03-25", "categories": []}
{"id": "2603.29755", "title": "CausalPulse: An Industrial-Grade Neurosymbolic Multi-Agent Copilot for Causal Diagnostics in Smart Manufacturing", "abstract": "Modern manufacturing environments demand real-time, trustworthy, and interpretable root-cause insights to sustain productivity and quality. Traditional analytics pipelines often treat anomaly detection, causal inference, and root-cause analysis as isolated stages, limiting scalability and explainability. In this work, we present CausalPulse, an industry-grade multi-agent copilot that automates causal diagnostics in smart manufacturing. It unifies anomaly detection, causal discovery, and reasoning through a neurosymbolic architecture built on standardized agentic protocols. CausalPulse is being deployed in a Robert Bosch manufacturing plant, integrating seamlessly with existing monitoring workflows and supporting real-time operation at production scale. Evaluations on both public (Future Factories) and proprietary (Planar Sensor Element) datasets show high reliability, achieving overall success rates of 98.0% and 98.73%. Per-criterion success rates reached 98.75% for planning and tool use, 97.3% for self-reflection, and 99.2% for collaboration. Runtime experiments report end-to-end latency of 50-60s per diagnostic workflow with near-linear scalability (R^2=0.97), confirming real-time readiness. Comparison with existing industrial copilots highlights distinct advantages in modularity, extensibility, and deployment maturity. These results demonstrate how CausalPulse's modular, human-in-the-loop design enables reliable, interpretable, and production-ready automation for next-generation manufacturing.", "date": "2026-03-31", "categories": []}
{"id": "2412.14019", "title": "Retrieving Classes of Causal Orders with Inconsistent Knowledge Bases", "abstract": "Traditional causal discovery methods often depend on strong, untestable assumptions, making them unreliable in real-world applications. In this context, Large Language Models (LLMs) have emerged as a promising alternative for extracting causal knowledge from text-based metadata, effectively consolidating domain expertise. However, LLMs are prone to hallucinations, necessitating strategies that account for these limitations. One effective approach is to use a consistency measure as a proxy of reliability. Moreover, LLMs do not clearly distinguish direct from indirect causal relationships, complicating the discovery of causal Directed Acyclic Graphs (DAGs), which are often sparse. This ambiguity is evident in the way informal sentences are formulated in various domains. For this reason, focusing on causal orders provides a more practical and direct task for LLMs. We propose a new method for deriving abstractions of causal orders that maximizes a consistency score obtained from an LLM. Our approach begins by computing pairwise consistency scores between variables, from which we construct a semi-complete partially directed graph that consolidates these scores into an abstraction. Using this structure, we identify both a maximally oriented partially directed acyclic graph and an optimal set of acyclic tournaments that maximize consistency across all configurations. We further demonstrate how both the abstraction and the class of causal orders can be used to estimate causal effects. We evaluate our method on a wide set of causal DAGs extracted from scientific literature in epidemiology and public health. Our results show that the proposed approach can effectively recover the correct causal order, providing a reliable and practical LLM-assisted causal framework.", "date": "2024-12-18", "categories": []}
{"id": "2508.01321", "title": "Flow IV: Counterfactual Inference In Nonseparable Outcome Models Using Instrumental Variables", "abstract": "To reach human level intelligence, learning algorithms need to incorporate causal reasoning. But identifying causality, and particularly counterfactual reasoning, remains elusive. In this paper, we make progress on counterfactual inference in nonseparable outcome models by utilizing instrumental variables (IVs). IVs are a classic tool for mitigating bias from unobserved confounders when estimating causal effects. While IV methods for effect estimation have been extended to nonseparable outcome models under different assumptions, existing IV approaches to counterfactual prediction typically assume one-dimensional outcomes and additive noise. In this paper, we show that under standard IV assumptions, along with the assumption that the outcome function is invertible and has a triangular structure, then the treatment-outcome relationship becomes identifiable from observed data. We furthermore propose a method to learn the outcome function utilizing normalizing flows. This outcome function estimator can then be used to perform counterfactual inference. We refer to the method as Flow IV.", "date": "2025-08-02", "categories": []}
{"id": "2603.27114", "title": "Maximin Learning of Individualized Treatment Effect on Multi-Domain Outcomes", "abstract": "Precision mental health requires treatment decisions that account for heterogeneous symptoms reflecting multiple clinical domains. However, existing methods for estimating individualized treatment effects (ITE) rely on a single summary outcome or a specific set of observed symptoms or measures, which are sensitive to symptom selection and limit generalizability to unmeasured yet clinically relevant domains. We propose DRIFT, a new maximin framework for estimating robust ITEs from high-dimensional item-level data by leveraging latent factor representations and adversarial learning. DRIFT learns latent constructs via generalized factor analysis, then constructs an anchored on-target uncertainty set that extrapolates beyond the observed measures to approximate the broader hyper-population of potential outcomes. By optimizing worst-case performance over this uncertainty set, DRIFT yields ITEs that are robust to underrepresented or unmeasured domains. We further show that DRIFT is invariant to admissible reparameterizations of the latent factors and admits a closed-form maximin solution, with theoretical guarantees for identification and convergence. In analyses of a randomized controlled trial for major depressive disorder (EMBARC), DRIFT demonstrates superior performance and improved generalizability to external multi-domain outcomes, including side effects and self-reported symptoms not used during training.", "date": "2026-03-28", "categories": []}
{"id": "2603.27510", "title": "Decomposing Discrimination: Causal Mediation Analysis for AI-Driven Credit Decisions", "abstract": "Statistical fairness metrics in AI-driven credit decisions conflate two causally distinct mechanisms: discrimination operating directly from a protected attribute to a credit outcome, and structural inequality propagating through legitimate financial features. We formalise this distinction using Pearl's framework of natural direct and indirect effects applied to the credit decision setting. Our primary theoretical contribution is an identification strategy for natural direct and indirect effects under treatment-induced confounding -- the prevalent setting in which protected attributes causally affect both financial mediators and the final decision, violating standard sequential ignorability. We show that interventional direct and indirect effects (IDE/IIE) are identified under the weaker Modified Sequential Ignorability assumption, and prove that IDE/IIE provide conservative bounds on the unidentified natural effects under monotone indirect treatment response. We propose a doubly-robust augmented inverse probability weighted (AIPW) estimator for IDE/IIE with semiparametric efficiency properties, implemented via cross-fitting. An E-value sensitivity analysis addresses residual confounding on the direct pathway. Empirical evaluation on 89,465 real HMDA conventional purchase mortgage applications from New York State (2022) demonstrates that approximately 77% of the observed 7.9 percentage-point racial denial disparity operates through financial mediators shaped by structural inequality, while the remaining 23% constitutes a conservative lower bound on direct discrimination. The open-source CausalFair Python package implements the full pipeline for deployment at resource-constrained financial institutions.", "date": "2026-03-29", "categories": []}
{"id": "2603.28597", "title": "Position: Explainable AI is Causality in Disguise", "abstract": "The demand for Explainable AI (XAI) has triggered an explosion of methods, producing a landscape so fragmented that we now rely on surveys of surveys. Yet, fundamental challenges persist: conflicting metrics, failed sanity checks, and unresolved debates over robustness and fairness. The only consensus on how to achieve explainability is a lack of one. This has led many to point to the absence of a ground truth for defining ``the'' correct explanation as the main culprit.\n  This position paper posits that the persistent discord in XAI arises not from an absent ground truth but from a ground truth that exists, albeit as an elusive and challenging target: the causal model that governs the relevant system. By reframing XAI queries about data, models, or decisions as causal inquiries, we prove the necessity and sufficiency of causal models for XAI. We contend that without this causal grounding, XAI remains unmoored. Ultimately, we encourage the community to converge around advanced concept and causal discovery to escape this entrenched uncertainty.", "date": "2026-03-30", "categories": []}
{"id": "2502.14950", "title": "Symmetric observations without symmetric causal explanations", "abstract": "Inferring causal models from observed correlations is a challenging task, crucial to many areas of science. In order to alleviate the computational effort when sifting through possible causal explanations for some given observations, it is important to know whether symmetries in the observations correspond to symmetries in the underlying realization so that one can quickly discard impossible explanations. Via an explicit example, we demonstrate that, in general, symmetries cannot be exploited to reduce the hypothesis space. We use a tripartite probability distribution over binary events that is realized by using three (different) independent sources of classical randomness. We prove that even removing the condition that the sources distribute systems described by classical physics, the requirements that (i) the sources distribute the same physical systems, (ii) these physical systems respect relativistic causality, and (iii) the correlations are the observed ones are incompatible.", "date": "2025-02-20", "categories": []}
{"id": "2511.02531", "title": "Causal Graph Neural Networks for Healthcare", "abstract": "Healthcare artificial intelligence systems often degrade in performance when deployed across institutions, with documented performance drops and perpetuation of discriminatory patterns embedded in data. This brittleness comes, in part, from learning statistical associations rather than causal mechanisms. Causal graph neural networks address this by combining graph-based representations of biomedical data with causal inference to learn invariant mechanisms instead of just spurious correlations. This Perspective reviews the methodology of structural causal models, disentangled causal representation learning, and techniques for interventional prediction and counterfactual reasoning on graphs. We discuss applications across psychiatric diagnosis and brain network analysis, cancer subtyping with multi-omics causal integration, continuous physiological monitoring, and drug recommendations. These methods provide building blocks for patient-specific Causal Digital Twins that could support in silico clinical experimentation. Remaining challenges include computational costs that preclude real-time deployment, validation challenges that go beyond standard cross-validation, and the risk of causal-washing where methods adopt causal terminology without rigorous evidentiary support. We propose a tiered framework distinguishing causally-inspired architectures from causally-validated discoveries and outline future directions, including scalable causal discovery, multi-modal data integration, and regulatory pathways for these methods. Making practical Causal Digital Twins possible will require an honest assessment of what current methods deliver, sustained collaboration across disciplines, and validation standards that match the strength of the causal claims being made.", "date": "2025-11-04", "categories": []}
{"id": "2603.25796", "title": "Beyond identifiability: Learning causal representations with few environments and finite samples", "abstract": "We provide explicit, finite-sample guarantees for learning causal representations from data with a sublinear number of environments. Causal representation learning seeks to provide a rigourous foundation for the general representation learning problem by bridging causal models with latent factor models in order to learn interpretable representations with causal semantics. Despite a blossoming theory of identifiability in causal representation learning, estimation and finite-sample bounds are less well understood. We show that causal representations can be learned with only a logarithmic number of unknown, multi-node interventions, and that the intervention targets need not be carefully designed in advance. Through a careful perturbation analysis, we provide a new analysis of this problem that guarantees consistent recovery of (a) the latent causal graph, (b) the mixing matrix and representations, and (c) \\emph{unknown} intervention targets.", "date": "2026-03-26", "categories": []}
{"id": "2603.26024", "title": "Identification of Bivariate Causal Directionality Based on Anticipated Asymmetric Geometries", "abstract": "Identification of causal directionality in bivariate numerical data is a fundamental research problem with important practical implications. This paper presents two alternative methods to identify direction of causation by considering conditional distributions: (1) Anticipated Asymmetric Geometries (AAG) and (2) Monotonicity Index. The AAG method compares the actual conditional distributions to anticipated ones along two variables. Different comparison metrics, such as correlation, cosine similarity, Jaccard index, K-L divergence, K-S distance, and mutual information have been evaluated. Anticipated distributions have been projected as normal based on dual response statistics: mean and standard deviation. The Monotonicity Index approach compares the calculated monotonicity indexes of the gradients of conditional distributions along two axes and exhibits counts of gradient sign changes. Both methods assume stochastic properties of the bivariate data and exploit anticipated unimodality of conditional distributions of the effect. It turns out that the tuned AAG method outperforms the Monotonicity Index and reaches a top accuracy of 77.9% compared to ANMs accuracy of 63 +/- 10% when classifying 95 pairs of real-world examples (Mooij et al, 2014). The described methods include a number of hyperparameters that impact accuracy of the identification. For a given set of hyperparameters, both the AAG or Monotonicity Index method provide a unique deterministic outcome of the solution. To address sensitivity to hyperparameters, tuning of hyperparameters has been done by utilizing a full factorial Design of Experiment. A decision tree has been fitted to distinguish misclassified cases using the input data's symmetrical bivariate statistics to address the question of: How decisive is the identification method of causal directionality?", "date": "2026-03-27", "categories": []}
{"id": "2410.20894", "title": "Working Paper: Active Causal Structure Learning with Latent Variables: Towards Learning to Detour in Autonomous Robots", "abstract": "Artificial General Intelligence (AGI) Agents and Robots must be able to cope with everchanging environments and tasks. They must be able to actively construct new internal causal models of their interactions with the environment when new structural changes take place in the environment. Thus, we claim that active causal structure learning with latent variables (ACSLWL) is a necessary component to build AGI agents and robots. This paper describes how a complex planning and expectation-based detour behavior can be learned by ACSLWL when, unexpectedly, and for the first time, the simulated robot encounters a sort of transparent barrier in its pathway towards its target. ACSWL consists of acting in the environment, discovering new causal relations, constructing new causal models, exploiting the causal models to maximize its expected utility, detecting possible latent variables when unexpected observations occur, and constructing new structures-internal causal models and optimal estimation of the associated parameters, to be able to cope efficiently with the new encountered situations. That is, the agent must be able to construct new causal internal models that transform a previously unexpected and inefficient (sub-optimal) situation, into a predictable situation with an optimal operating plan.", "date": "2024-10-28", "categories": []}
{"id": "2503.08371", "title": "Density Ratio-based Proxy Causal Learning Without Density Ratios", "abstract": "We address the setting of Proxy Causal Learning (PCL), which has the goal of estimating causal effects from observed data in the presence of hidden confounding. Proxy methods accomplish this task using two proxy variables related to the latent confounder: a treatment proxy (related to the treatment) and an outcome proxy (related to the outcome). Two approaches have been proposed to perform causal effect estimation given proxy variables; however only one of these has found mainstream acceptance, since the other was understood to require density ratio estimation - a challenging task in high dimensions. In the present work, we propose a practical and effective implementation of the second approach, which bypasses explicit density ratio estimation and is suitable for continuous and high-dimensional treatments. We employ kernel ridge regression to derive estimators, resulting in simple closed-form solutions for dose-response and conditional dose-response curves, along with consistency guarantees. Our methods empirically demonstrate superior or comparable performance to existing frameworks on synthetic and real-world datasets.", "date": "2025-03-11", "categories": []}
{"id": "2505.19807", "title": "Density Ratio-Free Doubly Robust Proxy Causal Learning", "abstract": "We study the problem of causal function estimation in the Proxy Causal Learning (PCL) framework, where confounders are not observed but proxies for the confounders are available. Two main approaches have been proposed: outcome bridge-based and treatment bridge-based methods. In this work, we propose two kernel-based doubly robust estimators that combine the strengths of both approaches, and naturally handle continuous and high-dimensional variables. Our identification strategy builds on a recent density ratio-free method for treatment bridge-based PCL; furthermore, in contrast to previous approaches, it does not require indicator functions or kernel smoothing over the treatment variable. These properties make it especially well-suited for continuous or high-dimensional treatments. By using kernel mean embeddings, we propose the first density-ratio free doubly robust estimators for proxy causal learning, which have closed form solutions and strong uniform consistency guarantees. Our estimators outperform existing methods on PCL benchmarks, including a prior doubly robust method that requires both kernel smoothing and density ratio estimation.", "date": "2025-05-26", "categories": []}
{"id": "2509.15199", "title": "CausalPre: Scalable and Effective Data Pre-Processing for Causal Fairness", "abstract": "Causal fairness in databases is crucial to preventing biased and inaccurate outcomes in downstream tasks. While most prior work assumes a known causal model, recent efforts relax this assumption by enforcing additional constraints. However, these approaches often fail to capture broader attribute relationships that are critical to maintaining utility. This raises a fundamental question: Can we harness the benefits of causal reasoning to design efficient and effective fairness solutions without relying on strong assumptions about the underlying causal model? In this paper, we seek to answer this question by introducing CausalPre, a scalable and effective causality-guided data pre-processing framework that guarantees justifiable fairness, a strong causal notion of fairness. CausalPre extracts causally fair relationships by reformulating the originally complex and computationally infeasible extraction task into a tailored distribution estimation problem. To ensure scalability, CausalPre adopts a carefully crafted variant of low-dimensional marginal factorization to approximate the joint distribution, complemented by a heuristic algorithm that efficiently tackles the associated computational challenge. Extensive experiments on benchmark datasets demonstrate that CausalPre is both effective and scalable, challenging the conventional belief that achieving causal fairness requires trading off relationship coverage for relaxed model assumptions.", "date": "2025-09-18", "categories": []}
{"id": "2512.10152", "title": "Rethinking Bivariate Causal Discovery Through the Lens of Exchangeability", "abstract": "Causal discovery methods have traditionally been developed under two different modeling assumptions: independent and identically distributed (i.i.d.) data and time series data. In this paper, we focus on the i.i.d. setting, arguing that it should be reframed in terms of exchangeability, a strictly more general symmetry principle. For that goal, we propose an exchangeable hierarchical model that builds upon the recent Causal de Finetti theorem. Using this model, we show that both the uncertainty regarding the causal mechanism and the uncertainty in the distribution of latent variables are better captured under the broader assumption of exchangeability. In fact, we argue that this is most often the case with real data, as supported by an in-depth analysis of the T\\\"ubingen dataset. Exploiting this insight, we introduce a novel synthetic dataset that mimics the generation process induced by the proposed exchangeable hierarchical model. We show that our exchangeable synthetic dataset mirrors the statistical and causal structure of the T\\\"ubingen dataset more closely than other i.i.d. synthetic datasets. Furthermore, we introduce SynthNN, a neural-network-based causal-discovery method trained exclusively on the proposed synthetic dataset. The fact that SynthNN performs competitively with other state-of-the-art methods on the real-world T\\\"ubingen dataset provides strong evidence for the realism of the underlying exchangeable generative model.", "date": "2025-12-10", "categories": []}
{"id": "2603.24618", "title": "Causal AI For AMS Circuit Design: Interpretable Parameter Effects Analysis", "abstract": "Analog-mixed-signal (AMS) circuits are highly non-linear and operate on continuous real-world signals, making them far more difficult to model with data-driven AI than digital blocks. To close the gap between structured design data (device dimensions, bias voltages, etc.) and real-world performance, we propose a causal-inference framework that first discovers a directed-acyclic graph (DAG) from SPICE simulation data and then quantifies parameter impact through Average Treatment Effect (ATE) estimation. The approach yields human-interpretable rankings of design knobs and explicit 'what-if' predictions, enabling designers to understand trade-offs in sizing and topology. We evaluate the pipeline on three operational-amplifier families (OTA, telescopic, and folded-cascode) implemented in TSMC 65nm and benchmark it against a baseline neural-network (NN) regressor. Across all circuits the causal model reproduces simulation-based ATEs with an average absolute error of less than 25%, whereas the neural network deviates by more than 80% and frequently predicts the wrong sign. These results demonstrate that causal AI provides both higher accuracy and explainability, paving the way for more efficient, trustworthy AMS design automation.", "date": "2026-03-24", "categories": []}
{"id": "2603.25126", "title": "MCLMR: A Model-Agnostic Causal Learning Framework for Multi-Behavior Recommendation", "abstract": "Multi-Behavior Recommendation (MBR) leverages multiple user interaction types (e.g., views, clicks, purchases) to enrich preference modeling and alleviate data sparsity issues in traditional single-behavior approaches. However, existing MBR methods face fundamental challenges: they lack principled frameworks to model complex confounding effects from user behavioral habits and item multi-behavior distributions, struggle with effective aggregation of heterogeneous auxiliary behaviors, and fail to align behavioral representations across semantic gaps while accounting for bias distortions. To address these limitations, we propose MCLMR, a novel model-agnostic causal learning framework that can be seamlessly integrated into various MBR architectures. MCLMR first constructs a causal graph to model confounding effects and performs interventions for unbiased preference estimation. Under this causal framework, it employs an Adaptive Aggregation module based on Mixture-of-Experts to dynamically fuse auxiliary behavior information and a Bias-aware Contrastive Learning module to align cross-behavior representations in a bias-aware manner. Extensive experiments on three real-world datasets demonstrate that MCLMR achieves significant performance improvements across various baseline models, validating its effectiveness and generality. All data and code will be made publicly available. For anonymous review, our code is available at the following the link: https://github.com/gitrxh/MCLMR.", "date": "2026-03-26", "categories": []}
{"id": "2603.25202", "title": "CIV-DG: Conditional Instrumental Variables for Domain Generalization in Medical Imaging", "abstract": "Cross-site generalizability in medical AI is fundamentally compromised by selection bias, a structural mechanism where patient demographics (e.g., age, severity) non-randomly dictate hospital assignment. Conventional Domain Generalization (DG) paradigms, which predominantly target image-level distribution shifts, fail to address the resulting spurious correlations between site-specific variations and diagnostic labels. To surmount this identifiability barrier, we propose CIV-DG, a causal framework that leverages Conditional Instrumental Variables to disentangle pathological semantics from scanner-induced artifacts. By relaxing the strict random assignment assumption of standard IV methods, CIV-DG accommodates complex clinical scenarios where hospital selection is endogenously driven by patient demographics. We instantiate this theory via a Deep Generalized Method of Moments (DeepGMM) architecture, employing a conditional critic to minimize moment violations and enforce instrument-error orthogonality within demographic strata. Extensive experiments on the Camelyon17 benchmark and large-scale Chest X-Ray datasets demonstrate that CIV-DG significantly outperforms leading baselines, validating the efficacy of conditional causal mechanisms in resolving structural confounding for robust medical AI.", "date": "2026-03-26", "categories": []}
{"id": "2603.25397", "title": "A Causal Framework for Evaluating ICU Discharge Strategies", "abstract": "In this applied paper, we address the difficult open problem of when to discharge patients from the Intensive Care Unit. This can be conceived as an optimal stopping scenario with three added challenges: 1) the evaluation of a stopping strategy from observational data is itself a complex causal inference problem, 2) the composite objective is to minimize the length of intervention and maximize the outcome, but the two cannot be collapsed to a single dimension, and 3) the recording of variables stops when the intervention is discontinued. Our contributions are two-fold. First, we generalize the implementation of the g-formula Python package, providing a framework to evaluate stopping strategies for problems with the aforementioned structure, including positivity and coverage checks. Second, with a fully open-source pipeline, we apply this approach to MIMIC-IV, a public ICU dataset, demonstrating the potential for strategies that improve upon current care.", "date": "2026-03-26", "categories": []}
{"id": "2603.25473", "title": "Causal-INSIGHT: Probing Temporal Models to Extract Causal Structure", "abstract": "Understanding directed temporal interactions in multivariate time series is essential for interpreting complex dynamical systems and the predictive models trained on them. We present Causal-INSIGHT, a model-agnostic, post-hoc interpretation framework for extracting model-implied (predictor-dependent), directed, time-lagged influence structure from trained temporal predictors. Rather than inferring causal structure at the level of the data-generating process, Causal-INSIGHT analyzes how a fixed, pre-trained predictor responds to systematic, intervention-inspired input clamping applied at inference time.\n  From these responses, we construct directed temporal influence signals that reflect the dependencies the predictor relies on for prediction, and introduce Qbic, a sparsity-aware graph selection criterion that balances predictive fidelity and structural complexity without requiring ground-truth graph labels. Experiments across synthetic, simulated, and realistic benchmarks show that Causal-INSIGHT generalizes across diverse backbone architectures, maintains competitive structural accuracy, and yields significant improvements in temporal delay localization when applied to existing predictors.", "date": "2026-03-26", "categories": []}
{"id": "2407.01111", "title": "Proximity Matters: Local Proximity Enhanced Balancing for Treatment Effect Estimation", "abstract": "Heterogeneous treatment effect (HTE) estimation from observational data poses significant challenges due to treatment selection bias. Existing methods address this bias by minimizing distribution discrepancies between treatment groups in latent space, focusing on global alignment. However, the fruitful aspect of local proximity, where similar units exhibit similar outcomes, is often overlooked. In this study, we propose Proximity-enhanced CounterFactual Regression (CFR-Pro) to exploit proximity for enhancing representation balancing within the HTE estimation context. Specifically, we introduce a pair-wise proximity regularizer based on optimal transport to incorporate the local proximity in discrepancy calculation. However, the curse of dimensionality renders the proximity measure and discrepancy estimation ineffective -- exacerbated by limited data availability for HTE estimation. To handle this problem, we further develop an informative subspace projector, which trades off minimal distance precision for improved sample complexity. Extensive experiments demonstrate that CFR-Pro accurately matches units across different treatment groups, effectively mitigates treatment selection bias, and significantly outperforms competitors. Code is available at https://github.com/HowardZJU/CFR-Pro.", "date": "2024-07-01", "categories": []}
{"id": "2510.08294", "title": "Counterfactual Identifiability via Dynamic Optimal Transport", "abstract": "We address the open question of counterfactual identification for high-dimensional multivariate outcomes from observational data. Pearl (2000) argues that counterfactuals must be identifiable (i.e., recoverable from the observed data distribution) to justify causal claims. A recent line of work on counterfactual inference shows promising results but lacks identification, undermining the causal validity of its estimates. To address this, we establish a foundation for multivariate counterfactual identification using continuous-time flows, including non-Markovian settings under standard criteria. We characterise the conditions under which flow matching yields a unique, monotone, and rank-preserving counterfactual transport map with tools from dynamic optimal transport, ensuring consistent inference. Building on this, we validate the theory in controlled scenarios with counterfactual ground-truth and demonstrate improvements in axiomatic counterfactual soundness on real images.", "date": "2025-10-09", "categories": []}
{"id": "2511.04568", "title": "Riesz Regression As Direct Density Ratio Estimation", "abstract": "This study clarifies the relationship between Riesz regression [Chernozhukov et al., 2021] and density ratio estimation (DRE) in causal inference problems, such as average treatment effect estimation. We first show that the Riesz representer can be written as a signed density ratio and then demonstrate that the Riesz regression objective coincides with the least-squares importance fitting criterion [Kanamori et al., 2009]. Although Riesz regression applies to a broad class of representer estimation problems, this equivalence with DRE allows us to transfer existing DRE results, including convergence rate analyses, generalizations based on Bregman divergence minimization, and regularization techniques for flexible models such as neural networks.", "date": "2025-11-06", "categories": []}
{"id": "2603.22714", "title": "PopResume: Causal Fairness Evaluation of LLM/VLM Resume Screeners with Population-Representative Dataset", "abstract": "We present PopResume, a population-representative resume dataset for causal fairness auditing of LLM- and VLM-based resume screening systems. Unlike existing benchmarks that rely on manually injected demographic information and outcome-level disparities, PopResume is grounded in population statistics and preserves natural attribute relationships, enabling path-specific effect (PSE)-based fairness evaluation. We decompose the effect of a protected attribute on resume scores into two paths: the business necessity path, mediated by job-relevant qualifications, and the redlining path, mediated by demographic proxies. This distinction allows auditors to separate legally permissible from impermissible sources of disparity. Evaluating four LLMs and four VLMs on PopResume's 60.8K resumes across five occupations, we identify five representative discrimination patterns that aggregate metrics fail to capture. Our results demonstrate that PSE-based evaluation reveals fairness issues masked by outcome-level measures, underscoring the need for causally-grounded auditing frameworks in AI-assisted hiring.", "date": "2026-03-24", "categories": []}
{"id": "2501.16562", "title": "C-HDNet: Hyperdimensional Computing for Causal Effect Estimation from Observational Data Under Network Interference", "abstract": "We address the problem of estimating causal effects from observational data in the presence of network confounding, a setting where both treatment assignment and observed outcomes of individuals may be influenced by their neighbors within a network structure, resulting in network interference. Traditional causal inference methods often fail to account for these dependencies, leading to biased estimates. To tackle this challenge, we introduce a novel matching-based approach that utilizes principles from hyperdimensional computing to effectively encode and incorporate structural network information. This enables more accurate identification of comparable individuals, thereby improving the reliability of causal effect estimates. Through extensive empirical evaluation on multiple benchmark datasets, we demonstrate that our method either outperforms or performs on par with existing state-of-the-art approaches, including several recent deep learning-based models that are significantly more computationally intensive. In addition to its strong empirical performance, our method offers substantial practical advantages, achieving nearly an order-of-magnitude reduction in runtime without compromising accuracy, making it particularly well-suited for large-scale or time-sensitive application", "date": "2025-01-27", "categories": []}
{"id": "2603.15182", "title": "Sequential Transport for Causal Mediation Analysis", "abstract": "We propose sequential transport (ST), a distributional framework for mediation analysis that combines optimal transport (OT) with a mediator directed acyclic graph (DAG). Instead of relying on cross-world counterfactual assumptions, ST constructs unit-level mediator counterfactuals by minimally transporting each mediator, either marginally or conditionally, toward its distribution under an alternative treatment while preserving the causal dependencies encoded by the DAG. For numerical mediators, ST uses monotone (conditional) OT maps based on conditional CDF/quantile estimators; for categorical mediators, it extends naturally via simplex-based transport. We establish consistency of the estimated transport maps and of the induced unit-level decompositions into mutatis mutandis direct and indirect effects under standard regularity and support conditions. When the treatment is randomized or ignorable (possibly conditional on covariates), these decompositions admit a causal interpretation; otherwise, they provide a principled distributional attribution of differences between groups aligned with the mediator structure. Gaussian examples show that ST recovers classical mediation formulas, while additional simulations confirm good performance in nonlinear and mixed-type settings. An application to the COMPAS dataset illustrates how ST yields deterministic, DAG-consistent counterfactual mediators and a fine-grained mediator-level attribution of disparities.", "date": "2026-03-16", "categories": []}
{"id": "2603.18404", "title": "Multi-Domain Empirical Bayes for Linearly-Mixed Causal Representations", "abstract": "Causal representation learning (CRL) aims to learn low-dimensional causal latent variables from high-dimensional observations. While identifiability has been extensively studied for CRL, estimation has been less explored. In this paper, we explore the use of empirical Bayes (EB) to estimate causal representations. In particular, we consider the problem of learning from data from multiple domains, where differences between domains are modeled by interventions in a shared underlying causal model. Multi-domain CRL naturally poses a simultaneous inference problem that EB is designed to tackle. Here, we propose an EB $f$-modeling algorithm that improves the quality of learned causal variables by exploiting invariant structure within and across domains. Specifically, we consider a linear measurement model and interventional priors arising from a shared acyclic SCM. When the graph and intervention targets are known, we develop an EM-style algorithm based on causally structured score matching. We further discuss EB $g$-modeling in the context of existing CRL approaches. In experiments on synthetic data, our proposed method achieves more accurate estimation than other methods for CRL.", "date": "2026-03-19", "categories": []}
{"id": "2603.20295", "title": "MARLIN: Multi-Agent Reinforcement Learning for Incremental DAG Discovery", "abstract": "Uncovering causal structures from observational data is crucial for understanding complex systems and making informed decisions. While reinforcement learning (RL) has shown promise in identifying these structures in the form of a directed acyclic graph (DAG), existing methods often lack efficiency, making them unsuitable for online applications. In this paper, we propose MARLIN, an efficient multi agent RL based approach for incremental DAG learning. MARLIN uses a DAG generation policy that maps a continuous real valued space to the DAG space as an intra batch strategy, then incorporates two RL agents state specific and state invariant to uncover causal relationships and integrates these agents into an incremental learning framework. Furthermore, the framework leverages a factored action space to enhance parallelization efficiency. Extensive experiments on synthetic and real datasets demonstrate that MARLIN outperforms state of the art methods in terms of both efficiency and effectiveness.", "date": "2026-03-19", "categories": []}
{"id": "2603.20585", "title": "RECLAIM: Cyclic Causal Discovery Amid Measurement Noise", "abstract": "Uncovering causal relationships is a fundamental problem across science and engineering. However, most existing causal discovery methods assume acyclicity and direct access to the system variables -- assumptions that fail to hold in many real-world settings. For instance, in genomics, cyclic regulatory networks are common, and measurements are often corrupted by instrumental noise. To address these challenges, we propose RECLAIM, a causal discovery framework that natively handles both cycles and measurement noise. RECLAIM learns the causal graph structure by maximizing the likelihood of the observed measurements via expectation-maximization (EM), using residual normalizing flows for tractable likelihood computation. We consider two measurement models: (i) Gaussian additive noise, and (ii) a linear measurement system with additive Gaussian noise. We provide theoretical consistency guarantees for both the settings. Experiments on synthetic data and real-world protein signaling datasets demonstrate the efficacy of the proposed method.", "date": "2026-03-21", "categories": []}
{"id": "2603.20791", "title": "Neural Autoregressive Flows for Markov Boundary Learning", "abstract": "Recovering Markov boundary -- the minimal set of variables that maximizes predictive performance for a response variable -- is crucial in many applications. While recent advances improve upon traditional constraint-based techniques by scoring local causal structures, they still rely on nonparametric estimators and heuristic searches, lacking theoretical guarantees for reliability. This paper investigates a framework for efficient Markov boundary discovery by integrating conditional entropy from information theory as a scoring criterion. We design a novel masked autoregressive network to capture complex dependencies. A parallelizable greedy search strategy in polynomial time is proposed, supported by analytical evidence. We also discuss how initializing a graph with learned Markov boundaries accelerates the convergence of causal discovery. Comprehensive evaluations on real-world and synthetic datasets demonstrate the scalability and superior performance of our method in both Markov boundary discovery and causal discovery tasks.", "date": "2026-03-21", "categories": []}
{"id": "2603.20842", "title": "A Knowledge-Informed Pretrained Model for Causal Discovery", "abstract": "Causal discovery has been widely studied, yet many existing methods rely on strong assumptions or fall into two extremes: either depending on costly interventional signals or partial ground truth as strong priors, or adopting purely data driven paradigms with limited guidance, which hinders practical deployment. Motivated by real-world scenarios where only coarse domain knowledge is available, we propose a knowledge-informed pretrained model for causal discovery that integrates weak prior knowledge as a principled middle ground. Our model adopts a dual source encoder-decoder architecture to process observational data in a knowledge-informed way. We design a diverse pretraining dataset and a curriculum learning strategy that smoothly adapts the model to varying prior strengths across mechanisms, graph densities, and variable scales. Extensive experiments on in-distribution, out-of distribution, and real-world datasets demonstrate consistent improvements over existing baselines, with strong robustness and practical applicability.", "date": "2026-03-21", "categories": []}
{"id": "2603.21104", "title": "CounterScene: Counterfactual Causal Reasoning in Generative World Models for Safety-Critical Closed-Loop Evaluation", "abstract": "Generating safety-critical driving scenarios requires understanding why dangerous interactions arise, rather than merely forcing collisions. However, existing methods rely on heuristic adversarial agent selection and unstructured perturbations, lacking explicit modeling of interaction dependencies and thus exhibiting a realism--adversarial trade-off. We present CounterScene, a framework that endows closed-loop generative BEV world models with structured counterfactual reasoning for safety-critical scenario generation. Given a safe scene, CounterScene asks: what if the causally critical agent had behaved differently? To answer this, we introduce causal adversarial agent identification to identify the critical agent and classify conflict types, and develop a conflict-aware interactive world model in which a causal interaction graph is used to explicitly model dynamic inter-agent dependencies. Building on this structure, stage-adaptive counterfactual guidance performs minimal interventions on the identified agent, removing its spatial and temporal safety margins while allowing risk to emerge through natural interaction propagation. Extensive experiments on nuScenes demonstrate that CounterScene achieves the strongest adversarial effectiveness while maintaining superior trajectory realism across all horizons, improving long-horizon collision rate from 12.3% to 22.7% over the strongest baseline with better realism (ADE 1.88 vs.2.09). Notably, this advantage further widens over longer rollouts, and CounterScene generalizes zero-shot to nuPlan with state-of-the-art realism.", "date": "2026-03-22", "categories": []}
{"id": "2603.21844", "title": "On the Number of Conditional Independence Tests in Constraint-based Causal Discovery", "abstract": "Learning causal relations from observational data is a fundamental problem with wide-ranging applications across many fields. Constraint-based methods infer the underlying causal structure by performing conditional independence tests. However, existing algorithms such as the prominent PC algorithm need to perform a large number of independence tests, which in the worst case is exponential in the maximum degree of the causal graph. Despite extensive research, it remains unclear if there exist algorithms with better complexity without additional assumptions. Here, we establish an algorithm that achieves a better complexity of $p^{\\mathcal{O}(s)}$ tests, where $p$ is the number of nodes in the graph and $s$ denotes the maximum undirected clique size of the underlying essential graph. Complementing this result, we prove that any constraint-based algorithm must perform at least $2^{\\Omega(s)}$ conditional independence tests, establishing that our proposed algorithm achieves exponent-optimality up to a logarithmic factor in terms of the number of conditional independence tests needed. Finally, we validate our theoretical findings through simulations, on semi-synthetic gene-expression data, and real-world data, demonstrating the efficiency of our algorithm compared to existing methods in terms of number of conditional independence tests needed.", "date": "2026-03-23", "categories": []}
{"id": "2406.15189", "title": "Causal Learning in Biomedical Applications: Krebs Cycle as a Benchmark", "abstract": "Learning causal relationships from time series data is an important but challenging problem. Existing synthetic datasets often contain hidden artifacts that can be exploited by causal discovery methods, reducing their usefulness for benchmarking. We present a new benchmark dataset based on simulations of the Krebs cycle, a key biochemical pathway. The data are generated using a particle-based simulator that models molecular interactions in a controlled environment. Four distinct scenarios are provided, varying in time series length, number of samples, and intervention settings. The benchmark includes ground-truth causal graphs for evaluation. It supports quantitative comparisons using metrics such as Structural Hamming Distance, Structural Intervention Distance, and F1-score. A comprehensive evaluation of 14 causal discovery methods from different modelling paradigms is presented. Performance is compared across datasets using multiple accuracy and efficiency measures. The dataset provides a reproducible, interpretable, and non-trivial benchmark for testing causal learning methods on time series data. It avoids common pitfalls such as residual structural patterns and supports interventions and evaluation with known causal ground truth. This makes it a useful tool for the development and comparison of causal discovery algorithms.", "date": "2024-06-21", "categories": []}
{"id": "2603.20184", "title": "Kolmogorov-Arnold causal generative models", "abstract": "Causal generative models provide a principled framework for answering observational, interventional, and counterfactual queries from observational data. However, many deep causal models rely on highly expressive architectures with opaque mechanisms, limiting auditability in high-stakes domains. We propose KaCGM, a causal generative model for mixed-type tabular data where each structural equation is parameterized by a Kolmogorov--Arnold Network (KAN). This decomposition enables direct inspection of learned causal mechanisms, including symbolic approximations and visualization of parent--child relationships, while preserving query-agnostic generative semantics. We introduce a validation pipeline based on distributional matching and independence diagnostics of inferred exogenous variables, allowing assessment using observational data alone. Experiments on synthetic and semi-synthetic benchmarks show competitive performance against state-of-the-art methods. A real-world cardiovascular case study further demonstrates the extraction of simplified structural equations and interpretable causal effects. These results suggest that expressive causal generative modeling and functional transparency can be achieved jointly, supporting trustworthy deployment in tabular decision-making settings. Code: https://github.com/aalmodovares/kacgm", "date": "2026-03-20", "categories": []}
{"id": "2312.03871", "title": "Hidden yet quantifiable: A lower bound for confounding strength using randomized trials", "abstract": "In the era of fast-paced precision medicine, observational studies play a major role in properly evaluating new treatments in clinical practice. Yet, unobserved confounding can significantly compromise causal conclusions drawn from non-randomized data. We propose a novel strategy that leverages randomized trials to quantify unobserved confounding. First, we design a statistical test to detect unobserved confounding with strength above a given threshold. Then, we use the test to estimate an asymptotically valid lower bound on the unobserved confounding strength. We evaluate the power and validity of our statistical test on several synthetic and semi-synthetic datasets. Further, we show how our lower bound can correctly identify the absence and presence of unobserved confounding in a real-world setting.", "date": "2023-12-06", "categories": []}
{"id": "2402.01972", "title": "Combining T-learning and DR-learning: a framework for oracle-efficient estimation of causal contrasts", "abstract": "We introduce efficient plug-in (EP) learning, a novel framework for the estimation of heterogeneous causal contrasts, such as the conditional average treatment effect and conditional relative risk. The EP-learning framework enjoys the same oracle efficiency as Neyman-orthogonal learning strategies, such as DR-learning and R-learning, while addressing some of their primary drawbacks: (i) their practical applicability can be hindered by non-convex loss functions; and (ii) they may suffer from poor performance and instability due to inverse probability weighting and pseudo-outcomes that violate bounds. To overcome these issues, the EP-learner leverages an efficient plug-in estimator of the population risk function for the causal contrast. In doing so, it inherits the stability of plug-in strategies such as T-learning, while improving on their efficiency. Under reasonable conditions, EP-learners based on empirical risk minimization are oracle-efficient, exhibiting asymptotic equivalence to the minimizer of an oracle-efficient one-step debiased estimator of the population risk function. In simulation experiments, we show that EP-learners of the conditional average treatment effect and conditional relative risk outperform state-of-the-art competitors, including the T-learner, R-learner, and DR-learner. Open-source implementations of the proposed methods are available in our \\texttt{R} package \\texttt{hte3}.", "date": "2024-02-03", "categories": []}
{"id": "2509.03636", "title": "CausalARC: Abstract Reasoning with Causal World Models", "abstract": "On-the-fly reasoning often requires adaptation to novel problems under limited data and distribution shift. This work introduces CausalARC: an experimental testbed for AI reasoning in low-data and out-of-distribution regimes, modeled after the Abstraction and Reasoning Corpus (ARC). Each CausalARC reasoning task is sampled from a fully specified causal world model, formally expressed as a structural causal model. Principled data augmentations provide observational, interventional, and counterfactual feedback about the world model in the form of few-shot, in-context learning demonstrations. As a proof-of-concept, we illustrate the use of CausalARC for four language model evaluation settings: (1) abstract reasoning with test-time training, (2) counterfactual reasoning with in-context learning, (3) program synthesis, and (4) causal discovery with logical reasoning. Within- and between-model performance varied heavily across tasks, indicating room for significant improvement in language model reasoning.", "date": "2025-09-03", "categories": []}
{"id": "2603.18736", "title": "CausalRM: Causal-Theoretic Reward Modeling for RLHF from Observational User Feedbacks", "abstract": "Despite the success of reinforcement learning from human feedback (RLHF) in aligning language models, current reward modeling heavily relies on experimental feedback data collected from human annotators under controlled and costly conditions. In this work, we introduce observational reward modeling -- learning reward models with observational user feedback (e.g., clicks, copies, and upvotes) -- as a scalable and cost-effective alternative. We identify two fundamental challenges in this setting: (1) observational feedback is noisy due to annotation errors, which deviates it from true user preference; (2) observational feedback is biased by user preference, where users preferentially provide feedback on responses they feel strongly about, which creats a distribution shift between training and inference data. To address these challenges, we propose CausalRM, a causal-theoretic reward modeling framework that aims to learn unbiased reward models from observational feedback. To tackle challenge (1), CausalRM introduces a noise-aware surrogate loss term that is provably equivalent to the primal loss under noise-free conditions by explicitly modeling the annotation error generation process. To tackle challenge (2), CausalRM uses propensity scores -- the probability of a user providing feedback for a given response -- to reweight training samples, yielding a loss function that eliminates user preference bias. Extensive experiments across diverse LLM backbones and benchmark datasets validate that CausalRM effectively learns accurate reward signals from noisy and biased observational feedback and delivers substantial performance improvements on downstream RLHF tasks -- including a 49.2% gain on WildGuardMix and a 32.7% improvement on HarmBench. Code is available on our project website.", "date": "2026-03-19", "categories": []}
{"id": "2603.18968", "title": "Teleological Inference in Structural Causal Models via Intentional Interventions", "abstract": "Structural causal models (SCMs) were conceived to formulate and answer causal questions. This paper shows that SCMs can also be used to formulate and answer teleological questions, concerning the intentions of a state-aware, goal-directed agent intervening in a causal system. We review limitations of previous approaches to modeling such agents, and then introduce intentional interventions, a new time-agnostic operator that induces a twin SCM we call a structural final model (SFM). SFMs treat observed values as the outcome of intentional interventions and relate them to the counterfactual conditions of those interventions (what would have happened had the agent not intervened). We show how SFMs can be used to empirically detect agents and to discover their intentions.", "date": "2026-03-19", "categories": []}
{"id": "2405.16924", "title": "Demystifying amortized causal discovery with transformers", "abstract": "Supervised learning for causal discovery from observational data often achieves competitive performance despite seemingly avoiding the explicit assumptions that traditional methods require for identifiability. In this work, we analyze CSIvA (Ke et al., 2023) on bivariate causal models, a transformer architecture for amortized inference promising to train on synthetic data and transfer to real ones. First, we bridge the gap with identifiability theory, showing that the training distribution implicitly defines a prior on the causal model of the test observations: consistent with classical approaches, good performance is achieved when we have a good prior on the test data, and the underlying model is identifiable. Second, we find that CSIvA can not generalize to classes of causal models unseen during training: to overcome this limitation, we theoretically and empirically analyze \\textit{when} training CSIvA on datasets generated by multiple identifiable causal models with different structural assumptions improves its generalization at test time. Overall, we find that amortized causal discovery with transformers still adheres to identifiability theory, violating the previous hypothesis from Lopez-Paz et al. (2015) that supervised learning methods could overcome its restrictions.", "date": "2024-05-27", "categories": []}
{"id": "2501.05007", "title": "Quantum-enhanced causal discovery for a small number of samples", "abstract": "The discovery of causal relations from observed data has attracted significant interest from disciplines such as economics, social sciences, and biology. In practical applications, considerable knowledge of the underlying systems is often unavailable, and real data are usually associated with nonlinear causal structures, which makes the direct use of most conventional causality analysis methods difficult. This study proposes a novel quantum Peter-Clark (qPC) algorithm for causal discovery that does not require any assumptions about the underlying model structures. Based on conditional independence tests in a class of reproducing kernel Hilbert spaces characterized by quantum circuits, the proposed algorithm can explore causal relations from the observed data drawn from arbitrary distributions. We conducted systematic experiments on fundamental graphs of causal structures, demonstrating that the qPC algorithm exhibits better performance, particularly with smaller sample sizes compared to its classical counterpart. Furthermore, we proposed a novel optimization approach based on Kernel Target Alignment (KTA) for determining hyperparameters of quantum kernels. This method effectively reduced the risk of false positives in causal discovery, enabling more reliable inference. Our theoretical and experimental results demonstrate that the quantum algorithm can empower classical algorithms for accurate inference in causal discovery, supporting them in regimes where classical algorithms typically fail. In addition, the effectiveness of this method was validated using the datasets on Boston housing prices, heart disease, and biological signaling systems as real-world applications. These findings highlight the potential of quantum-based causal discovery methods in addressing practical challenges, particularly in small-sample scenarios, where traditional approaches have shown significant limitations.", "date": "2025-01-09", "categories": []}
{"id": "2502.14719", "title": "How PC-based Methods Err: Towards Better Reporting of Assumption Violations and Small Sample Errors", "abstract": "Causal discovery methods based on the PC algorithm are proven to be sound if all structural assumptions are fulfilled and all conditional independence tests are correct. This idealized setting is rarely given in real data. In this work, we first analyze how local errors can propagate throughout the output graph of a PC-based method, highlighting how consequential seemingly innocuous errors can become. Next, we introduce coherency scores to find assumption violations and small sample errors in the absence of a ground truth. These scores do not require statistical tests beyond those already executed by the causal discovery algorithm. Errors detected by our approach extend the set of errors that can be detected by comparable existing methods. We place our computationally cheap global error detection and quantification scores as a bridge between computationally expensive global answer-set-programming-based methods and less expensive local error detection methods. The scores are analyzed on simulated and real-world datasets.", "date": "2025-02-20", "categories": []}
{"id": "2507.03681", "title": "Robust estimation of heterogeneous treatment effects in randomized trials leveraging external data", "abstract": "Randomized trials are typically designed to detect average treatment effects but often lack the statistical power to uncover individual-level treatment effect heterogeneity, limiting their value for personalized decision-making. To address this, we propose the QR-learner, a model-agnostic learner that estimates conditional average treatment effects (CATE) within the trial population by leveraging external data from other trials or observational studies. The proposed method is robust: it can reduce the mean squared error relative to a trial-only CATE learner, and is guaranteed to recover the true CATE even when the external data are not aligned with the trial. Moreover, we introduce a procedure that combines the QR-learner with a trial-only CATE learner and show that it asymptotically matches or exceeds both component learners in terms of mean squared error. We examine the performance of our approach in simulation studies and apply the methods to a real-world dataset, demonstrating improvements in both CATE estimation and statistical power for detecting heterogeneous effects.", "date": "2025-07-04", "categories": []}
{"id": "2602.19414", "title": "Federated Causal Representation Learning in State-Space Systems for Decentralized Counterfactual Reasoning", "abstract": "Networks of interdependent industrial assets (clients) are tightly coupled through physical processes and control inputs, raising a key question: how would the output of one client change if another client were operated differently? This is difficult to answer because client-specific data are high-dimensional and private, making centralization of raw data infeasible. Each client also maintains proprietary local models that cannot be modified. We propose a federated framework for causal representation learning in state-space systems that captures interdependencies among clients under these constraints. Each client maps high-dimensional observations into low-dimensional latent states that disentangle intrinsic dynamics from control-driven influences. A central server estimates the global state-transition and control structure. This enables decentralized counterfactual reasoning where clients predict how outputs would change under alternative control inputs at others while only exchanging compact latent states. We prove convergence to a centralized oracle and provide privacy guarantees. Our experiments demonstrate scalability, and accurate cross-client counterfactual inference on synthetic and real-world industrial control system datasets.", "date": "2026-02-23", "categories": []}
{"id": "2603.17384", "title": "Cohomological Obstructions to Global Counterfactuals: A Sheaf-Theoretic Foundation for Generative Causal Models", "abstract": "Current continuous generative models (e.g., Diffusion Models, Flow Matching) implicitly assume that locally consistent causal mechanisms naturally yield globally coherent counterfactuals. In this paper, we prove that this assumption fails fundamentally when the causal graph exhibits non-trivial homology (e.g., structural conflicts or hidden confounders). We formalize structural causal models as cellular sheaves over Wasserstein spaces, providing a strict algebraic topological definition of cohomological obstructions in measure spaces. To ensure computational tractability and avoid deterministic singularities (which we define as manifold tearing), we introduce entropic regularization and derive the Entropic Wasserstein Causal Sheaf Laplacian, a novel system of coupled non-linear Fokker-Planck equations. Crucially, we prove an entropic pullback lemma for the first variation of pushforward measures. By integrating this with the Implicit Function Theorem (IFT) on Sinkhorn optimality conditions, we establish a direct algorithmic bridge to automatic differentiation (VJP), achieving O(1)-memory reverse-mode gradients strictly independent of the iteration horizon. Empirically, our framework successfully leverages thermodynamic noise to navigate topological barriers (\"entropic tunneling\") in high-dimensional scRNA-seq counterfactuals. Finally, we invert this theoretical framework to introduce the Topological Causal Score, demonstrating that our Sheaf Laplacian acts as a highly sensitive algebraic detector for topology-aware causal discovery.", "date": "2026-03-18", "categories": []}
{"id": "2603.17385", "title": "The Causal Uncertainty Principle: Manifold Tearing and the Topological Limits of Counterfactual Interventions", "abstract": "Judea Pearl's do-calculus provides a foundation for causal inference, but its translation to continuous generative models remains fraught with geometric challenges. We establish the fundamental limits of such interventions. We define the Counterfactual Event Horizon and prove the Manifold Tearing Theorem: deterministic flows inevitably develop finite-time singularities under extreme interventions. We establish the Causal Uncertainty Principle for the trade-off between intervention extremity and identity preservation. Finally, we introduce Geometry-Aware Causal Flow (GACF), a scalable algorithm that utilizes a topological radar to bypass manifold tearing, validated on high-dimensional scRNA-seq data.", "date": "2026-03-18", "categories": []}
{"id": "2603.17405", "title": "Causal Representation Learning on High-Dimensional Data: Benchmarks, Reproducibility, and Evaluation Metrics", "abstract": "Causal representation learning (CRL) models aim to transform high-dimensional data into a latent space, enabling interventions to generate counterfactual samples or modify existing data based on the causal relationships among latent variables. To facilitate the development and evaluation of these models, a variety of synthetic and real-world datasets have been proposed, each with distinct advantages and limitations. For practical applications, CRL models must perform robustly across multiple evaluation directions, including reconstruction, disentanglement, causal discovery, and counterfactual reasoning, using appropriate metrics for each direction. However, this multi-directional evaluation can complicate model comparison, as a model may excel in some direction while under-performing in others. Another significant challenge in this field is reproducibility: the source code corresponding to published results must be publicly available, and repeated runs should yield performance consistent with the original reports. In this study, we critically analyzed the synthetic and real-world datasets currently employed in the literature, highlighting their limitations and proposing a set of essential characteristics for suitable datasets in CRL model development. We also introduce a single aggregate metric that consolidates performance across all evaluation directions, providing a comprehensive score for each model. Finally, we reviewed existing implementations from the literature and assessed them in terms of reproducibility, identifying gaps and best practices in the field.", "date": "2026-03-18", "categories": []}
{"id": "2603.17722", "title": "Predicting Trajectories of Long COVID in Adult Women: The Critical Role of Causal Disentanglement", "abstract": "Early prediction of Post-Acute Sequelae of SARS-CoV-2 severity is a critical challenge for women's health, particularly given the diagnostic overlap between PASC and common hormonal transitions such as menopause. Identifying and accounting for these confounding factors is essential for accurate long-term trajectory prediction. We conducted a retrospective study of 1,155 women (mean age 61) from the NIH RECOVER dataset. By integrating static clinical profiles with four weeks of longitudinal wearable data (monitoring cardiac activity and sleep), we developed a causal network based on a Large Language Model to predict future PASC scores. Our framework achieved a precision of 86.7\\% in clinical severity prediction. Our causal attribution analysis demonstrate the model's ability to differentiate between active pathology and baseline noise: direct indicators such as breathlessness and malaise reached maximum saliency (1.00), while confounding factors like menopause and diabetes were successfully suppressed with saliency scores below 0.27.", "date": "2026-03-18", "categories": []}
{"id": "2603.15926", "title": "Evaluating Causal Discovery Algorithms for Path-Specific Fairness and Utility in Healthcare", "abstract": "Causal discovery in health data faces evaluation challenges when ground truth is unknown. We address this by collaborating with experts to construct proxy ground-truth graphs, establishing benchmarks for synthetic Alzheimer's disease and heart failure clinical records data. We evaluate the Peter-Clark, Greedy Equivalence Search, and Fast Causal Inference algorithms on structural recovery and path-specific fairness decomposition, going beyond composite fairness scores. On synthetic data, Peter-Clark achieved the best structural recovery. On heart failure data, Fast Causal Inference achieved the highest utility. For path-specific effects, ejection fraction contributed 3.37 percentage points to the indirect effect in the ground truth. These differences drove variations in the fairness-utility ratio across algorithms. Our results highlight the need for graph-aware fairness evaluation and fine-grained path-specific analysis when deploying causal discovery in clinical applications.", "date": "2026-03-16", "categories": []}
{"id": "2603.16436", "title": "DISCOVER: A Solver for Distributional Counterfactual Explanations", "abstract": "Counterfactual explanations (CE) explain model decisions by identifying input modifications that lead to different predictions. Most existing methods operate at the instance level. Distributional Counterfactual Explanations (DCE) extend this setting by optimizing an optimal transport objective that balances proximity to a factual input distribution and alignment to a target output distribution, with statistical certification via chance constrained bounds. However, DCE relies on gradient based optimization, while many real-world tabular pipelines are dominated by non-differentiable models. We propose DISCOVER, a model-agnostic solver for distributional counterfactual explanations. DISCOVER preserves the original DCE objective and certification while replacing gradient descent with a sparse propose-and-select search paradigm. It exploits a sample-wise decomposition of the transport objective to compute per-row impact scores and enforce a top-$k$ intervention budget, focusing edits on the most influential samples. To guide candidate generation without predictor gradients, DISCOVER introduces an OT-guided cone sampling primitive driven by input-side transport geometry. Experiments on multiple tabular datasets demonstrate strong joint alignment of input and output distributions, extending distributional counterfactual reasoning to modern black box learning pipelines. A code repository is available at https://github.com/understanding-ml/DCE.", "date": "2026-03-17", "categories": []}
{"id": "2603.16829", "title": "Conditional Distributional Treatment Effects: Doubly Robust Estimation and Testing", "abstract": "Beyond conditional average treatment effects, treatments may impact the entire outcome distribution in covariate-dependent ways, for example, by altering the variance or tail risks for specific subpopulations. We propose a novel estimand to capture such conditional distributional treatment effects, and develop a doubly robust estimator that is minimax optimal in the local asymptotic sense. Using this, we develop a test for the global homogeneity of conditional potential outcome distributions that accommodates discrepancies beyond the maximum mean discrepancy (MMD), has provably valid type 1 error, and is consistent against fixed alternatives -- the first test, to our knowledge, with such guarantees in this setting. Furthermore, we derive exact closed-form expressions for two natural discrepancies (including the MMD), and provide a computationally efficient, permutation-free algorithm for our test.", "date": "2026-03-17", "categories": []}
{"id": "2410.21328", "title": "Deconfounded Time Series Forecasting: A Causal Inference Approach", "abstract": "Time series forecasting is a critical task in various domains, where accurate predictions can drive informed decision-making. Traditional forecasting methods often rely on current observations of variables to predict future outcomes, typically overlooking the influence of latent confounders, unobserved variables that simultaneously affect both the predictors and the target outcomes. This oversight can introduce bias and degrade the performance of predictive models. In this study, we address this challenge by proposing an enhanced forecasting approach that incorporates representations of latent confounders derived from historical data. By integrating these confounders into the predictive process, our method aims to improve the accuracy and robustness of time series forecasts. The proposed approach is demonstrated through its application to climate science data, showing significant improvements over traditional methods that do not account for confounders.", "date": "2024-10-27", "categories": []}
{"id": "2410.21474", "title": "Estimating Causal Effects of Text Interventions Leveraging LLMs", "abstract": "Quantifying the effects of textual interventions in social systems, such as reducing anger in social media posts to see its impact on engagement, is challenging. Real-world interventions are often infeasible, necessitating reliance on observational data. Traditional causal inference methods, typically designed for binary or discrete treatments, are inadequate for handling the complex, high-dimensional textual data. This paper addresses these challenges by proposing CausalDANN, a novel approach to estimate causal effects using text transformations facilitated by large language models (LLMs). Unlike existing methods, our approach accommodates arbitrary textual interventions and leverages text-level classifiers with domain adaptation ability to produce robust effect estimates against domain shifts, even when only the control group is observed. This flexibility in handling various text interventions is a key advancement in causal estimation for textual data, offering opportunities to better understand human behaviors and develop effective interventions within social systems.", "date": "2024-10-28", "categories": []}
{"id": "2510.01112", "title": "The causal structure of galactic astrophysics", "abstract": "Data-driven astrophysics currently relies on the detection and characterisation of correlations between objects' properties, which are then used to test physical theories that make predictions for them. This process fails to utilise information in the data that forms a crucial part of the theories' predictions, namely which variables are directly correlated (as opposed to accidentally correlated through others), the directions of these determinations, and the presence or absence of confounders that correlate variables in the dataset but are themselves absent from it. We propose to recover this information through causal discovery, a well-developed methodology for inferring the causal structure of datasets that is however almost entirely unknown to astrophysics. We develop a causal discovery algorithm suitable for large astrophysical datasets and illustrate it on $\\sim$4.5$\\times10^5$ nearby galaxies from the Nasa Sloan Atlas, demonstrating its ability to distinguish physical mechanisms that are degenerate on the basis of correlations alone.", "date": "2025-10-01", "categories": []}
{"id": "2603.07431", "title": "OrthoFormer: Instrumental Variable Estimation in Transformer Hidden States via Neural Control Functions", "abstract": "Transformer architectures excel at sequential modeling yet remain fundamentally limited by correlational learning - they capture spurious associations induced by latent confounders rather than invariant causal mechanisms. We identify this as an epistemological challenge: standard Transformers conflate static background factors (intrinsic identity, style, context) with dynamic causal flows (state evolution, mechanism), leading to catastrophic out-of-distribution failure. We propose OrthoFormer, a causally grounded architecture that embeds instrumental variable estimation directly into Transformer blocks via neural control functions. Our framework rests on four theoretical pillars: Structural Directionality (time-arrow enforcement), Representation Orthogonality (latent-noise separation), Causal Sparsity (Markov Blanket approximation), and End-to-End Consistency (gradient- detached stage separation). We prove that OrthoFormer achieves bias strictly less than OLS for any valid instrument lag, with residual bias decaying geometrically as O(\\r{ho}k ). We characterize the bias-variance-exogeneity trilemma inherent in self-instrumenting and identify the neural forbidden regression - where removing gradient detachment improves prediction loss while destroying causal validity. Experiments confirm all theoretical predictions. OrthoFormer represents a paradigm shift from correlational to causal sequence modeling, with implications for robustness, interpretability, and reliable decision-making under distribution shift.", "date": "2026-03-08", "categories": []}
{"id": "2603.08682", "title": "Structural Causal Bottleneck Models", "abstract": "We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models. At the core of SCBMs lies the assumption that causal effects between high-dimensional variables only depend on low-dimensional summary statistics, or bottlenecks, of the causes. SCBMs provide a flexible framework for task-specific dimension reduction while being estimable via standard, simple learning algorithms in practice. We analyse identifiability in SCBMs, connect them to information bottlenecks in the sense of Tishby & Zaslavsky (2015), and illustrate how to estimate them experimentally. We also demonstrate the benefit of bottlenecks for effect estimation in low-sample transfer learning settings. We argue that SCBMs provide an alternative to existing causal dimension reduction frameworks like causal representation learning or causal abstraction learning.", "date": "2026-03-09", "categories": []}
{"id": "2603.09288", "title": "Proxy-Guided Measurement Calibration", "abstract": "Aggregate outcome variables collected through surveys and administrative records are often subject to systematic measurement error. For instance, in disaster loss databases, county-level losses reported may differ from the true damages due to variations in on-the-ground data collection capacity, reporting practices, and event characteristics. Such miscalibration complicates downstream analysis and decision-making. We study the problem of outcome miscalibration and propose a framework guided by proxy variables for estimating and correcting the systematic errors. We model the data-generating process using a causal graph that separates latent content variables driving the true outcome from the latent bias variables that induce systematic errors. The key insight is that proxy variables that depend on the true outcome but are independent of the bias mechanism provide identifying information for quantifying the bias. Leveraging this structure, we introduce a two-stage approach that utilizes variational autoencoders to disentangle content and bias latents, enabling us to estimate the effect of bias on the outcome of interest. We analyze the assumptions underlying our approach and evaluate it on synthetic data, semi-synthetic datasets derived from randomized trials, and a real-world case study of disaster loss reporting.", "date": "2026-03-10", "categories": []}
{"id": "2603.13595", "title": "A Causal Framework for Mitigating Data Shifts in Healthcare", "abstract": "Developing predictive models that perform reliably across diverse patient populations and heterogeneous environments is a core aim of medical research. However, generalization is only possible if the learned model is robust to statistical differences between data used for training and data seen at the time and place of deployment. Domain generalization methods provide strategies to address data shifts, but each method comes with its own set of assumptions and trade-offs. To apply these methods in healthcare, we must understand how domain shifts arise, what assumptions we prefer to make, and what our design constraints are. This article proposes a causal framework for the design of predictive models to improve generalization. Causality provides a powerful language to characterize and understand diverse domain shifts, regardless of data modality. This allows us to pinpoint why models fail to generalize, leading to more principled strategies to prepare for and adapt to shifts. We recommend general mitigation strategies, discussing trade-offs and highlighting existing work. Our causality-based perspective offers a critical foundation for developing robust, interpretable, and clinically relevant AI solutions in healthcare, paving the way for reliable real-world deployment.", "date": "2026-03-13", "categories": []}
{"id": "2603.15335", "title": "Data Augmentation via Causal-Residual Bootstrapping", "abstract": "Data augmentation integrates domain knowledge into a dataset by making domain-informed modifications to existing data points. For example, image data can be augmented by duplicating images in different tints or orientations, thereby incorporating the knowledge that images may vary in these dimensions. Recent work by Teshima and Sugiyama has explored the integration of causal knowledge (e.g, A causes B causes C) up to conditional independence equivalence. We suggest a related approach for settings with additive noise that can incorporate information beyond a Markov equivalence class. The approach, built on the principle of independent mechanisms, permutes the residuals of models built on marginal probability distributions. Predictive models built on our augmented data demonstrate improved accuracy, for which we provide theoretical backing in linear Gaussian settings.", "date": "2026-03-16", "categories": []}
{"id": "2603.15542", "title": "InterveneBench: Benchmarking LLMs for Intervention Reasoning and Causal Study Design in Real Social Systems", "abstract": "Causal inference in social science relies on end-to-end, intervention-centered research-design reasoning grounded in real-world policy interventions, but current benchmarks fail to evaluate this capability of large language models (LLMs). We present InterveneBench, a benchmark designed to assess such reasoning in realistic social settings. Each instance in InterveneBench is derived from an empirical social science study and requires models to reason about policy interventions and identification assumptions without access to predefined causal graphs or structural equations. InterveneBench comprises 744 peer-reviewed studies across diverse policy domains. Experimental results show that state-of-the-art LLMs struggle under this setting. To address this limitation, we further propose a multi-agent framework, STRIDES. It achieves significant performance improvements over state-of-the-art reasoning models. Our code and data are available at https://github.com/Sii-yuning/STRIDES.", "date": "2026-03-16", "categories": []}
{"id": "2208.13701", "title": "Data-Driven Influence Functions for Optimization-Based Causal Inference", "abstract": "We study a constructive algorithm that approximates Gateaux derivatives for statistical functionals by finite differencing, with a focus on functionals that arise in causal inference. We study the case where probability distributions are not known a priori but need to be estimated from data. These estimated distributions lead to empirical Gateaux derivatives, and we study the relationships between empirical, numerical, and analytical Gateaux derivatives. Starting with a case study of the interventional mean (average potential outcome), we delineate the relationship between finite differences and the analytical Gateaux derivative. We then derive requirements on the rates of numerical approximation in perturbation and smoothing that preserve the statistical benefits of one-step adjustments, such as rate double robustness. We then study more complicated functionals such as dynamic treatment regimes, the linear-programming formulation for policy optimization in infinite-horizon Markov decision processes, and sensitivity analysis in causal inference. More broadly, we study optimization-based estimators, since this begets a class of estimands where identification via regression adjustment is straightforward but obtaining influence functions under minor variations thereof is not. The ability to approximate bias adjustments in the presence of arbitrary constraints illustrates the usefulness of constructive approaches for Gateaux derivatives. We also find that the statistical structure of the functional (rate double robustness) can permit less conservative rates for finite-difference approximation. This property, however, can be specific to particular functionals; e.g., it occurs for the average potential outcome (hence average treatment effect) but not the infinite-horizon MDP policy value.", "date": "2022-08-29", "categories": []}
{"id": "2508.21742", "title": "Orientability of Causal Relations in Time Series using Summary Causal Graphs and Faithful Distributions", "abstract": "Understanding causal relations between temporal variables is a central challenge in time series analysis, particularly when the full causal structure is unknown. Even when the full causal structure cannot be fully specified, experts often succeed in providing a high-level abstraction of the causal graph, known as a summary causal graph, which captures the main causal relations between different time series while abstracting away micro-level details. In this work, we present conditions that guarantee the orientability of micro-level edges between temporal variables given the background knowledge encoded in a summary causal graph and assuming having access to a faithful and causally sufficient distribution with respect to the true unknown graph. Our results provide theoretical guarantees for edge orientation at the micro-level, even in the presence of cycles or bidirected edges at the macro-level. These findings offer practical guidance for leveraging SCGs to inform causal discovery in complex temporal systems and highlight the value of incorporating expert knowledge to improve causal inference from observational time series data.", "date": "2025-08-29", "categories": []}
{"id": "2603.12305", "title": "HCP-DCNet: A Hierarchical Causal Primitive Dynamic Composition Network for Self-Improving Causal Understanding", "abstract": "The ability to understand and reason about cause and effect -- encompassing interventions, counterfactuals, and underlying mechanisms -- is a cornerstone of robust artificial intelligence. While deep learning excels at pattern recognition, it fundamentally lacks a model of causality, making systems brittle under distribution shifts and unable to answer ``what-if'' questions. This paper introduces the \\emph{Hierarchical Causal Primitive Dynamic Composition Network (HCP-DCNet)}, a unified framework that bridges continuous physical dynamics with discrete symbolic causal inference. Departing from monolithic representations, HCP-DCNet decomposes causal scenes into reusable, typed \\emph{causal primitives} organized into four abstraction layers: physical, functional, event, and rule. A dual-channel routing network dynamically composes these primitives into task-specific, fully differentiable \\emph{Causal Execution Graphs (CEGs)}. Crucially, the system employs a \\emph{causal-intervention-driven meta-evolution} strategy, enabling autonomous self-improvement through a constrained Markov decision process. We establish rigorous theoretical guarantees, including type-safe composition, routing convergence, and universal approximation of causal dynamics. Extensive experiments across simulated physical and social environments demonstrate that HCP-DCNet significantly outperforms state-of-the-art baselines in causal discovery, counterfactual reasoning, and compositional generalization. This work provides a principled, scalable, and interpretable architecture for building AI systems with human-like causal abstraction and continual self-refinement capabilities.", "date": "2026-03-12", "categories": []}
{"id": "2311.11321", "title": "Bounds on Representation-Induced Confounding Bias for Treatment Effect Estimation", "abstract": "State-of-the-art methods for conditional average treatment effect (CATE) estimation make widespread use of representation learning. Here, the idea is to reduce the variance of the low-sample CATE estimation by a (potentially constrained) low-dimensional representation. However, low-dimensional representations can lose information about the observed confounders and thus lead to bias, because of which the validity of representation learning for CATE estimation is typically violated. In this paper, we propose a new, representation-agnostic refutation framework for estimating bounds on the representation-induced confounding bias that comes from dimensionality reduction (or other constraints on the representations) in CATE estimation. First, we establish theoretically under which conditions CATE is non-identifiable given low-dimensional (constrained) representations. Second, as our remedy, we propose a neural refutation framework which performs partial identification of CATE or, equivalently, aims at estimating lower and upper bounds of the representation-induced confounding bias. We demonstrate the effectiveness of our bounds in a series of experiments. In sum, our refutation framework is of direct relevance in practice where the validity of CATE estimation is of importance.", "date": "2023-11-19", "categories": []}
{"id": "2411.03387", "title": "Quantifying Aleatoric Uncertainty of the Treatment Effect: A Novel Orthogonal Learner", "abstract": "Estimating causal quantities from observational data is crucial for understanding the safety and effectiveness of medical treatments. However, to make reliable inferences, medical practitioners require not only estimating averaged causal quantities, such as the conditional average treatment effect, but also understanding the randomness of the treatment effect as a random variable. This randomness is referred to as aleatoric uncertainty and is necessary for understanding the probability of benefit from treatment or quantiles of the treatment effect. Yet, the aleatoric uncertainty of the treatment effect has received surprisingly little attention in the causal machine learning community. To fill this gap, we aim to quantify the aleatoric uncertainty of the treatment effect at the covariate-conditional level, namely, the conditional distribution of the treatment effect (CDTE). Unlike average causal quantities, the CDTE is not point identifiable without strong additional assumptions. As a remedy, we employ partial identification to obtain sharp bounds on the CDTE and thereby quantify the aleatoric uncertainty of the treatment effect. We then develop a novel, orthogonal learner for the bounds on the CDTE, which we call AU-learner. We further show that our AU-learner has several strengths in that it satisfies Neyman-orthogonality and, thus, quasi-oracle efficiency. Finally, we propose a fully-parametric deep learning instantiation of our AU-learner.", "date": "2024-11-05", "categories": []}
{"id": "2411.12184", "title": "Testability of Instrumental Variables in Additive Nonlinear, Non-Constant Effects Models", "abstract": "We address the issue of the testability of instrumental variables derived from observational data. Most existing testable implications are centered on scenarios where the treatment is a discrete variable, e.g., instrumental inequality (Pearl, 1995), or where the effect is assumed to be constant, e.g., instrumental variables condition based on the principle of independent mechanisms (Burauel, 2023). However, treatments can often be continuous variables, such as drug dosages or nutritional content levels, and non-constant effects may occur in many real-world scenarios. In this paper, we consider an additive nonlinear, non-constant effects model with unmeasured confounders, in which treatments can be either discrete or continuous, and propose an Auxiliary-based Independence Test (AIT) condition to test whether a variable is a valid instrument. We first show that, under the completeness condition, if the candidate instrument is valid, then the AIT condition holds. Moreover, we illustrate the implications of the AIT condition and demonstrate that, under certain additional conditions, the AIT condition is necessary and sufficient to detect all invalid IVs. We also extend the AIT condition to include covariates and introduce a practical testing algorithm. Experimental results on both synthetic and three different real-world datasets show the effectiveness of our proposed condition.", "date": "2024-11-19", "categories": []}
{"id": "2601.15479", "title": "Benchmarking LLMs for Pairwise Causal Discovery in Biomedical and Multi-Domain Contexts", "abstract": "The safe deployment of large language models (LLMs) in high-stakes fields like biomedicine, requires them to be able to reason about cause and effect. We investigate this ability by testing 13 open-source LLMs on a fundamental task: pairwise causal discovery (PCD) from text. Our benchmark, using 12 diverse datasets, evaluates two core skills: 1) \\textbf{Causal Detection} (identifying if a text contains a causal link) and 2) \\textbf{Causal Extraction} (pulling out the exact cause and effect phrases). We tested various prompting methods, from simple instructions (zero-shot) to more complex strategies like Chain-of-Thought (CoT) and Few-shot In-Context Learning (FICL).\n  The results show major deficiencies in current models. The best model for detection, DeepSeek-R1-Distill-Llama-70B, only achieved a mean score of 49.57\\% ($C_{detect}$), while the best for extraction, Qwen2.5-Coder-32B-Instruct, reached just 47.12\\% ($C_{extract}$). Models performed best on simple, explicit, single-sentence relations. However, performance plummeted for more difficult (and realistic) cases, such as implicit relationships, links spanning multiple sentences, and texts containing multiple causal pairs. We provide a unified evaluation framework, built on a dataset validated with high inter-annotator agreement ($\\kappa \\ge 0.758$), and make all our data, code, and prompts publicly available to spur further research. \\href{https://github.com/sydneyanuyah/CausalDiscovery}{Code available here: https://github.com/sydneyanuyah/CausalDiscovery}", "date": "2026-01-21", "categories": []}
{"id": "2603.11546", "title": "Multi-Task Anti-Causal Learning for Reconstructing Urban Events from Residents' Reports", "abstract": "Many real-world machine learning tasks are anti-causal: they require inferring latent causes from observed effects. In practice, we often face multiple related tasks where part of the forward causal mechanism is invariant across tasks, while other components are task-specific. We propose Multi-Task Anti-Causal learning (MTAC), a framework for estimating causes from outcomes and confounders by explicitly exploiting such cross-task invariances. MTAC first performs causal discovery to learn a shared causal graph and then instantiates a structured multi-task structural equation model (SEM) that factorizes the outcome-generation process into (i) a task-invariant mechanism and (ii) task-specific mechanisms via a shared backbone with task-specific heads. Building on the learned forward model, MTAC performs maximum A posteriori (MAP)based inference to reconstruct causes by jointly optimizing latent mechanism variables and cause magnitudes under the learned causal structure. We evaluate MTAC on the application of urban event reconstruction from resident reports, spanning three tasks:parking violations, abandoned properties, and unsanitary conditions. On real-world data collected from Manhattan and the city of Newark, MTAC consistently improves reconstruction accuracy over strong baselines, achieving up to 34.61\\% MAE reduction and demonstrating the benefit of learning transferable causal mechanisms across tasks.", "date": "2026-03-12", "categories": []}
{"id": "2603.11565", "title": "CAETC: Causal Autoencoding and Treatment Conditioning for Counterfactual Estimation over Time", "abstract": "Counterfactual estimation over time is important in various applications, such as personalized medicine. However, time-dependent confounding bias in observational data still poses a significant challenge in achieving accurate and efficient estimation. We introduce causal autoencoding and treatment conditioning (CAETC), a novel method for this problem. Built on adversarial representation learning, our method leverages an autoencoding architecture to learn a partially invertible and treatment-invariant representation, where the outcome prediction task is cast as applying a treatment-specific conditioning on the representation. Our design is independent of the underlying sequence model and can be applied to existing architectures such as long short-term memories (LSTMs) or temporal convolution networks (TCNs). We conduct extensive experiments on synthetic, semi-synthetic, and real-world data to demonstrate that CAETC yields significant improvement in counterfactual estimation over existing methods.", "date": "2026-03-12", "categories": []}
{"id": "2603.11942", "title": "Causal Matrix Completion under Multiple Treatments via Mixed Synthetic Nearest Neighbors", "abstract": "Synthetic Nearest Neighbors (SNN) provides a principled solution to causal matrix completion under missing-not-at-random (MNAR) by exploiting local low-rank structure through fully observed anchor submatrices. However, its effectiveness critically relies on sufficient data availability within each treatment level, a condition that often fails in settings with multiple or complex treatments. In this work, we propose Mixed Synthetic Nearest Neighbors (MSNN), a new entry-wise causal identification estimator that integrates information across treatment levels. We show that MSNN retains the finite-sample error bounds and asymptotic normality guarantees of SNN, while enlarging the effective sample size available for estimation. Empirical results on synthetic and real-world datasets illustrate the efficacy of the proposed approach, especially under data-scarce treatment levels.", "date": "2026-03-12", "categories": []}
{"id": "2509.22953", "title": "GDR-learners: Orthogonal Learning of Generative Models for Potential Outcomes", "abstract": "Various deep generative models have been proposed to estimate potential outcomes distributions from observational data. However, none of them have the favorable theoretical property of general Neyman-orthogonality and, associated with it, quasi-oracle efficiency and double robustness. In this paper, we introduce a general suite of generative Neyman-orthogonal (doubly-robust) learners that estimate the conditional distributions of potential outcomes. Our proposed generative doubly-robust learners (GDR-learners) are flexible and can be instantiated with many state-of-the-art deep generative models. In particular, we develop GDR-learners based on (a) conditional normalizing flows (which we call GDR-CNFs), (b) conditional generative adversarial networks (GDR-CGANs), (c) conditional variational autoencoders (GDR-CVAEs), and (d) conditional diffusion models (GDR-CDMs). Unlike the existing methods, our GDR-learners possess the properties of quasi-oracle efficiency and rate double robustness, and are thus asymptotically optimal. In a series of (semi-)synthetic experiments, we demonstrate that our GDR-learners are very effective and outperform the existing methods in estimating the conditional distributions of potential outcomes.", "date": "2025-09-26", "categories": []}
{"id": "2509.24962", "title": "Overlap-Adaptive Regularization for Conditional Average Treatment Effect Estimation", "abstract": "The conditional average treatment effect (CATE) is widely used in personalized medicine to inform therapeutic decisions. However, state-of-the-art methods for CATE estimation (so-called meta-learners) often perform poorly in the presence of low overlap. In this work, we introduce a new approach to tackle this issue and improve the performance of existing meta-learners in the low-overlap regions. Specifically, we introduce Overlap-Adaptive Regularization (OAR) that regularizes target models proportionally to overlap weights so that, informally, the regularization is higher in regions with low overlap. To the best of our knowledge, our OAR is the first approach to leverage overlap weights in the regularization terms of the meta-learners. Our OAR approach is flexible and works with any existing CATE meta-learner: we demonstrate how OAR can be applied to both parametric and non-parametric second-stage models. Furthermore, we propose debiased versions of our OAR that preserve the Neyman-orthogonality of existing meta-learners and thus ensure more robust inference. Through a series of (semi-)synthetic experiments, we demonstrate that our OAR significantly improves CATE estimation in low-overlap settings in comparison to constant regularization.", "date": "2025-09-29", "categories": []}
{"id": "2511.04361", "title": "Causal Regime Detection in Energy Markets With Augmented Time Series Structural Causal Models", "abstract": "Energy markets exhibit complex causal relationships between weather patterns, generation technologies, and price formation, with regime changes occurring continuously rather than at discrete break points. Current approaches model electricity prices without explicit causal interpretation or counterfactual reasoning capabilities. We introduce Augmented Time Series Causal Models (ATSCM) for energy markets, extending counterfactual reasoning frameworks to multivariate temporal data with learned causal structure. Our approach models energy systems through interpretable factors (weather, generation mix, demand patterns), rich grid dynamics, and observable market variables. We integrate neural causal discovery to learn time-varying causal graphs without requiring ground truth DAGs. Applied to real-world electricity price data, ATSCM enables novel counterfactual queries such as \"What would prices be under different renewable generation scenarios?\".", "date": "2025-11-06", "categories": []}
{"id": "2603.10284", "title": "Copula-ResLogit: A Deep-Copula Framework for Unobserved Confounding Effects", "abstract": "A key challenge in travel demand analysis is the presence of unobserved factors that may generate non-causal dependencies, obscuring the true causal effects. To address the issue, the study introduces a novel deep learning based fully interpretable joint modelling framework, Copula-ResLogit, which integrates the flexibility of Residual Neural Network (ResNet) architectures with the dependence capturing capabilities of copula models. This hybrid structure enables us to first detect unobserved confounding through traditional copula function based joint modelling and then mitigate these hidden associations by incorporating deep learning components. The study applies this framework to two case studies, including the relationship between stress levels and wait time of pedestrians when crossing mid block in VR and the dependencies between travel mode choice and travel distance in London travel behaviour data. Results show that Copula-ResLogit substantially reduces or eliminates the dependencies, demonstrating the ability of residual layers to account for hidden confounding effects.", "date": "2026-03-11", "categories": []}
{"id": "2603.09532", "title": "What Do We Care About in Bandits with Noncompliance? BRACE: Bandits with Recommendations, Abstention, and Certified Effects", "abstract": "Bandits with noncompliance separate the learner's recommendation from the treatment actually delivered, so the learning target itself must be chosen. A platform may care about recommendation welfare in the current mediated workflow, treatment learning for a future direct-control regime, or anytime-valid uncertainty for one of those targets. These objectives need not agree. We formalize this objective-choice problem, identify the direct-control regime in which recommendation and treatment objectives collapse, and show by example that recommendation welfare can strictly exceed every learner-measurable treatment policy when downstream actors use private information. For finite-context square-IV problems we propose BRACE, a parameter-free phase-doubling algorithm that performs IV inversion only after matrix certification and otherwise returns full-range but honest structural intervals. BRACE delivers simultaneous policy-value validity, fixed-gap identification of the operationally optimal recommendation policy, and fixed-gap identification of the structurally optimal treatment policy under contextual homogeneity and invertibility. We complement the theory with a finite-context empirical benchmark spanning direct control, mediated present-versus-future tradeoffs, weak identification, homogeneity failure, and rectangular overidentification. The experiments show that safety appears as regret on easy problems, as abstention and wide valid intervals under weak identification, as a reason to prefer recommendation welfare under homogeneity failure, and as tighter structural uncertainty when extra instruments are available. For rich contexts, we also derive an orthogonal score whose conditional bias factorizes into compliance-model and outcome-model errors, clarifying what must be stabilized for anytime-valid semiparametric IV inference.", "date": "2026-03-10", "categories": []}
{"id": "2603.09563", "title": "Learning Bayesian and Markov Networks with an Unreliable Oracle", "abstract": "We study constraint-based structure learning of Markov networks and Bayesian networks in the presence of an unreliable conditional independence oracle that makes at most a bounded number of errors. For Markov networks, we observe that a low maximum number of vertex-wise disjoint paths implies that the structure is uniquely identifiable even if the number of errors is (moderately) exponential in the number of vertices. For Bayesian networks, however, we prove that one cannot tolerate any errors to always identify the structure even when many commonly used graph parameters like treewidth are bounded. Finally, we give algorithms for structure learning when the structure is uniquely identifiable.", "date": "2026-03-10", "categories": []}
{"id": "2406.14380", "title": "Estimating Treatment Effects under Algorithmic Interference: A Structured Neural Networks Approach", "abstract": "Online user-generated content platforms allocate billions of dollars of promotional traffic through algorithms in two-sided marketplaces. To evaluate updates to these algorithms, platforms frequently rely on creator-side randomized experiments. However, because treated and control creators compete for exposure, such experiments suffer from algorithmic interference: exposure outcomes depend on competitors' treatment status. We show that commonly used difference-in-means estimators can therefore be severely biased and may even recommend deploying inferior algorithms. To address this challenge, we develop a structured semiparametric framework that explicitly models the competitive allocation mechanism underlying exposure. Our approach combines an algorithm choice model that characterizes how exposure is allocated across competing content with a viewer response model that captures engagement conditional on exposure. We construct a debiased estimator grounded in the double machine learning framework to recover the global treatment effect of platform-wide rollout. Methodologically, we extend DML asymptotic theory to accommodate correlated samples arising from overlapping consideration sets. Using Monte Carlo simulations and a large-scale field experiment on a major short-video platform, we show that our estimator closely matches an interference-free benchmark obtained from a costly double-sided experimental design. In contrast, standard estimators exhibit substantial bias and, in some cases, even reverse the sign of the effect.", "date": "2024-06-20", "categories": []}
{"id": "2509.26429", "title": "An Orthogonal Learner for Individualized Outcomes in Markov Decision Processes", "abstract": "Predicting individualized potential outcomes in sequential decision-making is central for optimizing therapeutic decisions in personalized medicine (e.g., which dosing sequence to give to a cancer patient). However, predicting potential outcomes over long horizons is notoriously difficult. Existing methods that break the curse of the horizon typically lack strong theoretical guarantees such as orthogonality and quasi-oracle efficiency. In this paper, we revisit the problem of predicting individualized potential outcomes in sequential decision-making (i.e., estimating Q-functions in Markov decision processes with observational data) through a causal inference lens. In particular, we develop a comprehensive theoretical foundation for meta-learners in this setting with a focus on beneficial theoretical properties. As a result, we yield a novel meta-learner called DRQ-learner and establish that it is: (1) doubly robust (i.e., valid inference under the misspecification of one of the nuisances), (2) Neyman-orthogonal (i.e., insensitive to first-order estimation errors in the nuisance functions), and (3) achieves quasi-oracle efficiency (i.e., behaves asymptotically as if the ground-truth nuisance functions were known). Our DRQ-learner is applicable to settings with both discrete and continuous state spaces. Further, our DRQ-learner is flexible and can be used together with arbitrary machine learning models (e.g., neural networks). We validate our theoretical results through numerical experiments, thereby showing that our meta-learner outperforms state-of-the-art baselines.", "date": "2025-09-30", "categories": []}
{"id": "2511.22854", "title": "CRAwDAD: Causal Reasoning Augmentation with Dual-Agent Debate", "abstract": "When people reason about cause and effect, they often consider many competing \"what if\" scenarios before deciding which explanation fits best. Analogously, advanced language models capable of causal inference can consider multiple interventions and counterfactuals to judge the validity of causal claims. Crucially, this type of reasoning is less like a single calculation and more like an internal dialogue between alternative hypotheses. In this paper, we make this dialogue explicit through a dual-agent debate framework where one model provides a structured causal inference, and the other critically examines this reasoning for logical flaws. When disagreements arise, the agents attempt to persuade each other, challenging each other's logic and revising their conclusions until they converge on a mutually agreed answer. To take advantage of this deliberative process, we specifically use reasoning language models, whose strengths in both causal inference and adversarial debate remain under-explored relative to standard large language models. We evaluate our approach on the CLadder dataset, a benchmark linking natural language questions to formally defined causal graphs across all three rungs of Pearl's ladder of causation. With Qwen3 and DeepSeek-R1 as debater agents, we demonstrate that multi-agent debate improves DeepSeek-R1's overall accuracy in causal inference from 78.03% to 87.45%, with the counterfactual category specifically improving from 67.94% to 80.04% accuracy. Similarly, Qwen3's overall accuracy improves from 84.16% to 89.41%, and counterfactual questions from 71.53% to 80.35%, showing that even strong models can still benefit greatly from debate with weaker agents. Our results highlight the potential of reasoning models as building blocks for multi-agent systems in causal inference, and demonstrate the importance of diverse perspectives in causal problem-solving.", "date": "2025-11-28", "categories": []}
{"id": "2512.00969", "title": "Integrating a Causal Foundation Model into a Prescriptive Maintenance Framework for Optimising Production-Line OEE", "abstract": "The transition to prescriptive maintenance (PsM) in manufacturing is critically constrained by a dependence on predictive models. Such purely predictive models tend to capture statistical associations in the data without identifying the underlying causal drivers of failure, which can lead to costly misdiagnoses and ineffective measures. This fundamental limitation results in a key challenge: while we can predict that a failure may occur, we lack a systematic method to understand why a failure occurs. This paper proposes a model based on causal machine learning to bridge this gap. Our objective is to move beyond diagnosis to active prescription by simulating and evaluating potential fixes to optimise KPIs such as Overall Equipment Effectiveness (OEE). For this purpose, a pre-trained causal foundation model is used as a ``what-if'' simulator to estimate the effects of potential fixes. By estimating the causal effect of each intervention on system-level KPIs, specific actions can be recommended for the production line. This can help identify plausible root causes and quantify their operational impact. The model is evaluated using semi-synthetic manufacturing data and compared with non-causal and causal baseline machine learning models. This paper provides a technical basis for a human-centred approach, allowing engineers to test potential solutions in a causal environment to make more effective operational decisions and reduce costly downtimes.", "date": "2025-11-30", "categories": []}
{"id": "2603.08311", "title": "Sign Identifiability of Causal Effects in Stationary Stochastic Dynamical Systems", "abstract": "We study identifiability in continuous-time linear stationary stochastic differential equations with known causal structure. Unlike existing approaches, we relax the assumption of a known diffusion matrix, thereby respecting the model's intrinsic scale invariance. Rather than recovering drift coefficients themselves, we introduce edge-sign identifiability: for a given causal structure, we ask whether the sign of a given drift entry is uniquely determined across all observational covariance matrices induced by parametrizations compatible with that structure. Under a notion of faithfulness, we derive criteria for characterising identifiability, non-identifiability, and partial identifiability for general graphs. Applying our criteria to specific causal structures, both analogous to classical causal settings (e.g., instrumental variables) and novel cyclic settings, we determine their edge-sign identifiability and, in some cases, obtain explicit expressions for the sign of a target edge in terms of the observational covariance matrix.", "date": "2026-03-09", "categories": []}
{"id": "2309.12032", "title": "Expert-Aided Causal Discovery of Ancestral Graphs", "abstract": "Causal discovery (CD) is an important component of many scientific applications, yet most techniques produce unreliable point estimates that often contradict expert knowledge. To mitigate this, recent research has focused on ex-ante incorporation of background knowledge into the CD process, typically under an unrealistic causal sufficiency assumption. When probing experts is costly (e.g., hidden behind expensive LLM APIs), however, ex-post model refinement that maximizes query utility is preferable. Also, when independent experts provide conflicting but better-than-random feedback, a principled aggregation method is required. In this context, we introduce the first CD algorithm that enables (i) distributional inference over ancestral graphs (AGs), which represent causal systems under latent confounding, and (ii) integration of both ex-ante and uncertain ex-post expert knowledge. Briefly, our method is a diversity-seeking reinforcement learning algorithm, termed Ancestral GFlowNet (AGFN), whose policy we iteratively refine based on a Bayesian model of the noisy expert feedback. Importantly, we prove convergence to the true AG given sufficiently accurate responses. Through validation on synthetic and realistic datasets using simulated humans and LLMs, we show AGFN is competitive with or superior to strong baselines in terms of structural Hamming distance and Bayesian Information Criterion.", "date": "2023-09-21", "categories": []}
{"id": "2603.05568", "title": "Learning Optimal Distributionally Robust Individualized Treatment Rules Integrating Multi-Source Data", "abstract": "Integrative analysis of multiple datasets for estimating optimal individualized treatment rules (ITRs) can enhance decision efficiency. A central challenge is posterior shift, wherein the conditional distribution of potential outcomes given covariates differs between source and target populations. We propose a prior information-based distributionally robust ITR (PDRO-ITR) that maximizes the worst-case policy value over a covariate-dependent distributional uncertainty set, ensuring robust performance under posterior shift. The uncertainty set is constructed as an individualized combination of source distributions, with weights combining prior source-membership probabilities and deviation terms constrained to the probability simplex to accommodate posterior shift. We derive a closed-form solution for the PDRO-ITR and develop an adaptive procedure to tune the uncertainty level. We establish risk bounds for the PDRO-ITR estimator, which guarantees robust performance under the worst case. Extensive simulations and two real-data applications demonstrate that the proposed method achieves superior performance compared to existing approaches.", "date": "2026-03-05", "categories": []}
{"id": "2603.04473", "title": "Dictionary Based Pattern Entropy for Causal Direction Discovery", "abstract": "Discovering causal direction from temporal observational data is particularly challenging for symbolic sequences, where functional models and noise assumptions are often unavailable. We propose a novel \\emph{Dictionary Based Pattern Entropy ($DPE$)} framework that infers both the direction of causation and the specific subpatterns driving changes in the effect variable. The framework integrates \\emph{Algorithmic Information Theory} (AIT) and \\emph{Shannon Information Theory}. Causation is interpreted as the emergence of compact, rule based patterns in the candidate cause that systematically constrain the effect. $DPE$ constructs direction-specific dictionaries and quantifies their influence using entropy-based measures, enabling a principled link between deterministic pattern structure and stochastic variability. Causal direction is inferred via a minimum-uncertainty criterion, selecting the direction exhibiting stronger and more consistent pattern-driven organization. As summarized in Table 7, $DPE$ consistently achieves reliable performance across diverse synthetic systems, including delayed bit-flip perturbations, AR(1) coupling, 1D skew-tent maps, and sparse processes, outperforming or matching competing AIT-based methods ($ETC_E$, $ETC_P$, $LZ_P$). In biological and ecological datasets, performance is competitive, while alternative methods show advantages in specific genomic settings. Overall, the results demonstrate that minimizing pattern level uncertainty yields a robust, interpretable, and broadly applicable framework for causal discovery.", "date": "2026-03-04", "categories": []}
{"id": "2603.04534", "title": "Invariant Causal Routing for Governing Social Norms in Online Market Economies", "abstract": "Social norms are stable behavioral patterns that emerge endogenously within economic systems through repeated interactions among agents. In online market economies, such norms -- like fair exposure, sustained participation, and balanced reinvestment -- are critical for long-term stability. We aim to understand the causal mechanisms driving these emergent norms and to design principled interventions that can steer them toward desired outcomes. This is challenging because norms arise from countless micro-level interactions that aggregate into macro-level regularities, making causal attribution and policy transferability difficult. To address this, we propose \\textbf{Invariant Causal Routing (ICR)}, a causal governance framework that identifies policy-norm relations stable across heterogeneous environments. ICR integrates counterfactual reasoning with invariant causal discovery to separate genuine causal effects from spurious correlations and to construct interpretable, auditable policy rules that remain effective under distribution shift. In heterogeneous agent simulations calibrated with real data, ICR yields more stable norms, smaller generalization gaps, and more concise rules than correlation or coverage baselines, demonstrating that causal invariance offers a principled and interpretable foundation for governance.", "date": "2026-03-04", "categories": []}
{"id": "2603.04780", "title": "Distributional Equivalence in Linear Non-Gaussian Latent-Variable Cyclic Causal Models: Characterization and Learning", "abstract": "Causal discovery with latent variables is a fundamental task. Yet most existing methods rely on strong structural assumptions, such as enforcing specific indicator patterns for latents or restricting how they can interact with others. We argue that a core obstacle to a general, structural-assumption-free approach is the lack of an equivalence characterization: without knowing what can be identified, one generally cannot design methods for how to identify it. In this work, we aim to close this gap for linear non-Gaussian models. We establish the graphical criterion for when two graphs with arbitrary latent structure and cycles are distributionally equivalent, that is, they induce the same observed distribution set. Key to our approach is a new tool, edge rank constraints, which fills a missing piece in the toolbox for latent-variable causal discovery in even broader settings. We further provide a procedure to traverse the whole equivalence class and develop an algorithm to recover models from data up to such equivalence. To our knowledge, this is the first equivalence characterization with latent variables in any parametric setting without structural assumptions, and hence the first structural-assumption-free discovery method. Code and an interactive demo are available at https://equiv.cc.", "date": "2026-03-05", "categories": []}
{"id": "2603.05149", "title": "Federated Causal Discovery Across Heterogeneous Datasets under Latent Confounding", "abstract": "Causal discovery across multiple datasets is often constrained by data privacy regulations and cross-site heterogeneity, limiting the use of conventional methods that require a single, centralized dataset. To address these challenges, we introduce fedCI, a federated conditional independence test that rigorously handles heterogeneous datasets with non-identical sets of variables, site-specific effects, and mixed variable types, including continuous, ordinal, binary, and categorical variables. At its core, fedCI uses a federated Iteratively Reweighted Least Squares (IRLS) procedure to estimate the parameters of generalized linear models underlying likelihood-ratio tests for conditional independence. Building on this, we develop fedCI-IOD, a federated extension of the Integration of Overlapping Datasets (IOD) algorithm, that replaces its meta-analysis strategy and enables, for the fist time, federated causal discovery under latent confounding across distributed and heterogeneous datasets. By aggregating evidence federatively, fedCI-IOD not only preserves privacy but also substantially enhances statistical power, achieving performance comparable to fully pooled analyses and mitigating artifacts from low local sample sizes. Our tools are publicly available as the fedCI Python package, a privacy-preserving R implementation of IOD, and a web application for the fedCI-IOD pipeline, providing versatile, user-friendly solutions for federated conditional independence testing and causal discovery.", "date": "2026-03-05", "categories": []}
{"id": "2603.05288", "title": "Bayesian Supervised Causal Clustering", "abstract": "Finding patient subgroups with similar characteristics is crucial for personalized decision-making in various disciplines such as healthcare and policy evaluation. While most existing approaches rely on unsupervised clustering methods, there is a growing trend toward using supervised clustering methods that identify operationalizable subgroups in the context of a specific outcome of interest. We propose Bayesian Supervised Causal Clustering (BSCC), with treatment effect as outcome to guide the clustering process. BSCC identifies homogenous subgroups of individuals who are similar in their covariate profiles as well as their treatment effects. We evaluate BSCC on simulated datasets as well as real-world dataset from the third International Stroke Trial to assess the practical usefulness of the framework.", "date": "2026-03-05", "categories": []}
{"id": "2603.05370", "title": "Learning Causal Structure of Time Series using Best Order Score Search", "abstract": "Causal structure learning from observational data is central to many scientific and policy domains, but the time series setting common to many disciplines poses several challenges due to temporal dependence. In this paper we focus on score-based causal discovery for multivariate time series and introduce TS-BOSS, a time series extension of the recently proposed Best Order Score Search (BOSS) (Andrews et al. 2023). TS-BOSS performs a permutation-based search over dynamic Bayesian network structures while leveraging grow-shrink trees to cache intermediate score computations, preserving the scalability and strong empirical performance of BOSS in the static setting. We provide theoretical guarantees establishing the soundness of TS-BOSS under suitable assumptions, and we present an intermediate result that extends classical subgraph minimality results for permutation-based methods to the dynamic (time series) setting. Our experiments on synthetic data show that TS-BOSS is especially effective in high auto-correlation regimes, where it consistently achieves higher adjacency recall at comparable precision than standard constraint-based methods. Overall, TS-BOSS offers a high-performing, scalable approach for time series causal discovery and our results provide a principled bridge for extending sparsity-based, permutation-driven causal learning theory to dynamic settings.", "date": "2026-03-05", "categories": []}
{"id": "2603.05483", "title": "SurvHTE-Bench: A Benchmark for Heterogeneous Treatment Effect Estimation in Survival Analysis", "abstract": "Estimating heterogeneous treatment effects (HTEs) from right-censored survival data is critical in high-stakes applications such as precision medicine and individualized policy-making. Yet, the survival analysis setting poses unique challenges for HTE estimation due to censoring, unobserved counterfactuals, and complex identification assumptions. Despite recent advances, from Causal Survival Forests to survival meta-learners and outcome imputation approaches, evaluation practices remain fragmented and inconsistent. We introduce SurvHTE-Bench, the first comprehensive benchmark for HTE estimation with censored outcomes. The benchmark spans (i) a modular suite of synthetic datasets with known ground truth, systematically varying causal assumptions and survival dynamics, (ii) semi-synthetic datasets that pair real-world covariates with simulated treatments and outcomes, and (iii) real-world datasets from a twin study (with known ground truth) and from an HIV clinical trial. Across synthetic, semi-synthetic, and real-world settings, we provide the first rigorous comparison of survival HTE methods under diverse conditions and realistic assumption violations. SurvHTE-Bench establishes a foundation for fair, reproducible, and extensible evaluation of causal survival methods. The data and code of our benchmark are available at: https://github.com/Shahriarnz14/SurvHTE-Bench .", "date": "2026-03-05", "categories": []}
{"id": "2602.23541", "title": "Causal Identification from Counterfactual Data: Completeness and Bounding Results", "abstract": "Previous work establishing completeness results for counterfactual identification has been circumscribed to the setting where the input data belongs to observational or interventional distributions (Layers 1 and 2 of Pearl's Causal Hierarchy), since it was generally presumed impossible to obtain data from counterfactual distributions, which belong to Layer 3. However, recent work (Raghavan & Bareinboim, 2025) has formally characterized a family of counterfactual distributions which can be directly estimated via experimental methods - a notion they call counterfactual realizabilty. This leaves open the question of what additional counterfactual quantities now become identifiable, given this new access to (some) Layer 3 data. To answer this question, we develop the CTFIDU+ algorithm for identifying counterfactual queries from an arbitrary set of Layer 3 distributions, and prove that it is complete for this task. Building on this, we establish the theoretical limit of which counterfactuals can be identified from physically realizable distributions, thus implying the fundamental limit to exact causal inference in the non-parametric setting. Finally, given the impossibility of identifying certain critical types of counterfactuals, we derive novel analytic bounds for such quantities using realizable counterfactual data, and corroborate using simulations that counterfactual data helps tighten the bounds for non-identifiable quantities in practice.", "date": "2026-02-26", "categories": []}
{"id": "2603.03411", "title": "Scalable Contrastive Causal Discovery under Unknown Soft Interventions", "abstract": "Observational causal discovery is only identifiable up to the Markov equivalence class. While interventions can reduce this ambiguity, in practice interventions are often soft with multiple unknown targets. In many realistic scenarios, only a single intervention regime is observed. We propose a scalable causal discovery model for paired observational and interventional settings with shared underlying causal structure and unknown soft interventions. The model aggregates subset-level PDAGs and applies contrastive cross-regime orientation rules to construct a globally consistent maximal PDAG under Meek closure, enabling generalization to both in-distribution and out-of-distribution settings. Theoretically, we prove that our model is sound with respect to a restricted $\\Psi$ equivalence class induced solely by the information available in the subset-restricted setting. We further show that the model asymptotically recovers the corresponding identifiable PDAG and can orient additional edges compared to non-contrastive subset-restricted methods. Experiments on synthetic data demonstrate improved causal structure recovery, generalization to unseen graphs with held-out causal mechanisms, and scalability to larger graphs, with ablations supporting the theoretical results.", "date": "2026-03-03", "categories": []}
{"id": "2603.03587", "title": "Controllable Generative Sandbox for Causal Inference", "abstract": "Method validation and study design in causal inference rely on synthetic data with known counterfactuals. Existing simulators trade off distributional realism, the ability to capture mixed-type and multimodal tabular data, against causal controllability, including explicit control over overlap, unmeasured confounding, and treatment effect heterogeneity. We introduce CausalMix, a variational generative framework that closes this gap by coupling a mixture of Gaussian latent priors with data-type-specific decoders for continuous, binary, and categorical variables. The model incorporates explicit causal controls: an overlap regularizer shaping propensity-score distributions, alongside direct parameterizations of confounding strength and effect heterogeneity. This unified objective preserves fidelity to the observed data while enabling factorial manipulation of causal mechanisms, allowing overlap, confounding strength, and treatment effect heterogeneity to be varied independently at design time. Across benchmarks, CausalMix achieves state-of-the-art distributional metrics on mixed-type tables while providing stable, fine-grained causal control. We demonstrate practical utility in a comparative safety study of metastatic castration-resistant prostate cancer treatments, using CausalMix to compare estimators under calibrated data-generating processes, tune hyperparameters, and conduct simulation-based power analyses under targeted treatment effect heterogeneity scenarios.", "date": "2026-03-03", "categories": []}
{"id": "2603.03785", "title": "Observationally Informed Adaptive Causal Experimental Design", "abstract": "Randomized Controlled Trials (RCTs) represent the gold standard for causal inference yet remain a scarce resource. While large-scale observational data is often available, it is utilized only for retrospective fusion, and remains discarded in prospective trial design due to bias concerns. We argue this \"tabula rasa\" data acquisition strategy is fundamentally inefficient. In this work, we propose Active Residual Learning, a new paradigm that leverages the observational model as a foundational prior. This approach shifts the experimental focus from learning target causal quantities from scratch to efficiently estimating the residuals required to correct observational bias. To operationalize this, we introduce the R-Design framework. Theoretically, we establish two key advantages: (1) a structural efficiency gap, proving that estimating smooth residual contrasts admits strictly faster convergence rates than reconstructing full outcomes; and (2) information efficiency, where we quantify the redundancy in standard parameter-based acquisition (e.g., BALD), demonstrating that such baselines waste budget on task-irrelevant nuisance uncertainty. We propose R-EPIG (Residual Expected Predictive Information Gain), a unified criterion that directly targets the causal estimand, minimizing residual uncertainty for estimation or clarifying decision boundaries for policy. Experiments on synthetic and semi-synthetic benchmarks demonstrate that R-Design significantly outperforms baselines, confirming that repairing a biased model is far more efficient than learning one from scratch.", "date": "2026-03-04", "categories": []}
{"id": "2603.04276", "title": "Causality Elicitation from Large Language Models", "abstract": "Large language models (LLMs) are trained on enormous amounts of data and encode knowledge in their parameters. We propose a pipeline to elicit causal relationships from LLMs. Specifically, (i) we sample many documents from LLMs on a given topic, (ii) we extract an event list from from each document, (iii) we group events that appear across documents into canonical events, (iv) we construct a binary indicator vector for each document over canonical events, and (v) we estimate candidate causal graphs using causal discovery methods. Our approach does not guarantee real-world causality. Rather, it provides a framework for presenting the set of causal hypotheses that LLMs can plausibly assume, as an inspectable set of variables and candidate graphs.", "date": "2026-03-04", "categories": []}
{"id": "2603.02289", "title": "Topological Causal Effects", "abstract": "Estimating causal effects is particularly challenging when outcomes arise in complex, non-Euclidean spaces, where conventional methods often fail to capture meaningful structural variation. We develop a framework for topological causal inference that defines treatment effects through differences in the topological structure of potential outcomes, summarized by power-weighted silhouette functions of persistence diagrams. We develop an efficient, doubly robust estimator in a fully nonparametric model, establish functional weak convergence, and construct a formal test of the null hypothesis of no topological effect. Empirical studies illustrate that the proposed method reliably quantifies topological treatment effects across diverse complex outcome types.", "date": "2026-03-02", "categories": []}
{"id": "2603.02359", "title": "Estimating Visual Attribute Effects in Advertising from Observational Data: A Deepfake-Informed Double Machine Learning Approach", "abstract": "Digital advertising increasingly relies on visual content, yet marketers lack rigorous methods for understanding how specific visual attributes causally affect consumer engagement. This paper addresses a fundamental methodological challenge: estimating causal effects when the treatment, such as a model's skin tone, is an attribute embedded within the image itself. Standard approaches like Double Machine Learning (DML) fail in this setting because vision encoders entangle treatment information with confounding variables, producing severely biased estimates. We develop DICE-DML (Deepfake-Informed Control Encoder for Double Machine Learning), a framework that leverages generative AI to disentangle treatment from confounders. The approach combines three mechanisms: (1) deepfake-generated image pairs that isolate treatment variation; (2) DICE-Diff adversarial learning on paired difference vectors, where background signals cancel to reveal pure treatment fingerprints; and (3) orthogonal projection that geometrically removes treatment-axis components. In simulations with known ground truth, DICE-DML reduces root mean squared error by 73-97% compared to standard DML, with the strongest improvement (97.5%) at the null effect point, demonstrating robust Type I error control. Applying DICE-DML to 232,089 Instagram influencer posts, we estimate the causal effect of skin tone on engagement. Standard DML produces diagnostically invalid results (negative outcome R^2), while DICE-DML achieves valid confounding control (R^2 = 0.63) and estimates a marginally significant negative effect of darker skin tone (-522 likes; p = 0.062), substantially smaller than the biased standard estimate. Our framework provides a principled approach for causal inference with visual data when treatments and confounders coexist within images.", "date": "2026-03-02", "categories": []}
{"id": "2603.02902", "title": "Distributed Dynamic Invariant Causal Prediction in Environmental Time Series", "abstract": "The extraction of invariant causal relationships from time series data with environmental attributes is critical for robust decision-making in domains such as climate science and environmental monitoring. However, existing methods either emphasize dynamic causal analysis without leveraging environmental contexts or focus on static invariant causal inference, leaving a gap in distributed temporal settings. In this paper, we propose Distributed Dynamic Invariant Causal Prediction in Time-series (DisDy-ICPT), a novel framework that learns dynamic causal relationships over time while mitigating spatial confounding variables without requiring data communication. We theoretically prove that DisDy-ICPT recovers stable causal predictors within a bounded number of communication rounds under standard sampling assumptions. Empirical evaluations on synthetic benchmarks and environment-segmented real-world datasets show that DisDy-ICPT achieves superior predictive stability and accuracy compared to baseline methods A and B. Our approach offers promising applications in carbon monitoring and weather forecasting. Future work will extend DisDy-ICPT to online learning scenarios.", "date": "2026-03-03", "categories": []}
{"id": "2603.03035", "title": "Generalized Bayes for Causal Inference", "abstract": "Uncertainty quantification is central to many applications of causal machine learning, yet principled Bayesian inference for causal effects remains challenging. Standard Bayesian approaches typically require specifying a probabilistic model for the data-generating process, including high-dimensional nuisance components such as propensity scores and outcome regressions. Standard posteriors are thus vulnerable to strong modeling choices, including complex prior elicitation. In this paper, we propose a generalized Bayesian framework for causal inference. Our framework avoids explicit likelihood modeling; instead, we place priors directly on the causal estimands and update these using an identification-driven loss function, which yields generalized posteriors for causal effects. As a result, our framework turns existing loss-based causal estimators into estimators with full uncertainty quantification. Our framework is flexible and applicable to a broad range of causal estimands (e.g., ATE, CATE). Further, our framework can be applied on top of state-of-the-art causal machine learning pipelines (e.g., Neyman-orthogonal meta-learners). For Neyman-orthogonal losses, we show that the generalized posteriors converge to their oracle counterparts and remain robust to first-stage nuisance estimation error. With calibration, we thus obtain valid frequentist uncertainty even when nuisance estimators converge at slower-than-parametric rates. Empirically, we demonstrate that our proposed framework offers causal effect estimation with calibrated uncertainty across several causal inference settings. To the best of our knowledge, this is the first flexible framework for constructing generalized Bayesian posteriors for causal machine learning.", "date": "2026-03-03", "categories": []}
{"id": "2603.03207", "title": "I-CAM-UV: Integrating Causal Graphs over Non-Identical Variable Sets Using Causal Additive Models with Unobserved Variables", "abstract": "Causal discovery from observational data is a fundamental tool in various fields of science. While existing approaches are typically designed for a single dataset, we often need to handle multiple datasets with non-identical variable sets in practice. One straightforward approach is to estimate a causal graph from each dataset and construct a single causal graph by overlapping. However, this approach identifies limited causal relationships because unobserved variables in each dataset can be confounders, and some variable pairs may be unobserved in any dataset. To address this issue, we leverage Causal Additive Models with Unobserved Variables (CAM-UV) that provide causal graphs having information related to unobserved variables. We show that the ground truth causal graph has structural consistency with the information of CAM-UV on each dataset. As a result, we propose an approach named I-CAM-UV to integrate CAM-UV results by enumerating all consistent causal graphs. We also provide an efficient combinatorial search algorithm and demonstrate the usefulness of I-CAM-UV against existing methods.", "date": "2026-03-03", "categories": []}
{"id": "2402.06223", "title": "Beyond DAGs: A Latent Partial Causal Model for Multimodal Learning", "abstract": "Directed Acyclic Graphs (DAGs) are a standard tool in causal modeling, but their suitability for capturing the complexity of large-scale multimodal data is questionable. In practice, real-world multimodal datasets are often collected from heterogeneous generative processes that do not conform to a single DAG. Instead, they may involve multiple, and even opposing, DAG structures with inverse causal directions. To address this gap, in this work, we first propose a novel latent partial causal model tailored for multimodal data representation learning, featuring two latent coupled variables parts connected by an undirected edge, to represent the transfer of knowledge across modalities. Under specific statistical assumptions, we establish an identifiability result, demonstrating that representations learned by MultiModal Contrastive Learning (MMCL) correspond to the latent coupled variables up to a trivial transformation. This result deepens our understanding of the why MMCL works, highlights its potential for representation disentanglement, and expands the utility of pre-trained models like CLIP. Synthetic experiments confirm the robustness of our findings, even when the assumptions are partially violated. Most importantly, experiments on a pre-trained CLIP model embodies disentangled representations, enabling few-shot learning and improving domain generalization across diverse real-world datasets. Together, these contributions push the boundaries of MMCL, both in theory and in practical applications.", "date": "2024-02-09", "categories": []}
{"id": "2508.11727", "title": "Causal Structure Learning in Hawkes Processes with Complex Latent Confounder Networks", "abstract": "Multivariate Hawkes process provides a powerful framework for modeling temporal dependencies and event-driven interactions in complex systems. While existing methods primarily focus on uncovering causal structures among observed subprocesses, real-world systems are often only partially observed, with latent subprocesses posing significant challenges. In this paper, we show that continuous-time event sequences can be represented by a discrete-time causal model as the time interval shrinks, and we leverage this insight to establish necessary and sufficient conditions for identifying latent subprocesses and the causal influences. Accordingly, we propose a two-phase iterative algorithm that alternates between inferring causal relationships among discovered subprocesses and uncovering new latent subprocesses, guided by path-based conditions that guarantee identifiability. Experiments on both synthetic and real-world datasets show that our method effectively recovers causal structures despite the presence of latent subprocesses.", "date": "2025-08-15", "categories": []}
{"id": "2511.02137", "title": "DoFlow: Flow-based Generative Models for Interventional and Counterfactual Forecasting on Time Series", "abstract": "Time-series forecasting increasingly demands not only accurate observational predictions but also causal forecasting under interventional and counterfactual queries in multivariate systems. We present DoFlow, a flow-based generative model defined over a causal Directed Acyclic Graph (DAG) that delivers coherent observational and interventional predictions, as well as counterfactuals through the natural encoding-decoding mechanism of continuous normalizing flows (CNFs). We also provide a supporting counterfactual recovery theory under certain assumptions. Beyond forecasting, DoFlow provides explicit likelihoods of future trajectories, enabling principled anomaly detection. Experiments on synthetic datasets with various causal DAG structures and real-world hydropower and cancer-treatment time series show that DoFlow achieves accurate system-wide observational forecasting, enables causal forecasting over interventional and counterfactual queries, and effectively detects anomalies. This work contributes to the broader goal of unifying causal reasoning and generative modeling for complex dynamical systems.", "date": "2025-11-04", "categories": []}
{"id": "2601.18678", "title": "Counterfactual Explanations on Robust Perceptual Geodesics", "abstract": "Latent-space optimization methods for counterfactual explanations - framed as minimal semantic perturbations that change model predictions - inherit the ambiguity of Wachter et al.'s objective: the choice of distance metric dictates whether perturbations are meaningful or adversarial. Existing approaches adopt flat or misaligned geometries, leading to off-manifold artifacts, semantic drift, or adversarial collapse. We introduce Perceptual Counterfactual Geodesics (PCG), a method that constructs counterfactuals by tracing geodesics under a perceptually Riemannian metric induced from robust vision features. This geometry aligns with human perception and penalizes brittle directions, enabling smooth, on-manifold, semantically valid transitions. Experiments on three vision datasets show that PCG outperforms baselines and reveals failure modes hidden under standard metrics.", "date": "2026-01-26", "categories": []}
{"id": "2603.00855", "title": "Navigating Time's Possibilities: Plausible Counterfactual Explanations for Multivariate Time-Series Forecast through Genetic Algorithms", "abstract": "Counterfactual learning has become promising for understanding and modeling causality in complex and dynamic systems. This paper presents a novel method for counterfactual learning in the context of multivariate time series analysis and forecast. The primary objective is to uncover hidden causal relationships and identify potential interventions to achieve desired outcomes. The proposed methodology integrates genetic algorithms and rigorous causality tests to infer and validate counterfactual dependencies within temporal sequences. More specifically, we employ Granger causality to enhance the reliability of identified causal relationships, rigorously assessing their statistical significance. Then, genetic algorithms, in conjunction with quantile regression, are used to exploit these intricate causal relationships to project future scenarios. The synergy between genetic algorithms and causality tests ensures a thorough exploration of the temporal dynamics present in the data, revealing hidden dependencies and enabling the projection of outcomes under hypothetical interventions. We evaluate the performance of our algorithm on real-world data, showcasing its ability to handle complex causal relationships, revealing meaningful counterfactual insights, and allowing for the prediction of outcomes under hypothetical interventions.", "date": "2026-03-01", "categories": []}
{"id": "2603.01052", "title": "No More Maybe-Arrows: Resolving Causal Uncertainty by Breaking Symmetries", "abstract": "The recent works on causal discovery have followed a similar trend of learning partial ancestral graphs (PAGs) since observational data constrain the true causal directed acyclic graph (DAG) only up to a Markov equivalence class. This limits their application in the majority of downstream tasks, as uncertainty in causal relations remains unresolved. We propose a new refinement framework, CausalSAGE, for converting PAGs to DAGs while respecting the underlying causal relations. The framework expands discrete variables into state-level representations, constrains the search space using structural knowledge and soft priors, and applies a unified differentiable objective for joint optimization. The final DAG is obtained by aggregating the optimized structures and enforcing acyclicity when necessary. Our experimental evaluations show that the obtained DAGs preserve the underlying causal relations while also being efficient to obtain.", "date": "2026-03-01", "categories": []}
{"id": "2603.01119", "title": "Robust Weighted Triangulation of Causal Effects Under Model Uncertainty", "abstract": "A fundamental challenge in causal inference with observational data is correct specification of a causal model. When there is model uncertainty, analysts may seek to use estimates from multiple candidate models that rely on distinct, and possibly partially overlapping, sets of identifying assumptions to infer the causal effect, a process known as triangulation. Principled methods for triangulation, however, remain underdeveloped. Here, we develop a framework for causal effect triangulation that combines model testability methods from causal discovery with statistical inference methods from semiparametric theory, while avoiding explicit model selection and post-selection inference problems. We propose a triangulation functional that combines identified functionals from each model with data-driven measures of model validity. We provide a bound on the distance of the functional from the true causal effect along with conditions under which this distance can be taken to zero. Finally, we derive valid statistical inference for this functional. Our framework formalizes robustness under causal pluralism without requiring agreement across models or commitment to a single specification. We demonstrate its performance through simulations and an empirical application.", "date": "2026-03-01", "categories": []}
{"id": "2603.01140", "title": "Teacher-Guided Causal Interventions for Image Denoising: Orthogonal Content-Noise Disentanglement in Vision Transformers", "abstract": "Conventional image denoising models often inadvertently learn spurious correlations between environmental factors and noise patterns. Moreover, due to high-frequency ambiguity, they struggle to reliably distinguish subtle textures from stochastic noise, resulting in over-removed details or residual noise artifacts. We therefore revisit denoising via causal intervention, arguing that purely correlational fitting entangles intrinsic content with extrinsic noise, which directly degrades robustness under distribution shifts. Motivated by this, we propose the Teacher-Guided Causal Disentanglement Network (TCD-Net), which explicitly decomposes the generative mechanism via structured interventions on feature spaces within a Vision Transformer framework. Specifically, our method integrates three key components: (1) An Environmental Bias Adjustment (EBA) module projects features into a stable, de-centered subspace to suppress global environmental bias (de-confounding). (2) A dual-branch disentanglement head employs an orthogonality constraint to force a strict separation between content and noise representations, preventing information leakage. (3) To resolve structural ambiguity, we leverage Nano Banana Pro, Google's reasoning-guided AI image generation model, to guide a causal prior, effectively pulling content representations back onto the natural-image manifold. Extensive experiments demonstrate that TCD-Net outperforms mainstream methods across multiple benchmarks in both fidelity and efficiency, achieving a real-time speed of 104.2 FPS on a single RTX 5090 GPU.", "date": "2026-03-01", "categories": []}
{"id": "2603.01339", "title": "Causal Effects with Unobserved Unit Types in Interacting Human-AI Systems", "abstract": "We study experiments on interacting populations of humans and AI agents, where both unit types and the interaction network remain unobserved. Although causal effects propagate throughout the system, the goal is to estimate effects on humans. Examples include online platforms where human users interact alongside AI-driven accounts. We assume a human-AI prior that gives each unit a probability of being human. While humans cannot be distinguished at the unit level, the prior allows us to compute the average human composition within large subpopulations. We then model outcome dynamics through a causal message passing (CMP) framework and analyze sample-mean outcomes across subpopulations. We show that by constructing subpopulations that vary in expected human composition and treatment exposure, one can consistently recover human-specific causal effects. Our results characterize when distributional knowledge of population composition (without observing unit types or the interaction network) is sufficient for identification. We validate the approach on a simulated human-AI platform driven by behaviorally differentiated LLM agents. Together, these results provide a theoretical and practical framework for experimentation in emerging human-AI systems.", "date": "2026-03-02", "categories": []}
{"id": "2603.02015", "title": "CausalWrap: Model-Agnostic Causal Constraint Wrappers for Tabular Synthetic Data", "abstract": "Tabular synthetic data generators are typically trained to match observational distributions, which can yield high conventional utility (e.g., column correlations, predictive accuracy) yet poor preservation of structural relations relevant to causal analysis and out-of-distribution (OOD) reasoning. When the downstream use of synthetic data involves causal reasoning -- estimating treatment effects, evaluating policies, or testing mediation pathways -- merely matching the observational distribution is insufficient: structural fidelity and treatment-mechanism preservation become essential. We propose CausalWrap (CW), a model-agnostic wrapper that injects partial causal knowledge (PCK) -- trusted edges, forbidden edges, and qualitative/monotonic constraints -- into any pretrained base generator (GAN, VAE, or diffusion model), without requiring access to its internals. CW learns a lightweight, differentiable post-hoc correction map applied to samples from the base generator, optimized with causal penalty terms under an augmented-Lagrangian schedule. We provide theoretical results connecting penalty-based optimization to constraint satisfaction and relating approximate factorization to joint distributional control. We validate CW on simulated structural causal models (SCMs) with known ground-truth interventions, semi-synthetic causal benchmarks (IHDP and an ACIC-style suite), and a real-world ICU cohort (MIMIC-IV) with expert-elicited partial graphs. CW improves causal fidelity across diverse base generators -- e.g., reducing average treatment effect (ATE) error by up to 63% on ACIC and lifting ATE agreement from 0.00 to 0.38 on the intensive care unit (ICU) cohort -- while largely retaining conventional utility.", "date": "2026-03-02", "categories": []}
{"id": "2603.02174", "title": "De-paradox Tree: Breaking Down Simpson's Paradox via A Kernel-Based Partition Algorithm", "abstract": "Real-world observational datasets and machine learning have revolutionized data-driven decision-making, yet many models rely on empirical associations that may be misleading due to confounding and subgroup heterogeneity. Simpson's paradox exemplifies this challenge, where aggregated and subgroup-level associations contradict each other, leading to misleading conclusions. Existing methods provide limited support for detecting and interpreting such paradoxical associations, especially for practitioners without deep causal expertise. We introduce De-paradox Tree, an interpretable algorithm designed to uncover hidden subgroup patterns behind paradoxical associations under assumed causal structures involving confounders and effect heterogeneity. It employs novel split criteria and balancing-based procedures to adjust for confounders and homogenize heterogeneous effects through recursive partitioning. Compared to state-of-the-art methods, De-paradox Tree builds simpler, more interpretable trees, selects relevant covariates, and identifies nested opposite effects while ensuring robust estimation of causal effects when causally admissible variables are provided. Our approach addresses the limitations of traditional causal inference and machine learning methods by introducing an interpretable framework that supports non-expert practitioners while explicitly acknowledging causal assumptions and scope limitations, enabling more reliable and informed decision-making in complex observational data environments.", "date": "2026-03-02", "categories": []}
{"id": "2603.02204", "title": "Partial Causal Structure Learning for Valid Selective Conformal Inference under Interventions", "abstract": "Selective conformal prediction can yield substantially tighter uncertainty sets when we can identify calibration examples that are exchangeable with the test example. In interventional settings, such as perturbation experiments in genomics, exchangeability often holds only within subsets of interventions that leave a target variable \"unaffected\" (e.g., non-descendants of an intervened node in a causal graph). We study the practical regime where this invariance structure is unknown and must be learned from data. Our contributions are: (i) a contamination-robust conformal coverage theorem that quantifies how misclassification of \"unaffected\" calibration examples degrades coverage via an explicit function $g(\\delta,n)$ of the contamination fraction and calibration set size, providing a finite-sample lower bound that holds for arbitrary contaminating distributions; (ii) a task-driven partial causal learning formulation that estimates only the binary descendant indicators $Z_{a,i}=\\mathbf{1}\\{i\\in\\mathrm{desc}(a)\\}$ needed for selective calibration, rather than the full causal graph; and (iii) algorithms for descendant discovery via perturbation intersection patterns (differentially affected variable set intersections across interventions), and for approximate distance-to-intervention estimation via local invariant causal prediction. We provide recovery conditions under which contamination is controlled. Experiments on synthetic linear structural equation models (SEMs) validate the bound: under controlled contamination up to $\\delta=0.30$, the corrected procedure maintains $\\ge 0.95$ coverage while uncorrected selective CP degrades to $0.867$. A proof-of-concept on Replogle K562 CRISPR interference (CRISPRi) perturbation data demonstrates applicability to real genomic screens.", "date": "2026-03-02", "categories": []}
{"id": "2507.16467", "title": "Estimating Treatment Effects with Independent Component Analysis", "abstract": "Independent Component Analysis (ICA) uses a measure of non-Gaussianity to identify latent sources from data and estimate their mixing coefficients (Shimizu et al., 2006). Meanwhile, higher-order Orthogonal Machine Learning (OML) exploits non-Gaussian treatment noise to provide more accurate estimates of treatment effects in the presence of confounding nuisance effects (Mackey et al., 2018). Remarkably, we find that the two approaches rely on the same moment conditions for consistent estimation. We then seize upon this connection to show how ICA can be effectively used for treatment effect estimation. Specifically, we prove that linear ICA can consistently estimate multiple treatment effects, even in the presence of Gaussian confounders, and identify regimes in which ICA is provably more sample-efficient than OML for treatment effect estimation. Our synthetic demand estimation experiments confirm this theory and demonstrate that linear ICA can accurately estimate treatment effects even in the presence of nonlinear nuisance.", "date": "2025-07-22", "categories": []}
{"id": "2509.21021", "title": "Efficient Ensemble Conditional Independence Test Framework for Causal Discovery", "abstract": "Constraint-based causal discovery relies on numerous conditional independence tests (CITs), but its practical applicability is severely constrained by the prohibitive computational cost, especially as CITs themselves have high time complexity with respect to the sample size. To address this key bottleneck, we introduce the Ensemble Conditional Independence Test (E-CIT), a general-purpose and plug-and-play framework. E-CIT operates on an intuitive divide-and-aggregate strategy: it partitions the data into subsets, applies a given base CIT independently to each subset, and aggregates the resulting p-values using a novel method grounded in the properties of stable distributions. This framework reduces the computational complexity of a base CIT to linear in the sample size when the subset size is fixed. Moreover, our tailored p-value combination method offers theoretical consistency guarantees under mild conditions on the subtests. Experimental results demonstrate that E-CIT not only significantly reduces the computational burden of CITs and causal discovery but also achieves competitive performance. Notably, it exhibits an improvement in complex testing scenarios, particularly on real-world datasets.", "date": "2025-09-25", "categories": []}
{"id": "2510.04970", "title": "Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning", "abstract": "We present FLOP (Fast Learning of Order and Parents), a score-based causal discovery algorithm for linear models. It pairs fast parent selection with iterative Cholesky-based score updates, cutting run-times over prior algorithms. This makes it feasible to fully embrace discrete search, enabling iterated local search with principled order initialization to find graphs with scores at or close to the global optimum. The resulting structures are highly accurate across benchmarks, with near-perfect recovery in standard settings. This performance calls for revisiting discrete search over graphs as a reasonable approach to causal discovery.", "date": "2025-10-06", "categories": []}
{"id": "2602.22287", "title": "Multi-Level Causal Embeddings", "abstract": "Abstractions of causal models allow for the coarsening of models such that relations of cause and effect are preserved. Whereas abstractions focus on the relation between two models, in this paper we study a framework for causal embeddings which enable multiple detailed models to be mapped into sub-systems of a coarser causal model. We define causal embeddings as a generalization of abstraction, and present a generalized notion of consistency. By defining a multi-resolution marginal problem, we showcase the relevance of causal embeddings for both the statistical marginal problem and the causal marginal problem; furthermore, we illustrate its practical use in merging datasets coming from models with different representations. ", "date": "2026-02-25", "categories": []}
{"id": "2602.23611", "title": "Fairness under Graph Uncertainty: Achieving Interventional Fairness with Partially Known Causal Graphs over Clusters of Variables", "abstract": "Algorithmic decisions about individuals require predictions that are not only accurate but also fair with respect to sensitive attributes such as gender and race. Causal notions of fairness align with legal requirements, yet many methods assume access to detailed knowledge of the underlying causal graph, which is a demanding assumption in practice. We propose a learning framework that achieves interventional fairness by leveraging a causal graph over \\textit{clusters of variables}, which is substantially easier to estimate than a variable-level graph. With possible \\textit{adjustment cluster sets} identified from such a cluster causal graph, our framework trains a prediction model by reducing the worst-case discrepancy between interventional distributions across these sets. To this end, we develop a computationally efficient barycenter kernel maximum mean discrepancy (MMD) that scales favorably with the number of sensitive attribute values. Extensive experiments show that our framework strikes a better balance between fairness and accuracy than existing approaches, highlighting its effectiveness under limited causal graph knowledge.", "date": "2026-02-27", "categories": []}
{"id": "2602.23800", "title": "Operationalizing Longitudinal Causal Discovery Under Real-World Workflow Constraints", "abstract": "Causal discovery has achieved substantial theoretical progress, yet its deployment in large-scale longitudinal systems remains limited. A key obstacle is that operational data are generated under institutional workflows whose induced partial orders are rarely formalized, enlarging the admissible graph space in ways inconsistent with the recording process. We characterize a workflow-induced constraint class for longitudinal causal discovery that restricts the admissible directed acyclic graph space through protocol-derived structural masks and timeline-aligned indexing. Rather than introducing a new optimization algorithm, we show that explicitly encoding workflow-consistent partial orders reduces structural ambiguity, especially in mixed discrete--continuous panels where within-time orientation is weakly identified. The framework combines workflow-derived admissible-edge constraints, measurement-aligned time indexing and block structure, bootstrap-based uncertainty quantification for lagged total effects, and a dynamic representation supporting intervention queries. In a nationwide annual health screening cohort in Japan with 107,261 individuals and 429,044 person-years, workflow-constrained longitudinal LiNGAM yields temporally consistent within-time substructures and interpretable lagged total effects with explicit uncertainty. Sensitivity analyses using alternative exposure and body-composition definitions preserve the main qualitative patterns. We argue that formalizing workflow-derived constraint classes improves structural interpretability without relying on domain-specific edge specification, providing a reproducible bridge between operational workflows and longitudinal causal discovery under standard identifiability assumptions.", "date": "2026-02-27", "categories": []}
{"id": "2602.24266", "title": "Efficient Discovery of Approximate Causal Abstractions via Neural Mechanism Sparsification", "abstract": "Neural networks are hypothesized to implement interpretable causal mechanisms, yet verifying this requires finding a causal abstraction -- a simpler, high-level Structural Causal Model (SCM) faithful to the network under interventions. Discovering such abstractions is hard: it typically demands brute-force interchange interventions or retraining. We reframe the problem by viewing structured pruning as a search over approximate abstractions. Treating a trained network as a deterministic SCM, we derive an Interventional Risk objective whose second-order expansion yields closed-form criteria for replacing units with constants or folding them into neighbors. Under uniform curvature, our score reduces to activation variance, recovering variance-based pruning as a special case while clarifying when it fails. The resulting procedure efficiently extracts sparse, intervention-faithful abstractions from pretrained networks, which we validate via interchange interventions.", "date": "2026-02-27", "categories": []}
{"id": "2505.08371", "title": "Density Ratio-based Causal Discovery from Bivariate Continuous-Discrete Data", "abstract": "We address the problem of inferring the causal direction between a continuous variable $X$ and a discrete variable $Y$ from observational data. For the model $X \\to Y$, we adopt the threshold model used in prior work. For the model $Y \\to X$, we consider two cases: (1) the conditional distributions of $X$ given different values of $Y$ form a location-shift family, and (2) they are mixtures of generalized normal distributions with independently parameterized components. We establish identifiability of the causal direction through three theoretical results. First, we prove that under $X \\to Y$, the density ratio of $X$ conditioned on different values of $Y$ is monotonic. Second, we establish that under $Y \\to X$ with non-location-shift conditionals, monotonicity of the density ratio holds only on a set of Lebesgue measure zero in the parameter space. Third, we show that under $X \\to Y$, the conditional distributions forming a location-shift family requires a precise coordination between the causal mechanism and input distribution, which is non-generic under the principle of independent mechanisms. Together, these results imply that monotonicity of the density ratio characterizes the direction $X \\to Y$, whereas non-monotonicity or location-shift conditionals characterizes $Y \\to X$. Based on this, we propose Density Ratio-based Causal Discovery (DRCD), a method that determines causal direction by testing for location-shift conditionals and monotonicity of the estimated density ratio. Experiments on synthetic and real-world datasets demonstrate that DRCD outperforms existing methods.", "date": "2025-05-13", "categories": []}
{"id": "2602.22254", "title": "Causal Direction from Convergence Time: Faster Training in the True Causal Direction", "abstract": "We introduce Causal Computational Asymmetry (CCA), a principle for causal direction identification based on optimization dynamics in which one neural network is trained to predict $Y$ from $X$ and another to predict $X$ from $Y$, and the direction that converges faster is inferred to be causal. Under the additive noise model $Y = f(X) + \\varepsilon$ with $\\varepsilon \\perp X$ and $f$ nonlinear and injective, we establish a formal asymmetry: in the reverse direction, residuals remain statistically dependent on the input regardless of approximation quality, inducing a strictly higher irreducible loss floor and non-separable gradient noise in the optimization dynamics, so that the reverse model requires strictly more gradient steps in expectation to reach any fixed loss threshold; consequently, the forward (causal) direction converges in fewer expected optimization steps. CCA operates in optimization-time space, distinguishing it from methods such as RESIT, IGCI, and SkewScore that rely on statistical independence or distributional asymmetries, and proper z-scoring of both variables is required for valid comparison of convergence rates. On synthetic benchmarks, CCA achieves 26/30 correct causal identifications across six neural architectures, including 30/30 on sine and exponential data-generating processes. We further embed CCA into a broader framework termed Causal Compression Learning (CCL), which integrates graph structure learning, causal information compression, and policy optimization, with all theoretical guarantees formally proved and empirically validated on synthetic datasets.", "date": "2026-02-24", "categories": []}
{"id": "2602.22532", "title": "Coarse-to-Fine Learning of Dynamic Causal Structures", "abstract": "Learning the dynamic causal structure of time series is a challenging problem. Most existing approaches rely on distributional or structural invariance to uncover underlying causal dynamics, assuming stationary or partially stationary causality. However, these assumptions often conflict with the complex, time-varying causal relationships observed in real-world systems. This motivates the need for methods that address fully dynamic causality, where both instantaneous and lagged dependencies evolve over time. Such a setting poses significant challenges for the efficiency and stability of causal discovery. To address these challenges, we introduce DyCausal, a dynamic causal structure learning framework. DyCausal leverages convolutional networks to capture causal patterns within coarse-grained time windows, and then applies linear interpolation to refine causal structures at each time step, thereby recovering fine-grained and time-varying causal graphs. In addition, we propose an acyclic constraint based on matrix norm scaling, which improves efficiency while effectively constraining loops in evolving causal structures. Comprehensive evaluations on both synthetic and real-world datasets demonstrate that DyCausal achieves superior performance compared to existing methods, offering a stable and efficient approach for identifying fully dynamic causal structures from coarse to fine.", "date": "2026-02-26", "categories": []}
{"id": "2602.23303", "title": "Inferential Mechanics Part 1: Causal Mechanistic Theories of Machine Learning in Chemical Biology with Implications", "abstract": "Machine learning techniques are now routinely encountered in research laboratories across the globe. Impressive progress has been made through ML and AI techniques with regards to large data set processing. This progress has increased the ability of the experimenter to digest data and make novel predictions regarding phenomena of interest. However, machine learning predictors generated from data sets taken from the natural sciences are often treated as black boxes which are used broadly and generally without detailed consideration of the causal structure of the data set of interest. Work has been attempted to bring causality into discussions of machine learning models of natural phenomena; however, a firm and unified theoretical treatment is lacking. This series of three papers explores the union of chemical theory, biological theory, probability theory and causality that will correct current causal flaws of machine learning in the natural sciences. This paper, Part 1 of the series, provides the formal framework of the foundational causal structure of phenomena in chemical biology and is extended to machine learning through the novel concept of focus, defined here as the ability of a machine learning algorithm to narrow down to a hidden underpinning mechanism in large data sets. Initial proof of these principles on a family of Akt inhibitors is also provided. The second paper containing Part 2 will provide a formal exploration of chemical similarity, and Part 3 will present extensive experimental evidence of how hidden causal structures weaken all machine learning in chemical biology. This series serves to establish for chemical biology a new kind of mathematical framework for modeling mechanisms in Nature without the need for the tools of reductionism: inferential mechanics.", "date": "2026-02-26", "categories": []}
{"id": "2312.16307", "title": "Incentive-Aware Synthetic Control: Accurate Counterfactual Estimation via Incentivized Exploration", "abstract": "Synthetic control methods (SCMs) are a canonical approach used to estimate treatment effects from panel data in the internet economy. We shed light on a frequently overlooked but ubiquitous assumption made in SCMs of \"overlap\": a treated unit can be written as some combination -- typically, convex or linear -- of the units that remain under control. We show that if units select their own interventions, and there is sufficiently large heterogeneity between units that prefer different interventions, overlap will not hold. We address this issue by proposing a recommender system which incentivizes units with different preferences to take interventions they would not normally consider. Specifically, leveraging tools from information design and online learning, we propose an SCM that incentivizes exploration in panel data settings by providing incentive-compatible intervention recommendations to units. We establish this estimator obtains valid counterfactual estimates without the need for an a priori overlap assumption. We extend our results to the setting of synthetic interventions, where the goal is to produce counterfactual outcomes under all interventions, not just control. Finally, we provide two hypothesis tests for determining whether unit overlap holds for a given panel dataset.", "date": "2023-12-26", "categories": []}
{"id": "2410.18424", "title": "A Causal Graph-Enhanced Gaussian Process Regression for Modeling Engine-out NOx", "abstract": "The stringent regulatory requirements on nitrogen oxides (NOx) emissions from diesel compression ignition engines require accurate and reliable models for real time monitoring and diagnostics. Although traditional methods such as physical sensors and virtual engine control module (ECM) sensors provide essential data, they are only used for estimation. Ubiquitous literature primarily focuses on deterministic models with little emphasis on capturing the various uncertainties. The lack of probabilistic frameworks restricts the applicability of these models for robust diagnostics. The objective of this paper is to develop and validate a probabilistic model to predict engine-out NOx emissions using Gaussian process regression. Our approach is as follows. We employ three variants of Gaussian process models: the first with a standard radial basis function kernel with input window, the second incorporating a deep kernel using convolutional neural networks to capture temporal dependencies, and the third enriching the deep kernel with a causal graph derived via graph convolutional networks. The causal graph embeds physics knowledge into the learning process. All models are compared against a virtual ECM sensor using both quantitative and qualitative metrics. We conclude that our model provides an improvement in predictive performance when using an input window and a deep kernel structure. Even more compelling is the further enhancement achieved by the incorporation of a causal graph into the deep kernel. These findings are corroborated across different verification and validation datasets.", "date": "2024-10-24", "categories": []}
{"id": "2509.25800", "title": "Characterization and Learning of Causal Graphs with Latent Confounders and Post-treatment Selection from Interventional Data", "abstract": "Interventional causal discovery seeks to identify causal relations by leveraging distributional changes introduced by interventions, even in the presence of latent confounders. Beyond the spurious dependencies induced by latent confounders, we highlight a common yet often overlooked challenge in the problem due to post-treatment selection, in which samples are selectively included in datasets after interventions. This fundamental challenge widely exists in biological studies; for example, in gene expression analysis, both observational and interventional samples are retained only if they meet quality control criteria (e.g., highly active cells). Neglecting post-treatment selection may introduce spurious dependencies and distributional changes under interventions, which can mimic causal responses, thereby distorting causal discovery results and challenging existing causal formulations. To address this, we introduce a novel causal formulation that explicitly models post-treatment selection and reveals how its differential reactions to interventions can distinguish causal relations from selection patterns, allowing us to go beyond traditional equivalence classes toward the underlying true causal structure. We then characterize its Markov properties and propose a Fine-grained Interventional equivalence class, named FI-Markov equivalence, represented by a new graphical diagram, F-PAG. Finally, we develop a provably sound and complete algorithm, F-FCI, to identify causal relations, latent confounders, and post-treatment selection up to $\\mathcal{FI}$-Markov equivalence, using both observational and interventional data. Experimental results on synthetic and real-world datasets demonstrate that our method recovers causal relations despite the presence of both selection and latent confounders.", "date": "2025-09-30", "categories": []}
{"id": "2602.22083", "title": "Coarsening Bias from Variable Discretization in Causal Functionals", "abstract": "A class of causal effect functionals requires integration over conditional densities of continuous variables, as in mediation effects and nonparametric identification in causal graphical models. Estimating such densities and evaluating the resulting integrals can be statistically and computationally demanding. A common workaround is to discretize the variable and replace integrals with finite sums. Although convenient, discretization alters the population-level functional and can induce non-negligible approximation bias, even under correct identification. Under smoothness conditions, we show that this coarsening bias is first order in the bin width and arises at the level of the target functional, distinct from statistical estimation error. We propose a simple bias-reduced functional that evaluates the outcome regression at within-bin conditional means, eliminating the leading term and yielding a second-order approximation error. We derive plug-in and one-step estimators for the bias-reduced functional. Simulations demonstrate substantial bias reduction and near-nominal confidence interval coverage, even under coarse binning. Our results provide a simple framework for controlling the impact of variable discretization on parameter approximation and estimation.", "date": "2026-02-25", "categories": []}
{"id": "2506.10914", "title": "Foundation Models for Causal Inference via Prior-Data Fitted Networks", "abstract": "Prior-data fitted networks (PFNs) have recently been proposed as a promising way to train tabular foundation models. PFNs are transformers that are pre-trained on synthetic data generated from a prespecified prior distribution and that enable Bayesian inference through in-context learning. In this paper, we introduce CausalFM, a comprehensive framework for training PFN-based foundation models in various causal inference settings. First, we formalize the construction of Bayesian priors for causal inference based on structural causal models (SCMs) in a principled way and derive necessary criteria for the validity of such priors. Building on this, we propose a novel family of prior distributions using causality-inspired Bayesian neural networks that enable CausalFM to perform Bayesian causal inference in various settings, including for back-door, front-door, and instrumental variable adjustment. Finally, we instantiate CausalFM and explicitly train models to perform in-context learning in these settings. We show that CausalFM achieves competitive in-context learning performance even when compared to baselines that are specifically trained for the task at hand. In sum, our framework can be used as a general recipe to train foundation models for various causal inference settings. In contrast to the current state-of-the-art in causal inference, CausalFM offers a novel paradigm with the potential to fundamentally change how practitioners perform causal inference in medicine, economics, and other disciplines.", "date": "2025-06-12", "categories": []}
{"id": "2602.20329", "title": "CaDrift: A Time-dependent Causal Generator of Drifting Data Streams", "abstract": "This work presents Causal Drift Generator (CaDrift), a time-dependent synthetic data generator framework based on Structural Causal Models (SCMs). The framework produces a virtually infinite combination of data streams with controlled shift events and time-dependent data, making it a tool to evaluate methods under evolving data. CaDrift synthesizes various distributional and covariate shifts by drifting mapping functions of the SCM, which change underlying cause-and-effect relationships between features and the target. In addition, CaDrift models occasional perturbations by leveraging interventions in causal modeling. Experimental results show that, after distributional shift events, the accuracy of classifiers tends to drop, followed by a gradual retrieval, confirming the generator's effectiveness in simulating shifts. The framework has been made available on GitHub.", "date": "2026-02-23", "categories": []}
{"id": "2602.20333", "title": "DMCD: Semantic-Statistical Framework for Causal Discovery", "abstract": "We present DMCD (DataMap Causal Discovery), a two-phase causal discovery framework that integrates LLM-based semantic drafting from variable metadata with statistical validation on observational data. In Phase I, a large language model proposes a sparse draft DAG, serving as a semantically informed prior over the space of possible causal structures. In Phase II, this draft is audited and refined via conditional independence testing, with detected discrepancies guiding targeted edge revisions.\n  We evaluate our approach on three metadata-rich real-world benchmarks spanning industrial engineering, environmental monitoring, and IT systems analysis. Across these datasets, DMCD achieves competitive or leading performance against diverse causal discovery baselines, with particularly large gains in recall and F1 score. Probing and ablation experiments suggest that these improvements arise from semantic reasoning over metadata rather than memorization of benchmark graphs. Overall, our results demonstrate that combining semantic priors with principled statistical verification yields a high-performing and practically effective approach to causal structure learning.", "date": "2026-02-23", "categories": []}
{"id": "2602.20383", "title": "Detecting and Mitigating Group Bias in Heterogeneous Treatment Effects", "abstract": "Heterogeneous treatment effects (HTEs) are increasingly estimated using machine learning models that produce highly personalized predictions of treatment effects. In practice, however, predicted treatment effects are rarely interpreted, reported, or audited at the individual level but, instead, are often aggregated to broader subgroups, such as demographic segments, risk strata, or markets. We show that such aggregation can induce systematic bias of the group-level causal effect: even when models for predicting the individual-level conditional average treatment effect (CATE) are correctly specified and trained on data from randomized experiments, aggregating the predicted CATEs up to the group level does not, in general, recover the corresponding group average treatment effect (GATE). We develop a unified statistical framework to detect and mitigate this form of group bias in randomized experiments. We first define group bias as the discrepancy between the model-implied and experimentally identified GATEs, derive an asymptotically normal estimator, and then provide a simple-to-implement statistical test. For mitigation, we propose a shrinkage-based bias-correction, and show that the theoretically optimal and empirically feasible solutions have closed-form expressions. The framework is fully general, imposes minimal assumptions, and only requires computing sample moments. We analyze the economic implications of mitigating detected group bias for profit-maximizing personalized targeting, thereby characterizing when bias correction alters targeting decisions and profits, and the trade-offs involved. Applications to large-scale experimental data at major digital platforms validate our theoretical results and demonstrate empirical performance.", "date": "2026-02-23", "categories": []}
{"id": "2602.21036", "title": "Empirically Calibrated Conditional Independence Tests", "abstract": "Conditional independence tests (CIT) are widely used for causal discovery and feature selection. Even with false discovery rate (FDR) control procedures, they often fail to provide frequentist guarantees in practice. We highlight two common failure modes: (i) in small samples, asymptotic guarantees for many CITs can be inaccurate and even correctly specified models fail to estimate the noise levels and control the error, and (ii) when sample sizes are large but models are misspecified, unaccounted dependencies skew the test's behavior and fail to return uniform p-values under the null. We propose Empirically Calibrated Conditional Independence Tests (ECCIT), a method that measures and corrects for miscalibration. For a chosen base CIT (e.g., GCM, HRT), ECCIT optimizes an adversary that selects features and response functions to maximize a miscalibration metric. ECCIT then fits a monotone calibration map that adjusts the base-test p-values in proportion to the observed miscalibration. Across empirical benchmarks on synthetic and real data, ECCIT achieves valid FDR with higher power than existing calibration strategies while remaining test agnostic.", "date": "2026-02-24", "categories": []}
{"id": "2602.21168", "title": "Sequential Counterfactual Inference for Temporal Clinical Data: Addressing the Time Traveler Dilemma", "abstract": "Counterfactual inference enables clinicians to ask \"what if\" questions about patient outcomes, but standard methods assume feature independence and simultaneous modifiability -- assumptions violated by longitudinal clinical data. We introduce the Sequential Counterfactual Framework, which respects temporal dependencies in electronic health records by distinguishing immutable features (chronic diagnoses) from controllable features (lab values) and modeling how interventions propagate through time. Applied to 2,723 COVID-19 patients (383 Long COVID heart failure cases, 2,340 matched controls), we demonstrate that 38-67% of patients with chronic conditions would require biologically impossible counterfactuals under naive methods. We identify a cardiorenal cascade (CKD -> AKI -> HF) with relative risks of 2.27 and 1.19 at each step, illustrating temporal propagation that sequential -- but not naive -- counterfactuals can capture. Our framework transforms counterfactual explanation from \"what if this feature were different?\" to \"what if we had intervened earlier, and how would that propagate forward?\" --  yielding clinically actionable insights grounded in biological plausibility.", "date": "2026-02-24", "categories": []}
{"id": "2407.01875", "title": "Spatio-Temporal Graphical Counterfactuals: An Overview", "abstract": "Counterfactual thinking is a crucial yet challenging topic for artificial intelligence to learn knowledge from data and ultimately improve performance for new scenarios. Many research works, including the Potential Outcome Model (POM) and the Structural Causal Model (SCM), have been proposed to address this. However, their modeling, theoretical foundations, and application approaches often differ. Moreover, there is a lack of graphical approaches for inferring spatio-temporal counterfactuals, that account for spatial and temporal interactions among multiple units. Thus, in this work, we aim to present a survey that compares and discusses different counterfactual models, theories and approaches. Additionally, we propose a unified graphical causal framework to infer spatio-temporal counterfactuals.", "date": "2024-07-02", "categories": []}
{"id": "2410.19412", "title": "Robust Time Series Causal Discovery for Agent-Based Model Validation", "abstract": "Agent-Based Model (ABM) validation is crucial as it helps ensuring the reliability of simulations, and causal discovery has become a powerful tool in this context. However, current causal discovery methods often face accuracy and robustness challenges when applied to complex and noisy time series data, which is typical in ABM scenarios. This study addresses these issues by proposing a Robust Cross-Validation (RCV) approach to enhance causal structure learning for ABM validation. We develop RCV-VarLiNGAM and RCV-PCMCI, novel extensions of two prominent causal discovery algorithms. These aim to reduce the impact of noise better and give more reliable causal relation results, even with high-dimensional, time-dependent data. The proposed approach is then integrated into an enhanced ABM validation framework, which is designed to handle diverse data and model structures.\n  The approach is evaluated using synthetic datasets and a complex simulated fMRI dataset. The results demonstrate greater reliability in causal structure identification. The study examines how various characteristics of datasets affect the performance of established causal discovery methods. These characteristics include linearity, noise distribution, stationarity, and causal structure density. This analysis is then extended to the RCV method to see how it compares in these different situations. This examination helps confirm whether the results are consistent with existing literature and also reveals the strengths and weaknesses of the novel approaches.\n  By tackling key methodological challenges, the study aims to enhance ABM validation with a more resilient valuation framework presented. These improvements increase the reliability of model-driven decision making processes in complex systems analysis.", "date": "2024-10-25", "categories": []}
{"id": "2510.16703", "title": "On the Granularity of Causal Effect Identifiability", "abstract": "The classical notion of causal effect identifiability is defined in terms of treatment and outcome variables. In this paper, we consider the identifiability of state-based causal effects: how an intervention on a particular state of treatment variables affects a particular state of outcome variables. We demonstrate that state-based causal effects may be identifiable even when variable-based causal effects may not. Moreover, we show that this separation occurs only when additional knowledge -- such as context-specific independencies -- is available. We further examine knowledge that constrains the states of variables, and show that such knowledge can improve both variable-based and state-based identifiability when combined with other knowledge such as context-specific independencies. We finally propose an approach for identifying causal effects under these additional constraints, and conduct empirical studies to further illustrate the separations between the two levels of identifiability.", "date": "2025-10-19", "categories": []}
{"id": "2512.01149", "title": "A Benchmark of Causal vs. Correlation AI for Predictive Maintenance", "abstract": "Predictive maintenance in manufacturing environments presents a challenging optimization problem characterized by extreme cost asymmetry, where missed failures incur costs roughly fifty times higher than false alarms. Predictive maintenance in manufacturing environments presents a challenging optimization problem characterized by extreme cost asymmetry, where missed failures incur costs roughly fifty times higher than false alarms. Conventional machine learning approaches typically optimize statistical accuracy metrics that do not reflect this operational reality and cannot reliably distinguish causal relationships from spurious correlations. This study benchmarks eight predictive models, ranging from baseline statistical approaches to Bayesian structural causal methods, on a dataset of 10,000 CNC machines with a 3.3 percent failure prevalence. While ensemble correlation-based models such as Random Forest (L4) achieve the highest raw cost savings (70.8 percent reduction), the Bayesian Structural Causal Model (L7) delivers competitive financial performance (66.4 percent cost reduction) with an inherent ability of failure attribution, which correlation-based models do not readily provide. The model achieves perfect attribution for HDF, PWF, and OSF failure types. These results suggest that causal methods, when combined with domain knowledge and Bayesian inference, offer a potentially favorable trade-off between predictive performance and operational interpretability in predictive maintenance applications.", "date": "2025-11-30", "categories": []}
{"id": "2601.17160", "title": "Information-Theoretic Causal Bounds under Unmeasured Confounding", "abstract": "We develop a data-driven information-theoretic framework for sharp partial identification of causal effects under unmeasured confounding. Existing approaches often rely on restrictive assumptions, such as bounded or discrete outcomes; require external inputs (for example, instrumental variables, proxies, or user-specified sensitivity parameters); necessitate full structural causal model specifications; or focus solely on population-level averages while neglecting covariate-conditional effects. We overcome all four limitations simultaneously by establishing novel information-theoretic, data-driven divergence bounds. Our key theoretical contribution shows that the f-divergence between the observational distribution P(Y | A = a, X = x) and the interventional distribution P(Y | do(A = a), X = x) is upper bounded by a function of the propensity score alone. This result enables sharp partial identification of conditional causal effects directly from observational data, without requiring external sensitivity parameters, auxiliary variables, full structural specifications, or outcome boundedness assumptions. For practical implementation, we develop a semiparametric estimator satisfying Neyman orthogonality (Chernozhukov et al., 2018), which ensures root-n consistent inference even when nuisance functions are estimated via flexible machine learning methods. Simulation studies and real-world data applications, implemented in the GitHub repository (https://github.com/yonghanjung/Information-Theretic-Bounds), demonstrate that our framework provides tight and valid causal bounds across a wide range of data-generating processes.", "date": "2026-01-23", "categories": []}
{"id": "2602.00774", "title": "A Novel VAE-DML Fusion Framework for Causal Analysis of Greenwashing in the Mining Industry", "abstract": "Against the backdrop of the global green transition and \"dual carbon\" goals, mining industry chain enterprises are pivotal entities in terms of resource consumption and environmental impact. Their environmental performance directly affects regional ecological security and is closely tied to national resource strategies and green transformation outcomes. Ensuring the authenticity and reliability of their environmental disclosure is thus a core and urgent issue for sustainable development and national strategic objectives.From a corporate governance perspective, this study examines equity balance as a fundamental governance mechanism, investigating its inhibitory effect on greenwashing behavior among these enterprises and the underlying pathways involved. Methodologically, the paper innovatively employs a Variational Autoencoder (VAE) and a Double Machine Learning (DML) model to construct counterfactual scenarios, mitigating endogeneity concerns and precisely identifying the causal relationship between equity balance and greenwashing. The findings indicate, first, a significant negative causal relationship between equity balance and corporate greenwashing, confirming its substantive governance effect. Second, this inhibitory effect exhibits notable heterogeneity, manifesting more strongly in western regions, upstream segments of the industrial chain, and industries with high environmental sensitivity. Third, the governance effect demonstrates clear temporal dynamics, with the strongest impact occurring in the current period, followed by a diminishing yet statistically significant lagged effect, and ultimately a stable long-term cumulative influence. Finally, mechanism analysis reveals that equity balance operates through three distinct channels to curb greenwashing: alleviating management performance pressure, enhancing the stability of the executive team, and intensifying media scrutiny.", "date": "2026-01-31", "categories": []}
{"id": "2602.08535", "title": "Causal Schr\\\"odinger Bridges: Constrained Optimal Transport on Structural Manifolds", "abstract": "Generative modeling typically seeks the path of least action via deterministic flows (ODE). While effective for in-distribution tasks, we argue that these deterministic paths become brittle under causal interventions, which often require transporting probability mass across low-density regions (\"off-manifold\") where the vector field is ill-defined. This leads to numerical instability and the pathology of anticipatory control. In this work, we introduce the Causal Schrodinger Bridge (CSB), a framework that reformulates counterfactual inference as Entropic Optimal Transport. By leveraging diffusion processes (SDEs), CSB enables probability mass to robustly \"tunnel\" through support mismatches while strictly enforcing structural admissibility. We prove the Structural Decomposition Theorem, showing that the global high-dimensional bridge factorizes exactly into local, robust transitions. This theorem provides a principled resolution to the Information Bottleneck that plagues monolithic architectures in high dimensions. We empirically validate CSB on a full-rank causal system (d=10^5, intrinsic rank 10^5), where standard structure-blind MLPs fail to converge (MSE ~0.31). By physically implementing the structural decomposition, CSB achieves high-fidelity transport (MSE ~0.06) in just 73.73 seconds on a single GPU. This stands in stark contrast to structure-agnostic O(d^3) baselines, estimated to require over 6 years. Our results demonstrate that CSB breaks the Curse of Dimensionality through structural intelligence, offering a scalable foundation for high-stakes causal discovery in 10^5-node systems. Code is available at: https://github.com/cochran1/causal-schrodinger-bridge", "date": "2026-02-09", "categories": []}
{"id": "2602.18662", "title": "Large Causal Models for Temporal Causal Discovery", "abstract": "Causal discovery for both cross-sectional and temporal data has traditionally followed a dataset-specific paradigm, where a new model is fitted for each individual dataset. Such an approach limits the potential of multi-dataset pretraining. The concept of large causal models (LCMs) envisions a class of pre-trained neural architectures specifically designed for temporal causal discovery. Prior approaches are constrained to small variable counts, degrade with larger inputs, and rely heavily on synthetic data, limiting generalization. We propose a principled framework for LCMs, combining diverse synthetic generators with realistic time-series datasets, allowing learning at scale. Extensive experiments on synthetic, semi-synthetic and realistic benchmarks show that LCMs scale effectively to higher variable counts and deeper architectures while maintaining strong performance. Trained models achieve competitive or superior accuracy compared to classical and neural baselines, particularly in out-of-distribution settings, while enabling fast, single-pass inference. Results demonstrate LCMs as a promising foundation-model paradigm for temporal causal discovery. Experiments and model weights are available at https://github.com/kougioulis/LCM-paper/.", "date": "2026-02-20", "categories": []}
{"id": "2602.18762", "title": "Bounds and Identification of Joint Probabilities of Potential Outcomes and Observed Variables under Monotonicity Assumptions", "abstract": "Evaluating joint probabilities of potential outcomes and observed variables, and their linear combinations, is a fundamental challenge in causal inference. This paper addresses the bounding and identification of these probabilities in settings with discrete treatment and discrete ordinal outcome. We propose new families of monotonicity assumptions and formulate the bounding problem as a linear programming problem. We further introduce a new monotonicity assumption specifically to achieve identification. Finally, we present numerical experiments to validate our methods and demonstrate their application using real-world datasets.", "date": "2026-02-21", "categories": []}
{"id": "2602.19131", "title": "Test-Time Learning of Causal Structure from Interventional Data", "abstract": "Supervised causal learning has shown promise in causal discovery, yet it often struggles with generalization across diverse interventional settings, particularly when intervention targets are unknown. To address this, we propose TICL (Test-time Interventional Causal Learning), a novel method that synergizes Test-Time Training with Joint Causal Inference. Specifically, we design a self-augmentation strategy to generate instance-specific training data at test time, effectively avoiding distribution shifts. Furthermore, by integrating joint causal inference, we developed a PC-inspired two-phase supervised learning scheme, which effectively leverages self-augmented training data while ensuring theoretical identifiability. Extensive experiments on bnlearn benchmarks demonstrate TICL's superiority in multiple aspects of causal discovery and intervention target detection.", "date": "2026-02-22", "categories": []}
{"id": "2602.19158", "title": "DoAtlas-1: A Causal Compilation Paradigm for Clinical AI", "abstract": "Medical foundation models generate narrative explanations but cannot quantify intervention effects, detect evidence conflicts, or validate literature claims, limiting clinical auditability. We propose causal compilation, a paradigm that transforms medical evidence from narrative text into executable code. The paradigm standardizes heterogeneous research evidence into structured estimand objects, each explicitly specifying intervention contrast, effect scale, time horizon, and target population, supporting six executable causal queries: do-calculus, counterfactual reasoning, temporal trajectories, heterogeneous effects, mechanistic decomposition, and joint interventions. We instantiate this paradigm in DoAtlas-1, compiling 1,445 effect kernels from 754 studies through effect standardization, conflict-aware graph construction, and real-world validation (Human Phenotype Project, 10,000 participants). The system achieves 98.5% canonicalization accuracy and 80.5% query executability. This paradigm shifts medical AI from text generation to executable, auditable, and verifiable causal reasoning.", "date": "2026-02-22", "categories": []}
{"id": "2602.19782", "title": "Addressing Instrument-Outcome Confounding in Mendelian Randomization through Representation Learning", "abstract": "Mendelian Randomization (MR) is a prominent observational epidemiological research method designed to address unobserved confounding when estimating causal effects. However, core assumptions -- particularly the independence between instruments and unobserved confounders -- are often violated due to population stratification or assortative mating. Leveraging the increasing availability of multi-environment data, we propose a representation learning framework that exploits cross-environment invariance to recover latent exogenous components of genetic instruments. We provide theoretical guarantees for identifying these latent instruments under various mixing mechanisms and demonstrate the effectiveness of our approach through simulations and semi-synthetic experiments using data from the All of Us Research Hub.", "date": "2026-02-23", "categories": []}
{"id": "2602.19851", "title": "Orthogonal Uplift Learning with Permutation-Invariant Representations for Combinatorial Treatments", "abstract": "We study uplift estimation for combinatorial treatments. Uplift measures the pure incremental causal effect of an intervention (e.g., sending a coupon or a marketing message) on user behavior, modeled as a conditional individual treatment effect. Many real-world interventions are combinatorial: a treatment is a policy that specifies context-dependent action distributions rather than a single atomic label. Although recent work considers structured treatments, most methods rely on categorical or opaque encodings, limiting robustness and generalization to rare or newly deployed policies. We propose an uplift estimation framework that aligns treatment representation with causal semantics. Each policy is represented by the mixture it induces over contextaction components and embedded via a permutation-invariant aggregation. This representation is integrated into an orthogonalized low-rank uplift model, extending Robinson-style decompositions to learned, vector-valued treatments. We show that the resulting estimator is expressive for policy-induced causal effects, orthogonally robust to nuisance estimation errors, and stable under small policy perturbations. Experiments on large-scale randomized platform data demonstrate improved uplift accuracy and stability in long-tailed policy regimes", "date": "2026-02-23", "categories": []}
{"id": "2602.19903", "title": "Rethinking Chronological Causal Discovery with Signal Processing", "abstract": "Causal discovery problems use a set of observations to deduce causality between variables in the real world, typically to answer questions about biological or physical systems. These observations are often recorded at regular time intervals, determined by a user or a machine, depending on the experiment design. There is generally no guarantee that the timing of these recordings matches the timing of the underlying biological or physical events. In this paper, we examine the sensitivity of causal discovery methods to this potential mismatch. We consider empirical and theoretical evidence to understand how causal discovery performance is impacted by changes of sampling rate and window length. We demonstrate that both classical and recent causal discovery methods exhibit sensitivity to these hyperparameters, and we discuss how ideas from signal processing may help us understand these phenomena.", "date": "2026-02-23", "categories": []}
{"id": "2602.20094", "title": "CausalFlip: A Benchmark for LLM Causal Judgment Beyond Semantic Matching", "abstract": "As large language models (LLMs) witness increasing deployment in complex, high-stakes decision-making scenarios, it becomes imperative to ground their reasoning in causality rather than spurious correlations. However, strong performance on traditional reasoning benchmarks does not guarantee true causal reasoning ability of LLMs, as high accuracy may still arise from memorizing semantic patterns instead of analyzing the underlying true causal structures. To bridge this critical gap, we propose a new causal reasoning benchmark, CausalFlip, designed to encourage the development of new LLM paradigm or training algorithms that ground LLM reasoning in causality rather than semantic correlation. CausalFlip consists of causal judgment questions built over event triples that could form different confounder, chain, and collider relations. Based on this, for each event triple, we construct pairs of semantically similar questions that reuse the same events but yield opposite causal answers, where models that rely heavily on semantic matching are systematically driven toward incorrect predictions. To further probe models' reliance on semantic patterns, we introduce a noisy-prefix evaluation that prepends causally irrelevant text before intermediate causal reasoning steps without altering the underlying causal relations or the logic of the reasoning process. We evaluate LLMs under multiple training paradigms, including answer-only training, explicit Chain-of-Thought (CoT) supervision, and a proposed internalized causal reasoning approach that aims to mitigate explicit reliance on correlation in the reasoning process. Our results show that explicit CoT can still be misled by spurious semantic correlations, where internalizing reasoning steps yields substantially improved causal grounding, suggesting that it is promising to better elicit the latent causal reasoning capabilities of base LLMs.", "date": "2026-02-23", "categories": []}
{"id": "2411.08875", "title": "Causal Explanations for Image Classifiers", "abstract": "Existing algorithms for explaining the output of image classifiers use different definitions of explanations and a variety of techniques to find them. However, none of the existing tools use a principled approach based on formal definitions of cause and explanation.\n  In this paper we present a novel black-box approach to computing explanations grounded in the theory of actual causality. We prove relevant theoretical results and present an algorithm for computing approximate explanations based on these definitions. We prove termination of our algorithm and discuss its complexity and the amount of approximation compared to the precise definition.\n  We implemented the framework in a tool ReX and we present experimental results and a comparison with state-of-the-art tools. We demonstrate that ReX is the most efficient black-box tool and produces the smallest explanations, in addition to outperforming other black-box tools on standard quality measures.", "date": "2024-11-13", "categories": []}
{"id": "2501.00755", "title": "An AI-powered Bayesian generative modeling approach for causal inference in observational studies", "abstract": "Causal inference in observational studies with high-dimensional covariates presents significant challenges. We introduce CausalBGM, an AI-powered Bayesian generative modeling approach that captures the causal relationship among covariates, treatment, and outcome. The core innovation is to estimate the individual treatment effect (ITE) by learning the individual-specific distribution of a low-dimensional latent feature set (e.g., latent confounders) that drives changes in both treatment and outcome. This individualized posterior representation yields estimates of the individual treatment effect (ITE) together with well-calibrated posterior intervals while mitigating confounding effect. CausalBGM is fitted through an iterative algorithm to update the model parameters and the latent features until convergence. This framework leverages the power of AI to capture complex dependencies among variables while adhering to the Bayesian principles. Extensive experiments demonstrate that CausalBGM consistently outperforms state-of-the-art methods, particularly in scenarios with high-dimensional covariates and large-scale datasets. By addressing key limitations of existing methods, CausalBGM emerges as a robust and promising framework for advancing causal inference in a wide range of modern applications. The code for CausalBGM is available at https://github.com/liuq-lab/bayesgm. The document for using CausalBGM is available at https://bayesgm.readthedocs.io.", "date": "2025-01-01", "categories": []}
{"id": "2505.14825", "title": "Assimilative Causal Inference", "abstract": "Causal inference is fundamental across scientific disciplines, yet existing methods struggle to capture instantaneous, time-evolving causal relationships in complex, high-dimensional systems. In this paper, assimilative causal inference (ACI) is developed, which is a methodological framework that leverages Bayesian data assimilation to trace causes backward from observed effects. ACI solves the inverse problem rather than quantifying forward influence. It uniquely identifies dynamic causal interactions without requiring observations of candidate causes, accommodates short datasets, and, in principle, can be implemented in high-dimensional settings by employing efficient data assimilation algorithms. Crucially, it provides online tracking of causal roles that may reverse intermittently and facilitates a mathematically rigorous criterion for the causal influence range, revealing how far effects propagate. The effectiveness of ACI is demonstrated by complex dynamical systems showcasing intermittency and extreme events. ACI opens valuable pathways for studying complex systems, where transient causal structures are critical.", "date": "2025-05-20", "categories": []}
{"id": "2510.27512", "title": "When Distributions Shifts: Causal Generalization for Low-Resource Languages", "abstract": "Machine learning models often fail under distribution shifts, a problem exacerbated in low-resource settings where limited data restricts robust generalization. Domain generalization(DG) methods address this challenge by learning representations that remain invariant across domains, frequently leveraging causal principles. In this work, we study two causal DG approaches for low-resource natural language processing. First, we apply causal data augmentation using GPT-4o-mini to generate counterfactual paraphrases for sentiment classification on the NaijaSenti Twitter corpus in Yoruba and Igbo. Second, we investigate invariant causal representation learning with the Debiasing in Aspect Review (DINER) framework for aspect-based sentiment analysis. We extend DINER to a multilingual setting by introducing Afri-SemEval, a dataset of 17 languages translated from SemEval-2014 Task. Experiments show improved robustness to unseen domains, with consistent gains from counterfactual augmentation and enhanced out-of-distribution performance from causal representation learning across multiple languages.", "date": "2025-10-31", "categories": []}
{"id": "2602.17829", "title": "Causality by Abstraction: Symbolic Rule Learning in Multivariate Timeseries with Large Language Models", "abstract": "Inferring causal relations in timeseries data with delayed effects is a fundamental challenge, especially when the underlying system exhibits complex dynamics that cannot be captured by simple functional mappings. Traditional approaches often fail to produce generalized and interpretable explanations, as multiple distinct input trajectories may yield nearly indistinguishable outputs. In this work, we present ruleXplain, a framework that leverages Large Language Models (LLMs) to extract formal explanations for input-output relations in simulation-driven dynamical systems. Our method introduces a constrained symbolic rule language with temporal operators and delay semantics, enabling LLMs to generate verifiable causal rules through structured prompting. ruleXplain relies on the availability of a principled model (e.g., a simulator) that maps multivariate input time series to output time series. Within ruleXplain, the simulator is used to generate diverse counterfactual input trajectories that yield similar target output, serving as candidate explanations. Such counterfactual inputs are clustered and provided as context to the LLM, which is tasked with the generation of symbolic rules encoding the joint temporal trends responsible for the patterns observable in the output times series. A closed-loop refinement process ensures rule consistency and semantic validity. We validate the framework using the PySIRTEM epidemic simulator, mapping testing rate inputs to daily infection counts; and the EnergyPlus building energy simulator, observing temperature and solar irradiance inputs to electricity needs. For validation, we perform three classes of experiments: (1) the efficacy of the ruleset through input reconstruction; (2) ablation studies evaluating the causal encoding of the ruleset; and (3) generalization tests of the extracted rules across unseen output trends with varying phase dynamics.", "date": "2026-02-19", "categories": []}
{"id": "2602.17934", "title": "Causal Neighbourhood Learning for Invariant Graph Representations", "abstract": "Graph data often contain noisy and spurious correlations that mask the true causal relationships, which are essential for enabling graph models to make predictions based on the underlying causal structure of the data. Dependence on spurious connections makes it challenging for traditional Graph Neural Networks (GNNs) to generalize effectively across different graphs. Furthermore, traditional aggregation methods tend to amplify these spurious patterns, limiting model robustness under distribution shifts. To address these issues, we propose Causal Neighbourhood Learning with Graph Neural Networks (CNL-GNN), a novel framework that performs causal interventions on graph structure. CNL-GNN effectively identifies and preserves causally relevant connections and reduces spurious influences through the generation of counterfactual neighbourhoods and adaptive edge perturbation guided by learnable importance masking and an attention-based mechanism. In addition, by combining structural-level interventions with the disentanglement of causal features from confounding factors, the model learns invariant node representations that are robust and generalize well across different graph structures. Our approach improves causal graph learning beyond traditional feature-based methods, resulting in a robust classification model. Extensive experiments on four publicly available datasets, including multiple domain variants of one dataset, demonstrate that CNL-GNN outperforms state-of-the-art GNN models.", "date": "2026-02-20", "categories": []}
{"id": "2602.17941", "title": "Optimizing Graph Causal Classification Models: Estimating Causal Effects and Addressing Confounders", "abstract": "Graph data is becoming increasingly prevalent due to the growing demand for relational insights in AI across various domains. Organizations regularly use graph data to solve complex problems involving relationships and connections. Causal learning is especially important in this context, since it helps to understand cause-effect relationships rather than mere associations. Since many real-world systems are inherently causal, graphs can efficiently model these systems. However, traditional graph machine learning methods including graph neural networks (GNNs), rely on correlations and are sensitive to spurious patterns and distribution changes. On the other hand, causal models enable robust predictions by isolating true causal factors, thus making them more stable under such shifts. Causal learning also helps in identifying and adjusting for confounders, ensuring that predictions reflect true causal relationships and remain accurate even under interventions. To address these challenges and build models that are robust and causally informed, we propose CCAGNN, a Confounder-Aware causal GNN framework that incorporates causal reasoning into graph learning, supporting counterfactual reasoning and providing reliable predictions in real-world settings. Comprehensive experiments on six publicly available datasets from diverse domains show that CCAGNN consistently outperforms leading state-of-the-art models.", "date": "2026-02-20", "categories": []}
{"id": "2602.17978", "title": "Learning Optimal and Sample-Efficient Decision Policies with Guarantees", "abstract": "The paradigm of decision-making has been revolutionised by reinforcement learning and deep learning. Although this has led to significant progress in domains such as robotics, healthcare, and finance, the use of RL in practice is challenging, particularly when learning decision policies in high-stakes applications that may require guarantees. Traditional RL algorithms rely on a large number of online interactions with the environment, which is problematic in scenarios where online interactions are costly, dangerous, or infeasible. However, learning from offline datasets is hindered by the presence of hidden confounders. Such confounders can cause spurious correlations in the dataset and can mislead the agent into taking suboptimal or adversarial actions. Firstly, we address the problem of learning from offline datasets in the presence of hidden confounders. We work with instrumental variables (IVs) to identify the causal effect, which is an instance of a conditional moment restrictions (CMR) problem. Inspired by double/debiased machine learning, we derive a sample-efficient algorithm for solving CMR problems with convergence and optimality guarantees, which outperforms state-of-the-art algorithms. Secondly, we relax the conditions on the hidden confounders in the setting of (offline) imitation learning, and adapt our CMR estimator to derive an algorithm that can learn effective imitator policies with convergence rate guarantees. Finally, we consider the problem of learning high-level objectives expressed in linear temporal logic (LTL) and develop a provably optimal learning algorithm that improves sample efficiency over existing methods. Through evaluation on reinforcement learning benchmarks and synthetic and semi-synthetic datasets, we demonstrate the usefulness of the methods developed in this thesis in real-world decision making.", "date": "2026-02-20", "categories": []}
{"id": "2403.11332", "title": "Graph Machine Learning based Doubly Robust Estimator for Network Causal Effects", "abstract": "We address the challenge of inferring causal effects in social network data. This results in challenges due to interference -- where a unit's outcome is affected by neighbors' treatments -- and network-induced confounding factors. While there is extensive literature focusing on estimating causal effects in social network setups, a majority of them make prior assumptions about the form of network-induced confounding mechanisms. Such strong assumptions are rarely likely to hold especially in high-dimensional networks. We propose a novel methodology that combines graph machine learning approaches with the double machine learning framework to enable accurate and efficient estimation of direct and peer effects using a single observational social network. We demonstrate the semiparametric efficiency of our proposed estimator under mild regularity conditions, allowing for consistent uncertainty quantification. We demonstrate that our method is accurate, robust, and scalable via an extensive simulation study. We use our method to investigate the impact of Self-Help Group participation on financial risk tolerance.", "date": "2024-03-17", "categories": []}
{"id": "2507.23497", "title": "Sufficient, Necessary and Complete Causal Explanations in Image Classification", "abstract": "Existing algorithms for explaining the outputs of image classifiers are based on a variety of approaches and produce explanations that frequently lack formal rigour. On the other hand, logic-based explanations are formally and rigorously defined but their computability relies on strict assumptions about the model that do not hold on image classifiers.\n  In this paper, we show that causal explanations, in addition to being formally and rigorously defined, enjoy the same formal properties as logic-based ones, while still lending themselves to black-box algorithms and being a natural fit for image classifiers. We prove formal properties of causal explanations and their equivalence to logic-based explanations. We demonstrate how to subdivide an image into its sufficient and necessary components. We introduce $\\delta$-complete explanations, which have a minimum confidence threshold and 1-complete causal explanations, explanations that are classified with the same confidence as the original image.\n  We implement our definitions, and our experimental results demonstrate that different models have different patterns of sufficiency, necessity, and completeness. Our algorithms are efficiently computable, taking on average 6s per image on a ResNet model to compute all types of explanations, and are totally black-box, needing no knowledge of the model, no access to model internals, no access to gradient, nor requiring any properties, such as monotonicity, of the model.", "date": "2025-07-31", "categories": []}
{"id": "2602.16830", "title": "The Impact of Formations on Football Matches Using Double Machine Learning. Is it worth parking the bus?", "abstract": "This study addresses a central tactical dilemma for football coaches: whether to employ a defensive strategy, colloquially known as \"parking the bus\", or a more offensive one. Using an advanced Double Machine Learning (DML) framework, this project provides a robust and interpretable tool to estimate the causal impact of different formations on key match outcomes such as goal difference, possession, corners, and disciplinary actions. Leveraging a dataset of over 22,000 matches from top European leagues, formations were categorized into six representative types based on tactical structure and expert consultation. A major methodological contribution lies in the adaptation of DML to handle categorical treatments, specifically formation combinations, through a novel matrix-based residualization process, allowing for a detailed estimation of formation-versus-formation effects that can inform a coach's tactical decision-making. Results show that while offensive formations like 4-3-3 and 4-2-3-1 offer modest statistical advantages in possession and corners, their impact on goals is limited. Furthermore, no evidence supports the idea that defensive formations, commonly associated with parking the bus, increase a team's winning potential. Additionally, red cards appear unaffected by formation choice, suggesting other behavioral factors dominate. Although this approach does not fully capture all aspects of playing style or team strength, it provides a valuable framework for coaches to analyze tactical efficiency and sets a precedent for future research in sports analytics.", "date": "2026-02-18", "categories": []}
{"id": "2602.17070", "title": "General sample size analysis for probabilities of causation: a delta method approach", "abstract": "Probabilities of causation (PoCs), such as the probability of necessity and sufficiency (PNS), are important tools for decision making but are generally not point identifiable. Existing work has derived bounds for these quantities using combinations of experimental and observational data. However, there is very limited research on sample size analysis, namely, how many experimental and observational samples are required to achieve a desired margin of error. In this paper, we propose a general sample size framework based on the delta method. Our approach applies to settings in which the target bounds of PoCs can be expressed as finite minima or maxima of linear combinations of experimental and observational probabilities. Through simulation studies, we demonstrate that the proposed sample size calculations lead to stable estimation of these bounds.", "date": "2026-02-19", "categories": []}
{"id": "2602.17217", "title": "Continual learning and refinement of causal models through dynamic predicate invention", "abstract": "Efficiently navigating complex environments requires agents to internalize the underlying logic of their world, yet standard world modelling methods often struggle with sample inefficiency, lack of transparency, and poor scalability. We propose a framework for constructing symbolic causal world models entirely online by integrating continuous model learning and repair into the agent's decision loop, by leveraging the power of Meta-Interpretive Learning and predicate invention to find semantically meaningful and reusable abstractions, allowing an agent to construct a hierarchy of disentangled, high-quality concepts from its observations. We demonstrate that our lifted inference approach scales to domains with complex relational dynamics, where propositional methods suffer from combinatorial explosion, while achieving sample-efficiency orders of magnitude higher than the established PPO neural-network-based baseline.", "date": "2026-02-19", "categories": []}
{"id": "2602.17543", "title": "genriesz: A Python Package for Automatic Debiased Machine Learning with Generalized Riesz Regression", "abstract": "Efficient estimation of causal and structural parameters can be automated using the Riesz representation theorem and debiased machine learning (DML). We present genriesz, an open-source Python package that implements automatic DML and generalized Riesz regression, a unified framework for estimating Riesz representers by minimizing empirical Bregman divergences. This framework includes covariate balancing, nearest-neighbor matching, calibrated estimation, and density ratio estimation as special cases. A key design principle of the package is automatic regressor balancing (ARB): given a Bregman generator $g$ and a representer model class, genriesz} automatically constructs a compatible link function so that the generalized Riesz regression estimator satisfies balancing (moment-matching) optimality conditions in a user-chosen basis. The package provides a modulr interface for specifying (i) the target linear functional via a black-box evaluation oracle, (ii) the representer model via basis functions (polynomial, RKHS approximations, random forest leaf encodings, neural embeddings, and a nearest-neighbor catchment basis), and (iii) the Bregman generator, with optional user-supplied derivatives. It returns regression adjustment (RA), Riesz weighting (RW), augmented Riesz weighting (ARW), and TMLE-style estimators with cross-fitting, confidence intervals, and $p$-values. We highlight representative workflows for estimation problems such as the average treatment effect (ATE), ATE on treated (ATT), and average marginal effect estimation. The Python package is available at https://github.com/MasaKat0/genriesz and on PyPI.", "date": "2026-02-19", "categories": []}
{"id": "2208.14153", "title": "Identifying Weight-Variant Latent Causal Models", "abstract": "The task of causal representation learning aims to uncover latent higher-level causal variables that affect lower-level observations. Identifying the true latent causal variables from observed data, while allowing instantaneous causal relations among latent variables, remains a challenge, however. To this end, we start with the analysis of three intrinsic indeterminacies in identifying latent variables from observations: transitivity, permutation indeterminacy, and scaling indeterminacy. We find that transitivity acts as a key role in impeding the identifiability of latent causal variables. To address the unidentifiable issue due to transitivity, we introduce a novel identifiability condition where the underlying latent causal model satisfies a linear-Gaussian model, in which the causal coefficients and the distribution of Gaussian noise are modulated by an additional observed variable. Under certain assumptions, including the existence of a reference condition under which latent causal influences vanish, we can show that the latent causal variables can be identified up to trivial permutation and scaling, and that partial identifiability results can still be obtained when this reference condition is violated for a subset of latent variables. Furthermore, based on these theoretical results, we propose a novel method, termed Structural caUsAl Variational autoEncoder (SuaVE), which directly learns causal representations and causal relationships among them, together with the mapping from the latent causal variables to the observed ones. Experimental results on synthetic and real data demonstrate the identifiability and consistency results and the efficacy of SuaVE in learning causal representations.", "date": "2022-08-30", "categories": []}
{"id": "2502.08963", "title": "Modeling Time-evolving Causality over Data Streams", "abstract": "Given an extensive, semi-infinite collection of multivariate coevolving data sequences (e.g., sensor/web activity streams) whose observations influence each other, how can we discover the time-changing cause-and-effect relationships in co-evolving data streams? How efficiently can we reveal dynamical patterns that allow us to forecast future values? In this paper, we present a novel streaming method, ModePlait, which is designed for modeling such causal relationships (i.e., time-evolving causality) in multivariate co-evolving data streams and forecasting their future values. The solution relies on characteristics of the causal relationships that evolve over time in accordance with the dynamic changes of exogenous variables. ModePlait has the following properties: (a) Effective: it discovers the time-evolving causality in multivariate co-evolving data streams by detecting the transitions of distinct dynamical patterns adaptively. (b) Accurate: it enables both the discovery of time-evolving causality and the forecasting of future values in a streaming fashion. (c) Scalable: our algorithm does not depend on data stream length and thus is applicable to very large sequences. Extensive experiments on both synthetic and real-world datasets demonstrate that our proposed model outperforms state-of-the-art methods in terms of discovering the time-evolving causality as well as forecasting.", "date": "2025-02-13", "categories": []}
{"id": "2602.16481", "title": "Leveraging Large Language Models for Causal Discovery: a Constraint-based, Argumentation-driven Approach", "abstract": "Causal discovery seeks to uncover causal relations from data, typically represented as causal graphs, and is essential for predicting the effects of interventions. While expert knowledge is required to construct principled causal graphs, many statistical methods have been proposed to leverage observational data with varying formal guarantees. Causal Assumption-based Argumentation (ABA) is a framework that uses symbolic reasoning to ensure correspondence between input constraints and output graphs, while offering a principled way to combine data and expertise. We explore the use of large language models (LLMs) as imperfect experts for Causal ABA, eliciting semantic structural priors from variable names and descriptions and integrating them with conditional-independence evidence. Experiments on standard benchmarks and semantically grounded synthetic graphs demonstrate state-of-the-art performance, and we additionally introduce an evaluation protocol to mitigate memorisation bias when assessing LLMs for causal discovery.", "date": "2026-02-18", "categories": []}
{"id": "2602.16612", "title": "Causal and Compositional Abstraction", "abstract": "Abstracting from a low level to a more explanatory high level of description, and ideally while preserving causal structure, is fundamental to scientific practice, to causal inference problems, and to robust, efficient and interpretable AI. We present a general account of abstractions between low and high level models as natural transformations, focusing on the case of causal models. This provides a new formalisation of causal abstraction, unifying several notions in the literature, including constructive causal abstraction, Q-$\\tau$ consistency, abstractions based on interchange interventions, and `distributed' causal abstractions. Our approach is formalised in terms of category theory, and uses the general notion of a compositional model with a given set of queries and semantics in a monoidal, cd- or Markov category; causal models and their queries such as interventions being special cases. We identify two basic notions of abstraction: downward abstractions mapping queries from high to low level; and upward abstractions, mapping concrete queries such as Do-interventions from low to high. Although usually presented as the latter, we show how common causal abstractions may, more fundamentally, be understood in terms of the former. Our approach also leads us to consider a new stronger notion of `component-level' abstraction, applying to the individual components of a model. In particular, this yields a novel, strengthened form of constructive causal abstraction at the mechanism-level, for which we prove characterisation results. Finally, we show that abstraction can be generalised to further compositional models, including those with a quantum semantics implemented by quantum circuits, and we take first steps in exploring abstractions between quantum compositional circuit models and high-level classical causal models as a means to explainable quantum AI.", "date": "2026-02-18", "categories": []}
{"id": "2405.21012", "title": "IGC-Net for conditional average potential outcome estimation over time", "abstract": "Estimating potential outcomes for treatments over time based on observational data is important for personalized decision-making in medicine. However, many existing methods for this task fail to properly adjust for time-varying confounding and thus yield biased estimates. There are only a few neural methods with proper adjustments, but these have inherent limitations (e.g., division by propensity scores that are often close to zero), which result in poor performance. As a remedy, we introduce the iterative G-computation network (IGC-Net). Our IGC-Net is a novel, neural end-to-end model which adjusts for time-varying confounding in order to estimate conditional average potential outcomes (CAPOs) over time. Specifically, our IGC-Net is the first neural model to perform fully regression-based iterative G-computation for CAPOs in the time-varying setting. We evaluate the effectiveness of our IGC-Net across various experiments. In sum, this work represents a significant step towards personalized decision-making from electronic health records.", "date": "2024-05-31", "categories": []}
{"id": "2502.13022", "title": "Efficient and Sharp Off-Policy Learning under Unobserved Confounding", "abstract": "We develop a novel method for personalized off-policy learning in scenarios with unobserved confounding. Thereby, we address a key limitation of standard policy learning: standard policy learning assumes unconfoundedness, meaning that no unobserved factors influence both treatment assignment and outcomes. However, this assumption is often violated, because of which standard policy learning produces biased estimates and thus leads to policies that can be harmful. To address this limitation, we employ causal sensitivity analysis and derive a semi-parametrically efficient estimator for a sharp bound on the value function under unobserved confounding. Our estimator has three advantages: (1) Unlike existing works, our estimator avoids unstable minimax optimization based on inverse propensity weighted outcomes. (2) Our estimator is semi-parametrically efficient. (3) We prove that our estimator leads to the optimal confounding-robust policy. Finally, we extend our theory to the related task of policy improvement under unobserved confounding, i.e., when a baseline policy such as the standard of care is available. We show in experiments with synthetic and real-world data that our method outperforms simple plug-in approaches and existing baselines. Our method is highly relevant for decision-making where unobserved confounding can be problematic, such as in healthcare and public policy.", "date": "2025-02-18", "categories": []}
{"id": "2508.11060", "title": "Counterfactual Survival Q-learning via Buckley-James Boosting, with Applications to ACTG 175 and CALGB 8923", "abstract": "We propose a Buckley James (BJ) Boost Q learning framework for estimating optimal dynamic treatment regimes from right censored survival outcomes in longitudinal randomized clinical trials, motivated by the clinical need to support patient specific treatment decisions when follow up is incomplete and covariate effects may be nonlinear. The method combines accelerated failure time modeling with iterative boosting using flexible base learners, including componentwise least squares and regression trees, within a counterfactual Q learning framework. By modeling conditional survival time directly, BJ Boost Q learning avoids the proportional hazards assumption, yields clinically interpretable time scale contrasts, and enables estimation of stage specific Q functions and individualized decision rules under standard potential outcomes assumptions. In contrast to Cox based Q learning, which relies on hazard modeling and can be sensitive to nonproportional hazards and model misspecification, our approach provides a robust and flexible alternative for regime learning. Simulation studies and analyses of the ACTG175 HIV trial and the CALGB 8923 two stage leukemia trial show that BJ Boost Q learning improves treatment decision accuracy and produces more stable within participant counterfactual contrasts, particularly in multistage settings where estimation error and bias can compound across stages.", "date": "2025-08-14", "categories": []}
{"id": "2602.15306", "title": "Sparse Additive Model Pruning for Order-Based Causal Structure Learning", "abstract": "Causal structure learning, also known as causal discovery, aims to estimate causal relationships between variables as a form of a causal directed acyclic graph (DAG) from observational data. One of the major frameworks is the order-based approach that first estimates a topological order of the underlying DAG and then prunes spurious edges from the fully-connected DAG induced by the estimated topological order. Previous studies often focus on the former ordering step because it can dramatically reduce the search space of DAGs. In practice, the latter pruning step is equally crucial for ensuring both computational efficiency and estimation accuracy. Most existing methods employ a pruning technique based on generalized additive models and hypothesis testing, commonly known as CAM-pruning. However, this approach can be a computational bottleneck as it requires repeatedly fitting additive models for all variables. Furthermore, it may harm estimation quality due to multiple testing. To address these issues, we introduce a new pruning method based on sparse additive models, which enables direct pruning of redundant edges without relying on hypothesis testing. We propose an efficient algorithm for learning sparse additive models by combining the randomized tree embedding technique with group-wise sparse regression. Experimental results on both synthetic and real datasets demonstrated that our method is significantly faster than existing pruning methods while maintaining comparable or superior accuracy.", "date": "2026-02-17", "categories": []}
{"id": "2602.15546", "title": "CEPAE: Conditional Entropy-Penalized Autoencoders for Time Series Counterfactuals", "abstract": "The ability to accurately perform counterfactual inference on time series is crucial for decision-making in fields like finance, healthcare, and marketing, as it allows us to understand the impact of events or treatments on outcomes over time. In this paper, we introduce a new counterfactual inference approach tailored to time series data impacted by market events, which is motivated by an industrial application. Utilizing the abduction-action-prediction procedure and the Structural Causal Model framework, we first adapt methods based on variational autoencoders and adversarial autoencoders, both previously used in counterfactual literature although not in time series settings. Then, we present the Conditional Entropy-Penalized Autoencoder (CEPAE), a novel autoencoder-based approach for counterfactual inference, which employs an entropy penalization loss over the latent space to encourage disentangled data representations. We validate our approach both theoretically and experimentally on synthetic, semi-synthetic, and real-world datasets, showing that CEPAE generally outperforms the other approaches in the evaluated metrics.", "date": "2026-02-17", "categories": []}
{"id": "2602.15730", "title": "Causal Effect Estimation with Latent Textual Treatments", "abstract": "Understanding the causal effects of text on downstream outcomes is a central task in many applications. Estimating such effects requires researchers to run controlled experiments that systematically vary textual features. While large language models (LLMs) hold promise for generating text, producing and evaluating controlled variation requires more careful attention. In this paper, we present an end-to-end pipeline for the generation and causal estimation of latent textual interventions. Our work first performs hypothesis generation and steering via sparse autoencoders (SAEs), followed by robust causal estimation. Our pipeline addresses both computational and statistical challenges in text-as-treatment experiments. We demonstrate that naive estimation of causal effects suffers from significant bias as text inherently conflates treatment and covariate information. We describe the estimation bias induced in this setting and propose a solution based on covariate residualization. Our empirical results show that our pipeline effectively induces variation in target features and mitigates estimation error, providing a robust foundation for causal effect estimation in text-as-treatment settings.", "date": "2026-02-17", "categories": []}
{"id": "2402.02644", "title": "Permutation-based Inference for Variational Learning of Directed Acyclic Graphs", "abstract": "Estimating the structure of Bayesian networks as directed acyclic graphs (DAGs) from observational data is a fundamental challenge, particularly in causal discovery. Bayesian approaches excel by quantifying uncertainty and addressing identifiability, but key obstacles remain: (i) representing distributions over DAGs and (ii) estimating a posterior in the underlying combinatorial space. We introduce PIVID, a method that jointly infers a distribution over permutations and DAGs using variational inference and continuous relaxations of discrete distributions. Through experiments on synthetic and real-world datasets, we show that PIVID can outperform deterministic and Bayesian approaches, achieving superior accuracy-uncertainty trade-offs while scaling efficiently with the number of variables.", "date": "2024-02-04", "categories": []}
{"id": "2406.04955", "title": "Experimental Evaluation of ROS-Causal in Real-World Human-Robot Spatial Interaction Scenarios", "abstract": "Deploying robots in human-shared environments requires a deep understanding of how nearby agents and objects interact. Employing causal inference to model cause-and-effect relationships facilitates the prediction of human behaviours and enables the anticipation of robot interventions. However, a significant challenge arises due to the absence of implementation of existing causal discovery methods within the ROS ecosystem, the standard de-facto framework in robotics, hindering effective utilisation on real robots. To bridge this gap, in our previous work we proposed ROS-Causal, a ROS-based framework designed for onboard data collection and causal discovery in human-robot spatial interactions. In this work, we present an experimental evaluation of ROS-Causal both in simulation and on a new dataset of human-robot spatial interactions in a lab scenario, to assess its performance and effectiveness. Our analysis demonstrates the efficacy of this approach, showcasing how causal models can be extracted directly onboard by robots during data collection. The online causal models generated from the simulation are consistent with those from lab experiments. These findings can help researchers to enhance the performance of robotic systems in shared environments, firstly by studying the causal relations between variables in simulation without real people, and then facilitating the actual robot deployment in real human environments. ROS-Causal: https://lcastri.github.io/roscausal", "date": "2024-06-07", "categories": []}
{"id": "2601.00097", "title": "The Agentic Leash: Extracting Causal Feedback Fuzzy Cognitive Maps with LLMs", "abstract": "We design a large-language-model (LLM) agent system that extracts causal feedback fuzzy cognitive maps (FCMs) from raw text. The causal learning or extraction process is agentic both because of the LLM's semi-autonomy and because ultimately the FCM dynamical system's equilibria drive the LLM agents to fetch and process causal text. The fetched text can in principle modify the adaptive FCM causal structure and so modify the source of its quasi-autonomy$-$its equilibrium limit cycles and fixed-point attractors. This bidirectional process endows the evolving FCM dynamical system with a degree of autonomy while the system still stays on its agentic leash. We show in particular that a sequence of three system-instruction sets guide an LLM agent as it systematically extracts key nouns and noun phrases from text, as it extracts FCM concept nodes from among those nouns and noun phrases, and then as it extracts or infers partial or fuzzy causal edges between those FCM nodes. We test this FCM generation on a recent essay about the promise of AI from the late diplomat and political theorist Henry Kissinger and his colleagues. This three-step process produced FCM dynamical systems that converged to the same equilibrium limit cycles as did the human-generated FCMs even though the human-generated FCM differed in the number of nodes and edges. A final FCM mixed generated FCMs from separate Gemini and ChatGPT LLM agents. The mixed FCM absorbed the equilibria of its dominant mixture component but also created new equilibria of its own to better approximate the underlying causal dynamical system.", "date": "2025-12-31", "categories": []}
{"id": "2602.13339", "title": "An Integrated Causal Inference Framework for Traffic Safety Modeling with Semantic Street-View Visual Features", "abstract": "Macroscopic traffic safety modeling aims to identify critical risk factors for regional crashes, thereby informing targeted policy interventions for safety improvement. However, current approaches rely heavily on static sociodemographic and infrastructure metrics, frequently overlooking the impacts from drivers' visual perception of driving environment. Although visual environment features have been found to impact driving and traffic crashes, existing evidence remains largely observational, failing to establish the robust causality for traffic policy evaluation under complex spatial environment. To fill these gaps, we applied semantic segmentation on Google Street View imageries to extract visual environmental features and proposed a Double Machine Learning framework to quantify their causal effects on regional crashes. Meanwhile, we utilized SHAP values to characterize the nonlinear influence mechanisms of confounding variables in the models and applied causal forests to estimate conditional average treatment effects. Leveraging crash records from the Miami metropolitan area, Florida, and 220,000 street view images, evidence shows that greenery proportion exerts a significant and robust negative causal effect on traffic crashes (Average Treatment Effect = -6.38, p = 0.005). This protective effect exhibits spatial heterogeneity, being most pronounced in densely populated and socially vulnerable urban cores. While greenery significantly mitigates angle and rear-end crashes, its protective benefit for vulnerable road users (VRUs) remains limited. Our findings provide causal evidence for greening as a potential safety intervention, prioritizing hazardous visual environments while highlighting the need for distinct design optimizations to protect VRUs.", "date": "2026-02-12", "categories": []}
{"id": "2602.14274", "title": "Integrating Unstructured Text into Causal Inference: Empirical Evidence from Real Data", "abstract": "Causal inference, a critical tool for informing business decisions, traditionally relies heavily on structured data. However, in many real-world scenarios, such data can be incomplete or unavailable. This paper presents a framework that leverages transformer-based language models to perform causal inference using unstructured text. We demonstrate the effectiveness of our framework by comparing causal estimates derived from unstructured text against those obtained from structured data across population, group, and individual levels. Our findings show consistent results between the two approaches, validating the potential of unstructured text in causal inference tasks. Our approach extends the applicability of causal inference methods to scenarios where only textual data is available, enabling data-driven business decision-making when structured tabular data is scarce.", "date": "2026-02-15", "categories": []}
{"id": "2602.14456", "title": "Traceable Latent Variable Discovery Based on Multi-Agent Collaboration", "abstract": "Revealing the underlying causal mechanisms in the real world is crucial for scientific and technological progress. Despite notable advances in recent decades, the lack of high-quality data and the reliance of traditional causal discovery algorithms (TCDA) on the assumption of no latent confounders, as well as their tendency to overlook the precise semantics of latent variables, have long been major obstacles to the broader application of causal discovery. To address this issue, we propose a novel causal modeling framework, TLVD, which integrates the metadata-based reasoning capabilities of large language models (LLMs) with the data-driven modeling capabilities of TCDA for inferring latent variables and their semantics. Specifically, we first employ a data-driven approach to construct a causal graph that incorporates latent variables. Then, we employ multi-LLM collaboration for latent variable inference, modeling this process as a game with incomplete information and seeking its Bayesian Nash Equilibrium (BNE) to infer the possible specific latent variables. Finally, to validate the inferred latent variables across multiple real-world web-based data sources, we leverage LLMs for evidence exploration to ensure traceability. We comprehensively evaluate TLVD on three de-identified real patient datasets provided by a hospital and two benchmark datasets. Extensive experimental results confirm the effectiveness and reliability of TLVD, with average improvements of 32.67% in Acc, 62.21% in CAcc, and 26.72% in ECit across the five datasets.", "date": "2026-02-16", "categories": []}
{"id": "2602.14503", "title": "Bounding Probabilities of Causation with Partial Causal Diagrams", "abstract": "Probabilities of causation are fundamental to individual-level explanation and decision making, yet they are inherently counterfactual and not point-identifiable from data in general. Existing bounds either disregard available covariates, require complete causal graphs, or rely on restrictive binary settings, limiting their practical use. In real-world applications, causal information is often partial but nontrivial. This paper proposes a general framework for bounding probabilities of causation using partial causal information. We show how the available structural or statistical information can be systematically incorporated as constraints in a optimization programming formulation, yielding tighter and formally valid bounds without full identifiability. This approach extends the applicability of probabilities of causation to realistic settings where causal knowledge is incomplete but informative.", "date": "2026-02-16", "categories": []}
{"id": "2602.14791", "title": "Extending Multi-Source Bayesian Optimization With Causality Principles", "abstract": "Multi-Source Bayesian Optimization (MSBO) serves as a variant of the traditional Bayesian Optimization (BO) framework applicable to situations involving optimization of an objective black-box function over multiple information sources such as simulations, surrogate models, or real-world experiments. However, traditional MSBO assumes the input variables of the objective function to be independent and identically distributed, limiting its effectiveness in scenarios where causal information is available and interventions can be performed, such as clinical trials or policy-making. In the single-source domain, Causal Bayesian Optimization (CBO) extends standard BO with the principles of causality, enabling better modeling of variable dependencies. This leads to more accurate optimization, improved decision-making, and more efficient use of low-cost information sources. In this article, we propose a principled integration of the MSBO and CBO methodologies in the multi-source domain, leveraging the strengths of both to enhance optimization efficiency and reduce computational complexity in higher-dimensional problems. We present the theoretical foundations of both Causal and Multi-Source Bayesian Optimization, and demonstrate how their synergy informs our Multi-Source Causal Bayesian Optimization (MSCBO) algorithm. We compare the performance of MSCBO against its foundational counterparts for both synthetic and real-world datasets with varying levels of noise, highlighting the robustness and applicability of MSCBO. Based on our findings, we conclude that integrating MSBO with the causality principles of CBO facilitates dimensionality reduction and lowers operational costs, ultimately improving convergence speed, performance, and scalability.", "date": "2026-02-16", "categories": []}
{"id": "2602.14972", "title": "Use What You Know: Causal Foundation Models with Partial Graphs", "abstract": "Estimating causal quantities traditionally relies on bespoke estimators tailored to specific assumptions. Recently proposed Causal Foundation Models (CFMs) promise a more unified approach by amortising causal discovery and inference in a single step. However, in their current state, they do not allow for the incorporation of any domain knowledge, which can lead to suboptimal predictions. We bridge this gap by introducing methods to condition CFMs on causal information, such as the causal graph or more readily available ancestral information. When access to complete causal graph information is too strict a requirement, our approach also effectively leverages partial causal information. We systematically evaluate conditioning strategies and find that injecting learnable biases into the attention mechanism is the most effective method to utilise full and partial causal information. Our experiments show that this conditioning allows a general-purpose CFM to match the performance of specialised models trained on specific causal structures. Overall, our approach addresses a central hurdle on the path towards all-in-one causal foundation models: the capability to answer causal queries in a data-driven manner while effectively leveraging any amount of domain expertise.", "date": "2026-02-16", "categories": []}
{"id": "2409.16407", "title": "Towards Representation Learning for Weighting Problems in Design-Based Causal Inference", "abstract": "Reweighting a distribution to minimize a distance to a target distribution is a powerful and flexible strategy for estimating a wide range of causal effects, but can be challenging in practice because optimal weights typically depend on knowledge of the underlying data generating process. In this paper, we focus on design-based weights, which do not incorporate outcome information; prominent examples include prospective cohort studies, survey weighting, and the weighting portion of augmented weighting estimators. In such applications, we explore the central role of representation learning in finding desirable weights in practice. Unlike the common approach of assuming a well-specified representation, we highlight the error due to the choice of a representation and outline a general framework for finding suitable representations that minimize this error. Building on recent work that combines balancing weights and neural networks, we propose an end-to-end estimation procedure that learns a flexible representation, while retaining promising theoretical properties. We show that this approach is competitive in a range of common causal inference tasks.", "date": "2024-09-24", "categories": []}
{"id": "2511.21537", "title": "Context-Specific Causal Graph Discovery with Unobserved Contexts: Non-Stationarity, Regimes and Spatio-Temporal Patterns", "abstract": "Real-world problems, for example in climate applications, often require causal reasoning on spatially gridded time series data or data with comparable structure. While the underlying system is often believed to behave similarly at different Points in space and time, those variations that do exist are relevant twofold: They often encode important information in and of themselves. And they may negatively affect the stability and validity of results if not accounted for. We study the information encoded in changes of the causal graph, with stability in mind. Two core challenges arise, related to the complexity of encoding system-states and to statistical convergence properties in the presence of imperfectly recoverable non-stationary structure. We provide a framework realizing principles conceptually suitable to overcome these challenges - an interpretation supported by numerical experiments. Primarily, we modify constraint-based causal discovery approaches on the level of independence testing. This leads to a framework which is additionally highly modular, easily extensible and widely applicable. For example, it allows to leverage existing constraint-based causal discovery methods (demonstrated on PC, PC-stable, FCI, PCMCI, PCMCI+ and LPCMCI), and to systematically divide the problem into simpler subproblems that are easier to analyze and understand and relate more clearly to well-studied problems like change-point-detection, clustering, independence-testing and more. Code is available at https://github.com/martin-rabel/Causal_GLDF.", "date": "2025-11-26", "categories": []}
{"id": "2602.12592", "title": "Power Interpretable Causal ODE Networks: A Unified Model for Explainable Anomaly Detection and Root Cause Analysis in Power Systems", "abstract": "Anomaly detection and root cause analysis (RCA) are critical for ensuring the safety and resilience of cyber-physical systems such as power grids. However, existing machine learning models for time series anomaly detection often operate as black boxes, offering only binary outputs without any explanation, such as identifying anomaly type and origin. To address this challenge, we propose Power Interpretable Causality Ordinary Differential Equation (PICODE) Networks, a unified, causality-informed architecture that jointly performs anomaly detection along with the explanation why it is detected as an anomaly, including root cause localization, anomaly type classification, and anomaly shape characterization. Experimental results in power systems demonstrate that PICODE achieves competitive detection performance while offering improved interpretability and reduced reliance on labeled data or external causal graphs. We provide theoretical results demonstrating the alignment between the shape of anomaly functions and the changes in the weights of the extracted causal graphs.", "date": "2026-02-13", "categories": []}
{"id": "2602.11350", "title": "Structured Hybrid Mechanistic Models for Robust Estimation of Time-Dependent Intervention Outcomes", "abstract": "Estimating intervention effects in dynamical systems is crucial for outcome optimization. In medicine, such interventions arise in physiological regulation (e.g., cardiovascular system under fluid administration) and pharmacokinetics, among others. Propofol administration is an anesthetic intervention, where the challenge is to estimate the optimal dose required to achieve a target brain concentration for anesthesia, given patient characteristics, while avoiding under- or over-dosing. The pharmacokinetic state is characterized by drug concentrations across tissues, and its dynamics are governed by prior states, patient covariates, drug clearance, and drug administration. While data-driven models can capture complex dynamics, they often fail in out-of-distribution (OOD) regimes. Mechanistic models on the other hand are typically robust, but might be oversimplified. We propose a hybrid mechanistic-data-driven approach to estimate time-dependent intervention outcomes. Our approach decomposes the dynamical system's transition operator into parametric and nonparametric components, further distinguishing between intervention-related and unrelated dynamics. This structure leverages mechanistic anchors while learning residual patterns from data. For scenarios where mechanistic parameters are unknown, we introduce a two-stage procedure: first, pre-training an encoder on simulated data, and subsequently learning corrections from observed data. Two regimes with incomplete mechanistic knowledge are considered: periodic pendulum and Propofol bolus injections. Results demonstrate that our hybrid approach outperforms purely data-driven and mechanistic approaches, particularly OOD. This work highlights the potential of hybrid mechanistic-data-driven models for robust intervention optimization in complex, real-world dynamical systems.", "date": "2026-02-11", "categories": []}
{"id": "2602.11527", "title": "CausalAgent: A Conversational Multi-Agent System for End-to-End Causal Inference", "abstract": "Causal inference holds immense value in fields such as healthcare, economics, and social sciences. However, traditional causal analysis workflows impose significant technical barriers, requiring researchers to possess dual backgrounds in statistics and computer science, while manually selecting algorithms, handling data quality issues, and interpreting complex results. To address these challenges, we propose CausalAgent, a conversational multi-agent system for end-to-end causal inference. The system innovatively integrates Multi-Agent Systems (MAS), Retrieval-Augmented Generation (RAG), and the Model Context Protocol (MCP) to achieve automation from data cleaning and causal structure learning to bias correction and report generation through natural language interaction. Users need only upload a dataset and pose questions in natural language to receive a rigorous, interactive analysis report. As a novel user-centered human-AI collaboration paradigm, CausalAgent explicitly models the analysis workflow. By leveraging interactive visualizations, it significantly lowers the barrier to entry for causal analysis while ensuring the rigor and interpretability of the process.", "date": "2026-02-12", "categories": []}
{"id": "2510.19643", "title": "Overlap-weighted orthogonal meta-learner for treatment effect estimation over time", "abstract": "Estimating heterogeneous treatment effects (HTEs) in time-varying settings is particularly challenging, as the probability of observing certain treatment sequences decreases exponentially with longer prediction horizons. Thus, the observed data contain little support for many plausible treatment sequences, which creates severe overlap problems. Existing meta-learners for the time-varying setting typically assume adequate treatment overlap, and thus suffer from exploding estimation variance when the overlap is low. To address this problem, we introduce a novel overlap-weighted orthogonal (WO) meta-learner for estimating HTEs that targets regions in the observed data with high probability of receiving the interventional treatment sequences. This offers a fully data-driven approach through which our WO-learner can counteract instabilities as in existing meta-learners and thus obtain more reliable HTE estimates. Methodologically, we develop a novel Neyman-orthogonal population risk function that minimizes the overlap-weighted oracle risk. We show that our WO-learner has the favorable property of Neyman-orthogonality, meaning that it is robust against misspecification in the nuisance functions. Further, our WO-learner is fully model-agnostic and can be applied to any machine learning model. Through extensive experiments with both transformer and LSTM backbones, we demonstrate the benefits of our novel WO-learner.", "date": "2025-10-22", "categories": []}
{"id": "2601.21014", "title": "Efficient Causal Structure Learning via Modular Subgraph Integration", "abstract": "Learning causal structures from observational data remains a fundamental yet computationally intensive task, particularly in high-dimensional settings where existing methods face challenges such as the super-exponential growth of the search space and increasing computational demands. To address this, we introduce VISTA (Voting-based Integration of Subgraph Topologies for Acyclicity), a modular framework that decomposes the global causal structure learning problem into local subgraphs based on Markov Blankets. The global integration is achieved through a weighted voting mechanism that penalizes low-support edges via exponential decay, filters unreliable ones with an adaptive threshold, and ensures acyclicity using a Feedback Arc Set (FAS) algorithm. The framework is model-agnostic, imposing no assumptions on the inductive biases of base learners, is compatible with arbitrary data settings without requiring specific structural forms, and fully supports parallelization. We also theoretically establish finite-sample error bounds for VISTA, and prove its asymptotic consistency under mild conditions. Extensive experiments on both synthetic and real datasets consistently demonstrate the effectiveness of VISTA, yielding notable improvements in both accuracy and efficiency over a wide range of base learners.", "date": "2026-01-28", "categories": []}
{"id": "2602.10282", "title": "Linear-LLM-SCM: Benchmarking LLMs for Coefficient Elicitation in Linear-Gaussian Causal Models", "abstract": "Large language models (LLMs) have shown potential in identifying qualitative causal relations, but their ability to perform quantitative causal reasoning -- estimating effect sizes that parametrize functional relationships -- remains underexplored in continuous domains. We introduce Linear-LLM-SCM, a plug-and-play benchmarking framework for evaluating LLMs on linear Gaussian structural causal model (SCM) parametrization when the DAG is given. The framework decomposes a DAG into local parent-child sets and prompts an LLM to produce a regression-style structural equation per node, which is aggregated and compared against available ground-truth parameters. Our experiments show several challenges in such benchmarking tasks, namely, strong stochasticity in the results in some of the models and susceptibility to DAG misspecification via spurious edges in the continuous domains. Across models, we observe substantial variability in coefficient estimates for some settings and sensitivity to structural and semantic perturbations, highlighting current limitations of LLMs as quantitative causal parameterizers. We also open-sourced the benchmarking framework so that researchers can utilize their DAGs and any off-the-shelf LLMs plug-and-play for evaluation in their domains effortlessly.", "date": "2026-02-10", "categories": []}
{"id": "2602.10305", "title": "Confounding Robust Continuous Control via Automatic Reward Shaping", "abstract": "Reward shaping has been applied widely to accelerate Reinforcement Learning (RL) agents' training. However, a principled way of designing effective reward shaping functions, especially for complex continuous control problems, remains largely under-explained. In this work, we propose to automatically learn a reward shaping function for continuous control problems from offline datasets, potentially contaminated by unobserved confounding variables. Specifically, our method builds upon the recently proposed causal Bellman equation to learn a tight upper bound on the optimal state values, which is then used as the potentials in the Potential-Based Reward Shaping (PBRS) framework. Our proposed reward shaping algorithm is tested with Soft-Actor-Critic (SAC) on multiple commonly used continuous control benchmarks and exhibits strong performance guarantees under unobserved confounders. More broadly, our work marks a solid first step towards confounding robust continuous control from a causal perspective. Code for training our reward shaping functions can be found at https://github.com/mateojuliani/confounding_robust_cont_control.", "date": "2026-02-10", "categories": []}
{"id": "2505.17133", "title": "Learning Probabilities of Causation with Mask-Augmented Data", "abstract": "Probabilities of causation play a central role in modern decision making. Tian and Pearl first introduced formal definitions and derived tight bounds for three binary probabilities of causation, such as the probability of necessity and sufficiency (PNS). However, estimating these probabilities requires both experimental and observational distributions specific to each subpopulation, which are often unreliable or impractical to obtain from limited population-level data. To solve this problem, we propose two machine learning models: Exact-MLP and Mask-MLP, which are trained on a small set of reliable subpopulations and are able to predict PNS bounds for all other subpopulations. We validate our models across four Structural Causal Models (SCMs), each evaluated on population-level data with sample sizes between 100k and 200k. Our models achieve average mean absolute errors (MAEs) of roughly 0.03 on main tasks, reducing MAE by about 80% relative to the corresponding baselines. These results demonstrate both the feasibility of machine learning models for learning probabilities of causation and the effectiveness of the proposed approach.", "date": "2025-05-22", "categories": []}
{"id": "2507.05526", "title": "Estimating Interventional Distributions with Uncertain Causal Graphs through Meta-Learning", "abstract": "In scientific domains -- from biology to the social sciences -- many questions boil down to \\textit{What effect will we observe if we intervene on a particular variable?} If the causal relationships (e.g.~a causal graph) are known, it is possible to estimate the intervention distributions. In the absence of this domain knowledge, the causal structure must be discovered from the available observational data. However, observational data are often compatible with multiple causal graphs, making methods that commit to a single structure prone to overconfidence. A principled way to manage this structural uncertainty is via Bayesian inference, which averages over a posterior distribution on possible causal structures and functional mechanisms. Unfortunately, the number of causal structures grows super-exponentially with the number of nodes in the graph, making computations intractable. We propose to circumvent these challenges by using meta-learning to create an end-to-end model: the Model-Averaged Causal Estimation Transformer Neural Process (MACE-TNP). The model is trained to predict the Bayesian model-averaged interventional posterior distribution, and its end-to-end nature bypasses the need for expensive calculations. Empirically, we demonstrate that MACE-TNP outperforms strong Bayesian baselines. Our work establishes meta-learning as a flexible and scalable paradigm for approximating complex Bayesian causal inference, that can be scaled to increasingly challenging settings in the future.", "date": "2025-07-07", "categories": []}
{"id": "2601.07752", "title": "A Unified Framework for Debiased Machine Learning: Riesz Representer Fitting under Bregman Divergence", "abstract": "Estimating the Riesz representer is central to debiased machine learning for causal and structural parameter estimation. We propose generalized Riesz regression, a unified framework for estimating the Riesz representer by fitting a representer model via Bregman divergence minimization. This framework includes various divergences as special cases, such as the squared distance and the Kullback--Leibler (KL) divergence, where the former recovers Riesz regression and the latter recovers tailored loss minimization. Under suitable pairs of divergence and model specifications (link functions), the dual problems of the Riesz representer fitting problem correspond to covariate balancing, which we call automatic covariate balancing. Moreover, under the same specifications, the sample average of outcomes weighted by the estimated Riesz representer satisfies Neyman orthogonality even without estimating the regression function, a property we call automatic Neyman orthogonalization. This property not only reduces the estimation error of Neyman orthogonal scores but also clarifies a key distinction between debiased machine learning and targeted maximum likelihood estimation (TMLE). Our framework can also be viewed as a generalization of density ratio fitting under Bregman divergences to Riesz representer estimation, and it applies beyond density ratio estimation. We provide convergence analyses for both reproducing kernel Hilbert space (RKHS) and neural network model classes. A Python package for generalized Riesz regression is released as genriesz and is available at https://github.com/MasaKat0/genriesz.", "date": "2026-01-12", "categories": []}
{"id": "2602.09207", "title": "CausalGDP: Causality-Guided Diffusion Policies for Reinforcement Learning", "abstract": "Reinforcement learning (RL) has achieved remarkable success in a wide range of sequential decision-making problems. Recent diffusion-based policies further improve RL by modeling complex, high-dimensional action distributions. However, existing diffusion policies primarily rely on statistical associations and fail to explicitly account for causal relationships among states, actions, and rewards, limiting their ability to identify which action components truly cause high returns. In this paper, we propose Causality-guided Diffusion Policy (CausalGDP), a unified framework that integrates causal reasoning into diffusion-based RL. CausalGDP first learns a base diffusion policy and an initial causal dynamical model from offline data, capturing causal dependencies among states, actions, and rewards. During real-time interaction, the causal information is continuously updated and incorporated as a guidance signal to steer the diffusion process toward actions that causally influence future states and rewards. By explicitly considering causality beyond association, CausalGDP focuses policy optimization on action components that genuinely drive performance improvements. Experimental results demonstrate that CausalGDP consistently achieves competitive or superior performance over state-of-the-art diffusion-based and offline RL methods, especially in complex, high-dimensional control tasks.", "date": "2026-02-09", "categories": []}
{"id": "2510.18161", "title": "Beating the Winner's Curse via Inference-Aware Policy Optimization", "abstract": "There has been a surge of recent interest in automatically learning policies to target treatment decisions based on rich individual covariates. In addition, practitioners want confidence that the learned policy has better performance than the incumbent policy according to downstream policy evaluation. However, due to the winner's curse -- an issue where the policy optimization procedure exploits prediction errors rather than finding actual improvements -- predicted performance improvements are often not substantiated by downstream policy evaluation. To address this challenge, we propose a novel strategy called inference-aware policy optimization, which modifies policy optimization to account for how the policy will be evaluated downstream. Specifically, it optimizes not only for the estimated objective value, but also for the chances that the estimate of the policy's improvement passes a significance test during downstream policy evaluation. We mathematically characterize the Pareto frontier of policies according to the tradeoff of these two goals. Based on our characterization, we design a policy optimization algorithm that estimates the Pareto frontier using machine learning models; then, the decision-maker can select the policy that optimizes their desired tradeoff, after which policy evaluation can be performed on the test set as usual. Finally, we perform simulations to illustrate the effectiveness of our methodology.", "date": "2025-10-20", "categories": []}
{"id": "2602.07203", "title": "Exactly Computing do-Shapley Values", "abstract": "Structural Causal Models (SCM) are a powerful framework for describing complicated dynamics across the natural sciences. A particularly elegant way of interpreting SCMs is do-Shapley, a game-theoretic method of quantifying the average effect of $d$ variables across exponentially many interventions. Like Shapley values, computing do-Shapley values generally requires evaluating exponentially many terms. The foundation of our work is a reformulation of do-Shapley values in terms of the irreducible sets of the underlying SCM. Leveraging this insight, we can exactly compute do-Shapley values in time linear in the number of irreducible sets $r$, which itself can range from $d$ to $2^d$ depending on the graph structure of the SCM. Since $r$ is unknown a priori, we complement the exact algorithm with an estimator that, like general Shapley value estimators, can be run with any query budget. As the query budget approaches $r$, our estimators can produce more accurate estimates than prior methods by several orders of magnitude, and, when the budget reaches $r$, return the Shapley values up to machine precision. Beyond computational speed, we also reduce the identification burden: we prove that non-parametric identifiability of do-Shapley values requires only the identification of interventional effects for the $d$ singleton coalitions, rather than all classes.", "date": "2026-02-06", "categories": []}
{"id": "2602.07233", "title": "Extracting Root-Causal Brain Activity Driving Psychopathology from Resting State fMRI", "abstract": "Neuroimaging studies of psychiatric disorders often correlate imaging patterns with diagnostic labels or composite symptom scores, yielding diffuse associations that obscure underlying mechanisms. We instead seek to identify root-causal maps -- localized BOLD disturbances that initiate pathological cascades -- and to link them selectively to symptom dimensions. We introduce a bilevel structural causal model that connects between-subject symptom structure to within-subject resting-state fMRI via independent latent sources with localized direct effects. Based on this model, we develop SOURCE (Symptom-Oriented Uncovering of Root-Causal Elements), a procedure that links interpretable symptom axes to a parsimonious set of localized drivers. Experiments show that SOURCE recovers localized maps consistent with root-causal BOLD drivers and increases interpretability and anatomical specificity relative to existing comparators.", "date": "2026-02-06", "categories": []}
{"id": "2602.07765", "title": "Disentangled Instrumental Variables for Causal Inference with Networked Observational Data", "abstract": "Instrumental variables (IVs) are crucial for addressing unobservable confounders, yet their stringent exogeneity assumptions pose significant challenges in networked data. Existing methods typically rely on modelling neighbour information when recovering IVs, thereby inevitably mixing shared environment-induced endogenous correlations and individual-specific exogenous variation, leading the resulting IVs to inherit dependence on unobserved confounders and to violate exogeneity. To overcome this challenge, we propose $\\underline{Dis}$entangled $\\underline{I}$nstrumental $\\underline{V}$ariables (DisIV) framework, a novel method for causal inference based on networked observational data with latent confounders. DisIV exploits network homogeneity as an inductive bias and employs a structural disentanglement mechanism to extract individual-specific components that serve as latent IVs. The causal validity of the extracted IVs is constrained through explicit orthogonality and exclusion conditions. Extensive semi-synthetic experiments on real-world datasets demonstrate that DisIV consistently outperforms state-of-the-art baselines in causal effect estimation under network-induced confounding.", "date": "2026-02-08", "categories": []}
{"id": "2602.08171", "title": "A Causal Machine Learning Framework for Treatment Personalization in Clinical Trials: Application to Ulcerative Colitis", "abstract": "Randomized controlled trials estimate average treatment effects, but treatment response heterogeneity motivates personalized approaches. A critical question is whether statistically detectable heterogeneity translates into improved treatment decisions -- these are distinct questions that can yield contradictory answers. We present a modular causal machine learning framework that evaluates each question separately: permutation importance identifies which features predict heterogeneity, best linear predictor (BLP) testing assesses statistical significance, and doubly robust policy evaluation measures whether acting on the heterogeneity improves patient outcomes. We apply this framework to patient-level data from the UNIFI maintenance trial of ustekinumab in ulcerative colitis, comparing placebo, standard-dose ustekinumab every 12 weeks, and dose-intensified ustekinumab every 8 weeks, using cross-fitted X-learner models with baseline demographics, medication history, week-8 clinical scores, laboratory biomarkers, and video-derived endoscopic features. BLP testing identified strong associations between endoscopic features and treatment effect heterogeneity for ustekinumab versus placebo, yet doubly robust policy evaluation showed no improvement in expected remission from incorporating endoscopic features, and out-of-fold multi-arm evaluation showed worse performance. Diagnostic comparison of prognostic contribution against policy value revealed that endoscopic scores behaved as disease severity markers -- improving outcome prediction in untreated patients but adding noise to treatment selection -- while clinical variables (fecal calprotectin, age, CRP) captured the decision-relevant variation. These results demonstrate that causal machine learning applications to clinical trials should include policy-level evaluation alongside heterogeneity testing.", "date": "2026-02-09", "categories": []}
{"id": "2602.08315", "title": "Fast Flow Matching based Conditional Independence Tests for Causal Discovery", "abstract": "Constraint-based causal discovery methods require a large number of conditional independence (CI) tests, which severely limits their practical applicability due to high computational complexity. Therefore, it is crucial to design an algorithm that accelerates each individual test. To this end, we propose the Flow Matching-based Conditional Independence Test (FMCIT). The proposed test leverages the high computational efficiency of flow matching and requires the model to be trained only once throughout the entire causal discovery procedure, substantially accelerating causal discovery. According to numerical experiments, FMCIT effectively controls type-I error and maintains high testing power under the alternative hypothesis, even in the presence of high-dimensional conditioning sets. In addition, we further integrate FMCIT into a two-stage guided PC skeleton learning framework, termed GPC-FMCIT, which combines fast screening with guided, budgeted refinement using FMCIT. This design yields explicit bounds on the number of CI queries while maintaining high statistical power. Experiments on synthetic and real-world causal discovery tasks demonstrate favorable accuracy-efficiency trade-offs over existing CI testing methods and PC variants.", "date": "2026-02-09", "categories": []}
{"id": "2602.08340", "title": "Effect-Level Validation for Causal Discovery", "abstract": "Causal discovery is increasingly applied to large-scale telemetry data to estimate the effects of user-facing interventions, yet its reliability for decision-making in feedback-driven systems with strong self-selection remains unclear. In this paper, we propose an effect-centric, admissibility-first framework that treats discovered graphs as structural hypotheses and evaluates them by identifiability, stability, and falsification rather than by graph recovery accuracy alone. Empirically, we study the effect of early exposure to competitive gameplay on short-term retention using real-world game telemetry. We find that many statistically plausible discovery outputs do not admit point-identified causal queries once minimal temporal and semantic constraints are enforced, highlighting identifiability as a critical bottleneck for decision support. When identification is possible, several algorithm families converge to similar, decision-consistent effect estimates despite producing substantially different graph structures, including cases where the direct treatment-outcome edge is absent and the effect is preserved through indirect causal pathways. These converging estimates survive placebo, subsampling, and sensitivity refutation. In contrast, other methods exhibit sporadic admissibility and threshold-sensitive or attenuated effects due to endpoint ambiguity. These results suggest that graph-level metrics alone are inadequate proxies for causal reliability for a given target query. Therefore, trustworthy causal conclusions in telemetry-driven systems require prioritizing admissibility and effect-level validation over causal structural recovery alone.", "date": "2026-02-09", "categories": []}
{"id": "2602.08461", "title": "Estimating Aleatoric Uncertainty in the Causal Treatment Effect", "abstract": "Previous work on causal inference has primarily focused on averages and conditional averages of treatment effects, with significantly less attention on variability and uncertainty in individual treatment responses. In this paper, we introduce the variance of the treatment effect (VTE) and conditional variance of treatment effect (CVTE) as the natural measure of aleatoric uncertainty inherent in treatment responses, and we demonstrate that these quantities are identifiable from observed data under mild assumptions, even in the presence of unobserved confounders. We further propose nonparametric kernel-based estimators for VTE and CVTE, and our theoretical analysis establishes their convergence. We also test the performance of our method through extensive empirical experiments on both synthetic and semi-simulated datasets, where it demonstrates superior or comparable performance to naive baselines.", "date": "2026-02-09", "categories": []}
{"id": "2602.08629", "title": "CauScale: Neural Causal Discovery at Scale", "abstract": "Causal discovery is essential for advancing data-driven fields such as scientific AI and data analysis, yet existing approaches face significant time- and space-efficiency bottlenecks when scaling to large graphs. To address this challenge, we present CauScale, a neural architecture designed for efficient causal discovery that scales inference to graphs with up to 1000 nodes. CauScale improves time efficiency via a reduction unit that compresses data embeddings and improves space efficiency by adopting tied attention weights to avoid maintaining axis-specific attention maps. To keep high causal discovery accuracy, CauScale adopts a two-stream design: a data stream extracts relational evidence from high-dimensional observations, while a graph stream integrates statistical graph priors and preserves key structural signals. CauScale successfully scales to 500-node graphs during training, where prior work fails due to space limitations. Across testing data with varying graph scales and causal mechanisms, CauScale achieves 99.6% mAP on in-distribution data and 84.4% on out-of-distribution data, while delivering 4-13,000 times inference speedups over prior methods. Our project page is at https://github.com/OpenCausaLab/CauScale.", "date": "2026-02-09", "categories": []}
{"id": "2405.15167", "title": "ProDAG: Projected Variational Inference for Directed Acyclic Graphs", "abstract": "Directed acyclic graph (DAG) learning is a central task in structure discovery and causal inference. Although the field has witnessed remarkable advances over the past few years, it remains statistically and computationally challenging to learn a single (point estimate) DAG from data, let alone provide uncertainty quantification. We address the difficult task of quantifying graph uncertainty by developing a Bayesian variational inference framework based on novel, provably valid distributions that have support directly on the space of sparse DAGs. These distributions, which we use to define our prior and variational posterior, are induced by a projection operation that maps an arbitrary continuous distribution onto the space of sparse weighted acyclic adjacency matrices. While this projection is combinatorial, it can be solved efficiently using recent continuous reformulations of acyclicity constraints. We empirically demonstrate that our method, ProDAG, can outperform state-of-the-art alternatives in both accuracy and uncertainty quantification.", "date": "2024-05-24", "categories": []}
{"id": "2602.06337", "title": "Can Post-Training Transform LLMs into Causal Reasoners?", "abstract": "Causal inference is essential for decision-making but remains challenging for non-experts. While large language models (LLMs) show promise in this domain, their precise causal estimation capabilities are still limited, and the impact of post-training on these abilities is insufficiently explored. This paper examines the extent to which post-training can enhance LLMs' capacity for causal inference. We introduce CauGym, a comprehensive dataset comprising seven core causal tasks for training and five diverse test sets. Using this dataset, we systematically evaluate five post-training approaches: SFT, DPO, KTO, PPO, and GRPO. Across five in-domain and four existing benchmarks, our experiments demonstrate that appropriate post-training enables smaller LLMs to perform causal inference competitively, often surpassing much larger models. Our 14B parameter model achieves 93.5% accuracy on the CaLM benchmark, compared to 55.4% by OpenAI o3. Furthermore, the post-trained LLMs exhibit strong generalization and robustness under real-world conditions such as distribution shifts and noisy data. Collectively, these findings provide the first systematic evidence that targeted post-training can produce reliable and robust LLM-based causal reasoners. Our data and GRPO-model are available at https://github.com/OpenCausaLab/CauGym.", "date": "2026-02-06", "categories": []}
{"id": "2602.06611", "title": "Adaptive-CaRe: Adaptive Causal Regularization for Robust Outcome Prediction", "abstract": "Accurate prediction of outcomes is crucial for clinical decision-making and personalized patient care. Supervised machine learning algorithms, which are commonly used for outcome prediction in the medical domain, optimize for predictive accuracy, which can result in models latching onto spurious correlations instead of robust predictors. Causal structure learning methods on the other hand have the potential to provide robust predictors for the target, but can be too conservative because of algorithmic and data assumptions, resulting in loss of diagnostic precision. Therefore, we propose a novel model-agnostic regularization strategy, Adaptive-CaRe, for generalized outcome prediction in the medical domain. Adaptive-CaRe strikes a balance between both predictive value and causal robustness by incorporating a penalty that is proportional to the difference between the estimated statistical contribution and estimated causal contribution of the input features for model predictions. Our experiments on synthetic data establish the efficacy of the proposed Adaptive-CaRe regularizer in finding robust predictors for the target while maintaining competitive predictive accuracy. With experiments on a standard causal benchmark, we provide a blueprint for navigating the trade-off between predictive accuracy and causal robustness by tweaking the regularization strength, $\\lambda$. Validation using real-world dataset confirms that the results translate to practical, real-domain settings. Therefore, Adaptive-CaRe provides a simple yet effective solution to the long-standing trade-off between predictive accuracy and causal robustness in the medical domain. Future work would involve studying alternate causal structure learning frameworks and complex classification models to provide deeper insights at a larger scale.", "date": "2026-02-06", "categories": []}
{"id": "2602.06899", "title": "Sample Complexity of Causal Identification with Temporal Heterogeneity", "abstract": "Recovering a unique causal graph from observational data is an ill-posed problem because multiple generating mechanisms can lead to the same observational distribution. This problem becomes solvable only by exploiting specific structural or distributional assumptions. While recent work has separately utilized time-series dynamics or multi-environment heterogeneity to constrain this problem, we integrate both as complementary sources of heterogeneity. This integration yields unified necessary identifiability conditions and enables a rigorous analysis of the statistical limits of recovery under thin versus heavy-tailed noise. In particular, temporal structure is shown to effectively substitute for missing environmental diversity, possibly achieving identifiability even under insufficient heterogeneity. Extending this analysis to heavy-tailed (Student's t) distributions, we demonstrate that while geometric identifiability conditions remain invariant, the sample complexity diverges significantly from the Gaussian baseline. Explicit information-theoretic bounds quantify this cost of robustness, establishing the fundamental limits of covariance-based causal graph recovery methods in realistic non-stationary systems. This work shifts the focus from whether causal structure is identifiable to whether it is statistically recoverable in practice.", "date": "2026-02-06", "categories": []}
{"id": "2406.05014", "title": "Root Cause Analysis of Outliers with Missing Structural Knowledge", "abstract": "The goal of Root Cause Analysis (RCA) is to explain why an anomaly occurred by identifying where the fault originated. Several recent works model the anomalous event as resulting from a change in the causal mechanism at the root cause, i.e., as a soft intervention. RCA is then the task of identifying which causal mechanism changed. In real-world applications, one often has either few or only a single sample from the post-intervention distribution: a severe limitation for most methods, which assume one knows or can estimate the distribution. However, even those that do not are statistically ill-posed due to the need to probe regression models in regions of low probability density. In this paper, we propose simple, efficient methods to overcome both difficulties in the case where there is a single root cause and the causal graph is a polytree. When one knows the causal graph, we give guarantees for a traversal algorithm that requires only marginal anomaly scores and does not depend on specifying an arbitrary anomaly score cut-off. When one does not know the causal graph, we show that the heuristic of identifying root causes as the variables with the highest marginal anomaly scores is causally justified. To this end, we prove that anomalies with small scores are unlikely to cause those with larger scores in polytrees and give upper bounds for the likelihood of causal pathways with non-monotonic anomaly scores.", "date": "2024-06-07", "categories": []}
{"id": "2510.22031", "title": "Differentiable Constraint-Based Causal Discovery", "abstract": "Causal discovery from observational data is a fundamental task in artificial intelligence, with far-reaching implications for decision-making, predictions, and interventions. Despite significant advances, existing methods can be broadly categorized as constraint-based or score-based approaches. Constraint-based methods offer rigorous causal discovery but are often hindered by small sample sizes, while score-based methods provide flexible optimization but typically forgo explicit conditional independence testing. This work explores a third avenue: developing differentiable $d$-separation scores, obtained through a percolation theory using soft logic. This enables the implementation of a new type of causal discovery method: gradient-based optimization of conditional independence constraints. Empirical evaluations demonstrate the robust performance of our approach in low-sample regimes, surpassing traditional constraint-based and score-based baselines on a real-world dataset. Code and data of the proposed method are publicly available at https://github$.$com/PurdueMINDS/DAGPA.", "date": "2025-10-24", "categories": []}
{"id": "2602.05033", "title": "Causal Representation Meets Stochastic Modeling under Generic Geometry", "abstract": "Learning meaningful causal representations from observations has emerged as a crucial task for facilitating machine learning applications and driving scientific discoveries in fields such as climate science, biology, and physics. This process involves disentangling high-level latent variables and their causal relationships from low-level observations. Previous work in this area that achieves identifiability typically focuses on cases where the observations are either i.i.d. or follow a latent discrete-time process. Nevertheless, many real-world settings require identifying latent variables that are continuous-time stochastic processes (e.g., multivariate point processes). To this end, we develop identifiable causal representation learning for continuous-time latent stochastic point processes. We study its identifiability by analyzing the geometry of the parameter space. Furthermore, we develop MUTATE, an identifiable variational autoencoder framework with a time-adaptive transition module to infer stochastic dynamics. Across simulated and empirical studies, we find that MUTATE can effectively answer scientific questions, such as the accumulation of mutations in genomics and the mechanisms driving neuron spike triggers in response to time-varying dynamics.", "date": "2026-02-04", "categories": []}
{"id": "2602.05997", "title": "Causal Inference on Stopped Random Walks in Online Advertising", "abstract": "We consider a causal inference problem frequently encountered in online advertising systems, where a publisher (e.g., Instagram, TikTok) interacts repeatedly with human users and advertisers by sporadically displaying to each user an advertisement selected through an auction. Each treatment corresponds to a parameter value of the advertising mechanism (e.g., auction reserve-price), and we want to estimate through experiments the corresponding long-term treatment effect (e.g., annual advertising revenue). In our setting, the treatment affects not only the instantaneous revenue from showing an ad, but also changes each user's interaction-trajectory, and each advertiser's bidding policy -- as the latter is constrained by a finite budget. In particular, each a treatment may even affect the size of the population, since users interact longer with a tolerable advertising mechanism. We drop the classical i.i.d. assumption and model the experiment measurements (e.g., advertising revenue) as a stopped random walk, and use a budget-splitting experimental design, the Anscombe Theorem, a Wald-like equation, and a Central Limit Theorem to construct confidence intervals for the long-term treatment effect.", "date": "2026-02-05", "categories": []}
{"id": "2602.03914", "title": "Causal Discovery for Cross-Sectional Data Based on Super-Structure and Divide-and-Conquer", "abstract": "This paper tackles a critical bottleneck in Super-Structure-based divide-and-conquer causal discovery: the high computational cost of constructing accurate Super-Structures--particularly when conditional independence (CI) tests are expensive and domain knowledge is unavailable. We propose a novel, lightweight framework that relaxes the strict requirements on Super-Structure construction while preserving the algorithmic benefits of divide-and-conquer. By integrating weakly constrained Super-Structures with efficient graph partitioning and merging strategies, our approach substantially lowers CI test overhead without sacrificing accuracy. We instantiate the framework in a concrete causal discovery algorithm and rigorously evaluate its components on synthetic data. Comprehensive experiments on Gaussian Bayesian networks, including magic-NIAB, ECOLI70, and magic-IRRI, demonstrate that our method matches or closely approximates the structural accuracy of PC and FCI while drastically reducing the number of CI tests. Further validation on the real-world China Health and Retirement Longitudinal Study (CHARLS) dataset confirms its practical applicability. Our results establish that accurate, scalable causal discovery is achievable even under minimal assumptions about the initial Super-Structure, opening new avenues for applying divide-and-conquer methods to large-scale, knowledge-scarce domains such as biomedical and social science research.", "date": "2026-02-03", "categories": []}
{"id": "2602.04077", "title": "Efficient Subgroup Analysis via Optimal Trees with Global Parameter Fusion", "abstract": "Identifying and making statistical inferences on differential treatment effects (commonly known as subgroup analysis in clinical research) is central to precision health. Subgroup analysis allows practitioners to pinpoint populations for whom a treatment is especially beneficial or protective, thereby advancing targeted interventions. Tree based recursive partitioning methods are widely used for subgroup analysis due to their interpretability. Nevertheless, these approaches encounter significant limitations, including suboptimal partitions induced by greedy heuristics and overfitting from locally estimated splits, especially under limited sample sizes. To address these limitations, we propose a fused optimal causal tree method that leverages mixed integer optimization (MIO) to facilitate precise subgroup identification. Our approach ensures globally optimal partitions and introduces a parameter fusion constraint to facilitate information sharing across related subgroups. This design substantially improves subgroup discovery accuracy and enhances statistical efficiency. We provide theoretical guarantees by rigorously establishing out of sample risk bounds and comparing them with those of classical tree based methods. Empirically, our method consistently outperforms popular baselines in simulations. Finally, we demonstrate its practical utility through a case study on the Health and Aging Brain Study Health Disparities (HABS-HD) dataset, where our approach yields clinically meaningful insights.", "date": "2026-02-03", "categories": []}
{"id": "2602.04457", "title": "Journey to the Centre of Cluster: Harnessing Interior Nodes for A/B Testing under Network Interference", "abstract": "A/B testing on platforms often faces challenges from network interference, where a unit's outcome depends not only on its own treatment but also on the treatments of its network neighbors. To address this, cluster-level randomization has become standard, enabling the use of network-aware estimators. These estimators typically trim the data to retain only a subset of informative units, achieving low bias under suitable conditions but often suffering from high variance. In this paper, we first demonstrate that the interior nodes - units whose neighbors all lie within the same cluster - constitute the vast majority of the post-trimming subpopulation. In light of this, we propose directly averaging over the interior nodes to construct the mean-in-interior (MII) estimator, which circumvents the delicate reweighting required by existing network-aware estimators and substantially reduces variance in classical settings. However, we show that interior nodes are often not representative of the full population, particularly in terms of network-dependent covariates, leading to notable bias. We then augment the MII estimator with a counterfactual predictor trained on the entire network, allowing us to adjust for covariate distribution shifts between the interior nodes and full population. By rearranging the expression, we reveal that our augmented MII estimator embodies an analytical form of the point estimator within prediction-powered inference framework. This insight motivates a semi-supervised lens, wherein interior nodes are treated as labeled data subject to selection bias. Extensive and challenging simulation studies demonstrate the outstanding performance of our augmented MII estimator across various settings.", "date": "2026-02-04", "categories": []}
{"id": "2602.04667", "title": "Causal explanations of outliers in systems with lagged time-dependencies", "abstract": "Root-cause analysis in controlled time dependent systems poses a major challenge in applications. Especially energy systems are difficult to handle as they exhibit instantaneous as well as delayed effects and if equipped with storage, do have a memory. In this paper we adapt the causal root-cause analysis method of Budhathoki et al. [2022] to general time-dependent systems, as it can be regarded as a strictly causal definition of the term \"root-cause\". Particularly, we discuss two truncation approaches to handle the infinite dependency graphs present in time-dependent systems. While one leaves the causal mechanisms intact, the other approximates the mechanisms at the start nodes. The effectiveness of the different approaches is benchmarked using a challenging data generation process inspired by a problem in factory energy management: the avoidance of peaks in the power consumption. We show that given enough lags our extension is able to localize the root-causes in the feature and time domain. Further the effect of mechanism approximation is discussed.", "date": "2026-02-04", "categories": []}
{"id": "2602.04736", "title": "Conditional Counterfactual Mean Embeddings: Doubly Robust Estimation and Learning Rates", "abstract": "A complete understanding of heterogeneous treatment effects involves characterizing the full conditional distribution of potential outcomes. To this end, we propose the Conditional Counterfactual Mean Embeddings (CCME), a framework that embeds conditional distributions of counterfactual outcomes into a reproducing kernel Hilbert space (RKHS). Under this framework, we develop a two-stage meta-estimator for CCME that accommodates any RKHS-valued regression in each stage. Based on this meta-estimator, we develop three practical CCME estimators: (1) Ridge Regression estimator, (2) Deep Feature estimator that parameterizes the feature map by a neural network, and (3) Neural-Kernel estimator that performs RKHS-valued regression, with the coefficients parameterized by a neural network. We provide finite-sample convergence rates for all estimators, establishing that they possess the double robustness property. Our experiments demonstrate that our estimators accurately recover distributional features including multimodal structure of conditional counterfactual distributions.", "date": "2026-02-04", "categories": []}
{"id": "2301.12412", "title": "Contextual Causal Bayesian Optimisation", "abstract": "We introduce a unified framework for contextual and causal Bayesian optimisation, which aims to design intervention policies maximising the expectation of a target variable. Our approach leverages both observed contextual information and known causal graph structures to guide the search. Within this framework, we propose a novel algorithm that jointly optimises over policies and the sets of variables on which these policies are defined. This thereby extends and unifies two previously distinct approaches: Causal Bayesian Optimisation and Contextual Bayesian Optimisation, while also addressing their limitations in scenarios that yield suboptimal results. We derive worst-case and instance-dependent high-probability regret bounds for our algorithm. We report experimental results across diverse environments, corroborating that our approach achieves sublinear regret and reduces sample complexity in high-dimensional settings.", "date": "2023-01-29", "categories": []}
{"id": "2407.03094", "title": "Conformal Prediction for Causal Effects of Continuous Treatments", "abstract": "Uncertainty quantification of causal effects is crucial for safety-critical applications such as personalized medicine. A powerful approach for this is conformal prediction, which has several practical benefits due to model-agnostic finite-sample guarantees. Yet, existing methods for conformal prediction of causal effects are limited to binary/discrete treatments and make highly restrictive assumptions such as known propensity scores. In this work, we provide a novel conformal prediction method for potential outcomes of continuous treatments. We account for the additional uncertainty introduced through propensity estimation so that our conformal prediction intervals are valid even if the propensity score is unknown. Our contributions are three-fold: (1) We derive finite-sample prediction intervals for potential outcomes of continuous treatments. (2) We provide an algorithm for calculating the derived intervals. (3) We demonstrate the effectiveness of the conformal prediction intervals in experiments on synthetic and real-world datasets. To the best of our knowledge, we are the first to propose conformal prediction for continuous treatments when the propensity score is unknown and must be estimated from data.", "date": "2024-07-03", "categories": []}
{"id": "2512.01834", "title": "Mitigating Gender Bias in Depression Detection via Counterfactual Inference", "abstract": "Audio-based depression detection models have demonstrated promising performance but often suffer from gender bias due to imbalanced training data. Epidemiological statistics show a higher prevalence of depression in females, leading models to learn spurious correlations between gender and depression. Consequently, models tend to over-diagnose female patients while underperforming on male patients, raising significant fairness concerns. To address this, we propose a novel Counterfactual Debiasing Framework grounded in causal inference. We construct a causal graph to model the decision-making process and identify gender bias as the direct causal effect of gender on the prediction. During inference, we employ counterfactual inference to estimate and subtract this direct effect, ensuring the model relies primarily on authentic acoustic pathological features. Extensive experiments on the DAIC-WOZ dataset using two advanced acoustic backbones demonstrate that our framework not only significantly reduces gender bias but also improves overall detection performance compared to existing debiasing strategies.", "date": "2025-12-01", "categories": []}
{"id": "2602.01155", "title": "Multi-Agent Causal Reasoning System for Error Pattern Rule Automation in Vehicles", "abstract": "Modern vehicles generate thousands of different discrete events known as Diagnostic Trouble Codes (DTCs). Automotive manufacturers use Boolean combinations of these codes, called error patterns (EPs), to characterize system faults and ensure vehicle safety. Yet, EP rules are still manually handcrafted by domain experts, a process that is expensive and prone to errors as vehicle complexity grows. This paper introduces CAREP (Causal Automated Reasoning for Error Patterns), a multi-agent system that automatizes the generation of EP rules from high-dimensional event sequences of DTCs. CAREP combines a causal discovery agent that identifies potential DTC-EP relations, a contextual information agent that integrates metadata and descriptions, and an orchestrator agent that synthesizes candidate boolean rules together with interpretable reasoning traces. Evaluation on a large-scale automotive dataset with over 29,100 unique DTCs and 474 error patterns demonstrates that CAREP can automatically and accurately discover the unknown EP rules, outperforming LLM-only baselines while providing transparent causal explanations. By uniting practical causal discovery and agent-based reasoning, CAREP represents a step toward fully automated fault diagnostics, enabling scalable, interpretable, and cost-efficient vehicle maintenance.", "date": "2026-02-01", "categories": []}
{"id": "2602.02601", "title": "CaST: Causal Discovery via Spatio-Temporal Graphs in Disaster Tweets", "abstract": "Understanding causality between real-world events from social media is essential for situational awareness, yet existing causal discovery methods often overlook the interplay between semantic, spatial, and temporal contexts. We propose CaST: Causal Discovery via Spatio-Temporal Graphs, a unified framework for causal discovery in disaster domain that integrates semantic similarity and spatio-temporal proximity using Large Language Models (LLMs) pretrained on disaster datasets. CaST constructs an event graph for each window of tweets. Each event extracted from tweets is represented as a node embedding enriched with its contextual semantics, geographic coordinates, and temporal features. These event nodes are then connected to form a spatio-temporal event graph, which is processed using a multi-head Graph Attention Network (GAT) \\cite{gat} to learn directed causal relationships. We construct an in-house dataset of approximately 167K disaster-related tweets collected during Hurricane Harvey and annotated following the MAVEN-ERE schema. Experimental results show that CaST achieves superior performance over both traditional and state-of-the-art methods. Ablation studies further confirm that incorporating spatial and temporal signals substantially improves both recall and stability during training. Overall, CaST demonstrates that integrating spatio-temporal reasoning into event graphs enables more robust and interpretable causal discovery in disaster-related social media text.", "date": "2026-02-01", "categories": []}
{"id": "2602.02623", "title": "Learning Consistent Causal Abstraction Networks", "abstract": "Causal artificial intelligence aims to enhance explainability, trustworthiness, and robustness in AI by leveraging structural causal models (SCMs). In this pursuit, recent advances formalize network sheaves and cosheaves of causal knowledge. Pushing in the same direction, we tackle the learning of consistent causal abstraction network (CAN), a sheaf-theoretic framework where (i) SCMs are Gaussian, (ii) restriction maps are transposes of constructive linear causal abstractions (CAs) adhering to the semantic embedding principle, and (iii) edge stalks correspond--up to permutation--to the node stalks of more detailed SCMs. Our problem formulation separates into edge-specific local Riemannian problems and avoids nonconvex objectives. We propose an efficient search procedure, solving the local problems with SPECTRAL, our iterative method with closed-form updates and suitable for positive definite and semidefinite covariance matrices. Experiments on synthetic data show competitive performance in the CA learning task, and successful recovery of diverse CAN structures.", "date": "2026-02-02", "categories": []}
{"id": "2602.03353", "title": "Causal Graph Learning via Distributional Invariance of Cause-Effect Relationship", "abstract": "This paper introduces a new framework for recovering causal graphs from observational data, leveraging the observation that the distribution of an effect, conditioned on its causes, remains invariant to changes in the prior distribution of those causes. This insight enables a direct test for potential causal relationships by checking the variance of their corresponding effect-cause conditional distributions across multiple downsampled subsets of the data. These subsets are selected to reflect different prior cause distributions, while preserving the effect-cause conditional relationships. Using this invariance test and exploiting an (empirical) sparsity of most causal graphs, we develop an algorithm that efficiently uncovers causal relationships with quadratic complexity in the number of observational variables, reducing the processing time by up to 25x compared to state-of-the-art methods. Our empirical experiments on a varied benchmark of large-scale datasets show superior or equivalent performance compared to existing works, while achieving enhanced scalability.", "date": "2026-02-03", "categories": []}
{"id": "2602.03459", "title": "Causal Inference on Networks under Misspecified Exposure Mappings: A Partial Identification Framework", "abstract": "Estimating treatment effects in networks is challenging, as each potential outcome depends on the treatments of all other nodes in the network. To overcome this difficulty, existing methods typically impose an exposure mapping that compresses the treatment assignments in the network into a low-dimensional summary. However, if this mapping is misspecified, standard estimators for direct and spillover effects can be severely biased. We propose a novel partial identification framework for causal inference on networks to assess the robustness of treatment effects under misspecifications of the exposure mapping. Specifically, we derive sharp upper and lower bounds on direct and spillover effects under such misspecifications. As such, our framework presents a novel application of causal sensitivity analysis to exposure mappings. We instantiate our framework for three canonical exposure settings widely used in practice: (i) weighted means of the neighborhood treatments, (ii) threshold-based exposure mappings, and (iii) truncated neighborhood interference in the presence of higher-order spillovers. Furthermore, we develop orthogonal estimators for these bounds and prove that the resulting bound estimates are valid, sharp, and efficient. Our experiments show the bounds remain informative and provide reliable conclusions under misspecification of exposure mappings.", "date": "2026-02-03", "categories": []}
{"id": "2602.03823", "title": "Preference-based Conditional Treatment Effects and Policy Learning", "abstract": "We introduce a new preference-based framework for conditional treatment effect estimation and policy learning, built on the Conditional Preference-based Treatment Effect (CPTE). CPTE requires only that outcomes be ranked under a preference rule, unlocking flexible modeling of heterogeneous effects with multivariate, ordinal, or preference-driven outcomes. This unifies applications such as conditional probability of necessity and sufficiency, conditional Win Ratio, and Generalized Pairwise Comparisons. Despite the intrinsic non-identifiability of comparison-based estimands, CPTE provides interpretable targets and delivers new identifiability conditions for previous unidentifiable estimands. We present estimation strategies via matching, quantile, and distributional regression, and further design efficient influence-function estimators to correct plug-in bias and maximize policy value. Synthetic and semi-synthetic experiments demonstrate clear performance gains and practical impact.", "date": "2026-02-03", "categories": []}
{"id": "2503.04363", "title": "Causally Reliable Concept Bottleneck Models", "abstract": "Concept-based models are an emerging paradigm in deep learning that constrains the inference process to operate through human-interpretable variables, facilitating explainability and human interaction. However, these architectures, on par with popular opaque neural models, fail to account for the true causal mechanisms underlying the target phenomena represented in the data. This hampers their ability to support causal reasoning tasks, limits out-of-distribution generalization, and hinders the implementation of fairness constraints. To overcome these issues, we propose Causally reliable Concept Bottleneck Models (C$^2$BMs), a class of concept-based architectures that enforce reasoning through a bottleneck of concepts structured according to a model of the real-world causal mechanisms. We also introduce a pipeline to automatically learn this structure from observational data and unstructured background knowledge (e.g., scientific literature). Experimental evidence suggests that C$^2$BMs are more interpretable, causally reliable, and improve responsiveness to interventions w.r.t. standard opaque and concept-based models, while maintaining their accuracy.", "date": "2025-03-06", "categories": []}
{"id": "2504.18881", "title": "TSCAN: Context-Aware Uplift Modeling via Two-Stage Training for Online Merchant Business Diagnosis", "abstract": "A primary challenge in ITE estimation is sample selection bias. Traditional approaches utilize treatment regularization techniques such as the Integral Probability Metrics (IPM), re-weighting, and propensity score modeling to mitigate this bias. However, these regularizations may introduce undesirable information loss and limit the performance of the model. Furthermore, treatment effects vary across different external contexts, and the existing methods are insufficient in fully interacting with and utilizing these contextual features. To address these issues, we propose a Context-Aware uplift model based on the Two-Stage training approach (TSCAN), comprising CAN-U and CAN-D sub-models. In the first stage, we train an uplift model, called CAN-U, which includes the treatment regularizations of IPM and propensity score prediction, to generate a complete dataset with counterfactual uplift labels. In the second stage, we train a model named CAN-D, which utilizes an isotonic output layer to directly model uplift effects, thereby eliminating the reliance on the regularization components. CAN-D adaptively corrects the errors estimated by CAN-U through reinforcing the factual samples, while avoiding the negative impacts associated with the aforementioned regularizations. Additionally, we introduce a Context-Aware Attention Layer throughout the two-stage process to manage the interactions between treatment, merchant, and contextual features, thereby modeling the varying treatment effect in different contexts. We conduct extensive experiments on two real-world datasets to validate the effectiveness of TSCAN. Ultimately, the deployment of our model for real-world merchant diagnosis on one of China's largest online food ordering platforms validates its practical utility and impact.", "date": "2025-04-26", "categories": []}
{"id": "2505.12801", "title": "Transportability without Graphs: A Bayesian Approach to Identifying s-Admissible Backdoor Sets", "abstract": "Transporting causal information across populations is a critical challenge in clinical decision-making. Causal modeling provides criteria for identifiability and transportability, but these require knowledge of the causal graph, which rarely holds in practice. We propose a Bayesian method that combines observational data from the target domain with experimental data from a different domain to identify s-admissible backdoor sets, which enable unbiased estimation of causal effects across populations, without requiring the causal graph. We prove that if such a set exists, we can always find one within the Markov boundary of the outcome, narrowing the search space, and we establish asymptotic convergence guarantees for our method. We develop a greedy algorithm that reframes transportability as a feature selection problem, selecting conditioning sets that maximize the marginal likelihood of experimental data given observational data. In simulated and semi-synthetic data, our method correctly identifies transportability bias, improves causal effect estimation, and performs favorably against alternatives.", "date": "2025-05-19", "categories": []}
{"id": "2505.18410", "title": "On Theoretical Identifiability of Discrete Latent Causal Graphical Models", "abstract": "This paper considers a challenging problem of identifying a causal graphical model under the presence of latent variables. While various identifiability conditions have been proposed in the literature, they often require multiple pure children per latent variable or restrictions on the latent causal graph. Furthermore, it is common for all observed variables to exhibit the same modality. Consequently, the existing identifiability conditions are often too stringent for complex real-world data. We consider a general nonparametric measurement model with arbitrary observed variable types and binary latent variables, and propose a double triangular graphical condition that guarantees identifiability of the entire causal graphical model. The proposed condition significantly relaxes the popular pure children condition. We also establish necessary conditions for identifiability and provide valuable insights into fundamental limits of identifiability. Simulation studies verify that latent structures satisfying our conditions can be accurately estimated from data.", "date": "2025-05-23", "categories": []}
{"id": "2506.13680", "title": "Hybrid Meta-learners for Estimating Heterogeneous Treatment Effects", "abstract": "Estimating conditional average treatment effects (CATE) from observational data involves modeling decisions that differ from supervised learning, particularly concerning how to regularize model complexity. Previous approaches can be grouped into two primary \"meta-learner\" paradigms that impose distinct inductive biases. Indirect meta-learners first fit and regularize separate potential outcome (PO) models and then estimate CATE by taking their difference, whereas direct meta-learners construct and directly regularize estimators for the CATE function itself. Neither approach consistently outperforms the other across all scenarios: indirect learners perform well when the PO functions are simple, while direct learners outperform when the CATE is simpler than individual PO functions. In this paper, we introduce the Hybrid Learner (H-learner), a novel regularization strategy that interpolates between the direct and indirect regularizations depending on the dataset at hand. The H-learner achieves this by learning intermediate functions whose difference closely approximates the CATE without necessarily requiring accurate individual approximations of the POs themselves. We demonstrate that intentionally allowing suboptimal fits to the POs improves the bias-variance tradeoff in estimating CATE. Experiments conducted on semi-synthetic and real-world benchmark datasets illustrate that the H-learner consistently operates at the Pareto frontier, effectively combining the strengths of both direct and indirect meta-learners.", "date": "2025-06-16", "categories": []}
{"id": "2602.00190", "title": "From Gameplay Traces to Game Mechanics: Causal Induction with Large Language Models", "abstract": "Deep learning agents can achieve high performance in complex game domains without often understanding the underlying causal game mechanics. To address this, we investigate Causal Induction: the ability to infer governing laws from observational data, by tasking Large Language Models (LLMs) with reverse-engineering Video Game Description Language (VGDL) rules from gameplay traces. To reduce redundancy, we select nine representative games from the General Video Game AI (GVGAI) framework using semantic embeddings and clustering. We compare two approaches to VGDL generation: direct code generation from observations, and a two-stage method that first infers a structural causal model (SCM) and then translates it into VGDL. Both approaches are evaluated across multiple prompting strategies and controlled context regimes, varying the amount and form of information provided to the model, from just raw gameplay observations to partial VGDL specifications. Results show that the SCM-based approach more often produces VGDL descriptions closer to the ground truth than direct generation, achieving preference win rates of up to 81\\% in blind evaluations and yielding fewer logically inconsistent rules. These learned SCMs can be used for downstream use cases such as causal reinforcement learning, interpretable agents, and procedurally generating novel but logically consistent games.", "date": "2026-01-30", "categories": []}
{"id": "2602.00427", "title": "Topological Residual Asymmetry for Bivariate Causal Direction", "abstract": "Inferring causal direction from purely observational bivariate data is fragile: many methods commit to a direction even in ambiguous or near non-identifiable regimes. We propose Topological Residual Asymmetry (TRA), a geometry-based criterion for additive-noise models. TRA compares the shapes of two cross-fitted regressor-residual clouds after rank-based copula standardization: in the correct direction, residuals are approximately independent, producing a two-dimensional bulk, while in the reverse direction -- especially under low noise -- the cloud concentrates near a one-dimensional tube. We quantify this bulk-tube contrast using a 0D persistent-homology functional, computed efficiently from Euclidean MST edge-length profiles. We prove consistency in a triangular-array small-noise regime, extend the method to fixed noise via a binned variant (TRA-s), and introduce TRA-C, a confounding-aware abstention rule calibrated by a Gaussian-copula plug-in bootstrap. Extensive experiments across many challenging synthetic and real-data scenarios demonstrate the method's superiority.", "date": "2026-01-31", "categories": []}
{"id": "2602.00788", "title": "Multi-Objective Multi-Fidelity Bayesian Optimization with Causal Priors", "abstract": "Multi-fidelity Bayesian optimization (MFBO) accelerates the search for the global optimum of black-box functions by integrating inexpensive, low-fidelity approximations. The central task of an MFBO policy is to balance the cost-efficiency of low-fidelity proxies against their reduced accuracy to ensure effective progression toward the high-fidelity optimum. Existing MFBO methods primarily capture associational dependencies between inputs, fidelities, and objectives, rather than causal mechanisms, and can perform poorly when lower-fidelity proxies are poorly aligned with the target fidelity. We propose RESCUE (REducing Sampling cost with Causal Understanding and Estimation), a multi-objective MFBO method that incorporates causal calculus to systematically address this challenge. RESCUE learns a structural causal model capturing causal relationships between inputs, fidelities, and objectives, and uses it to construct a probabilistic multi-fidelity (MF) surrogate that encodes intervention effects. Exploiting the causal structure, we introduce a causal hypervolume knowledge-gradient acquisition strategy to select input-fidelity pairs that balance expected multi-objective improvement and cost. We show that RESCUE improves sample efficiency over state-of-the-art MF optimization methods on synthetic and real-world problems in robotics, machine learning (AutoML), and healthcare.", "date": "2026-01-31", "categories": []}
{"id": "2602.02161", "title": "Generating Causal Temporal Interaction Graphs for Counterfactual Validation of Temporal Link Prediction", "abstract": "Temporal link prediction (TLP) models are commonly evaluated based on predictive accuracy, yet such evaluations do not assess whether these models capture the causal mechanisms that govern temporal interactions. In this work, we propose a framework for counterfactual validation of TLP models by generating causal temporal interaction graphs (CTIGs) with known ground-truth causal structure. We first introduce a structural equation model for continuous-time event sequences that supports both excitatory and inhibitory effects, and then extend this mechanism to temporal interaction graphs. To compare causal models, we propose a distance metric based on cross-model predictive error, and empirically validate the hypothesis that predictors trained on one causal model degrade when evaluated on sufficiently distant models. Finally, we instantiate counterfactual evaluation under (i) controlled causal shifts between generating models and (ii) timestamp shuffling as a stochastic distortion with measurable causal distance. Our framework provides a foundation for causality-aware benchmarking.", "date": "2026-02-02", "categories": []}
{"id": "2602.02371", "title": "C-kNN-LSH: A Nearest-Neighbor Algorithm for Sequential Counterfactual Inference", "abstract": "Estimating causal effects from longitudinal trajectories is central to understanding the progression of complex conditions and optimizing clinical decision-making, such as comorbidities and long COVID recovery. We introduce \\emph{C-kNN--LSH}, a nearest-neighbor framework for sequential causal inference designed to handle such high-dimensional, confounded situations. By utilizing locality-sensitive hashing, we efficiently identify ``clinical twins'' with similar covariate histories, enabling local estimation of conditional treatment effects across evolving disease states. To mitigate bias from irregular sampling and shifting patient recovery profiles, we integrate neighborhood estimator with a doubly-robust correction.\n  Theoretical analysis guarantees our estimator is consistent and second-order robust to nuisance error.\n  Evaluated on a real-world Long COVID cohort with 13,511 participants, \\emph{C-kNN-LSH} demonstrates superior performance in capturing recovery heterogeneity and estimating policy values compared to existing baselines.", "date": "2026-02-02", "categories": []}
{"id": "2602.02451", "title": "Active Causal Experimentalist (ACE): Learning Intervention Strategies via Direct Preference Optimization", "abstract": "Discovering causal relationships requires controlled experiments, but experimentalists face a sequential decision problem: each intervention reveals information that should inform what to try next. Traditional approaches such as random sampling, greedy information maximization, and round-robin coverage treat each decision in isolation, unable to learn adaptive strategies from experience. We propose Active Causal Experimentalist (ACE), which learns experimental design as a sequential policy. Our key insight is that while absolute information gains diminish as knowledge accumulates (making value-based RL unstable), relative comparisons between candidate interventions remain meaningful throughout. ACE exploits this via Direct Preference Optimization, learning from pairwise intervention comparisons rather than non-stationary reward magnitudes. Across synthetic benchmarks, physics simulations, and economic data, ACE achieves 70-71% improvement over baselines at equal intervention budgets (p < 0.001, Cohen's d ~ 2). Notably, the learned policy autonomously discovers that collider mechanisms require concentrated interventions on parent variables, a theoretically-grounded strategy that emerges purely from experience. This suggests preference-based learning can recover principled experimental strategies, complementing theory with learned domain adaptation.", "date": "2026-02-02", "categories": []}
{"id": "2402.02277", "title": "Causal Bayesian Optimization via Exogenous Distribution Learning", "abstract": "Maximizing a target variable as an operational objective within a structural causal model is a fundamental problem. Causal Bayesian Optimization (CBO) approaches typically achieve this either by performing interventions that modify the causal structure to increase the reward or by introducing action nodes to endogenous variables, thereby adjusting the data-generating mechanisms to meet the objective. In this paper, we propose a novel method that learns the distribution of exogenous variables-an aspect often ignored or marginalized through expectation in existing CBO frameworks. By modeling the exogenous distribution, we enhance the approximation fidelity of the data-generating structural causal models (SCMs) used in surrogate models, which are commonly trained on limited observational data. Furthermore, the ability to recover exogenous variables enables the application of our approach to more general causal structures beyond the confines of Additive Noise Models (ANMs) and single-mode Gaussian, allowing the use of more expressive priors for context noise. We incorporate the learned exogenous distribution into a new CBO method, demonstrating its advantages across diverse datasets and application scenarios.", "date": "2024-02-03", "categories": []}
{"id": "2402.17233", "title": "Hybrid$^2$ Neural ODE Causal Modeling and an Application to Glycemic Response", "abstract": "Hybrid models composing mechanistic ODE-based dynamics with flexible and expressive neural network components have grown rapidly in popularity, especially in scientific domains where such ODE-based modeling offers important interpretability and validated causal grounding (e.g., for counterfactual reasoning). The incorporation of mechanistic models also provides inductive bias in standard blackbox modeling approaches, critical when learning from small datasets or partially observed, complex systems. Unfortunately, as the hybrid models become more flexible, the causal grounding provided by the mechanistic model can quickly be lost. We address this problem by leveraging another common source of domain knowledge: \\emph{ranking} of treatment effects for a set of interventions, even if the precise treatment effect is unknown. We encode this information in a \\emph{causal loss} that we combine with the standard predictive loss to arrive at a \\emph{hybrid loss} that biases our learning towards causally valid hybrid models. We demonstrate our ability to achieve a win-win, state-of-the-art predictive performance \\emph{and} causal validity, in the challenging task of modeling glucose dynamics post-exercise in individuals with type 1 diabetes.", "date": "2024-02-27", "categories": []}
{"id": "2502.07656", "title": "Causal Imitation Learning under Expert-Observable and Expert-Unobservable Confounding", "abstract": "We propose a general framework for causal Imitation Learning (IL) with hidden confounders, which subsumes several existing settings. Our framework accounts for two types of hidden confounders: (a) variables observed by the expert but not by the imitator, and (b) confounding noise hidden from both. By leveraging trajectory histories as instruments, we reformulate causal IL in our framework into a Conditional Moment Restriction (CMR) problem. We propose DML-IL, an algorithm that solves this CMR problem via instrumental variable regression, and upper bound its imitation gap. Empirical evaluation on continuous state-action environments, including Mujoco tasks, demonstrates that DML-IL outperforms existing causal IL baselines.", "date": "2025-02-11", "categories": []}
{"id": "2505.15274", "title": "Identification of Probabilities of Causation: from Recursive to Closed-Form Bounds", "abstract": "Probabilities of causation (PoCs) are fundamental quantities for counterfactual analysis and personalized decision making. However, existing analytical results are largely confined to binary settings. This paper extends PoCs to multi-valued treatments and outcomes by deriving closed form bounds for a representative family of discrete PoCs within Structural Causal Models, using standard experimental and observational distributions. We introduce the notion of equivalence classes of PoCs, which reduces arbitrary discrete PoCs to this family, and establish a replaceability principle that transfers bounds across value permutations. For the resulting bounds, we prove soundness in all dimensions and empirically verify tightness in low dimensional cases via Balke's linear programming method; we further conjecture that this tightness extends to all dimensions. Simulations indicate that our closed form bounds consistently tighten recent recursive bounds while remaining simpler to compute. Finally, we illustrate the practical relevance of our results through toy examples.", "date": "2025-05-21", "categories": []}
{"id": "2505.19589", "title": "Model Agnostic Differentially Private Causal Inference", "abstract": "Estimating causal effects from observational data is essential in fields such as medicine, economics and social sciences, where privacy concerns are paramount. We propose a general, model-agnostic framework for differentially private estimation of average treatment effects (ATE) that avoids strong structural assumptions on the data-generating process or the models used to estimate propensity scores and conditional outcomes. In contrast to prior work, which enforces differential privacy by directly privatizing these nuisance components, our approach decouples nuisance estimation from privacy protection. This separation allows the use of flexible, state-of-the-art black-box models, while differential privacy is achieved by perturbing only predictions and aggregation steps within a fold-splitting scheme with ensemble techniques. We instantiate the framework for three classical estimators -- the G-Formula, inverse propensity weighting (IPW), and augmented IPW (AIPW) -- and provide formal utility and privacy guarantees, together with privatized confidence intervals. Empirical results on synthetic and real data show that our methods maintain competitive performance under realistic privacy budgets.", "date": "2025-05-26", "categories": []}
{"id": "2509.22122", "title": "Direct Bias-Correction Term Estimation for Average Treatment Effect Estimation", "abstract": "This study considers the estimation of the direct bias-correction term for estimating the average treatment effect (ATE). Let $\\{(X_i, D_i, Y_i)\\}_{i=1}^{n}$ be the observations, where $X_i$ denotes $K$-dimensional covariates, $D_i \\in \\{0, 1\\}$ denotes a binary treatment assignment indicator, and $Y_i$ denotes an outcome. In ATE estimation, $h_0(D_i, X_i) = \\frac{1[D_i = 1]}{e_0(X_i)} - \\frac{1[D_i = 0]}{1 - e_0(X_i)}$ is called the bias-correction term, where $e_0(X_i)$ is the propensity score. The bias-correction term is also referred to as the Riesz representer or clever covariates, depending on the literature, and plays an important role in construction of efficient ATE estimators. In this study, we propose estimating $h_0$ by directly minimizing the Bregman divergence between its model and $h_0$, which includes squared error and Kullback--Leibler divergence as special cases. Our proposed method is inspired by direct density ratio estimation methods and generalizes existing bias-correction term estimation methods, such as covariate balancing weights, Riesz regression, and nearest neighbor matching. Importantly, under specific choices of bias-correction term models and Bregman divergence, we can automatically ensure the covariate balancing property. Thus, our study provides a practical modeling and estimation approach through a generalization of existing methods.", "date": "2025-09-26", "categories": []}
{"id": "2510.25128", "title": "An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation", "abstract": "The technique of data augmentation (DA) is often used in machine learning for regularization purposes to better generalize under i.i.d. settings. In this work, we present a unifying framework with topics in causal inference to make a case for the use of DA beyond just the i.i.d. setting, but for generalization across interventions as well. Specifically, we argue that when the outcome generating mechanism is invariant to our choice of DA, then such augmentations can effectively be thought of as interventions on the treatment generating mechanism itself. This can potentially help to reduce bias in causal effect estimation arising from hidden confounders. In the presence of such unobserved confounding we typically make use of instrumental variables (IVs) -- sources of treatment randomization that are conditionally independent of the outcome. However, IVs may not be as readily available as DA for many applications, which is the main motivation behind this work. By appropriately regularizing IV based estimators, we introduce the concept of IV-like (IVL) regression for mitigating confounding bias and improving predictive performance across interventions even when certain IV properties are relaxed. Finally, we cast parameterized DA as an IVL regression problem and show that when used in composition can simulate a worst-case application of such DA, further improving performance on causal estimation and generalization tasks beyond what simple DA may offer. This is shown both theoretically for the population case and via simulation experiments for the finite sample case using a simple linear example. We also present real data experiments to support our case.", "date": "2025-10-29", "categories": []}
{"id": "2512.20523", "title": "ScoreMatchingRiesz: Score Matching for Debiased Machine Learning and Policy Path Estimation", "abstract": "We propose ScoreMatchingRiesz, a family of Riesz representer estimators based on score matching. The Riesz representer is a key nuisance component in debiased machine learning, enabling $\\sqrt{n}$-consistent and asymptotically efficient estimation of causal and structural targets via Neyman-orthogonal scores. We formulate Riesz representer estimation as a score estimation problem. This perspective stabilizes representer estimation by allowing us to leverage denoising score matching and telescoping density ratio estimation. We also introduce the policy path, a parameter that captures how policy effects evolve under continuous treatments. We show that the policy path can be estimated via score matching by smoothly connecting average marginal effect (AME) and average policy effect (APE) estimation, which improves the interpretability of policy effects.", "date": "2025-12-23", "categories": []}
{"id": "2601.22206", "title": "Causal Imitation Learning Under Measurement Error and Distribution Shift", "abstract": "We study offline imitation learning (IL) when part of the decision-relevant state is observed only through noisy measurements and the distribution may change between training and deployment. Such settings induce spurious state-action correlations, so standard behavioral cloning (BC) -- whether conditioning on raw measurements or ignoring them -- can converge to systematically biased policies under distribution shift. We propose a general framework for IL under measurement error, inspired by explicitly modeling the causal relationships among the variables, yielding a target that retains a causal interpretation and is robust to distribution shift. Building on ideas from proximal causal inference, we introduce \\texttt{CausIL}, which treats noisy state observations as proxy variables, and we provide identification conditions under which the target policy is recoverable from demonstrations without rewards or interactive expert queries. We develop estimators for both discrete and continuous state spaces; for continuous settings, we use an adversarial procedure over RKHS function classes to learn the required parameters. We evaluate \\texttt{CausIL} on semi-simulated longitudinal data from the PhysioNet/Computing in Cardiology Challenge 2019 cohort and demonstrate improved robustness to distribution shift compared to BC baselines.", "date": "2026-01-29", "categories": []}
{"id": "2601.22399", "title": "Score-based Integrated Gradient for Root Cause Explanations of Outliers", "abstract": "Identifying the root causes of outliers is a fundamental problem in causal inference and anomaly detection. Traditional approaches based on heuristics or counterfactual reasoning often struggle under uncertainty and high-dimensional dependencies. We introduce SIREN, a novel and scalable method that attributes the root causes of outliers by estimating the score functions of the data likelihood. Attribution is computed via integrated gradients that accumulate score contributions along paths from the outlier toward the normal data distribution. Our method satisfies three of the four classic Shapley value axioms - dummy, efficiency, and linearity - as well as an asymmetry axiom derived from the underlying causal structure. Unlike prior work, SIREN operates directly on the score function, enabling tractable and uncertainty-aware root cause attribution in nonlinear, high-dimensional, and heteroscedastic causal models. Extensive experiments on synthetic random graphs and real-world cloud service and supply chain datasets show that SIREN outperforms state-of-the-art baselines in both attribution accuracy and computational efficiency.", "date": "2026-01-29", "categories": []}
{"id": "2403.15711", "title": "Towards Identifiable Latent Additive Noise Models", "abstract": "Causal representation learning (CRL) offers the promise of uncovering the underlying causal model by which observed data was generated, but the practical applicability of existing methods remains limited by the strong assumptions required for identifiability and by challenges in applying them to real-world settings. Most current approaches are applicable only to relatively restrictive model classes, such as linear or polynomial models, which limits their flexibility and robustness in practice. One promising approach to this problem seeks to address these issues by leveraging changes in causal influences among latent variables. In this vein we propose a more general and relaxed framework than typically applied, formulated by imposing constraints on the function classes applied. Within this framework, we establish partial identifiability results under weaker conditions, including scenarios where only a subset of causal influences change. We then extend our analysis to a broader class of latent post-nonlinear models. Building on these theoretical insights, we develop a flexible method for learning latent causal representations. We demonstrate the effectiveness of our approach on synthetic and semi-synthetic datasets, and further showcase its applicability in a case study on human motion analysis, a complex real-world domain that also highlights the potential to broaden the practical reach of identifiable CRL models.", "date": "2024-03-23", "categories": []}
{"id": "2601.20090", "title": "Should I Have Expressed a Different Intent? Counterfactual Generation for LLM-Based Autonomous Control", "abstract": "Large language model (LLM)-powered agents can translate high-level user intents into plans and actions in an environment. Yet after observing an outcome, users may wonder: What if I had phrased my intent differently? We introduce a framework that enables such counterfactual reasoning in agentic LLM-driven control scenarios, while providing formal reliability guarantees. Our approach models the closed-loop interaction between a user, an LLM-based agent, and an environment as a structural causal model (SCM), and leverages test-time scaling to generate multiple candidate counterfactual outcomes via probabilistic abduction. Through an offline calibration phase, the proposed conformal counterfactual generation (CCG) yields sets of counterfactual outcomes that are guaranteed to contain the true counterfactual outcome with high probability. We showcase the performance of CCG on a wireless network control use case, demonstrating significant advantages compared to naive re-execution baselines.", "date": "2026-01-27", "categories": []}
{"id": "2601.21135", "title": "TRACE: Trajectory Recovery for Continuous Mechanism Evolution in Causal Representation Learning", "abstract": "Temporal causal representation learning methods assume that causal mechanisms switch instantaneously between discrete domains, yet real-world systems often exhibit continuous mechanism transitions. For example, a vehicle's dynamics evolve gradually through a turning maneuver, and human gait shifts smoothly from walking to running. We formalize this setting by modeling transitional mechanisms as convex combinations of finitely many atomic mechanisms, governed by time-varying mixing coefficients. Our theoretical contributions establish that both the latent causal variables and the continuous mixing trajectory are jointly identifiable. We further propose TRACE, a Mixture-of-Experts framework where each expert learns one atomic mechanism during training, enabling recovery of mechanism trajectories at test time. This formulation generalizes to intermediate mechanism states never observed during training. Experiments on synthetic and real-world data demonstrate that TRACE recovers mixing trajectories with up to 0.99 correlation, substantially outperforming discrete-switching baselines.", "date": "2026-01-29", "categories": []}
{"id": "2601.21567", "title": "FlexCausal: Flexible Causal Disentanglement via Structural Flow Priors and Manifold-Aware Interventions", "abstract": "Causal Disentangled Representation Learning(CDRL) aims to learn and disentangle low dimensional representations and their underlying causal structure from observations. However, existing disentanglement methods rely on a standard mean-field approximation with a diagonal posterior covariance, which decorrelates all latent dimensions. Additionally, these methods often assume isotropic Gaussian priors for exogenous noise, failing to capture the complex, non-Gaussian statistical properties prevalent in real-world causal factors. Therefore, we propose FlexCausal, a novel CDRL framework based on a block-diagonal covariance VAE. FlexCausal utilizes a Factorized Flow-based Prior to realistically model the complex densities of exogenous noise, effectively decoupling the learning of causal mechanisms from distributional statistics. By integrating supervised alignment objectives with counterfactual consistency constraints, our framework ensures a precise structural correspondence between the learned latent subspaces and the ground-truth causal relations. Finally, we introduce a manifold-aware relative intervention strategy to ensure high-fidelity generation. Experimental results on both synthetic and real-world datasets demonstrate that FlexCausal significantly outperforms other methods.", "date": "2026-01-29", "categories": []}
{"id": "2410.14485", "title": "CaTs and DAGs: Integrating Directed Acyclic Graphs with Transformers for Causally Constrained Predictions", "abstract": "Artificial Neural Networks (ANNs), including fully-connected networks and transformers, are highly flexible and powerful function approximators, widely applied in fields like computer vision and natural language processing. However, their inability to inherently respect causal structures can limit their robustness, making them vulnerable to covariate shift and difficult to interpret/explain. This poses significant challenges for their reliability in real-world applications. In this paper, we introduce Causal Transformers (CaTs), a general model class designed to operate under predefined causal constraints, as specified by a Directed Acyclic Graph (DAG). CaTs retain the powerful function approximation abilities of traditional neural networks while adhering to the underlying structural constraints, improving robustness, reliability, and interpretability at inference time. This approach opens new avenues for deploying neural networks in more demanding, real-world scenarios where robustness and explainability is critical.", "date": "2024-10-18", "categories": []}
{"id": "2502.19741", "title": "Causal Effect Estimation under Networked Interference without Networked Unconfoundedness Assumption", "abstract": "Estimating causal effects under networked interference from observational data is a crucial yet challenging problem. Most existing methods mainly rely on the networked unconfoundedness assumption, which guarantees the identification of networked effects. However, this assumption is often violated due to the latent confounders inherent in observational data, thereby hindering the identification of networked effects. To address this issue, we leverage the rich interaction patterns between units in networks, which provide valuable information for recovering these latent confounders. Building on this insight, we develop a confounder recovery framework that explicitly characterizes three categories of latent confounders in networked settings: those affecting only the unit, those affecting only the unit's neighbors, and those influencing both. Based on this framework, we design a networked effect estimator using identifiable representation learning techniques. From a theoretical standpoint, we prove the identifiability of all three types of latent confounders and, by leveraging the recovered confounders, establish a formal identification result for networked effects. Extensive experiments validate our theoretical findings and demonstrate the effectiveness of the proposed method.", "date": "2025-02-27", "categories": []}
{"id": "2504.11609", "title": "Towards Interpretable Deep Generative Models via Causal Representation Learning", "abstract": "Recent developments in generative artificial intelligence (AI) rely on machine learning techniques such as deep learning and generative modeling to achieve state-of-the-art performance across wide-ranging domains. These methods' surprising performance is due in part to their ability to learn implicit \"representations\" of complex, multi-modal data. Unfortunately, deep neural networks are notoriously black boxes that obscure these representations, making them difficult to interpret or analyze. To resolve these difficulties, one approach is to build new interpretable neural network models from the ground up. This is the goal of the emerging field of causal representation learning (CRL) that uses causality as a vector for building flexible, interpretable, and transferable generative AI. CRL can be seen as a synthesis of three intrinsically statistical ideas: (i) latent variable models such as factor analysis; (ii) causal graphical models with latent variables; and (iii) nonparametric statistics and deep learning. This paper introduces CRL from a statistical perspective, focusing on connections to classical models as well as statistical and causal identifiability results. We also highlights key application areas, implementation strategies, and open statistical questions.", "date": "2025-04-15", "categories": []}
{"id": "2506.12371", "title": "Path-specific effects for pulse-oximetry guided decisions in critical care", "abstract": "Identifying and measuring biases associated with sensitive attributes is a crucial consideration in healthcare to prevent treatment disparities. One prominent issue is inaccurate pulse oximeter readings, which tend to overestimate oxygen saturation for dark-skinned patients and misrepresent supplemental oxygen needs. Most existing research has revealed statistical disparities linking device measurement errors to patient outcomes in intensive care units (ICUs) without causal formalization. This study causally investigates how racial discrepancies in oximetry measurements affect invasive ventilation in ICU settings. We employ a causal inference-based approach using path-specific effects to isolate the impact of bias by race on clinical decision-making. To estimate these effects, we leverage a doubly robust estimator, propose its self-normalized variant for improved sample efficiency, and provide novel finite-sample guarantees. Our methodology is validated on semi-synthetic data and applied to two large real-world health datasets: MIMIC-IV and eICU. Contrary to prior work, our analysis reveals minimal impact of racial discrepancies on invasive ventilation rates. However, path-specific effects mediated by oxygen saturation disparity are more pronounced on ventilation duration, and the severity differs across datasets. Our work provides a novel pipeline for investigating potential disparities in clinical decision-making and, more importantly, highlights the necessity of causal methods to robustly assess fairness in healthcare.", "date": "2025-06-14", "categories": []}
{"id": "2511.08091", "title": "Gateways to Tractability for Satisfiability in Pearl's Causal Hierarchy", "abstract": "Pearl's Causal Hierarchy (PCH) is a central framework for reasoning about probabilistic, interventional, and counterfactual statements, yet the satisfiability problem for PCH formulas is computationally intractable in almost all classical settings. We revisit this challenge through the lens of parameterized complexity and identify the first gateways to tractability. Our results include fixed-parameter and XP-algorithms for satisfiability in key probabilistic and counterfactual fragments, using parameters such as primal treewidth and the number of variables, together with matching hardness results that map the limits of tractability. Technically, we depart from the dynamic programming paradigm typically employed for treewidth-based algorithms and instead exploit structural characterizations of well-formed causal models, providing a new algorithmic toolkit for causal reasoning.", "date": "2025-11-11", "categories": []}
{"id": "2601.16249", "title": "Ordering-based Causal Discovery via Generalized Score Matching", "abstract": "Learning DAG structures from purely observational data remains a long-standing challenge across scientific domains. An emerging line of research leverages the score of the data distribution to initially identify a topological order of the underlying DAG via leaf node detection and subsequently performs edge pruning for graph recovery. This paper extends the score matching framework for causal discovery, which is originally designated for continuous data, and introduces a novel leaf discriminant criterion based on the discrete score function. Through simulated and real-world experiments, we demonstrate that our theory enables accurate inference of true causal orders from observed discrete data and the identified ordering can significantly boost the accuracy of existing causal discovery baselines on nearly all of the settings.", "date": "2026-01-22", "categories": []}
{"id": "2601.18118", "title": "LungCRCT: Causal Representation based Lung CT Processing for Lung Cancer Treatment", "abstract": "Due to silence in early stages, lung cancer has been one of the most leading causes of mortality in cancer patients world-wide. Moreover, major symptoms of lung cancer are hard to differentiate with other respiratory disease symptoms such as COPD, further leading patients to overlook cancer progression in early stages. Thus, to enhance survival rates in lung cancer, early detection from consistent proactive respiratory system monitoring becomes crucial. One of the most prevalent and effective methods for lung cancer monitoring would be low-dose computed tomography(LDCT) chest scans, which led to remarkable enhancements in lung cancer detection or tumor classification tasks under rapid advancements and applications of computer vision based AI models such as EfficientNet or ResNet in image processing. However, though advanced CNN models under transfer learning or ViT based models led to high performing lung cancer detections, due to its intrinsic limitations in terms of correlation dependence and low interpretability due to complexity, expansions of deep learning models to lung cancer treatment analysis or causal intervention analysis simulations are still limited. Therefore, this research introduced LungCRCT: a latent causal representation learning based lung cancer analysis framework that retrieves causal representations of factors within the physical causal mechanism of lung cancer progression. With the use of advanced graph autoencoder based causal discovery algorithms with distance Correlation disentanglement and entropy-based image reconstruction refinement, LungCRCT not only enables causal intervention analysis for lung cancer treatments, but also leads to robust, yet extremely light downstream models in malignant tumor classification tasks with an AUC score of 93.91%.", "date": "2026-01-26", "categories": []}
{"id": "2601.18189", "title": "Smooth, Sparse, and Stable: Finite-Time Exact Skeleton Recovery via Smoothed Proximal Gradients", "abstract": "Continuous optimization has significantly advanced causal discovery, yet existing methods (e.g., NOTEARS) generally guarantee only asymptotic convergence to a stationary point. This often yields dense weighted matrices that require arbitrary post-hoc thresholding to recover a DAG. This gap between continuous optimization and discrete graph structures remains a fundamental challenge. In this paper, we bridge this gap by proposing the Hybrid-Order Acyclicity Constraint (AHOC) and optimizing it via the Smoothed Proximal Gradient (SPG-AHOC). Leveraging the Manifold Identification Property of proximal algorithms, we provide a rigorous theoretical guarantee: the Finite-Time Oracle Property. We prove that under standard identifiability assumptions, SPG-AHOC recovers the exact DAG support (structure) in finite iterations, even when optimizing a smoothed approximation. This result eliminates structural ambiguity, as our algorithm returns graphs with exact zero entries without heuristic truncation. Empirically, SPG-AHOC achieves state-of-the-art accuracy and strongly corroborates the finite-time identification theory.", "date": "2026-01-26", "categories": []}
{"id": "2402.03379", "title": "Entire Chain Uplift Modeling with Context-Enhanced Learning for Intelligent Marketing", "abstract": "Uplift modeling, vital in online marketing, seeks to accurately measure the impact of various strategies, such as coupons or discounts, on different users by predicting the Individual Treatment Effect (ITE). In an e-commerce setting, user behavior follows a defined sequential chain, including impression, click, and conversion. Marketing strategies exert varied uplift effects at each stage within this chain, impacting metrics like click-through and conversion rate. Despite its utility, existing research has neglected to consider the inter-task across all stages impacts within a specific treatment and has insufficiently utilized the treatment information, potentially introducing substantial bias into subsequent marketing decisions. We identify these two issues as the chain-bias problem and the treatment-unadaptive problem. This paper introduces the Entire Chain UPlift method with context-enhanced learning (ECUP), devised to tackle these issues. ECUP consists of two primary components: 1) the Entire Chain-Enhanced Network, which utilizes user behavior patterns to estimate ITE throughout the entire chain space, models the various impacts of treatments on each task, and integrates task prior information to enhance context awareness across all stages, capturing the impact of treatment on different tasks, and 2) the Treatment-Enhanced Network, which facilitates fine-grained treatment modeling through bit-level feature interactions, thereby enabling adaptive feature adjustment. Extensive experiments on public and industrial datasets validate ECUPs effectiveness. Moreover, ECUP has been deployed on the Meituan food delivery platform, serving millions of daily active users, with the related dataset released for future research.", "date": "2024-02-04", "categories": []}
{"id": "2408.08023", "title": "Causal Discovery from Time-Series Data with Short-Term Invariance-Based Convolutional Neural Networks", "abstract": "Causal discovery from time-series data aims to capture both intra-slice (contemporaneous) and inter-slice (time-lagged) causality between variables within the temporal chain, which is crucial for various scientific disciplines. Compared to causal discovery from non-time-series data, causal discovery from time-series data necessitates more serialized samples with a larger amount of observed time steps. To address the challenges, we propose a novel gradient-based causal discovery approach STIC, which focuses on \\textbf{S}hort-\\textbf{T}erm \\textbf{I}nvariance using \\textbf{C}onvolutional neural networks to uncover the causal relationships from time-series data. Specifically, STIC leverages both the short-term time and mechanism invariance of causality within each window observation, which possesses the property of independence, to enhance sample efficiency. Furthermore, we construct two causal convolution kernels, which correspond to the short-term time and mechanism invariance respectively, to estimate the window causal graph. To demonstrate the necessity of convolutional neural networks for causal discovery from time-series data, we theoretically derive the equivalence between convolution and the underlying generative principle of time-series data under the assumption that the additive noise model is identifiable. Experimental evaluations conducted on both synthetic and FMRI benchmark datasets demonstrate that our STIC outperforms baselines significantly and achieves the state-of-the-art performance, particularly when the datasets contain a limited number of observed time steps. Code is available at \\url{https://github.com/HITshenrj/STIC}.", "date": "2024-08-15", "categories": []}
{"id": "2508.08450", "title": "Differentiable Cyclic Causal Discovery Under Unmeasured Confounders", "abstract": "Understanding causal relationships between variables is fundamental across scientific disciplines. Most causal discovery algorithms rely on two key assumptions: (i) all variables are observed, and (ii) the underlying causal graph is acyclic. While these assumptions simplify theoretical analysis, they are often violated in real-world systems, such as biological networks. Existing methods that account for confounders either assume linearity or struggle with scalability. To address these limitations, we propose DCCD-CONF, a novel framework for differentiable learning of nonlinear cyclic causal graphs in the presence of unmeasured confounders using interventional data. Our approach alternates between optimizing the graph structure and estimating the confounder distribution by maximizing the log-likelihood of the data. Through experiments on synthetic data and real-world gene perturbation datasets, we show that DCCD-CONF outperforms state-of-the-art methods in both causal graph recovery and confounder identification. Additionally, we also provide consistency guarantees for our framework, reinforcing its theoretical soundness.", "date": "2025-08-11", "categories": []}
{"id": "2601.16675", "title": "I Guess That's Why They Call it the Blues: Causal Analysis for Audio Classifiers", "abstract": "It is well-known that audio classifiers often rely on non-musically relevant features and spurious correlations to classify audio. Hence audio classifiers are easy to manipulate or confuse, resulting in wrong classifications. While inducing a misclassification is not hard, until now the set of features that the classifiers rely on was not well understood.\n  In this paper we introduce a new method that uses causal reasoning to discover features of the frequency space that are sufficient and necessary for a given classification. We describe an implementation of this algorithm in the tool FreqReX and provide experimental results on a number of standard benchmark datasets. Our experiments show that causally sufficient and necessary subsets allow us to manipulate the outputs of the models in a variety of ways by changing the input very slightly. Namely, a change to one out of 240,000 frequencies results in a change in classification 58% of the time, and the change can be so small that it is practically inaudible. These results show that causal analysis is useful for understanding the reasoning process of audio classifiers and can be used to successfully manipulate their outputs.", "date": "2026-01-23", "categories": []}
{"id": "2508.21804", "title": "Considerations for Estimating Causal Effects of Informatively Timed Treatments", "abstract": "Epidemiological studies are often concerned with estimating causal effects of a sequence of treatment decisions on survival outcomes. In many settings, treatment decisions do not occur at fixed, pre-specified followup times. Rather, timing varies across subjects in ways that may be informative of subsequent treatment decisions and potential outcomes. Awareness of the issue and its potential solutions is lacking in the literature, which motivate this work. Here, we formalize the issue of informative timing, problems associated with ignoring it, and show how g-methods can be used to analyze sequential treatments that are informatively timed. As we describe, in such settings, the waiting times between successive treatment decisions may be properly viewed as a time-varying confounders. Using synthetic examples, we illustrate how g-methods that do not adjust for these waiting times may be biased and how adjustment can be done in scenarios where patients may die or be censored in between treatments. We draw connections between adjustment and identification with discrete-time versus continuous-time models. Finally, we provide implementation guidance and examples using publicly available software. Our concluding message is that 1) considering timing is important for valid inference and 2) correcting for informative timing can be done with g-methods that adjust for waiting times between treatments as time-varying confounders.", "date": "2025-08-29", "categories": []}
{"id": "2510.04842", "title": "Distributionally Robust Causal Abstractions", "abstract": "Causal Abstraction (CA) theory provides a principled framework for relating causal models that describe the same system at different levels of granularity while ensuring interventional consistency between them. Recent methods for learning CAs, however, assume fixed and well-specified exogenous distributions, leaving them vulnerable to environmental shifts and model misspecification. In this work, we address these limitations by introducing the first class of distributionally robust CAs and their associated learning algorithms. The latter cast robust causal abstraction learning as a constrained min-max optimization problem with Wasserstein ambiguity sets. We provide theoretical guarantees for both empirical and Gaussian environments, enabling principled selection of ambiguity set radii and establish quantitative guarantees on worst-case abstraction error. Furthermore, we present empirical evidence across different problems and CA learning methods, demonstrating our framework's robustness not only to environmental shifts but also to structural and intervention mapping misspecification.", "date": "2025-10-06", "categories": []}
{"id": "2601.15360", "title": "Robust X-Learner: Breaking the Curse of Imbalance and Heavy Tails via Robust Cross-Imputation", "abstract": "Estimating Heterogeneous Treatment Effects (HTE) in industrial applications such as AdTech and healthcare presents a dual challenge: extreme class imbalance and heavy-tailed outcome distributions. While the X-Learner framework effectively addresses imbalance through cross-imputation, we demonstrate that it is fundamentally vulnerable to \"Outlier Smearing\" when reliant on Mean Squared Error (MSE) minimization. In this failure mode, the bias from a few extreme observations (\"whales\") in the minority group is propagated to the entire majority group during the imputation step, corrupting the estimated treatment effect structure. To resolve this, we propose the Robust X-Learner (RX-Learner). This framework integrates a redescending {\\gamma}-divergence objective -- structurally equivalent to the Welsch loss under Gaussian assumptions -- into the gradient boosting machinery. We further stabilize the non-convex optimization using a Proxy Hessian strategy grounded in Majorization-Minimization (MM) principles. Empirical evaluation on a semi-synthetic Criteo Uplift dataset demonstrates that the RX-Learner reduces the Precision in Estimation of Heterogeneous Effect (PEHE) metric by 98.6% compared to the standard X-Learner, effectively decoupling the stable \"Core\" population from the volatile \"Periphery\".", "date": "2026-01-21", "categories": []}
{"id": "2601.04110", "title": "Causal Data Augmentation for Robust Fine-Tuning of Tabular Foundation Models", "abstract": "Fine-tuning tabular foundation models (TFMs) under data scarcity is challenging, as early stopping on even scarcer validation data often fails to capture true generalization performance. We propose CausalMixFT, a method that enhances fine-tuning robustness and downstream performance by generating structurally consistent synthetic samples using Structural Causal Models (SCMs) fitted on the target dataset. This approach augments limited real data with causally informed synthetic examples, preserving feature dependencies while expanding training diversity. Evaluated across 33 classification datasets from TabArena and over 2300 fine-tuning runs, our CausalMixFT method consistently improves median normalized ROC-AUC from 0.10 (standard fine-tuning) to 0.12, outperforming purely statistical generators such as CTGAN (-0.01), TabEBM (-0.04), and TableAugment (-0.09). Moreover, it narrows the median validation-test performance correlation gap from 0.67 to 0.30, enabling more reliable validation-based early stopping, a key step toward improving fine-tuning stability under data scarcity. These results demonstrate that incorporating causal structure into data augmentation provides an effective and principled route to fine-tuning tabular foundation models in low-data regimes.", "date": "2026-01-07", "categories": []}
{"id": "2601.15254", "title": "Many Experiments, Few Repetitions, Unpaired Data, and Sparse Effects: Is Causal Inference Possible?", "abstract": "We study the problem of estimating causal effects under hidden confounding in the following unpaired data setting: we observe some covariates $X$ and an outcome $Y$ under different experimental conditions (environments) but do not observe them jointly; we either observe $X$ or $Y$. Under appropriate regularity conditions, the problem can be cast as an instrumental variable (IV) regression with the environment acting as a (possibly high-dimensional) instrument. When there are many environments but only a few observations per environment, standard two-sample IV estimators fail to be consistent. We propose a GMM-type estimator based on cross-fold sample splitting of the instrument-covariate sample and prove that it is consistent as the number of environments grows but the sample size per environment remains constant. We further extend the method to sparse causal effects via $\\ell_1$-regularized estimation and post-selection refitting.", "date": "2026-01-21", "categories": []}
{"id": "2405.15325", "title": "On the Identification of Temporally Causal Representation with Instantaneous Dependence", "abstract": "Temporally causal representation learning aims to identify the latent causal process from time series observations, but most methods require the assumption that the latent causal processes do not have instantaneous relations. Although some recent methods achieve identifiability in the instantaneous causality case, they require either interventions on the latent variables or grouping of the observations, which are in general difficult to obtain in real-world scenarios. To fill this gap, we propose an \\textbf{ID}entification framework for instantane\\textbf{O}us \\textbf{L}atent dynamics (\\textbf{IDOL}) by imposing a sparse influence constraint that the latent causal processes have sparse time-delayed and instantaneous relations. Specifically, we establish identifiability results of the latent causal process based on sufficient variability and the sparse influence constraint by employing contextual information of time series data. Based on these theories, we incorporate a temporally variational inference architecture to estimate the latent variables and a gradient-based sparsity regularization to identify the latent causal process. Experimental results on simulation datasets illustrate that our method can identify the latent causal process. Furthermore, evaluations on multiple human motion forecasting benchmarks with instantaneous dependencies indicate the effectiveness of our method in real-world settings.", "date": "2024-05-24", "categories": []}
{"id": "2510.25005", "title": "Cyclic Counterfactuals under Shift-Scale Interventions", "abstract": "Most counterfactual inference frameworks traditionally assume acyclic structural causal models (SCMs), i.e. directed acyclic graphs (DAGs). However, many real-world systems (e.g. biological systems) contain feedback loops or cyclic dependencies that violate acyclicity. In this work, we study counterfactual inference in cyclic SCMs under shift-scale interventions, i.e., soft, policy-style changes that rescale and/or shift a variable's mechanism.", "date": "2025-10-28", "categories": []}
{"id": "2512.10032", "title": "Cluster-Dags as Powerful Background Knowledge For Causal Discovery", "abstract": "Finding cause-effect relationships is of key importance in science. Causal discovery aims to recover a graph from data that succinctly describes these cause-effect relationships. However, current methods face several challenges, especially when dealing with high-dimensional data and complex dependencies. Incorporating prior knowledge about the system can aid causal discovery. In this work, we leverage Cluster-DAGs as a prior knowledge framework to warm-start causal discovery. We show that Cluster-DAGs offer greater flexibility than existing approaches based on tiered background knowledge and introduce two modified constraint-based algorithms, Cluster-PC and Cluster-FCI, for causal discovery in the fully and partially observed setting, respectively. Empirical evaluation on simulated data demonstrates that Cluster-PC and Cluster-FCI outperform their respective baselines without prior knowledge.", "date": "2025-12-10", "categories": []}
{"id": "2601.09841", "title": "A pipeline for enabling path-specific causal fairness in observational health data", "abstract": "When training machine learning (ML) models for potential deployment in a healthcare setting, it is essential to ensure that they do not replicate or exacerbate existing healthcare biases. Although many definitions of fairness exist, we focus on path-specific causal fairness, which allows us to better consider the social and medical contexts in which biases occur (e.g., direct discrimination by a clinician or model versus bias due to differential access to the healthcare system) and to characterize how these biases may appear in learned models. In this work, we map the structural fairness model to the observational healthcare setting and create a generalizable pipeline for training causally fair models. The pipeline explicitly considers specific healthcare context and disparities to define a target \"fair\" model. Our work fills two major gaps: first, we expand on characterizations of the \"fairness-accuracy\" tradeoff by detangling direct and indirect sources of bias and jointly presenting these fairness considerations alongside considerations of accuracy in the context of broadly known biases. Second, we demonstrate how a foundation model trained without fairness constraints on observational health data can be leveraged to generate causally fair downstream predictions in tasks with known social and medical disparities. This work presents a model-agnostic pipeline for training causally fair machine learning models that address both direct and indirect forms of healthcare bias.", "date": "2026-01-14", "categories": []}
{"id": "2601.10700", "title": "LIBERTy: A Causal Framework for Benchmarking Concept-Based Explanations of LLMs with Structural Counterfactuals", "abstract": "Concept-based explanations quantify how high-level concepts (e.g., gender or experience) influence model behavior, which is crucial for decision-makers in high-stakes domains. Recent work evaluates the faithfulness of such explanations by comparing them to reference causal effects estimated from counterfactuals. In practice, existing benchmarks rely on costly human-written counterfactuals that serve as an imperfect proxy. To address this, we introduce a framework for constructing datasets containing structural counterfactual pairs: LIBERTy (LLM-based Interventional Benchmark for Explainability with Reference Targets). LIBERTy is grounded in explicitly defined Structured Causal Models (SCMs) of the text generation, interventions on a concept propagate through the SCM until an LLM generates the counterfactual. We introduce three datasets (disease detection, CV screening, and workplace violence prediction) together with a new evaluation metric, order-faithfulness. Using them, we evaluate a wide range of methods across five models and identify substantial headroom for improving concept-based explanations. LIBERTy also enables systematic analysis of model sensitivity to interventions: we find that proprietary LLMs show markedly reduced sensitivity to demographic concepts, likely due to post-training mitigation. Overall, LIBERTy provides a much-needed benchmark for developing faithful explainability methods.", "date": "2026-01-15", "categories": []}
{"id": "2601.13365", "title": "CausationEntropy: Pythonic Optimal Causation Entropy", "abstract": "Optimal Causation Entropy (oCSE) is a robust causal network modeling technique that reveals causal networks from dynamical systems and coupled oscillators, distinguishing direct from indirect paths. CausationEntropy is a Python package that implements oCSE and several of its significant optimizations and methodological extensions. In this paper, we introduce the version 1.1 release of CausationEntropy, which includes new synthetic data generators, plotting tools, and several advanced information-theoretical causal network discovery algorithms with criteria for estimating Gaussian, k-nearest neighbors (kNN), geometric k-nearest neighbors (geometric-kNN), kernel density (KDE) and Poisson entropic estimators. The package is easy to install from the PyPi software repository, is thoroughly documented, supplemented with extensive code examples, and is modularly structured to support future additions. The entire codebase is released under the MIT license and is available on GitHub and through PyPi Repository. We expect this package to serve as a benchmark tool for causal discovery in complex dynamical systems.", "date": "2026-01-19", "categories": []}
{"id": "2601.13614", "title": "CauScientist: Teaching LLMs to Respect Data for Causal Discovery", "abstract": "Causal discovery is fundamental to scientific understanding and reliable decision-making. Existing approaches face critical limitations: purely data-driven methods suffer from statistical indistinguishability and modeling assumptions, while recent LLM-based methods either ignore statistical evidence or incorporate unverified priors that can mislead result. To this end, we propose CauScientist, a collaborative framework that synergizes LLMs as hypothesis-generating \"data scientists\" with probabilistic statistics as rigorous \"verifiers\". CauScientist employs hybrid initialization to select superior starting graphs, iteratively refines structures through LLM-proposed modifications validated by statistical criteria, and maintains error memory to guide efficient search space. Experiments demonstrate that CauScientist substantially outperforms purely data-driven baselines, achieving up to 53.8% F1 score improvement and enhancing recall from 35.0% to 100.0%. Notably, while standalone LLM performance degrades with graph complexity, CauScientist reduces structural hamming distance (SHD) by 44.0% compared to Qwen3-32B on 37-node graphs. Our project page is at https://github.com/OpenCausaLab/CauScientist.", "date": "2026-01-20", "categories": []}
{"id": "2601.14099", "title": "Causal feature selection framework for stable soft sensor modeling based on time-delayed cross mapping", "abstract": "Soft sensor modeling plays a crucial role in process monitoring. Causal feature selection can enhance the performance of soft sensor models in industrial applications. However, existing methods ignore two critical characteristics of industrial processes. Firstly, causal relationships between variables always involve time delays, whereas most causal feature selection methods investigate causal relationships in the same time dimension. Secondly, variables in industrial processes are often interdependent, which contradicts the decorrelation assumption of traditional causal inference methods. Consequently, soft sensor models based on existing causal feature selection approaches often lack sufficient accuracy and stability. To overcome these challenges, this paper proposes a causal feature selection framework based on time-delayed cross mapping. Time-delayed cross mapping employs state space reconstruction to effectively handle interdependent variables in causality analysis, and considers varying causal strength across time delay. Time-delayed convergent cross mapping (TDCCM) is introduced for total causal inference, and time-delayed partial cross mapping (TDPCM) is developed for direct causal inference. Then, in order to achieve automatic feature selection, an objective feature selection strategy is presented. The causal threshold is automatically determined based on the model performance on the validation set, and the causal features are then selected. Two real-world case studies show that TDCCM achieves the highest average performance, while TDPCM improves soft sensor stability and performance in the worst scenario. The code is publicly available at https://github.com/dirge1/TDPCM.", "date": "2026-01-20", "categories": []}
{"id": "2509.19814", "title": "Causal Inference under Threshold Manipulation: Bayesian Mixture Modeling and Heterogeneous Treatment Effects", "abstract": "Many marketing applications, including credit card incentive programs, offer rewards to customers who exceed specific spending thresholds to encourage increased consumption. Quantifying the causal effect of these thresholds on customers is crucial for effective marketing strategy design. Although regression discontinuity design is a standard method for such causal inference tasks, its assumptions can be violated when customers, aware of the thresholds, strategically manipulate their spending to qualify for the rewards. To address this issue, we propose a novel framework for estimating the causal effect under threshold manipulation. The main idea is to model the observed spending distribution as a mixture of two distributions: one representing customers strategically affected by the threshold, and the other representing those unaffected. To fit the mixture model, we adopt a two-step Bayesian approach consisting of modeling non-bunching customers and fitting a mixture model to a sample around the threshold. We show posterior contraction of the resulting posterior distribution of the causal effect under large samples. Furthermore, we extend this framework to a hierarchical Bayesian setting to estimate heterogeneous causal effects across customer subgroups, allowing for stable inference even with small subgroup sample sizes. We demonstrate the effectiveness of our proposed methods through simulation studies and illustrate their practical implications using a real-world marketing dataset.", "date": "2025-09-24", "categories": []}
{"id": "2601.10137", "title": "Step-by-Step Causality: Transparent Causal Discovery with Multi-Agent Tree-Query and Adversarial Confidence Estimation", "abstract": "Causal discovery aims to recover ``what causes what'', but classical constraint-based methods (e.g., PC, FCI) suffer from error propagation, and recent LLM-based causal oracles often behave as opaque, confidence-free black boxes. This paper introduces Tree-Query, a tree-structured, multi-expert LLM framework that reduces pairwise causal discovery to a short sequence of queries about backdoor paths, (in)dependence, latent confounding, and causal direction, yielding interpretable judgments with robustness-aware confidence scores. Theoretical guarantees are provided for asymptotic identifiability of four pairwise relations. On data-free benchmarks derived from Mooij et al. and UCI causal graphs, Tree-Query improves structural metrics over direct LLM baselines, and a diet--weight case study illustrates confounder screening and stable, high-confidence causal conclusions. Tree-Query thus offers a principled way to obtain data-free causal priors from LLMs that can complement downstream data-driven causal discovery. Code is available at https://anonymous.4open.science/r/Repo-9B3E-4F96.", "date": "2026-01-15", "categories": []}
{"id": "2509.25599", "title": "Coupling Generative Modeling and an Autoencoder with the Causal Bridge", "abstract": "We consider inferring the causal effect of a treatment (intervention) on an outcome of interest in situations where there is potentially an unobserved confounder influencing both the treatment and the outcome. This is achievable by assuming access to two separate sets of control (proxy) measurements associated with treatment and outcomes, which are used to estimate treatment effects through a function termed the em causal bridge (CB). We present a new theoretical perspective, associated assumptions for when estimating treatment effects with the CB is feasible, and a bound on the average error of the treatment effect when the CB assumptions are violated. From this new perspective, we then demonstrate how coupling the CB with an autoencoder architecture allows for the sharing of statistical strength between observed quantities (proxies, treatment, and outcomes), thus improving the quality of the CB estimates. Experiments on synthetic and real-world data demonstrate the effectiveness of the proposed approach in relation to the state-of-the-art methodology for proxy measurements.", "date": "2025-09-29", "categories": []}
{"id": "2601.09579", "title": "Constraint- and Score-Based Nonlinear Granger Causality Discovery with Kernels", "abstract": "Kernel-based methods are used in the context of Granger Causality to enable the identification of nonlinear causal relationships between time series variables. In this paper, we show that two state of the art kernel-based Granger Causality (GC) approaches can be theoretically unified under the framework of Kernel Principal Component Regression (KPCR), and introduce a method based on this unification, demonstrating that this approach can improve causal identification. Additionally, we introduce a Gaussian Process score-based model with Smooth Information Criterion penalisation on the marginal likelihood, and demonstrate improved performance over existing state of the art time-series nonlinear causal discovery methods. Furthermore, we propose a contemporaneous causal identification algorithm, fully based on GC, using the proposed score-based $GP_{SIC}$ method, and compare its performance to a state of the art contemporaneous time series causal discovery algorithm.", "date": "2026-01-14", "categories": []}
{"id": "2309.12450", "title": "A Convex Framework for Confounding Robust Inference", "abstract": "We study policy evaluation of offline contextual bandits subject to unobserved confounders. Sensitivity analysis methods are commonly used to estimate the policy value under the worst-case confounding over a given uncertainty set. However, existing work often resorts to some coarse relaxation of the uncertainty set for the sake of tractability, leading to overly conservative estimation of the policy value. In this paper, we propose a general estimator that provides a sharp lower bound of the policy value using convex programming. The generality of our estimator enables various extensions such as sensitivity analysis with f-divergence, model selection with cross validation and information criterion, and robust policy learning with the sharp lower bound. Furthermore, our estimation method can be reformulated as an empirical risk minimization problem thanks to the strong duality, which enables us to provide strong theoretical guarantees of the proposed estimator using techniques of the M-estimation.", "date": "2023-09-21", "categories": []}
{"id": "2509.20211", "title": "Practical do-Shapley Explanations with Estimand-Agnostic Causal Inference", "abstract": "Among explainability techniques, SHAP stands out as one of the most popular, but often overlooks the causal structure of the problem. In response, do-SHAP employs interventional queries, but its reliance on estimands hinders its practical application. To address this problem, we propose the use of estimand-agnostic approaches, which allow for the estimation of any identifiable query from a single model, making do-SHAP feasible on complex graphs. We also develop a novel algorithm to significantly accelerate its computation at a negligible cost, as well as a method to explain inaccessible Data Generating Processes. We demonstrate the estimation and computational performance of our approach, and validate it on two real-world datasets, highlighting its potential in obtaining reliable explanations.", "date": "2025-09-24", "categories": []}
{"id": "2601.06140", "title": "Causal and Federated Multimodal Learning for Cardiovascular Risk Prediction under Heterogeneous Populations", "abstract": "Cardiovascular disease (CVD) continues to be the major cause of death globally, calling for predictive models that not only handle diverse and high-dimensional biomedical signals but also maintain interpretability and privacy. We create a single multimodal learning framework that integrates cross modal transformers with graph neural networks and causal representation learning to measure personalized CVD risk. The model combines genomic variation, cardiac MRI, ECG waveforms, wearable streams, and structured EHR data to predict risk while also implementing causal invariance constraints across different clinical subpopulations.\n  To maintain transparency, we employ SHAP based feature attribution, counterfactual explanations and causal latent alignment for understandable risk factors. Besides, we position the design in a federated, privacy, preserving optimization protocol and establish rules for convergence, calibration and uncertainty quantification under distributional shift. Experimental studies based on large-scale biobank and multi institutional datasets reveal state discrimination and robustness, exhibiting fair performance across demographic strata and clinically distinct cohorts. This study paves the way for a principled approach to clinically trustworthy, interpretable and privacy respecting CVD prediction at the population level.", "date": "2026-01-05", "categories": []}
{"id": "2601.06782", "title": "Dimension-reduced outcome-weighted learning for estimating individualized treatment regimes in observational studies", "abstract": "Individualized treatment regimes (ITRs) aim to improve clinical outcomes by assigning treatment based on patient-specific characteristics. However, existing methods often struggle with high-dimensional covariates, limiting accuracy, interpretability, and real-world applicability. We propose a novel sufficient dimension reduction approach that directly targets the contrast between potential outcomes and identifies a low-dimensional subspace of the covariates capturing treatment effect heterogeneity. This reduced representation enables more accurate estimation of optimal ITRs through outcome-weighted learning. To accommodate observational data, our method incorporates kernel-based covariate balancing, allowing treatment assignment to depend on the full covariate set and avoiding the restrictive assumption that the subspace sufficient for modeling heterogeneous treatment effects is also sufficient for confounding adjustment. We show that the proposed method achieves universal consistency, i.e., its risk converges to the Bayes risk, under mild regularity conditions. We demonstrate its finite sample performance through simulations and an analysis of intensive care unit sepsis patient data to determine who should receive transthoracic echocardiography.", "date": "2026-01-11", "categories": []}
{"id": "2601.05474", "title": "Efficient Differentiable Causal Discovery via Reliable Super-Structure Learning", "abstract": "Recently, differentiable causal discovery has emerged as a promising approach to improve the accuracy and efficiency of existing methods. However, when applied to high-dimensional data or data with latent confounders, these methods, often based on off-the-shelf continuous optimization algorithms, struggle with the vast search space, the complexity of the objective function, and the nontrivial nature of graph-theoretical constraints. As a result, there has been a surge of interest in leveraging super-structures to guide the optimization process. Nonetheless, learning an appropriate super-structure at the right level of granularity, and doing so efficiently across various settings, presents significant challenges.\n  In this paper, we propose ALVGL, a novel and general enhancement to the differentiable causal discovery pipeline. ALVGL employs a sparse and low-rank decomposition to learn the precision matrix of the data. We design an ADMM procedure to optimize this decomposition, identifying components in the precision matrix that are most relevant to the underlying causal structure. These components are then combined to construct a super-structure that is provably a superset of the true causal graph. This super-structure is used to initialize a standard differentiable causal discovery method with a more focused search space, thereby improving both optimization efficiency and accuracy.\n  We demonstrate the versatility of ALVGL by instantiating it across a range of structural causal models, including both Gaussian and non-Gaussian settings, with and without unmeasured confounders. Extensive experiments on synthetic and real-world datasets show that ALVGL not only achieves state-of-the-art accuracy but also significantly improves optimization efficiency, making it a reliable and effective solution for differentiable causal discovery.", "date": "2026-01-09", "categories": []}
{"id": "2601.05597", "title": "Good Allocations from Bad Estimates", "abstract": "Conditional average treatment effect (CATE) estimation is the de facto gold standard for targeting a treatment to a heterogeneous population. The method estimates treatment effects up to an error $\\epsilon > 0$ in each of $M$ different strata of the population, targeting individuals in decreasing order of estimated treatment effect until the budget runs out. In general, this method requires $O(M/\\epsilon^2)$ samples. This is best possible if the goal is to estimate all treatment effects up to an $\\epsilon$ error. In this work, we show how to achieve the same total treatment effect as CATE with only $O(M/\\epsilon)$ samples for natural distributions of treatment effects. The key insight is that coarse estimates suffice for near-optimal treatment allocations. In addition, we show that budget flexibility can further reduce the sample complexity of allocation. Finally, we evaluate our algorithm on various real-world RCT datasets. In all cases, it finds nearly optimal treatment allocations with surprisingly few samples. Our work highlights the fundamental distinction between treatment effect estimation and treatment allocation: the latter requires far fewer samples.", "date": "2026-01-09", "categories": []}
{"id": "2601.05647", "title": "Transformer Is Inherently a Causal Learner", "abstract": "We reveal that transformers trained in an autoregressive manner naturally encode time-delayed causal structures in their learned representations. When predicting future values in multivariate time series, the gradient sensitivities of transformer outputs with respect to past inputs directly recover the underlying causal graph, without any explicit causal objectives or structural constraints. We prove this connection theoretically under standard identifiability conditions and develop a practical extraction method using aggregated gradient attributions. On challenging cases such as nonlinear dynamics, long-term dependencies, and non-stationary systems, this approach greatly surpasses the performance of state-of-the-art discovery algorithms, especially as data heterogeneity increases, exhibiting scaling potential where causal accuracy improves with data volume and heterogeneity, a property traditional methods lack. This unifying view lays the groundwork for a future paradigm where causal discovery operates through the lens of foundation models, and foundation models gain interpretability and enhancement through the lens of causality.", "date": "2026-01-09", "categories": []}
{"id": "2405.15120", "title": "A Counterfactual Analysis of the Dishonest Casino", "abstract": "The dishonest casino is a well-known hidden Markov model (HMM) often used in education to introduce HMMs and graphical models. A sequence of die rolls is observed with the casino switching between a fair and a loaded die. Instead of recovering the latent regime through filtering, smoothing, or the Viterbi algorithm, we ask a counterfactual question: how much of the gambler's winnings are caused by the casino's cheating? We introduce a class of structural causal models (SCMs) consistent with the HMM and define the expected winnings attributable to cheating (EWAC). Because EWAC is only partially identifiable, we bound it via linear programs (LPs). Numerical experiments help to develop intuition using benchmark SCMs based on independence, comonotonic, and countermonotonic copulas. Imposing a time homogeneity condition on the SCM yields tighter bounds, whereas relaxing it produces looser bounds that admit an explicit LP solution. Domain knowledge such as pathwise monotonicity or counterfactual stability can be incorporated through additional linear constraints. Finally, we show the time-averaged EWAC becomes fully identifiable as the number of time periods tends to infinity. Our work is the first to develop LP bounds for counterfactuals in an HMM setting, benefiting educational contexts where counterfactual inference is taught.", "date": "2024-05-24", "categories": []}
{"id": "2601.04223", "title": "Beyond Interaction Effects: Two Logics for Studying Population Inequalities", "abstract": "When sociologists and other social scientist ask whether the return to college differs by race and gender, they face a choice between two fundamentally different modes of inquiry. Traditional interaction models follow deductive logic: the researcher specifies which variables moderate effects and tests these hypotheses. Machine learning methods follow inductive logic: algorithms search across vast combinatorial spaces to discover patterns of heterogeneity. This article develops a framework for navigating between these approaches. We show that the choice between deduction and induction reflects a tradeoff between interpretability and flexibility, and we demonstrate through simulation when each approach excels. Our framework is particularly relevant for inequality research, where understanding how treatment effects vary across intersecting social subpopulation is substantively central.", "date": "2025-12-26", "categories": []}
{"id": "2601.04361", "title": "Causally-Aware Information Bottleneck for Domain Adaptation", "abstract": "We tackle a common domain adaptation setting in causal systems. In this setting, the target variable is observed in the source domain but is entirely missing in the target domain. We aim to impute the target variable in the target domain from the remaining observed variables under various shifts. We frame this as learning a compact, mechanism-stable representation. This representation preserves information relevant for predicting the target while discarding spurious variation. For linear Gaussian causal models, we derive a closed-form Gaussian Information Bottleneck (GIB) solution. This solution reduces to a canonical correlation analysis (CCA)-style projection and offers Directed Acyclic Graph (DAG)-aware options when desired. For nonlinear or non-Gaussian data, we introduce a Variational Information Bottleneck (VIB) encoder-predictor. This approach scales to high dimensions and can be trained on source data and deployed zero-shot to the target domain. Across synthetic and real datasets, our approach consistently attains accurate imputations, supporting practical use in high-dimensional causal models and furnishing a unified, lightweight toolkit for causal domain adaptation.", "date": "2026-01-07", "categories": []}
{"id": "2601.04673", "title": "Estimating Causal Effects in Gaussian Linear SCMs with Finite Data", "abstract": "Estimating causal effects from observational data remains a fundamental challenge in causal inference, especially in the presence of latent confounders. This paper focuses on estimating causal effects in Gaussian Linear Structural Causal Models (GL-SCMs), which are widely used due to their analytical tractability. However, parameter estimation in GL-SCMs is often infeasible with finite data, primarily due to overparameterization. To address this, we introduce the class of Centralized Gaussian Linear SCMs (CGL-SCMs), a simplified yet expressive subclass where exogenous variables follow standardized distributions. We show that CGL-SCMs are equally expressive in terms of causal effect identifiability from observational distributions and present a novel EM-based estimation algorithm that can learn CGL-SCM parameters and estimate identifiable causal effects from finite observational samples. Our theoretical analysis is validated through experiments on synthetic data and benchmark causal graphs, demonstrating that the learned models accurately recover causal distributions.", "date": "2026-01-08", "categories": []}
{"id": "2412.11850", "title": "Causal Invariance Learning via Efficient Nonconvex Optimization", "abstract": "Identifying the causal relationship among variables from observational data is an important yet challenging task. This work focuses on identifying the direct causes of an outcome and estimating their magnitude, i.e., learning the causal outcome model. Data from multiple environments provide valuable opportunities to uncover causality by exploiting the invariance principle that the causal outcome model holds across heterogeneous environments. Based on the invariance principle, we propose the Negative Weighted Distributionally Robust Optimization (NegDRO) framework to learn an invariant prediction model. NegDRO minimizes the worst-case combination of risks across multiple environments and enforces invariance by allowing potential negative weights. Under the additive interventions regime, we establish three major contributions: (i) On the statistical side, we provide sufficient and nearly necessary identification conditions under which the invariant prediction model coincides with the causal outcome model; (ii) On the optimization side, despite the nonconvexity of NegDRO, we establish its benign optimization landscape, where all stationary points lie close to the true causal outcome model; (iii) On the computational side, we develop a gradient-based algorithm that provably converges to the causal outcome model, with non-asymptotic convergence rates in both sample size and gradient-descent iterations. In particular, our method avoids exhaustive combinatorial searches over exponentially many subsets of covariates found in the literature, ensuring scalability even when the dimension of the covariates is large. To our knowledge, this is the first causal invariance learning method that finds the approximate global optimality for a nonconvex optimization problem efficiently.", "date": "2024-12-16", "categories": []}
{"id": "2506.12227", "title": "Uncovering Bias Paths with LLM-guided Causal Discovery: An Active Learning and Dynamic Scoring Approach", "abstract": "Ensuring fairness in machine learning requires understanding how sensitive attributes like race or gender causally influence outcomes. Existing causal discovery (CD) methods often struggle to recover fairness-relevant pathways in the presence of noise, confounding, or data corruption. Large language models (LLMs) offer a complementary signal by leveraging semantic priors from variable metadata. We propose a hybrid LLM-guided CD framework that extends a breadth-first search strategy with active learning and dynamic scoring. Variable pairs are prioritized for querying using a composite score combining mutual information, partial correlation, and LLM confidence, enabling more efficient and robust structure discovery. To evaluate fairness sensitivity, we introduce a semi-synthetic benchmark based on the UCI Adult dataset, embedding domain-informed bias pathways alongside noise and latent confounders. We assess how well CD methods recover both global graph structure and fairness-critical paths (e.g., sex-->education-->income). Our results demonstrate that LLM-guided methods, including our active, dynamically scored variant, outperform baselines in recovering fairness-relevant structure under noisy conditions. We analyze when LLM-driven insights complement statistical dependencies and discuss implications for fairness auditing in high-stakes domains.", "date": "2025-06-13", "categories": []}
{"id": "2511.22842", "title": "CausalProfiler: Generating Synthetic Benchmarks for Rigorous and Transparent Evaluation of Causal Machine Learning", "abstract": "Causal machine learning (Causal ML) aims to answer \"what if\" questions using machine learning algorithms, making it a promising tool for high-stakes decision-making. Yet, empirical evaluation practices in Causal ML remain limited. Existing benchmarks often rely on a handful of hand-crafted or semi-synthetic datasets, leading to brittle, non-generalizable conclusions. To bridge this gap, we introduce CausalProfiler, a synthetic benchmark generator for Causal ML methods. Based on a set of explicit design choices about the class of causal models, queries, and data considered, the CausalProfiler randomly samples causal models, data, queries, and ground truths constituting the synthetic causal benchmarks. In this way, Causal ML methods can be rigorously and transparently evaluated under a variety of conditions. This work offers the first random generator of synthetic causal benchmarks with coverage guarantees and transparent assumptions operating on the three levels of causal reasoning: observation, intervention, and counterfactual. We demonstrate its utility by evaluating several state-of-the-art methods under diverse conditions and assumptions, both in and out of the identification regime, illustrating the types of analyses and insights the CausalProfiler enables.", "date": "2025-11-28", "categories": []}
{"id": "2601.03459", "title": "An Expectation-Maximization Algorithm for Domain Adaptation in Gaussian Causal Models", "abstract": "We study the problem of imputing a designated target variable that is systematically missing in a shifted deployment domain, when a Gaussian causal DAG is available from a fully observed source domain. We propose a unified EM-based framework that combines source and target data through the DAG structure to transfer information from observed variables to the missing target. On the methodological side, we formulate a population EM operator in the DAG parameter space and introduce a first-order (gradient) EM update that replaces the costly generalized least-squares M-step with a single projected gradient step. Under standard local strong-concavity and smoothness assumptions and a BWY-style \\cite{Balakrishnan2017EM} gradient-stability (bounded missing-information) condition, we show that this first-order EM operator is locally contractive around the true target parameters, yielding geometric convergence and finite-sample guarantees on parameter error and the induced target-imputation error in Gaussian SEMs under covariate shift and local mechanism shifts. Algorithmically, we exploit the known causal DAG to freeze source-invariant mechanisms and re-estimate only those conditional distributions directly affected by the shift, making the procedure scalable to higher-dimensional models. In experiments on a synthetic seven-node SEM, the 64-node MAGIC-IRRI genetic network, and the Sachs protein-signaling data, the proposed DAG-aware first-order EM algorithm improves target imputation accuracy over a fit-on-source Bayesian network and a Kiiveri-style EM baseline, with the largest gains under pronounced domain shift.", "date": "2026-01-06", "categories": []}
{"id": "2409.20412", "title": "Conformal Prediction for Dose-Response Models with Continuous Treatments", "abstract": "Understanding the dose-response relation between a continuous treatment and the outcome for an individual can greatly drive decision-making, particularly in areas like personalized drug dosing and personalized healthcare interventions. Point estimates are often insufficient in these high-risk environments, highlighting the need for uncertainty quantification to support informed decisions. Conformal prediction, a distribution-free and model-agnostic method for uncertainty quantification, has seen limited application in continuous treatments or dose-response models. To address this gap, we propose a novel methodology that frames the causal dose-response problem as a covariate shift, leveraging weighted conformal prediction. By incorporating propensity estimation, conformal predictive systems, and likelihood ratios, we present a practical solution for generating prediction intervals for dose-response models. Additionally, our method approximates local coverage for every treatment value by applying kernel functions as weights in weighted conformal prediction. Finally, we use a new synthetic benchmark dataset to demonstrate the significance of covariate shift assumptions in achieving robust prediction intervals for dose-response models.", "date": "2024-09-30", "categories": []}
{"id": "2510.14073", "title": "Exploratory Causal Inference in SAEnce", "abstract": "Randomized Controlled Trials are one of the pillars of science; nevertheless, they rely on hand-crafted hypotheses and expensive analysis. Such constraints prevent causal effect estimation at scale, potentially anchoring on popular yet incomplete hypotheses. We propose to discover the unknown effects of a treatment directly from data. For this, we turn unstructured data from a trial into meaningful representations via pretrained foundation models and interpret them via a sparse autoencoder. However, discovering significant causal effects at the neural level is not trivial due to multiple-testing issues and effects entanglement. To address these challenges, we introduce Neural Effect Search, a novel recursive procedure solving both issues by progressive stratification. After assessing the robustness of our algorithm on semi-synthetic experiments, we showcase, in the context of experimental ecology, the first successful unsupervised causal effect identification on a real-world scientific trial.", "date": "2025-10-15", "categories": []}
{"id": "2601.02400", "title": "Detecting and Mitigating Treatment Leakage in Text-Based Causal Inference: Distillation and Sensitivity Analysis", "abstract": "Text-based causal inference increasingly employs textual data as proxies for unobserved confounders, yet this approach introduces a previously undertheorized source of bias: treatment leakage. Treatment leakage occurs when text intended to capture confounding information also contains signals predictive of treatment status, thereby inducing post-treatment bias in causal estimates. Critically, this problem can arise even when documents precede treatment assignment, as authors may employ future-referencing language that anticipates subsequent interventions. Despite growing recognition of this issue, no systematic methods exist for identifying and mitigating treatment leakage in text-as-confounder applications. This paper addresses this gap through three contributions. First, we provide formal statistical and set-theoretic definitions of treatment leakage that clarify when and why bias occurs. Second, we propose four text distillation methods -- similarity-based passage removal, distant supervision classification, salient feature removal, and iterative nullspace projection -- designed to eliminate treatment-predictive content while preserving confounder information. Third, we validate these methods through simulations using synthetic text and an empirical application examining International Monetary Fund structural adjustment programs and child mortality. Our findings indicate that moderate distillation optimally balances bias reduction against confounder retention, whereas overly stringent approaches degrade estimate precision.", "date": "2025-12-30", "categories": []}
{"id": "2601.03032", "title": "Causal Manifold Fairness: Enforcing Geometric Invariance in Representation Learning", "abstract": "Fairness in machine learning is increasingly critical, yet standard approaches often treat data as static points in a high-dimensional space, ignoring the underlying generative structure. We posit that sensitive attributes (e.g., race, gender) do not merely shift data distributions but causally warp the geometry of the data manifold itself. To address this, we introduce Causal Manifold Fairness (CMF), a novel framework that bridges causal inference and geometric deep learning. CMF learns a latent representation where the local Riemannian geometry, defined by the metric tensor and curvature, remains invariant under counterfactual interventions on sensitive attributes. By enforcing constraints on the Jacobian and Hessian of the decoder, CMF ensures that the rules of the latent space (distances and shapes) are preserved across demographic groups. We validate CMF on synthetic Structural Causal Models (SCMs), demonstrating that it effectively disentangles sensitive geometric warping while preserving task utility, offering a rigorous quantification of the fairness-utility trade-off via geometric metrics.", "date": "2026-01-06", "categories": []}
{"id": "2601.03099", "title": "Time-Aware Synthetic Control", "abstract": "The synthetic control (SC) framework is widely used for observational causal inference with time-series panel data. SC has been successful in diverse applications, but existing methods typically treat the ordering of pre-intervention time indices interchangeable. This invariance means they may not fully take advantage of temporal structure when strong trends are present. We propose Time-Aware Synthetic Control (TASC), which employs a state-space model with a constant trend while preserving a low-rank structure of the signal. TASC uses the Kalman filter and Rauch-Tung-Striebel smoother: it first fits a generative time-series model with expectation-maximization and then performs counterfactual inference. We evaluate TASC on both simulated and real-world datasets, including policy evaluation and sports prediction. Our results suggest that TASC offers advantages in settings with strong temporal trends and high levels of observation noise.", "date": "2026-01-06", "categories": []}
{"id": "2601.03203", "title": "Counterfactual Fairness with Graph Uncertainty", "abstract": "Evaluating machine learning (ML) model bias is key to building trustworthy and robust ML systems. Counterfactual Fairness (CF) audits allow the measurement of bias of ML models with a causal framework, yet their conclusions rely on a single causal graph that is rarely known with certainty in real-world scenarios. We propose CF with Graph Uncertainty (CF-GU), a bias evaluation procedure that incorporates the uncertainty of specifying a causal graph into CF. CF-GU (i) bootstraps a Causal Discovery algorithm under domain knowledge constraints to produce a bag of plausible Directed Acyclic Graphs (DAGs), (ii) quantifies graph uncertainty with the normalized Shannon entropy, and (iii) provides confidence bounds on CF metrics. Experiments on synthetic data show how contrasting domain knowledge assumptions support or refute audits of CF, while experiments on real-world data (COMPAS and Adult datasets) pinpoint well-known biases with high confidence, even when supplied with minimal domain knowledge constraints.", "date": "2026-01-06", "categories": []}
{"id": "2207.05067", "title": "On the Representation of Pairwise Causal Background Knowledge and Its Applications in Causal Inference", "abstract": "Pairwise causal background knowledge about the existence or absence of causal edges and paths is frequently encountered in observational studies. Such constraints allow the shared directed and undirected edges in the constrained subclass of Markov equivalent DAGs to be represented as a causal maximally partially directed acyclic graph (MPDAG). In this paper, we first provide a sound and complete graphical characterization of causal MPDAGs and introduce a minimal representation of a causal MPDAG. Then, we give a unified representation for three types of pairwise causal background knowledge, including direct, ancestral and non-ancestral causal knowledge, by introducing a novel concept called direct causal clause (DCC). Using DCCs, we study the consistency and equivalence of pairwise causal background knowledge and show that any pairwise causal background knowledge set can be uniquely and equivalently decomposed into the causal MPDAG representing the refined Markov equivalence class and a minimal residual set of DCCs. Polynomial-time algorithms are also provided for checking consistency and equivalence, as well as for finding the decomposed MPDAG and the residual DCCs. Finally, with pairwise causal background knowledge, we prove a sufficient and necessary condition to identify causal effects and surprisingly find that the identifiability of causal effects only depends on the decomposed MPDAG. We also develop a local IDA-type algorithm to estimate the possible values of an unidentifiable effect. Simulations suggest that pairwise causal background knowledge can significantly improve the identifiability of causal effects.", "date": "2022-07-10", "categories": []}
{"id": "2410.16100", "title": "ExDBN: Learning Dynamic Bayesian Networks using Extended Mixed-Integer Programming Formulations", "abstract": "Causal learning from data has received much attention recently. Bayesian networks can be used to capture causal relationships. There, one recovers a weighted directed acyclic graph in which random variables are represented by vertices, and the weights associated with each edge represent the strengths of the causal relationships between them. This concept is extended to capture dynamic effects by introducing a dependency on past data, which may be captured by the structural equation model. This formalism is utilized in the present contribution to propose a score-based learning algorithm. A mixed-integer quadratic program is formulated and an algorithmic solution proposed, in which the pre-generation of exponentially many acyclicity constraints is avoided by utilizing the so-called branch-and-cut (``lazy constraint'') method. Comparing the novel approach to the state-of-the-art, we show that the proposed approach turns out to produce more accurate results when applied to small and medium-sized synthetic instances containing up to 80 time series. Lastly, two interesting applications in bioscience and finance, to which the method is directly applied, further stress the importance of developing highly accurate, globally convergent solvers that can handle instances of modest size.", "date": "2024-10-21", "categories": []}
{"id": "2510.24639", "title": "Causal Ordering for Structure Learning from Time Series", "abstract": "Predicting causal structure from time series data is crucial for understanding complex phenomena in physiology, brain connectivity, climate dynamics, and socio-economic behaviour. Causal discovery in time series is hindered by the combinatorial complexity of identifying true causal relationships, especially as the number of variables and time points grow. A common approach to simplify the task is the so-called ordering-based methods. Traditional ordering methods inherently limit the representational capacity of the resulting model. In this work, we fix this issue by leveraging multiple valid causal orderings, instead of a single one as standard practice. We propose DOTS (Diffusion Ordered Temporal Structure), using diffusion-based causal discovery for temporal data. By integrating multiple orderings, DOTS effectively recovers the transitive closure of the underlying directed acyclic graph, mitigating spurious artifacts inherent in single-ordering approaches. We formalise the problem under standard assumptions such as stationarity and the additive noise model, and leverage score matching with diffusion processes to enable efficient Hessian estimation. Extensive experiments validate the approach. Empirical evaluations on synthetic and real-world datasets demonstrate that DOTS outperforms state-of-the-art baselines, offering a scalable and robust approach to temporal causal discovery. On synthetic benchmarks ($d{=}\\!3-\\!6$ variables, $T{=}200\\!-\\!5{,}000$ samples), DOTS improves mean window-graph $F1$ from $0.63$ (best baseline) to $0.81$. On the CausalTime real-world benchmark ($d{=}20\\!-\\!36$), while baselines remain the best on individual datasets, DOTS attains the highest average summary-graph $F1$ while halving runtime relative to graph-optimisation methods. These results establish DOTS as a scalable and accurate solution for temporal causal discovery.", "date": "2025-10-28", "categories": []}
{"id": "2512.17341", "title": "Sharp Structure-Agnostic Lower Bounds for General Linear Functional Estimation", "abstract": "We establish a general statistical optimality theory for estimation problems where the target parameter is a linear functional of an unknown nuisance component that must be estimated from data. This formulation covers many causal and predictive parameters and has applications to numerous disciplines. We adopt the structure-agnostic framework introduced by \\citet{balakrishnan2023fundamental}, which poses no structural properties on the nuisance functions other than access to black-box estimators that achieve some statistical estimation rate. This framework is particularly appealing when one is only willing to consider estimation strategies that use non-parametric regression and classification oracles as black-box sub-processes. Within this framework, we first prove the statistical optimality of the celebrated and widely used doubly robust estimators for the Average Treatment Effect (ATE), the most central parameter in causal inference. We then characterize the minimax optimal rate under the general formulation. Notably, we differentiate between two regimes in which double robustness can and cannot be achieved and in which first-order debiasing yields different error rates. Our result implies that first-order debiasing is simultaneously optimal in both regimes. We instantiate our theory by deriving optimal error rates that recover existing results and extend to various settings of interest, including the case when the nuisance is defined by generalized regressions and when covariate shift exists for training and test distribution.", "date": "2025-12-19", "categories": []}
{"id": "2512.24478", "title": "HOLOGRAPH: Active Causal Discovery via Sheaf-Theoretic Alignment of Large Language Model Priors", "abstract": "Causal discovery from observational data remains fundamentally limited by identifiability constraints. Recent work has explored leveraging Large Language Models (LLMs) as sources of prior causal knowledge, but existing approaches rely on heuristic integration that lacks theoretical grounding. We introduce HOLOGRAPH, a framework that formalizes LLM-guided causal discovery through sheaf theory--representing local causal beliefs as sections of a presheaf over variable subsets. Our key insight is that coherent global causal structure corresponds to the existence of a global section, while topological obstructions manifest as non-vanishing sheaf cohomology. We propose the Algebraic Latent Projection to handle hidden confounders and Natural Gradient Descent on the belief manifold for principled optimization. Experiments on synthetic and real-world benchmarks demonstrate that HOLOGRAPH provides rigorous mathematical foundations while achieving competitive performance on causal discovery tasks with 50-100 variables. Our sheaf-theoretic analysis reveals that while Identity, Transitivity, and Gluing axioms are satisfied to numerical precision (<10^{-6}), the Locality axiom fails for larger graphs, suggesting fundamental non-local coupling in latent variable projections. Code is available at [https://github.com/hyunjun1121/holograph](https://github.com/hyunjun1121/holograph).", "date": "2025-12-30", "categories": []}
{"id": "2601.01368", "title": "Causal discovery for linear causal model with correlated noise: an Adversarial Learning Approach", "abstract": "Causal discovery from data with unmeasured confounding factors is a challenging problem. This paper proposes an approach based on the f-GAN framework, learning the binary causal structure independent of specific weight values. We reformulate the structure learning problem as minimizing Bayesian free energy and prove that this problem is equivalent to minimizing the f-divergence between the true data distribution and the model-generated distribution. Using the f-GAN framework, we transform this objective into a min-max adversarial optimization problem. We implement the gradient search in the discrete graph space using Gumbel-Softmax relaxation.", "date": "2026-01-04", "categories": []}
{"id": "2601.01511", "title": "Reading Between the Lines: Deconfounding Causal Estimates using Text Embeddings and Deep Learning", "abstract": "Estimating causal treatment effects in observational settings is frequently compromised by selection bias arising from unobserved confounders. While traditional econometric methods struggle when these confounders are orthogonal to structured covariates, high-dimensional unstructured text often contains rich proxies for these latent variables. This study proposes a Neural Network-Enhanced Double Machine Learning (DML) framework designed to leverage text embeddings for causal identification. Using a rigorous synthetic benchmark, we demonstrate that unstructured text embeddings capture critical confounding information that is absent from structured tabular data. However, we show that standard tree-based DML estimators retain substantial bias (+24%) due to their inability to model the continuous topology of embedding manifolds. In contrast, our deep learning approach reduces bias to -0.86% with optimized architectures, effectively recovering the ground-truth causal parameter. These findings suggest that deep learning architectures are essential for satisfying the unconfoundedness assumption when conditioning on high-dimensional natural language data", "date": "2026-01-04", "categories": []}
{"id": "2601.02314", "title": "Project Ariadne: A Structural Causal Framework for Auditing Faithfulness in LLM Agents", "abstract": "As Large Language Model (LLM) agents are increasingly tasked with high-stakes autonomous decision-making, the transparency of their reasoning processes has become a critical safety concern. While \\textit{Chain-of-Thought} (CoT) prompting allows agents to generate human-readable reasoning traces, it remains unclear whether these traces are \\textbf{faithful} generative drivers of the model's output or merely \\textbf{post-hoc rationalizations}. We introduce \\textbf{Project Ariadne}, a novel XAI framework that utilizes Structural Causal Models (SCMs) and counterfactual logic to audit the causal integrity of agentic reasoning. Unlike existing interpretability methods that rely on surface-level textual similarity, Project Ariadne performs \\textbf{hard interventions} ($do$-calculus) on intermediate reasoning nodes -- systematically inverting logic, negating premises, and reversing factual claims -- to measure the \\textbf{Causal Sensitivity} ($\\phi$) of the terminal answer. Our empirical evaluation of state-of-the-art models reveals a persistent \\textit{Faithfulness Gap}. We define and detect a widespread failure mode termed \\textbf{Causal Decoupling}, where agents exhibit a violation density ($\\rho$) of up to $0.77$ in factual and scientific domains. In these instances, agents arrive at identical conclusions despite contradictory internal logic, proving that their reasoning traces function as \"Reasoning Theater\" while decision-making is governed by latent parametric priors. Our findings suggest that current agentic architectures are inherently prone to unfaithful explanation, and we propose the Ariadne Score as a new benchmark for aligning stated logic with model action.", "date": "2026-01-05", "categories": []}
{"id": "2505.16037", "title": "Causal LLM Routing: End-to-End Regret Minimization from Observational Data", "abstract": "LLM routing aims to select the most appropriate model for each query, balancing competing performance metrics such as accuracy and cost across a pool of language models. Prior approaches typically adopt a decoupled strategy, where the metrics are first predicted and the model is then selected based on these estimates. This setup is prone to compounding errors and often relies on full-feedback data, where each query is evaluated by all candidate models, which is costly to obtain and maintain in practice. In contrast, we learn from observational data, which records only the outcome of the model actually deployed. We propose a causal end-to-end framework that learns routing policies by minimizing decision-making regret from observational data. To enable efficient optimization, we introduce two theoretically grounded surrogate objectives: a classification-based upper bound, and a softmax-weighted regret approximation shown to recover the optimal policy at convergence. We further extend our framework to handle heterogeneous cost preferences via an interval-conditioned architecture. Experiments on public benchmarks show that our method outperforms existing baselines, achieving state-of-the-art performance across different embedding models.", "date": "2025-05-21", "categories": []}
{"id": "2601.00200", "title": "Detecting Unobserved Confounders: A Kernelized Regression Approach", "abstract": "Detecting unobserved confounders is crucial for reliable causal inference in observational studies. Existing methods require either linearity assumptions or multiple heterogeneous environments, limiting applicability to nonlinear single-environment settings. To bridge this gap, we propose Kernel Regression Confounder Detection (KRCD), a novel method for detecting unobserved confounding in nonlinear observational data under single-environment conditions. KRCD leverages reproducing kernel Hilbert spaces to model complex dependencies. By comparing standard and higherorder kernel regressions, we derive a test statistic whose significant deviation from zero indicates unobserved confounding. Theoretically, we prove two key results: First, in infinite samples, regression coefficients coincide if and only if no unobserved confounders exist. Second, finite-sample differences converge to zero-mean Gaussian distributions with tractable variance. Extensive experiments on synthetic benchmarks and the Twins dataset demonstrate that KRCD not only outperforms existing baselines but also achieves superior computational efficiency.", "date": "2026-01-01", "categories": []}
{"id": "2512.24564", "title": "CPR: Causal Physiological Representation Learning for Robust ECG Analysis under Distribution Shifts", "abstract": "Deep learning models for Electrocardiogram (ECG) diagnosis have achieved remarkable accuracy but exhibit fragility against adversarial perturbations, particularly Smooth Adversarial Perturbations (SAP) that mimic biological morphology. Existing defenses face a critical dilemma: Adversarial Training (AT) provides robustness but incurs a prohibitive computational burden, while certified methods like Randomized Smoothing (RS) introduce significant inference latency, rendering them impractical for real-time clinical monitoring. We posit that this vulnerability stems from the models' reliance on non-robust spurious correlations rather than invariant pathological features. To address this, we propose Causal Physiological Representation Learning (CPR). Unlike standard denoising approaches that operate without semantic constraints, CPR incorporates a Physiological Structural Prior within a causal disentanglement framework. By modeling ECG generation via a Structural Causal Model (SCM), CPR enforces a structural intervention that strictly separates invariant pathological morphology (P-QRS-T complex) from non-causal artifacts. Empirical results on PTB-XL demonstrate that CPR significantly outperforms standard clinical preprocessing methods. Specifically, under SAP attacks, CPR achieves an F1 score of 0.632, surpassing Median Smoothing (0.541 F1) by 9.1%. Crucially, CPR matches the certified robustness of Randomized Smoothing while maintaining single-pass inference efficiency, offering a superior trade-off between robustness, efficiency, and clinical interpretability.", "date": "2025-12-31", "categories": []}
{"id": "2512.24696", "title": "Causal Discovery with Mixed Latent Confounding via Precision Decomposition", "abstract": "We study causal discovery from observational data in linear Gaussian systems affected by \\emph{mixed latent confounding}, where some unobserved factors act broadly across many variables while others influence only small subsets. This setting is common in practice and poses a challenge for existing methods: differentiable and score-based DAG learners can misinterpret global latent effects as causal edges, while latent-variable graphical models recover only undirected structure.\n  We propose \\textsc{DCL-DECOR}, a modular, precision-led pipeline that separates these roles. The method first isolates pervasive latent effects by decomposing the observed precision matrix into a structured component and a low-rank component. The structured component corresponds to the conditional distribution after accounting for pervasive confounders and retains only local dependence induced by the causal graph and localized confounding. A correlated-noise DAG learner is then applied to this deconfounded representation to recover directed edges while modeling remaining structured error correlations, followed by a simple reconciliation step to enforce bow-freeness.\n  We provide identifiability results that characterize the recoverable causal target under mixed confounding and show how the overall problem reduces to well-studied subproblems with modular guarantees. Synthetic experiments that vary the strength and dimensionality of pervasive confounding demonstrate consistent improvements in directed edge recovery over applying correlated-noise DAG learning directly to the confounded data.", "date": "2025-12-31", "categories": []}
{"id": "2505.08128", "title": "Beyond Basic A/B testing: Improving Statistical Efficiency for Business Growth", "abstract": "The standard A/B testing approaches are mostly based on t-test in large scale industry applications. These standard approaches however suffers from low statistical power in business settings, due to nature of small sample-size or non-Gaussian distribution or return-on-investment (ROI) consideration. In this paper, we (i) show the statistical efficiency of using estimating equation and U statistics, which can address these issues separately; and (ii) propose a novel doubly robust generalized U that allows flexible definition of treatment effect, and can handles small samples, distribution robustness, ROI and confounding consideration in one framework. We provide theoretical results on asymptotics and efficiency bounds, together with insights on the efficiency gain from theoretical analysis. We further conduct comprehensive simulation studies, apply the methods to multiple real A/B tests at LinkedIn, and share results and learnings that are broadly useful.", "date": "2025-05-13", "categories": []}
{"id": "2510.25924", "title": "Transferring Causal Effects using Proxies", "abstract": "We consider the problem of estimating a causal effect in a multi-domain setting. The causal effect of interest is confounded by an unobserved confounder and can change between the different domains. We assume that we have access to a proxy of the hidden confounder and that all variables are discrete or categorical. We propose methodology to estimate the causal effect in the target domain, where we assume to observe only the proxy variable. Under these conditions, we prove identifiability (even when treatment and response variables are continuous). We introduce two estimation techniques, prove consistency, and derive confidence intervals. The theoretical results are supported by simulation studies and a real-world example studying the causal effect of website rankings on consumer choices.", "date": "2025-10-29", "categories": []}
{"id": "2512.22150", "title": "Towards Unsupervised Causal Representation Learning via Latent Additive Noise Model Causal Autoencoders", "abstract": "Unsupervised representation learning seeks to recover latent generative factors, yet standard methods relying on statistical independence often fail to capture causal dependencies. A central challenge is identifiability: as established in disentangled representation learning and nonlinear ICA literature, disentangling causal variables from observational data is impossible without supervision, auxiliary signals, or strong inductive biases. In this work, we propose the Latent Additive Noise Model Causal Autoencoder (LANCA) to operationalize the Additive Noise Model (ANM) as a strong inductive bias for unsupervised discovery. Theoretically, we prove that while the ANM constraint does not guarantee unique identifiability in the general mixing case, it resolves component-wise indeterminacy by restricting the admissible transformations from arbitrary diffeomorphisms to the affine class. Methodologically, arguing that the stochastic encoding inherent to VAEs obscures the structural residuals required for latent causal discovery, LANCA employs a deterministic Wasserstein Auto-Encoder (WAE) coupled with a differentiable ANM Layer. This architecture transforms residual independence from a passive assumption into an explicit optimization objective. Empirically, LANCA outperforms state-of-the-art baselines on synthetic physics benchmarks (Pendulum, Flow), and on photorealistic environments (CANDLE), where it demonstrates superior robustness to spurious correlations arising from complex background scenes.", "date": "2025-12-15", "categories": []}
{"id": "2512.22290", "title": "When Algorithms Manage Humans: A Double Machine Learning Approach to Estimating Nonlinear Effects of Algorithmic Control on Gig Worker Performance and Wellbeing", "abstract": "A central question for the future of work is whether person centered management can survive when algorithms take on managerial roles. Standard tools often miss what is happening because worker responses to algorithmic systems are rarely linear. We use a Double Machine Learning framework to estimate a moderated mediation model without imposing restrictive functional forms. Using survey data from 464 gig workers, we find a clear nonmonotonic pattern. Supportive HR practices improve worker wellbeing, but their link to performance weakens in a murky middle where algorithmic oversight is present yet hard to interpret. The relationship strengthens again when oversight is transparent and explainable. These results show why simple linear specifications can miss the pattern and sometimes suggest the opposite conclusion. For platform design, the message is practical: control that is only partly defined creates confusion, but clear rules and credible recourse can make strong oversight workable. Methodologically, the paper shows how Double Machine Learning can be used to estimate conditional indirect effects in organizational research without forcing the data into a linear shape.", "date": "2025-12-25", "categories": []}
{"id": "2512.22846", "title": "Causal-Policy Forest for End-to-End Policy Learning", "abstract": "This study proposes an end-to-end algorithm for policy learning in causal inference. We observe data consisting of covariates, treatment assignments, and outcomes, where only the outcome corresponding to the assigned treatment is observed. The goal of policy learning is to train a policy from the observed data, where a policy is a function that recommends an optimal treatment for each individual, to maximize the policy value. In this study, we first show that maximizing the policy value is equivalent to minimizing the mean squared error for the conditional average treatment effect (CATE) under $\\{-1, 1\\}$ restricted regression models. Based on this finding, we modify the causal forest, an end-to-end CATE estimation algorithm, for policy learning. We refer to our algorithm as the causal-policy forest. Our algorithm has three advantages. First, it is a simple modification of an existing, widely used CATE estimation method, therefore, it helps bridge the gap between policy learning and CATE estimation in practice. Second, while existing studies typically estimate nuisance parameters for policy learning as a separate task, our algorithm trains the policy in a more end-to-end manner. Third, as in standard decision trees and random forests, we train the models efficiently, avoiding computational intractability.", "date": "2025-12-28", "categories": []}
{"id": "2512.23408", "title": "Probabilistic Modelling is Sufficient for Causal Inference", "abstract": "Causal inference is a key research area in machine learning, yet confusion reigns over the tools needed to tackle it. There are prevalent claims in the machine learning literature that you need a bespoke causal framework or notation to answer causal questions. In this paper, we want to make it clear that you \\emph{can} answer any causal inference question within the realm of probabilistic modelling and inference, without causal-specific tools or notation. Through concrete examples, we demonstrate how causal questions can be tackled by writing down the probability of everything. Lastly, we reinterpret causal tools as emerging from standard probabilistic modelling and inference, elucidating their necessity and utility.", "date": "2025-12-29", "categories": []}
{"id": "2507.19855", "title": "Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning", "abstract": "Large Language Models (LLMs), despite their advanced linguistic capabilities, fundamentally lack an intuitive understanding of physical dynamics, which limits their effectiveness in real-world scenarios that require causal reasoning. In this paper, we introduce Causal World Model Induction (CWMI), a novel framework designed to embed an explicit model of causal physics within an LLM. Our approach incorporates a dedicated Causal Physics Module (CPM) and a new training objective called Causal Intervention Loss, encouraging the model to learn cause-and-effect relationships from multimodal data. By training the model to predict the outcomes of hypothetical interventions instead of merely capturing statistical correlations, CWMI develops a robust internal representation of physical laws. Experimental results show that CWMI significantly outperforms state-of-the-art LLMs on zero-shot physical reasoning tasks, including the PIQA benchmark and our newly proposed PhysiCa-Bench dataset. These findings demonstrate that inducing a causal world model is a critical step toward more reliable and generalizable AI systems.", "date": "2025-07-26", "categories": []}
{"id": "2512.21379", "title": "Sensitivity Analysis of the Consistency Assumption", "abstract": "Sensitivity analysis informs causal inference by assessing the sensitivity of conclusions to departures from assumptions. The consistency assumption states that there are no hidden versions of treatment and that the outcome arising naturally equals the outcome arising from intervention. When reasoning about the possibility of consistency violations, it can be helpful to distinguish between covariates and versions of treatment. In the context of surgery, for example, genomic variables are covariates and the skill of a particular surgeon is a version of treatment. There may be hidden versions of treatment, and this paper addresses that concern with a new kind of sensitivity analysis. Whereas many methods for sensitivity analysis are focused on confounding by unmeasured covariates, the methodology of this paper is focused on confounding by hidden versions of treatment. In this paper, new mathematical notation is introduced to support the novel method, and example applications are described.", "date": "2025-12-24", "categories": []}
{"id": "2512.21746", "title": "A Model of Causal Explanation on Neural Networks for Tabular Data", "abstract": "The problem of explaining the results produced by machine learning methods continues to attract attention. Neural network (NN) models, along with gradient boosting machines, are expected to be utilized even in tabular data with high prediction accuracy. This study addresses the related issues of pseudo-correlation, causality, and combinatorial reasons for tabular data in NN predictors. We propose a causal explanation method, CENNET, and a new explanation power index using entropy for the method. CENNET provides causal explanations for predictions by NNs and uses structural causal models (SCMs) effectively combined with the NNs although SCMs are usually not used as predictive models on their own in terms of predictive accuracy. We show that CEN-NET provides such explanations through comparative experiments with existing methods on both synthetic and quasi-real data in classification tasks.", "date": "2025-12-25", "categories": []}
{"id": "2512.13725", "title": "Compressed Causal Reasoning: Quantization and GraphRAG Effects on Interventional and Counterfactual Accuracy", "abstract": "Causal reasoning in Large Language Models spanning association, intervention, and counterfactual inference is essential for reliable decision making in high stakes settings. As deployment shifts toward edge and resource constrained environments, quantized models such as INT8 and NF4 are becoming standard. Yet the impact of precision reduction on formal causal reasoning is poorly understood. To our knowledge, this is the first study to systematically evaluate quantization effects across all three levels of Pearls Causal Ladder. Using a 3000 sample stratified CLadder benchmark, we find that rung level accuracy in Llama 3 8B remains broadly stable under quantization, with NF4 showing less than one percent overall degradation. Interventional queries at rung 2 are the most sensitive to precision loss, whereas counterfactual reasoning at rung 3 is comparatively stable but exhibits heterogeneous weaknesses across query types such as collider bias and backdoor adjustment. Experiments on the CRASS benchmark show near identical performance across precisions, indicating that existing commonsense counterfactual datasets lack the structural sensitivity needed to reveal quantization induced reasoning drift. We further evaluate Graph Retrieval Augmented Generation using ground truth causal graphs and observe a consistent improvement in NF4 interventional accuracy of plus 1.7 percent, partially offsetting compression related degradation. These results suggest that causal reasoning is unexpectedly robust to four bit quantization, graph structured augmentation can selectively reinforce interventional reasoning, and current counterfactual benchmarks fail to capture deeper causal brittleness. This work provides an initial empirical map of compressed causal reasoning and practical guidance for deploying efficient and structurally supported causal AI systems.", "date": "2025-12-13", "categories": []}
{"id": "2512.21211", "title": "Causal-driven attribution (CDA): Estimating channel influence without user-level data", "abstract": "Attribution modelling lies at the heart of marketing effectiveness, yet most existing approaches depend on user-level path data, which are increasingly inaccessible due to privacy regulations and platform restrictions. This paper introduces a Causal-Driven Attribution (CDA) framework that infers channel influence using only aggregated impression-level data, avoiding any reliance on user identifiers or click-path tracking. CDA integrates temporal causal discovery (using PCMCI) with causal effect estimation via a Structural Causal Model to recover directional channel relationships and quantify their contributions to conversions. Using large-scale synthetic data designed to replicate real marketing dynamics, we show that CDA achieves an average relative RMSE of 9.50% when given the true causal graph, and 24.23% when using the predicted graph, demonstrating strong accuracy under correct structure and meaningful signal recovery even under structural uncertainty. CDA captures cross-channel interdependencies while providing interpretable, privacy-preserving attribution insights, offering a scalable and future-proof alternative to traditional path-based models.", "date": "2025-12-24", "categories": []}
{"id": "2310.05805", "title": "Boosted Control Functions: Distribution generalization and invariance in confounded models", "abstract": "Modern machine learning methods and the availability of large-scale data have significantly advanced our ability to predict target quantities from large sets of covariates. However, these methods often struggle under distributional shifts, particularly in the presence of hidden confounding. While the impact of hidden confounding is well-studied in causal effect estimation, e.g., instrumental variables, its implications for prediction tasks under shifting distributions remain underexplored. This work addresses this gap by introducing a strong notion of invariance that, unlike existing weaker notions, allows for distribution generalization even in the presence of nonlinear, non-identifiable structural functions. Central to this framework is the Boosted Control Function (BCF), a novel, identifiable target of inference that satisfies the proposed strong invariance notion and is provably worst-case optimal under distributional shifts. The theoretical foundation of our work lies in Simultaneous Equation Models for Distribution Generalization (SIMDGs), which bridge machine learning with econometrics by describing data-generating processes under distributional shifts. To put these insights into practice, we propose the ControlTwicing algorithm to estimate the BCF using nonparametric machine-learning techniques and study its generalization performance on synthetic and real-world datasets compared to robust and empirical risk minimization approaches.", "date": "2023-10-09", "categories": []}
{"id": "2512.18315", "title": "On Efficient Adjustment for Micro Causal Effects in Summary Causal Graphs", "abstract": "Observational studies in fields such as epidemiology often rely on covariate adjustment to estimate causal effects. Classical graphical criteria, like the back-door criterion and the generalized adjustment criterion, are powerful tools for identifying valid adjustment sets in directed acyclic graphs (DAGs). However, these criteria are not directly applicable to summary causal graphs (SCGs), which are abstractions of DAGs commonly used in dynamic systems. In SCGs, each node typically represents an entire time series and may involve cycles, making classical criteria inapplicable for identifying causal effects. Recent work established complete conditions for determining whether the micro causal effect of a treatment or an exposure $X_{t-\\gamma}$ on an outcome $Y_t$ is identifiable via covariate adjustment in SCGs, under the assumption of no hidden confounding. However, these identifiability conditions have two main limitations. First, they are complex, relying on cumbersome definitions and requiring the enumeration of multiple paths in the SCG, which can be computationally expensive. Second, when these conditions are satisfied, they only provide two valid adjustment sets, limiting flexibility in practical applications. In this paper, we propose an equivalent but simpler formulation of those identifiability conditions and introduce a new criterion that identifies a broader class of valid adjustment sets in SCGs. Additionally, we characterize the quasi-optimal adjustment set among these, i.e., the one that minimizes the asymptotic variance of the causal effect estimator. Our contributions offer both theoretical advancement and practical tools for more flexible and efficient causal inference in abstracted causal graphs.", "date": "2025-12-20", "categories": []}
{"id": "2512.19746", "title": "Robust Causal Directionality Inference in Quantum Inference under MNAR Observation and High-Dimensional Noise", "abstract": "In quantum mechanics, observation actively shapes the system, paralleling the statistical notion of Missing Not At Random (MNAR). This study introduces a unified framework for \\textbf{robust causal directionality inference} in quantum engineering, determining whether relations are system$\\to$observation, observation$\\to$system, or bidirectional.\n  The method integrates CVAE-based latent constraints, MNAR-aware selection models, GEE-stabilized regression, penalized empirical likelihood, and Bayesian optimization. It jointly addresses quantum and classical noise while uncovering causal directionality, with theoretical guarantees for double robustness, perturbation stability, and oracle inequalities.\n  Simulation and real-data analyses (TCGA gene expression, proteomics) show that the proposed MNAR-stabilized CVAE+GEE+AIPW+PEL framework achieves lower bias and variance, near-nominal coverage, and superior quantum-specific diagnostics. This establishes robust causal directionality inference as a key methodological advance for reliable quantum engineering.", "date": "2025-12-18", "categories": []}
{"id": "2501.10190", "title": "Temporal Causal Reasoning with (Non-Recursive) Structural Equation Models", "abstract": "Structural Equation Models (SEM) are the standard approach to representing causal dependencies between variables in causal models. In this paper we propose a new interpretation of SEMs when reasoning about Actual Causality, in which SEMs are viewed as mechanisms transforming the dynamics of exogenous variables into the dynamics of endogenous variables. This allows us to combine counterfactual causal reasoning with existing temporal logic formalisms, and to introduce a temporal logic, CPLTL, for causal reasoning about such structures. We show that the standard restriction to so-called \\textit{recursive} models (with no cycles in the dependency graph) is not necessary in our approach, allowing us to reason about mutually dependent processes and feedback loops. Finally, we introduce new notions of model equivalence for temporal causal models, and show that CPLTL has an efficient model-checking procedure.", "date": "2025-01-17", "categories": []}
{"id": "2512.18083", "title": "Causal Inference as Distribution Adaptation: Optimizing ATE Risk under Propensity Uncertainty", "abstract": "Standard approaches to causal inference, such as Outcome Regression and Inverse Probability Weighted Regression Adjustment (IPWRA), are typically derived through the lens of missing data imputation and identification theory. In this work, we unify these methods from a Machine Learning perspective, reframing ATE estimation as a \\textit{domain adaptation problem under distribution shift}. We demonstrate that the canonical Hajek estimator is a special case of IPWRA restricted to a constant hypothesis class, and that IPWRA itself is fundamentally Importance-Weighted Empirical Risk Minimization designed to correct for the covariate shift between the treated sub-population and the target population.\n  Leveraging this unified framework, we critically examine the optimization objectives of Doubly Robust estimators. We argue that standard methods enforce \\textit{sufficient but not necessary} conditions for consistency by requiring outcome models to be individually unbiased. We define the true \"ATE Risk Function\" and show that minimizing it requires only that the biases of the treated and control models structurally cancel out. Exploiting this insight, we propose the \\textbf{Joint Robust Estimator (JRE)}. Instead of treating propensity estimation and outcome modeling as independent stages, JRE utilizes bootstrap-based uncertainty quantification of the propensity score to train outcome models jointly. By optimizing for the expected ATE risk over the distribution of propensity scores, JRE leverages model degrees of freedom to achieve robustness against propensity misspecification. Simulation studies demonstrate that JRE achieves up to a 15\\% reduction in MSE compared to standard IPWRA in finite-sample regimes with misspecified outcome models.", "date": "2025-12-19", "categories": []}
{"id": "2512.18135", "title": "Unifying Causal Reinforcement Learning: Survey, Taxonomy, Algorithms and Applications", "abstract": "Integrating causal inference (CI) with reinforcement learning (RL) has emerged as a powerful paradigm to address critical limitations in classical RL, including low explainability, lack of robustness and generalization failures. Traditional RL techniques, which typically rely on correlation-driven decision-making, struggle when faced with distribution shifts, confounding variables, and dynamic environments. Causal reinforcement learning (CRL), leveraging the foundational principles of causal inference, offers promising solutions to these challenges by explicitly modeling cause-and-effect relationships. In this survey, we systematically review recent advancements at the intersection of causal inference and RL. We categorize existing approaches into causal representation learning, counterfactual policy optimization, offline causal RL, causal transfer learning, and causal explainability. Through this structured analysis, we identify prevailing challenges, highlight empirical successes in practical applications, and discuss open problems. Finally, we provide future research directions, underscoring the potential of CRL for developing robust, generalizable, and interpretable artificial intelligence systems.", "date": "2025-12-19", "categories": []}
{"id": "2512.18737", "title": "PIPCFR: Pseudo-outcome Imputation with Post-treatment Variables for Individual Treatment Effect Estimation", "abstract": "The estimation of individual treatment effects (ITE) focuses on predicting the outcome changes that result from a change in treatment. A fundamental challenge in observational data is that while we need to infer outcome differences under alternative treatments, we can only observe each individual's outcome under a single treatment. Existing approaches address this limitation either by training with inferred pseudo-outcomes or by creating matched instance pairs. However, recent work has largely overlooked the potential impact of post-treatment variables on the outcome. This oversight prevents existing methods from fully capturing outcome variability, resulting in increased variance in counterfactual predictions. This paper introduces Pseudo-outcome Imputation with Post-treatment Variables for Counterfactual Regression (PIPCFR), a novel approach that incorporates post-treatment variables to improve pseudo-outcome imputation. We analyze the challenges inherent in utilizing post-treatment variables and establish a novel theoretical bound for ITE risk that explicitly connects post-treatment variables to ITE estimation accuracy. Unlike existing methods that ignore these variables or impose restrictive assumptions, PIPCFR learns effective representations that preserve informative components while mitigating bias. Empirical evaluations on both real-world and simulated datasets demonstrate that PIPCFR achieves significantly lower ITE errors compared to existing methods.", "date": "2025-12-21", "categories": []}
{"id": "2512.19194", "title": "Causal Heterogeneous Graph Learning Method for Chronic Obstructive Pulmonary Disease Prediction", "abstract": "Due to the insufficient diagnosis and treatment capabilities at the grassroots level, there are still deficiencies in the early identification and early warning of acute exacerbation of Chronic obstructive pulmonary disease (COPD), often resulting in a high prevalence rate and high burden, but the screening rate is relatively low. In order to gradually improve this situation. In this paper, this study develop a Causal Heterogeneous Graph Representation Learning (CHGRL) method for COPD comorbidity risk prediction method that: a) constructing a heterogeneous Our dataset includes the interaction between patients and diseases; b) A cause-aware heterogeneous graph learning architecture has been constructed, combining causal inference mechanisms with heterogeneous graph learning, which can support heterogeneous graph causal learning for different types of relationships; and c) Incorporate the causal loss function in the model design, and add counterfactual reasoning learning loss and causal regularization loss on the basis of the cross-entropy classification loss. We evaluate our method and compare its performance with strong GNN baselines. Following experimental evaluation, the proposed model demonstrates high detection accuracy.", "date": "2025-12-22", "categories": []}
{"id": "2512.19349", "title": "VIGOR+: Iterative Confounder Generation and Validation via LLM-CEVAE Feedback Loop", "abstract": "Hidden confounding remains a fundamental challenge in causal inference from observational data. Recent advances leverage Large Language Models (LLMs) to generate plausible hidden confounders based on domain knowledge, yet a critical gap exists: LLM-generated confounders often exhibit semantic plausibility without statistical utility. We propose VIGOR+ (Variational Information Gain for iterative cOnfounder Refinement), a novel framework that closes the loop between LLM-based confounder generation and CEVAE-based statistical validation. Unlike prior approaches that treat generation and validation as separate stages, VIGOR+ establishes an iterative feedback mechanism: validation signals from CEVAE (including information gain, latent consistency metrics, and diagnostic messages) are transformed into natural language feedback that guides subsequent LLM generation rounds. This iterative refinement continues until convergence criteria are met. We formalize the feedback mechanism, prove convergence properties under mild assumptions, and provide a complete algorithmic framework.", "date": "2025-12-22", "categories": []}
{"id": "2503.21526", "title": "Constraint-based causal discovery with tiered background knowledge and latent variables in single or overlapping datasets", "abstract": "In this paper we consider the use of tiered background knowledge within constraint based causal discovery. Our focus is on settings relaxing causal sufficiency, i.e. allowing for latent variables which may arise because relevant information could not be measured at all, or not jointly, as in the case of multiple overlapping datasets. We first present novel insights into the properties of the 'tiered FCI' (tFCI) algorithm. Building on this, we introduce a new extension of the IOD (integrating overlapping datasets) algorithm incorporating tiered background knowledge, the 'tiered IOD' (tIOD) algorithm. We show that under full usage of the tiered background knowledge tFCI and tIOD are sound, while simple versions of the tIOD and tFCI are sound and complete. We further show that the tIOD algorithm can often be expected to be considerably more efficient and informative than the IOD algorithm even beyond the obvious restriction of the Markov equivalence classes. We provide a formal result on the conditions for this gain in efficiency and informativeness. Our results are accompanied by a series of examples illustrating the exact role and usefulness of tiered background knowledge.", "date": "2025-03-27", "categories": []}
{"id": "2502.08282", "title": "Individualised Treatment Effects Estimation with Composite Treatments and Composite Outcomes", "abstract": "Estimating individualised treatment effect (ITE) -- that is the causal effect of a set of variables (also called exposures, treatments, actions, policies, or interventions), referred to as \\textit{composite treatments}, on a set of outcome variables of interest, referred to as \\textit{composite outcomes}, for a unit from observational data -- remains a fundamental problem in causal inference with applications across disciplines, such as healthcare, economics, education, social science, marketing, and computer science. Previous work in causal machine learning for ITE estimation is limited to simple settings, like single treatments and single outcomes. This hinders their use in complex real-world scenarios; for example, consider studying the effect of different ICU interventions, such as beta-blockers and statins for a patient admitted for heart surgery, on different outcomes of interest such as atrial fibrillation and in-hospital mortality. The limited research into composite treatments and outcomes is primarily due to data scarcity for all treatments and outcomes. To address the above challenges, we propose a novel and innovative hypernetwork-based approach, called \\emph{H-Learner}, to solve ITE estimation under composite treatments and composite outcomes, which tackles the data scarcity issue by dynamically sharing information across treatments and outcomes. Our empirical analysis with binary and arbitrary composite treatments and outcomes demonstrates the effectiveness of the proposed approach compared to existing methods.", "date": "2025-02-12", "categories": []}
{"id": "2511.11357", "title": "KarmaTS: A Universal Simulation Platform for Multivariate Time Series with Functional Causal Dynamics", "abstract": "We introduce KarmaTS, an interactive framework for constructing lag-indexed, executable spatiotemporal causal graphical models for multivariate time series (MTS) simulation. Motivated by the challenge of access-restricted physiological data, KarmaTS generates synthetic MTS with known causal dynamics and augments real-world datasets with expert knowledge. The system constructs a discrete-time structural causal process (DSCP) by combining expert knowledge and algorithmic proposals in a mixed-initiative, human-in-the-loop workflow. The resulting DSCP supports simulation and causal interventions, including those under user-specified distribution shifts. KarmaTS handles mixed variable types, contemporaneous and lagged edges, and modular edge functionals ranging from parameterizable templates to neural network models. Together, these features enable flexible validation and benchmarking of causal discovery algorithms through expert-informed simulation.", "date": "2025-11-14", "categories": []}
{"id": "2512.16046", "title": "CauSTream: Causal Spatio-Temporal Representation Learning for Streamflow Forecasting", "abstract": "Streamflow forecasting is crucial for water resource management and risk mitigation. While deep learning models have achieved strong predictive performance, they often overlook underlying physical processes, limiting interpretability and generalization. Recent causal learning approaches address these issues by integrating domain knowledge, yet they typically rely on fixed causal graphs that fail to adapt to data. We propose CauStream, a unified framework for causal spatiotemporal streamflow forecasting. CauSTream jointly learns (i) a runoff causal graph among meteorological forcings and (ii) a routing graph capturing dynamic dependencies across stations. We further establish identifiability conditions for these causal structures under a nonparametric setting. We evaluate CauSTream on three major U.S. river basins across three forecasting horizons. The model consistently outperforms prior state-of-the-art methods, with performance gaps widening at longer forecast windows, indicating stronger generalization to unseen conditions. Beyond forecasting, CauSTream also learns causal graphs that capture relationships among hydrological factors and stations. The inferred structures align closely with established domain knowledge, offering interpretable insights into watershed dynamics. CauSTream offers a principled foundation for causal spatiotemporal modeling, with the potential to extend to a wide range of scientific and environmental applications.", "date": "2025-12-18", "categories": []}
{"id": "2512.16233", "title": "DAG Learning from Zero-Inflated Count Data Using Continuous Optimization", "abstract": "We address network structure learning from zero-inflated count data by casting each node as a zero-inflated generalized linear model and optimizing a smooth, score-based objective under a directed acyclic graph constraint. Our Zero-Inflated Continuous Optimization (ZICO) approach uses node-wise likelihoods with canonical links and enforces acyclicity through a differentiable surrogate constraint combined with sparsity regularization. ZICO achieves superior performance with faster runtimes on simulated data. It also performs comparably to or better than common algorithms for reverse engineering gene regulatory networks. ZICO is fully vectorized and mini-batched, enabling learning on larger variable sets with practical runtimes in a wide range of domains.", "date": "2025-12-18", "categories": []}
{"id": "2512.16489", "title": "Advantages and limitations in the use of transfer learning for individual treatment effects in causal machine learning", "abstract": "Generalizing causal knowledge across diverse environments is challenging, especially when estimates from large-scale datasets must be applied to smaller or systematically different contexts, where external validity is critical. Model-based estimators of individual treatment effects (ITE) from machine learning require large sample sizes, limiting their applicability in domains such as behavioral sciences with smaller datasets. We demonstrate how estimation of ITEs with Treatment Agnostic Representation Networks (TARNet; Shalit et al., 2017) can be improved by leveraging knowledge from source datasets and adapting it to new settings via transfer learning (TL-TARNet; Aloui et al., 2023). In simulations that vary source and sample sizes and consider both randomized and non-randomized intervention target settings, the transfer-learning extension TL-TARNet improves upon standard TARNet, reducing ITE error and attenuating bias when a large unbiased source is available and target samples are small. In an empirical application using the India Human Development Survey (IHDS-II), we estimate the effect of mothers' firewood collection time on children's weekly study time; transfer learning pulls the target mean ITEs toward the source ITE estimate, reducing bias in the estimates obtained without transfer. These results suggest that transfer learning for causal models can improve the estimation of ITE in small samples.", "date": "2025-12-18", "categories": []}
{"id": "2512.02491", "title": "Stress-Testing Causal Claims via Cardinality Repairs", "abstract": "Causal analyses derived from observational data underpin high-stakes decisions in domains such as healthcare, public policy, and economics. Yet such conclusions can be surprisingly fragile: even minor data errors - duplicate records, or entry mistakes - may drastically alter causal relationships. This raises a fundamental question: how robust is a causal claim to small, targeted modifications in the data? Addressing this question is essential for ensuring the reliability, interpretability, and reproducibility of empirical findings. We introduce SubCure, a framework for robustness auditing via cardinality repairs. Given a causal query and a user-specified target range for the estimated effect, SubCure identifies a small set of tuples or subpopulations whose removal shifts the estimate into the desired range. This process not only quantifies the sensitivity of causal conclusions but also pinpoints the specific regions of the data that drive those conclusions. We formalize this problem under both tuple- and pattern-level deletion settings and show both are NP-complete. To scale to large datasets, we develop efficient algorithms that incorporate machine unlearning techniques to incrementally update causal estimates without retraining from scratch. We evaluate SubCure across four real-world datasets covering diverse application domains. In each case, it uncovers compact, high-impact subsets whose removal significantly shifts the causal conclusions, revealing vulnerabilities that traditional methods fail to detect. Our results demonstrate that cardinality repair is a powerful and general-purpose tool for stress-testing causal analyses and guarding against misleading claims rooted in ordinary data imperfections.", "date": "2025-12-02", "categories": []}
{"id": "2512.11653", "title": "Causal Inference in Energy Demand Prediction", "abstract": "Energy demand prediction is critical for grid operators, industrial energy consumers, and service providers. Energy demand is influenced by multiple factors, including weather conditions (e.g. temperature, humidity, wind speed, solar radiation), and calendar information (e.g. hour of day and month of year), which further affect daily work and life schedules. These factors are causally interdependent, making the problem more complex than simple correlation-based learning techniques satisfactorily allow for. We propose a structural causal model that explains the causal relationship between these variables. A full analysis is performed to validate our causal beliefs, also revealing important insights consistent with prior studies. For example, our causal model reveals that energy demand responds to temperature fluctuations with season-dependent sensitivity. Additionally, we find that energy demand exhibits lower variance in winter due to the decoupling effect between temperature changes and daily activity patterns. We then build a Bayesian model, which takes advantage of the causal insights we learned as prior knowledge. The model is trained and tested on unseen data and yields state-of-the-art performance in the form of a 3.84 percent MAPE on the test set. The model also demonstrates strong robustness, as the cross-validation across two years of data yields an average MAPE of 3.88 percent.", "date": "2025-12-12", "categories": []}
{"id": "2512.14764", "title": "Scaling Causal Mediation for Complex Systems: A Framework for Root Cause Analysis", "abstract": "Modern operational systems ranging from logistics and cloud infrastructure to industrial IoT, are governed by complex, interdependent processes. Understanding how interventions propagate through such systems requires causal inference methods that go beyond direct effects to quantify mediated pathways. Traditional mediation analysis, while effective in simple settings, fails to scale to the high-dimensional directed acyclic graphs (DAGs) encountered in practice, particularly when multiple treatments and mediators interact. In this paper, we propose a scalable mediation analysis framework tailored for large causal DAGs involving multiple treatments and mediators. Our approach systematically decomposes total effects into interpretable direct and indirect components. We demonstrate its practical utility through applied case studies in fulfillment center logistics, where complex dependencies and non-controllable factors often obscure root causes.", "date": "2025-12-16", "categories": []}
{"id": "2410.10021", "title": "Online Multi-modal Root Cause Identification in Microservice Systems", "abstract": "Root Cause Analysis (RCA) is essential for pinpointing the root causes of failures in microservice systems. Traditional data-driven RCA methods are typically limited to offline applications due to high computational demands, and existing online RCA methods handle only single-modal data, overlooking complex interactions in multi-modal systems. In this paper, we introduce OCEAN, a novel online multi-modal causal structure learning method for root cause localization. OCEAN employs a dilated convolutional neural network to capture long-term temporal dependencies and graph neural networks to learn causal relationships among system entities and key performance indicators. We further design a multi-factor attention mechanism to analyze and reassess the relationships among different metrics and log indicators/attributes for enhanced online causal graph learning. Additionally, a contrastive mutual information maximization-based graph fusion module is developed to effectively model the relationships across various modalities. Extensive experiments on three real-world datasets demonstrate the effectiveness and efficiency of our proposed method.", "date": "2024-10-13", "categories": []}
{"id": "2503.17037", "title": "Unitless Unrestricted Markov-Consistent SCM Generation: Better Benchmark Datasets for Causal Discovery", "abstract": "Causal discovery aims to extract qualitative causal knowledge in the form of causal graphs from data. Because causal ground truth is rarely known in the real world, simulated data plays a vital role in evaluating the performance of the various causal discovery algorithms proposed in the literature. But recent work highlighted certain artifacts of commonly used data generation techniques for a standard class of structural causal models (SCM) that may be nonphysical, including var- and R2-sortability, where the variables' variance and coefficients of determination (R2) after regressing on all other variables, respectively, increase along the causal order. Some causal methods exploit such artifacts, leading to unrealistic expectations for their performance on real-world data. Some modifications have been proposed to remove these artifacts; notably, the internally-standardized structural causal model (iSCM) avoids varsortability and largely alleviates R2-sortability on sparse causal graphs, but exhibits a reversed R2-sortability pattern for denser graphs not featured in their work. We analyze which sortability patterns we expect to see in real data, and propose a method for drawing coefficients that we argue more effectively samples the space of SCMs. Finally, we propose a novel extension of our SCM generation method to the time series setting.", "date": "2025-03-21", "categories": []}
{"id": "2512.09368", "title": "CFLight: Enhancing Safety with Traffic Signal Control through Counterfactual Learning", "abstract": "Traffic accidents result in millions of injuries and fatalities globally, with a significant number occurring at intersections each year. Traffic Signal Control (TSC) is an effective strategy for enhancing safety at these urban junctures. Despite the growing popularity of Reinforcement Learning (RL) methods in optimizing TSC, these methods often prioritize driving efficiency over safety, thus failing to address the critical balance between these two aspects. Additionally, these methods usually need more interpretability. CounterFactual (CF) learning is a promising approach for various causal analysis fields. In this study, we introduce a novel framework to improve RL for safety aspects in TSC. This framework introduces a novel method based on CF learning to address the question: ``What if, when an unsafe event occurs, we backtrack to perform alternative actions, and will this unsafe event still occur in the subsequent period?'' To answer this question, we propose a new structure causal model to predict the result after executing different actions, and we propose a new CF module that integrates with additional ``X'' modules to promote safe RL practices. Our new algorithm, CFLight, which is derived from this framework, effectively tackles challenging safety events and significantly improves safety at intersections through a near-zero collision control strategy. Through extensive numerical experiments on both real-world and synthetic datasets, we demonstrate that CFLight reduces collisions and improves overall traffic performance compared to conventional RL methods and the recent safe RL model. Moreover, our method represents a generalized and safe framework for RL methods, opening possibilities for applications in other domains. The data and code are available in the github https://github.com/AdvancedAI-ComplexSystem/SmartCity/tree/main/CFLight.", "date": "2025-12-10", "categories": []}
{"id": "2512.14361", "title": "Causal Structure Learning for Dynamical Systems with Theoretical Score Analysis", "abstract": "Real world systems evolve in continuous-time according to their underlying causal relationships, yet their dynamics are often unknown. Existing approaches to learning such dynamics typically either discretize time -- leading to poor performance on irregularly sampled data -- or ignore the underlying causality. We propose CaDyT, a novel method for causal discovery on dynamical systems addressing both these challenges. In contrast to state-of-the-art causal discovery methods that model the problem using discrete-time Dynamic Bayesian networks, our formulation is grounded in Difference-based causal models, which allow milder assumptions for modeling the continuous nature of the system. CaDyT leverages exact Gaussian Process inference for modeling the continuous-time dynamics which is more aligned with the underlying dynamical process. We propose a practical instantiation that identifies the causal structure via a greedy search guided by the Algorithmic Markov Condition and Minimum Description Length principle. Our experiments show that CaDyT outperforms state-of-the-art methods on both regularly and irregularly-sampled data, discovering causal networks closer to the true underlying dynamics.", "date": "2025-12-16", "categories": []}
{"id": "2507.05412", "title": "Incorporating Interventional Independence Improves Robustness against Interventional Distribution Shift", "abstract": "We study the problem of learning robust discriminative representations of causally related latent variables given the underlying causal graph and a training set comprising passively collected observational data and interventional data obtained through targeted interventions on some of these latent variables. We desire to learn representations that are robust against the resulting interventional distribution shifts. Existing approaches treat observational and interventional data alike, ignoring the independence relations arising from these interventions, even with known underlying causal models. As a result, their representations lead to large predictive performance disparities between observational and interventional data. This performance disparity worsens when interventional training data is scarce. In this paper, (1) we first identify a strong correlation between this performance disparity and the representations' violation of statistical independence induced during interventions. (2) For linear models, we derive sufficient conditions on the proportion of interventional training data, for which enforcing statistical independence between representations of the intervened node and its non-descendants during interventions lowers the test-time error on interventional data. Combining these insights, (3) we propose RepLIn, a training algorithm that explicitly enforces this statistical independence between interventional representations. We demonstrate the utility of RepLIn on a synthetic dataset, and on real image and text datasets on facial attribute classification and toxicity detection, respectively, with semi-synthetic causal structures. Our experiments show that RepLIn is scalable with the number of nodes in the causal graph and is suitable to improve robustness against interventional distribution shifts of both continuous and discrete latent variables compared to the ERM baselines.", "date": "2025-07-07", "categories": []}
{"id": "2508.01341", "title": "Debiasing Machine Learning Predictions for Causal Inference Without Additional Ground Truth Data: \"One Map, Many Trials\" in Satellite-Driven Poverty Analysis", "abstract": "Machine learning models trained on Earth observation data, such as satellite imagery, have demonstrated significant promise in predicting household-level wealth indices, enabling the creation of high-resolution wealth maps that can be leveraged across multiple causal trials while addressing chronic data scarcity in global development research. However, because standard training objectives prioritize overall predictive accuracy, these predictions often suffer from shrinkage toward the mean, leading to attenuated estimates of causal treatment effects and limiting their utility in policy evaluations. Existing debiasing methods, such as Prediction-Powered Inference (PPI), can handle this attenuation bias but require additional fresh ground-truth data at the downstream stage of causal inference, which restricts their applicability in data-scarce environments. We introduce and evaluate two post-hoc correction methods -- Linear Calibration Correction (LCC) and a Tweedie's correction approach -- that substantially reduce shrinkage-induced prediction bias without relying on newly collected labeled data. LCC applies a simple linear transformation estimated on a held-out calibration split; Tweedie's method locally de-shrink predictions using density score estimates and a noise scale learned upstream. We provide practical diagnostics for when a correction is warranted and discuss practical limitations. Across analytical results, simulations, and experiments with Demographic and Health Surveys (DHS) data, both approaches reduce attenuation; Tweedie's correction yields nearly unbiased treatment-effect estimates, enabling a \"one map, many trials\" paradigm. Although we demonstrate on EO-ML wealth mapping, the methods are not geospatial-specific: they apply to any setting where imputed outcomes are reused downstream (e.g., pollution indices, population density, or LLM-derived indicators).", "date": "2025-08-02", "categories": []}
{"id": "2512.11830", "title": "CR3G: Causal Reasoning for Patient-Centric Explanations in Radiology Report Generation", "abstract": "Automatic chest X-ray report generation is an important area of research aimed at improving diagnostic accuracy and helping doctors make faster decisions. Current AI models are good at finding correlations (or patterns) in medical images. Still, they often struggle to understand the deeper cause-and-effect relationships between those patterns and a patient condition. Causal inference is a powerful approach that goes beyond identifying patterns to uncover why certain findings in an X-ray relate to a specific diagnosis. In this paper, we will explore the prompt-driven framework Causal Reasoning for Patient-Centric Explanations in radiology Report Generation (CR3G) that is applied to chest X-ray analysis to improve understanding of AI-generated reports by focusing on cause-and-effect relationships, reasoning and generate patient-centric explanation. The aim to enhance the quality of AI-driven diagnostics, making them more useful and trustworthy in clinical practice. CR3G has shown better causal relationship capability and explanation capability for 2 out of 5 abnormalities.", "date": "2025-12-03", "categories": []}
{"id": "2512.11840", "title": "Amortized Causal Discovery with Prior-Fitted Networks", "abstract": "In recent years, differentiable penalized likelihood methods have gained popularity, optimizing the causal structure by maximizing its likelihood with respect to the data. However, recent research has shown that errors in likelihood estimation, even on relatively large sample sizes, disallow the discovery of proper structures. We propose a new approach to amortized causal discovery that addresses the limitations of likelihood estimator accuracy. Our method leverages Prior-Fitted Networks (PFNs) to amortize data-dependent likelihood estimation, yielding more reliable scores for structure learning. Experiments on synthetic, simulated, and real-world datasets show significant gains in structure recovery compared to standard baselines. Furthermore, we demonstrate directly that PFNs provide more accurate likelihood estimates than conventional neural network-based approaches.", "date": "2025-12-03", "categories": []}
{"id": "2512.12605", "title": "Causal inference and model explainability tools for retail", "abstract": "Most major retailers today have multiple divisions focused on various aspects, such as marketing, supply chain, online customer experience, store customer experience, employee productivity, and vendor fulfillment. They also regularly collect data corresponding to all these aspects as dashboards and weekly/monthly/quarterly reports. Although several machine learning and statistical techniques have been in place to analyze and predict key metrics, such models typically lack interpretability. Moreover, such techniques also do not allow the validation or discovery of causal links. In this paper, we aim to provide a recipe for applying model interpretability and causal inference for deriving sales insights. In this paper, we review the existing literature on causal inference and interpretability in the context of problems in e-commerce and retail, and apply them to a real-world dataset. We find that an inherently explainable model has a lower variance of SHAP values, and show that including multiple confounders through a double machine learning approach allows us to get the correct sign of causal effect.", "date": "2025-12-14", "categories": []}
{"id": "2512.12804", "title": "Causal Counterfactuals Reconsidered", "abstract": "I develop a novel semantics for probabilities of counterfactuals that generalizes the standard Pearlian semantics: it applies to probabilistic causal models that cannot be extended into realistic structural causal models and are therefore beyond the scope of Pearl's semantics. This generalization is needed because, as I show, such probabilistic causal models arise even in simple settings. My semantics offer a natural compromize in the long-standing debate between Pearl and Dawid over counterfactuals: I agree with Dawid that universal causal determinism and unrealistic variables should be rejected, but I agree with Pearl that a general semantics of counterfactuals is nonetheless possible. I restrict attention to causal models that satisfy the Markov condition, only contain realistic variables, and are causally complete. Although I formulate my proposal using structural causal models, as does Pearl, I refrain from using so-called response variables. Moreover, I prove that my semantics is equivalent to two other recent proposals that do not involve structural causal models, and that it is in line with various comments on stochastic counterfactuals that have appeared in the literature more broadly. Throughout I also reflect on the universality of the Markov condition and explore a novel generalization of causal abstractions", "date": "2025-12-14", "categories": []}
{"id": "2512.13290", "title": "LINA: Learning INterventions Adaptively for Physical Alignment and Generalization in Diffusion Models", "abstract": "Diffusion models (DMs) have achieved remarkable success in image and video generation. However, they still struggle with (1) physical alignment and (2) out-of-distribution (OOD) instruction following. We argue that these issues stem from the models' failure to learn causal directions and to disentangle causal factors for novel recombination. We introduce the Causal Scene Graph (CSG) and the Physical Alignment Probe (PAP) dataset to enable diagnostic interventions. This analysis yields three key insights. First, DMs struggle with multi-hop reasoning for elements not explicitly determined in the prompt. Second, the prompt embedding contains disentangled representations for texture and physics. Third, visual causal structure is disproportionately established during the initial, computationally limited denoising steps. Based on these findings, we introduce LINA (Learning INterventions Adaptively), a novel framework that learns to predict prompt-specific interventions, which employs (1) targeted guidance in the prompt and visual latent spaces, and (2) a reallocated, causality-aware denoising schedule. Our approach enforces both physical alignment and OOD instruction following in image and video DMs, achieving state-of-the-art performance on challenging causal generation tasks and the Winoground dataset. Our project page is at https://opencausalab.github.io/LINA.", "date": "2025-12-15", "categories": []}
{"id": "2410.04996", "title": "Assumption-Lean Post-Integrated Inference with Surrogate Control Outcomes", "abstract": "Data integration methods aim to extract low-dimensional embeddings from high-dimensional outcomes to remove unwanted variations, such as batch effects and unmeasured covariates, across heterogeneous datasets. However, multiple hypothesis testing after integration can be biased due to data-dependent processes. We introduce a robust post-integrated inference method that accounts for latent heterogeneity by utilizing control outcomes. Leveraging causal interpretations, we derive nonparametric identifiability of the direct effects using negative control outcomes. By utilizing surrogate control outcomes as an extension of negative control outcomes, we develop semiparametric inference on projected direct effect estimands, accounting for hidden mediators, confounders, and moderators. These estimands remain statistically meaningful under model misspecifications and with error-prone embeddings. We provide bias quantifications and finite-sample linear expansions with uniform concentration bounds. The proposed doubly robust estimators are consistent and efficient under minimal assumptions and potential misspecification, facilitating data-adaptive estimation with machine learning algorithms. Our proposal is evaluated using random forests through simulations and analysis of single-cell CRISPR perturbed datasets, which may contain potential unmeasured confounders.", "date": "2024-10-07", "categories": []}
{"id": "2509.19112", "title": "Towards Practical Multi-label Causal Discovery in High-Dimensional Event Sequences via One-Shot Graph Aggregation", "abstract": "Understanding causality in event sequences where outcome labels such as diseases or system failures arise from preceding events like symptoms or error codes is critical. Yet remains an unsolved challenge across domains like healthcare or vehicle diagnostics. We introduce CARGO, a scalable multi-label causal discovery method for sparse, high-dimensional event sequences comprising of thousands of unique event types. Using two pretrained causal Transformers as domain-specific foundation models for event sequences. CARGO infers in parallel, per sequence one-shot causal graphs and aggregates them using an adaptive frequency fusion to reconstruct the global Markov boundaries of labels. This two-stage approach enables efficient probabilistic reasoning at scale while bypassing the intractable cost of full-dataset conditional independence testing. Our results on a challenging real-world automotive fault prediction dataset with over 29,100 unique event types and 474 imbalanced labels demonstrate CARGO's ability to perform structured reasoning.", "date": "2025-09-23", "categories": []}
{"id": "2511.06634", "title": "CaberNet: Causal Representation Learning for Cross-Domain HVAC Energy Prediction", "abstract": "Cross-domain HVAC energy prediction is essential for scalable building energy management, particularly because collecting extensive labeled data for every new building is both costly and impractical. Yet, this task remains highly challenging due to the scarcity and heterogeneity of data across different buildings, climate zones, and seasonal patterns. In particular, buildings situated in distinct climatic regions introduce variability that often leads existing methods to overfit to spurious correlations, rely heavily on expert intervention, or compromise on data diversity. To address these limitations, we propose CaberNet, a causal and interpretable deep sequence model that learns invariant (Markov blanket) representations for robust cross-domain prediction. In a purely data-driven fashion and without requiring any prior knowledge, CaberNet integrates i) a global feature gate trained with a self-supervised Bernoulli regularization to distinguish superior causal features from inferior ones, and ii) a domain-wise training scheme that balances domain contributions, minimizes cross-domain loss variance, and promotes latent factor independence. We evaluate CaberNet on real-world datasets collected from three buildings located in three climatically diverse cities, and it consistently outperforms all baselines, achieving a 22.9% reduction in normalized mean squared error (NMSE) compared to the best benchmark. Our code is available at https://github.com/SusCom-Lab/CaberNet-CRL.", "date": "2025-11-10", "categories": []}
{"id": "2512.11219", "title": "Latent Variable Causal Discovery under Selection Bias", "abstract": "Addressing selection bias in latent variable causal discovery is important yet underexplored, largely due to a lack of suitable statistical tools: While various tools beyond basic conditional independencies have been developed to handle latent variables, none have been adapted for selection bias. We make an attempt by studying rank constraints, which, as a generalization to conditional independence constraints, exploits the ranks of covariance submatrices in linear Gaussian models. We show that although selection can significantly complicate the joint distribution, interestingly, the ranks in the biased covariance matrices still preserve meaningful information about both causal structures and selection mechanisms. We provide a graph-theoretic characterization of such rank constraints. Using this tool, we demonstrate that the one-factor model, a classical latent variable model, can be identified under selection bias. Simulations and real-world experiments confirm the effectiveness of using our rank constraints.", "date": "2025-12-12", "categories": []}
{"id": "2512.10156", "title": "Inference for Batched Adaptive Experiments", "abstract": "The advantages of adaptive experiments have led to their rapid adoption in economics, other fields, as well as among practitioners. However, adaptive experiments pose challenges for causal inference. This note suggests a BOLS (batched ordinary least squares) test statistic for inference of treatment effects in adaptive experiments. The statistic provides a precision-equalizing aggregation of per-period treatment-control differences under heteroskedasticity. The combined test statistic is a normalized average of heteroskedastic per-period z-statistics and can be used to construct asymptotically valid confidence intervals. We provide simulation results comparing rejection rates in the typical case with few treatment periods and few (or many) observations per batch.", "date": "2025-12-10", "categories": []}
{"id": "2512.10669", "title": "Learning by Analogy: A Causal Framework for Composition Generalization", "abstract": "Compositional generalization -- the ability to understand and generate novel combinations of learned concepts -- enables models to extend their capabilities beyond limited experiences. While effective, the data structures and principles that enable this crucial capability remain poorly understood. We propose that compositional generalization fundamentally requires decomposing high-level concepts into basic, low-level concepts that can be recombined across similar contexts, similar to how humans draw analogies between concepts. For example, someone who has never seen a peacock eating rice can envision this scene by relating it to their previous observations of a chicken eating rice.\n  In this work, we formalize these intuitive processes using principles of causal modularity and minimal changes. We introduce a hierarchical data-generating process that naturally encodes different levels of concepts and their interaction mechanisms. Theoretically, we demonstrate that this approach enables compositional generalization supporting complex relations between composed concepts, advancing beyond prior work that assumes simpler interactions like additive effects. Critically, we also prove that this latent hierarchical structure is provably recoverable (identifiable) from observable data like text-image pairs, a necessary step for learning such a generative process. To validate our theory, we apply insights from our theoretical framework and achieve significant improvements on benchmark datasets.", "date": "2025-12-11", "categories": []}
{"id": "2211.07264", "title": "Partial counterfactual identification and uplift modeling: theoretical results and real-world assessment", "abstract": "Counterfactuals are central in causal human reasoning and the scientific discovery process. The uplift, also called conditional average treatment effect, measures the causal effect of some action, or treatment, on the outcome of an individual. This paper discusses how it is possible to derive bounds on the probability of counterfactual statements based on uplift terms. First, we derive some original bounds on the probability of counterfactuals and we show that tightness of such bounds depends on the information of the feature set on the uplift term. Then, we propose a point estimator based on the assumption of conditional independence between the counterfactual outcomes. The quality of the bounds and the point estimators are assessed on synthetic data and a large real-world customer data set provided by a telecom company, showing significant improvement over the state of the art.", "date": "2022-11-14", "categories": []}
{"id": "2408.03463", "title": "Identifying treatment response subgroups in observational time-to-event data", "abstract": "Identifying patient subgroups with different treatment responses is an important task to inform medical recommendations, guidelines, and the design of future clinical trials. Existing approaches for treatment effect estimation primarily rely on Randomised Controlled Trials (RCTs), which tend to feature more homogeneous patient groups, making them less relevant for uncovering subgroups in the population encountered in real-world clinical practice. Subgroup analyses established for RCTs suffer from significant statistical biases when applied to observational studies, which benefit from larger and more representative populations. Our work introduces a novel, outcome-guided, subgroup analysis strategy for identifying subgroups of treatment response in both RCTs and observational studies alike. It hence positions itself in-between individualised and average treatment effect estimation to uncover patient subgroups with distinct treatment responses, critical for actionable insights that may influence treatment guidelines. In experiments, our approach significantly outperforms the current state-of-the-art method for subgroup analysis in both randomised and observational treatment regimes.", "date": "2024-08-06", "categories": []}
{"id": "2512.08805", "title": "Identifying counterfactual probabilities using bivariate distributions and uplift modeling", "abstract": "Uplift modeling estimates the causal effect of an intervention as the difference between potential outcomes under treatment and control, whereas counterfactual identification aims to recover the joint distribution of these potential outcomes (e.g., \"Would this customer still have churned had we given them a marketing offer?\"). This joint counterfactual distribution provides richer information than the uplift but is harder to estimate. However, the two approaches are synergistic: uplift models can be leveraged for counterfactual estimation. We propose a counterfactual estimator that fits a bivariate beta distribution to predicted uplift scores, yielding posterior distributions over counterfactual outcomes. Our approach requires no causal assumptions beyond those of uplift modeling. Simulations show the efficacy of the approach, which can be applied, for example, to the problem of customer churn in telecom, where it reveals insights unavailable to standard ML or uplift models alone.", "date": "2025-12-09", "categories": []}
{"id": "2506.20347", "title": "Emergent Granger Causality in Neural Networks: Can Prediction Alone Reveal Structure?", "abstract": "Granger Causality (GC) offers an elegant statistical framework to study the association between multivariate time series data. Vector autoregressive models (VAR) are simple and easy to fit, but have limited application because of their inherent inability to capture more complex (e.g., non-linear) associations. Numerous attempts have already been made in the literature that exploit the functional approximation power of deep neural networks (DNNs) for GC. However, these methods treat GC as a variable selection problem. We present a novel paradigm for investigating the learned GC from a single neural network used for joint modeling of all components of multivariate time series data, which is essentially linked with prediction and assessing the distribution shift in residuals. A deep learning model, with proper regularization, may learn the true GC structure when jointly used for all components of the time series when there is sufficient training data. We propose to uncover the learned GC structure by comparing the model uncertainty or distribution of the residuals when the past of everything is used as compared to the one where a specific time series component is dropped from the model. We also compare the effect of input layer dropout on the ability of a neural network to learn GC. We show that a well-regularized model can learn the true GC structure from the data without explicitly adding terms in the loss function that guide the model to select variables or perform sparse regression under specific settings. We also provide a comparison of deep learning architectures such as CNN, LSTM and transformer models on their ability to discover Granger Causality. The numerical experiments demonstrate that, compared to sparse regression models, a simple joint model is a strong baseline for learning the true GC which has the advantage that it does not require tuning of many extra hyper-parameters.", "date": "2025-06-25", "categories": []}
{"id": "2512.07267", "title": "Non-negative DAG Learning from Time-Series Data", "abstract": "This work aims to learn the directed acyclic graph (DAG) that captures the instantaneous dependencies underlying a multivariate time series. The observed data follow a linear structural vector autoregressive model (SVARM) with both instantaneous and time-lagged dependencies, where the instantaneous structure is modeled by a DAG to reflect potential causal relationships. While recent continuous relaxation approaches impose acyclicity through smooth constraint functions involving powers of the adjacency matrix, they lead to non-convex optimization problems that are challenging to solve. In contrast, we assume that the underlying DAG has only non-negative edge weights, and leverage this additional structure to impose acyclicity via a convex constraint. This enables us to cast the problem of non-negative DAG recovery from multivariate time-series data as a convex optimization problem in abstract form, which we solve using the method of multipliers. Crucially, the convex formulation guarantees global optimality of the solution. Finally, we assess the performance of the proposed method on synthetic time-series data, where it outperforms existing alternatives.", "date": "2025-12-08", "categories": []}
{"id": "2512.07528", "title": "Model-Based Reinforcement Learning Under Confounding", "abstract": "We investigate model-based reinforcement learning in contextual Markov decision processes (C-MDPs) in which the context is unobserved and induces confounding in the offline dataset. In such settings, conventional model-learning methods are fundamentally inconsistent, as the transition and reward mechanisms generated under a behavioral policy do not correspond to the interventional quantities required for evaluating a state-based policy. To address this issue, we adapt a proximal off-policy evaluation approach that identifies the confounded reward expectation using only observable state-action-reward trajectories under mild invertibility conditions on proxy variables. When combined with a behavior-averaged transition model, this construction yields a surrogate MDP whose Bellman operator is well defined and consistent for state-based policies, and which integrates seamlessly with the maximum causal entropy (MaxCausalEnt) model-learning framework. The proposed formulation enables principled model learning and planning in confounded environments where contextual information is unobserved, unavailable, or impractical to collect.", "date": "2025-12-08", "categories": []}
{"id": "2512.07796", "title": "Large Causal Models from Large Language Models", "abstract": "We introduce a new paradigm for building large causal models (LCMs) that exploits the enormous potential latent in today's large language models (LLMs). We describe our ongoing experiments with an implemented system called DEMOCRITUS (Decentralized Extraction of Manifold Ontologies of Causal Relations Integrating Topos Universal Slices) aimed at building, organizing, and visualizing LCMs that span disparate domains extracted from carefully targeted textual queries to LLMs. DEMOCRITUS is methodologically distinct from traditional narrow domain and hypothesis centered causal inference that builds causal models from experiments that produce numerical data. A high-quality LLM is used to propose topics, generate causal questions, and extract plausible causal statements from a diverse range of domains. The technical challenge is then to take these isolated, fragmented, potentially ambiguous and possibly conflicting causal claims, and weave them into a coherent whole, converting them into relational causal triples and embedding them into a LCM. Addressing this technical challenge required inventing new categorical machine learning methods, which we can only briefly summarize in this paper, as it is focused more on the systems side of building DEMOCRITUS. We describe the implementation pipeline for DEMOCRITUS comprising of six modules, examine its computational cost profile to determine where the current bottlenecks in scaling the system to larger models. We describe the results of using DEMOCRITUS over a wide range of domains, spanning archaeology, biology, climate change, economics, medicine and technology. We discuss the limitations of the current DEMOCRITUS system, and outline directions for extending its capabilities.", "date": "2025-12-08", "categories": []}
{"id": "2509.22467", "title": "CausalKANs: interpretable treatment effect estimation with Kolmogorov-Arnold networks", "abstract": "Deep neural networks achieve state-of-the-art performance in estimating heterogeneous treatment effects, but their opacity limits trust and adoption in sensitive domains such as medicine, economics, and public policy. Building on well-established and high-performing causal neural architectures, we propose causalKANs, a framework that transforms neural estimators of conditional average treatment effects (CATEs) into Kolmogorov--Arnold Networks (KANs). By incorporating pruning and symbolic simplification, causalKANs yields interpretable closed-form formulas while preserving predictive accuracy. Experiments on benchmark datasets demonstrate that causalKANs perform on par with neural baselines in CATE error metrics, and that even simple KAN variants achieve competitive performance, offering a favorable accuracy--interpretability trade-off. By combining reliability with analytic accessibility, causalKANs provide auditable estimators supported by closed-form expressions and interpretable plots, enabling trustworthy individualized decision-making in high-stakes settings. We release the code for reproducibility at https://github.com/aalmodovares/causalkans .", "date": "2025-09-26", "categories": []}
{"id": "2512.05373", "title": "Text Rationalization for Robust Causal Effect Estimation", "abstract": "Recent advances in natural language processing have enabled the increasing use of text data in causal inference, particularly for adjusting confounding factors in treatment effect estimation. Although high-dimensional text can encode rich contextual information, it also poses unique challenges for causal identification and estimation. In particular, the positivity assumption, which requires sufficient treatment overlap across confounder values, is often violated at the observational level, when massive text is represented in feature spaces. Redundant or spurious textual features inflate dimensionality, producing extreme propensity scores, unstable weights, and inflated variance in effect estimates. We address these challenges with Confounding-Aware Token Rationalization (CATR), a framework that selects a sparse necessary subset of tokens using a residual-independence diagnostic designed to preserve confounding information sufficient for unconfoundedness. By discarding irrelevant texts while retaining key signals, CATR mitigates observational-level positivity violations and stabilizes downstream causal effect estimators. Experiments on synthetic data and a real-world study using the MIMIC-III database demonstrate that CATR yields more accurate, stable, and interpretable causal effect estimates than existing baselines.", "date": "2025-12-05", "categories": []}
{"id": "2505.11085", "title": "A Fast Kernel-based Conditional Independence test with Application to Causal Discovery", "abstract": "Kernel-based conditional independence (KCI) testing is a powerful nonparametric method commonly employed in causal discovery tasks. Despite its flexibility and statistical reliability, cubic computational complexity limits its application to large datasets. To address this computational bottleneck, we propose \\textit{FastKCI}, a scalable and parallelizable kernel-based conditional independence test that utilizes a mixture-of-experts approach inspired by embarrassingly parallel inference techniques for Gaussian processes. By partitioning the dataset based on a Gaussian mixture model over the conditioning variables, FastKCI conducts local KCI tests in parallel, aggregating the results using an importance-weighted sampling scheme. Experiments on synthetic datasets and benchmarks on real-world production data validate that FastKCI maintains the statistical power of the original KCI test while achieving substantial computational speedups. FastKCI thus represents a practical and efficient solution for conditional independence testing in causal inference on large-scale data.", "date": "2025-05-16", "categories": []}
{"id": "2509.21423", "title": "Near-Optimal Experiment Design in Linear non-Gaussian Cyclic Models", "abstract": "We study the problem of causal structure learning from a combination of observational and interventional data generated by a linear non-Gaussian structural equation model that might contain cycles. Recent results show that using mere observational data identifies the causal graph only up to a permutation-equivalence class. We obtain a combinatorial characterization of this class by showing that each graph in an equivalence class corresponds to a perfect matching in a bipartite graph. This bipartite representation allows us to analyze how interventions modify or constrain the matchings. Specifically, we show that each atomic intervention reveals one edge of the true matching and eliminates all incompatible causal graphs. Consequently, we formalize the optimal experiment design task as an adaptive stochastic optimization problem over the set of equivalence classes with a natural reward function that quantifies how many graphs are eliminated from the equivalence class by an intervention. We show that this reward function is adaptive submodular and provide a greedy policy with a provable near-optimal performance guarantee. A key technical challenge is to efficiently estimate the reward function without having to explicitly enumerate all the graphs in the equivalence class. We propose a sampling-based estimator using random matchings and analyze its bias and concentration behavior. Our simulation results show that performing a small number of interventions guided by our stochastic optimization framework recovers the true underlying causal structure.", "date": "2025-09-25", "categories": []}
{"id": "2512.04980", "title": "Learning Causality for Longitudinal Data", "abstract": "This thesis develops methods for causal inference and causal representation learning (CRL) in high-dimensional, time-varying data.\n  The first contribution introduces the Causal Dynamic Variational Autoencoder (CDVAE), a model for estimating Individual Treatment Effects (ITEs) by capturing unobserved heterogeneity in treatment response driven by latent risk factors that affect only outcomes. CDVAE comes with theoretical guarantees on valid latent adjustment and generalization bounds for ITE error. Experiments on synthetic and real datasets show that CDVAE outperforms baselines, and that state-of-the-art models greatly improve when augmented with its latent substitutes, approaching oracle performance without access to true adjustment variables.\n  The second contribution proposes an efficient framework for long-term counterfactual regression based on RNNs enhanced with Contrastive Predictive Coding (CPC) and InfoMax. It captures long-range dependencies under time-varying confounding while avoiding the computational cost of transformers, achieving state-of-the-art results and introducing CPC into causal inference.\n  The third contribution advances CRL by addressing how latent causes manifest in observed variables. We introduce a model-agnostic interpretability layer based on the geometry of the decoder Jacobian. A sparse self-expression prior induces modular, possibly overlapping groups of observed features aligned with shared latent influences. We provide recovery guarantees in both disjoint and overlapping settings and show that meaningful latent-to-observed structure can be recovered without anchor features or single-parent assumptions. Scalable Jacobian-based regularization techniques are also developed.", "date": "2025-12-04", "categories": []}
{"id": "2512.03060", "title": "A Large Scale Heterogeneous Treatment Effect Estimation Framework and Its Applications of Users' Journey at Snap", "abstract": "Heterogeneous Treatment Effect (HTE) and Conditional Average Treatment Effect (CATE) models relax the assumption that treatment effects are the same for every user. We present a large scale industrial framework for estimating HTE using experimental data from hundreds of millions of Snapchat users. By combining results across many experiments, the framework uncovers latent user characteristics that were previously unmeasurable and produces stable treatment effect estimates at scale.\n  We describe the core components that enabled this system, including experiment selection, base learner design, and incremental training. We also highlight two applications: user influenceability to ads and user sensitivity to ads. An online A/B test using influenceability scores for targeting showed an improvement on key business metrics that is more than six times larger than what is typically considered significant.", "date": "2025-11-25", "categories": []}
{"id": "2512.03428", "title": "GaussDetect-LiNGAM:Causal Direction Identification without Gaussianity test", "abstract": "We propose GaussDetect-LiNGAM, a novel approach for bivariate causal discovery that eliminates the need for explicit Gaussianity tests by leveraging a fundamental equivalence between noise Gaussianity and residual independence in the reverse regression. Under the standard LiNGAM assumptions of linearity, acyclicity, and exogeneity, we prove that the Gaussianity of the forward-model noise is equivalent to the independence between the regressor and residual in the reverse model. This theoretical insight allows us to replace fragile and sample-sensitive Gaussianity tests with robust kernel-based independence tests. Experimental results validate the equivalence and demonstrate that GaussDetect-LiNGAM maintains high consistency across diverse noise types and sample sizes, while reducing the number of tests per decision (TPD). Our method enhances both the efficiency and practical applicability of causal inference, making LiNGAM more accessible and reliable in real-world scenarios.", "date": "2025-12-03", "categories": []}
{"id": "2511.15679", "title": "Front-door Reducibility: Reducing ADMGs to the Standard Front-door Setting via a Graphical Criterion", "abstract": "Front-door adjustment gives a simple closed-form identification formula under the classical front-door criterion, but its applicability is often viewed as narrow. By contrast, the general ID algorithm can identify many more causal effects in arbitrary graphs, yet typically outputs algebraically complex expressions that are hard to estimate and interpret. We show that many such graphs can in fact be reduced to a standard front-door setting via front-door reducibility (FDR), a graphical condition on acyclic directed mixed graphs that aggregates variables into super-nodes $(\\boldsymbol{X}^{*},\\boldsymbol{Y}^{*},\\boldsymbol{M}^{*})$. We characterize the FDR criterion, prove it is equivalent (at the graph level) to the existence of an FDR adjustment, and present FDR-TID, an exact algorithm that finds an admissible FDR triple with correctness, completeness, and finite-termination guarantees. Empirical examples show that many graphs far outside the textbook front-door setting are FDR, yielding simple, estimable adjustments where general ID expressions would be cumbersome. FDR therefore complements existing identification methods by prioritizing interpretability and computational simplicity without sacrificing generality across mixed graphs.", "date": "2025-11-19", "categories": []}
{"id": "2512.02060", "title": "From 'What-is' to 'What-if' in Human-Factor Analysis: A Post-Occupancy Evaluation Case", "abstract": "Human-factor analysis typically employs correlation analysis and significance testing to identify relationships between variables. However, these descriptive ('what-is') methods, while effective for identifying associations, are often insufficient for answering causal ('what-if') questions. Their application in such contexts often overlooks confounding and colliding variables, potentially leading to bias and suboptimal or incorrect decisions.\n  We advocate for explicitly distinguishing descriptive from interventional questions in human-factor analysis, and applying causal inference frameworks specifically to these problems to prevent methodological mismatches. This approach disentangles complex variable relationships and enables counterfactual reasoning. Using post-occupancy evaluation (POE) data from the Center for the Built Environment's (CBE) Occupant Survey as a demonstration case, we show how causal discovery reveals intervention hierarchies and directional relationships that traditional associational analysis misses. The systematic distinction between causally associated and independent variables, combined with intervention prioritization capabilities, offers broad applicability to complex human-centric systems, for example, in building science or ergonomics, where understanding intervention effects is critical for optimization and decision-making.", "date": "2025-11-28", "categories": []}
{"id": "2512.00048", "title": "Causal Reinforcement Learning based Agent-Patient Interaction with Clinical Domain Knowledge", "abstract": "Reinforcement Learning (RL) faces significant challenges in adaptive healthcare interventions, such as dementia care, where data is scarce, decisions require interpretability, and underlying patient-state dynamic are complex and causal in nature. In this work, we present a novel framework called Causal structure-aware Reinforcement Learning (CRL) that explicitly integrates causal discovery and reasoning into policy optimization. This method enables an agent to learn and exploit a directed acyclic graph (DAG) that describes the causal dependencies between human behavioral states and robot actions, facilitating more efficient, interpretable, and robust decision-making. We validate our approach in a simulated robot-assisted cognitive care scenario, where the agent interacts with a virtual patient exhibiting dynamic emotional, cognitive, and engagement states. The experimental results show that CRL agents outperform conventional model-free RL baselines by achieving higher cumulative rewards, maintaining desirable patient states more consistently, and exhibiting interpretable, clinically-aligned behavior. We further demonstrate that CRL's performance advantage remains robust across different weighting strategies and hyperparameter settings. In addition, we demonstrate a lightweight LLM-based deployment: a fixed policy is embedded into a system prompt that maps inferred states to actions, producing consistent, supportive dialogue without LLM finetuning. Our work illustrates the promise of causal reinforcement learning for human-robot interaction applications, where interpretability, adaptiveness, and data efficiency are paramount.", "date": "2025-11-17", "categories": []}
{"id": "2512.00836", "title": "Assessing model error in counterfactual worlds", "abstract": "Counterfactual scenario modeling exercises that ask \"what would happen if?\" are one of the most common ways we plan for the future. Despite their ubiquity in planning and decision making, scenario projections are rarely evaluated retrospectively. Differences between projections and observations come from two sources: scenario deviation and model miscalibration. We argue the latter is most important for assessing the value of models in decision making, but requires estimating model error in counterfactual worlds. Here we present and contrast three approaches for estimating this error, and demonstrate the benefits and limitations of each in a simulation experiment. We provide recommendations for the estimation of counterfactual error and discuss the components of scenario design that are required to make scenario projections evaluable.", "date": "2025-11-30", "categories": []}
{"id": "2512.00839", "title": "ARCADIA: Scalable Causal Discovery for Corporate Bankruptcy Analysis Using Agentic AI", "abstract": "This paper introduces ARCADIA, an agentic AI framework for causal discovery that integrates large-language-model reasoning with statistical diagnostics to construct valid, temporally coherent causal structures. Unlike traditional algorithms, ARCADIA iteratively refines candidate DAGs through constraint-guided prompting and causal-validity feedback, leading to stable and interpretable models for real-world high-stakes domains. Experiments on corporate bankruptcy data show that ARCADIA produces more reliable causal graphs than NOTEARS, GOLEM, and DirectLiNGAM while offering a fully explainable, intervention-ready pipeline. The framework advances AI by demonstrating how agentic LLMs can participate in autonomous scientific modeling and structured causal inference.", "date": "2025-11-30", "categories": []}
{"id": "2512.01827", "title": "CauSight: Learning to Supersense for Visual Causal Discovery", "abstract": "Causal thinking enables humans to understand not just what is seen, but why it happens. To replicate this capability in modern AI systems, we introduce the task of visual causal discovery. It requires models to infer cause-and-effect relations among visual entities across diverse scenarios instead of merely perceiving their presence. To this end, we first construct the Visual Causal Graph dataset (VCG-32K), a large-scale collection of over 32,000 images annotated with entity-level causal graphs, and further develop CauSight, a novel vision-language model to perform visual causal discovery through causally aware reasoning. Our training recipe integrates three components: (1) training data curation from VCG-32K, (2) Tree-of-Causal-Thought (ToCT) for synthesizing reasoning trajectories, and (3) reinforcement learning with a designed causal reward to refine the reasoning policy. Experiments show that CauSight outperforms GPT-4.1 on visual causal discovery, achieving over a threefold performance boost (21% absolute gain). Our code, model, and dataset are fully open-sourced at project page: https://github.com/OpenCausaLab/CauSight.", "date": "2025-12-01", "categories": []}
{"id": "2508.04182", "title": "COPO: Causal-Oriented Policy Optimization for Hallucinations of MLLMs", "abstract": "Despite Multimodal Large Language Models (MLLMs) having shown impressive capabilities, they may suffer from hallucinations. Empirically, we find that MLLMs attend disproportionately to task-irrelevant background regions compared with text-only LLMs, implying spurious background-answer correlations. We claim and analyze that (i) outcome-based rewards can be an important factor leading to spurious correlations, and (ii) spurious correlations can be an important factor leading to hallucinations. Based on these results, we propose Causal-Oriented Policy Optimization (COPO) to mitigate these spurious correlations, thus addressing the issue of hallucinations. It imposes token-level sufficiency and necessity constraints to measure each inference token's causal contribution, thus ensuring correct and evidence-grounded output. Specifically, we first evaluate each token's causal contribution via a newly proposed causal completeness reward. This reward is then used to construct a causally informed advantage function within the GRPO optimization framework, encouraging the model to focus on tokens that are causally sufficient and necessary for accurate generation. Experimental results across various benchmarks demonstrate the advantages of COPO.", "date": "2025-08-06", "categories": []}
{"id": "2511.20558", "title": "Spatio-Temporal Hierarchical Causal Models", "abstract": "The abundance of fine-grained spatio-temporal data, such as traffic sensor networks, offers vast opportunities for scientific discovery. However, inferring causal relationships from such observational data remains challenging, particularly due to unobserved confounders that are specific to units (e.g., geographical locations) yet influence outcomes over time. Most existing methods for spatio-temporal causal inference assume that all confounders are observed, an assumption that is often violated in practice. In this paper, we introduce Spatio-Temporal Hierarchical Causal Models (ST-HCMs), a novel graphical framework that extends hierarchical causal modeling to the spatio-temporal domain. At the core of our approach is the Spatio-Temporal Collapse Theorem, which shows that a complex ST-HCM converges to a simpler flat causal model as the amount of subunit data increases. This theoretical result enables a general procedure for causal identification, allowing ST-HCMs to recover causal effects even in the presence of unobserved, time-invariant unit-level confounders, a scenario where standard non-hierarchical models fail. We validate the effectiveness of our framework on both synthetic and real-world datasets, demonstrating its potential for robust causal inference in complex dynamic systems.", "date": "2025-11-25", "categories": []}
{"id": "2506.10899", "title": "Demystifying Spectral Feature Learning for Instrumental Variable Regression", "abstract": "We address the problem of causal effect estimation in the presence of hidden confounders, using nonparametric instrumental variable (IV) regression. A leading strategy employs spectral features - that is, learned features spanning the top eigensubspaces of the operator linking treatments to instruments. We derive a generalization error bound for a two-stage least squares estimator based on spectral features, and gain insights into the method's performance and failure modes. We show that performance depends on two key factors, leading to a clear taxonomy of outcomes. In a good scenario, the approach is optimal. This occurs with strong spectral alignment, meaning the structural function is well-represented by the top eigenfunctions of the conditional operator, coupled with this operator's slow eigenvalue decay, indicating a strong instrument. Performance degrades in a bad scenario: spectral alignment remains strong, but rapid eigenvalue decay (indicating a weaker instrument) demands significantly more samples for effective feature learning. Finally, in the ugly scenario, weak spectral alignment causes the method to fail, regardless of the eigenvalues' characteristics. Our synthetic experiments empirically validate this taxonomy. We further introduce a practical procedure to estimate these spectral properties from data, allowing practitioners to diagnose which regime a given problem falls into. We apply this method to the dSprites dataset, demonstrating its utility.", "date": "2025-06-12", "categories": []}
{"id": "2511.21636", "title": "Bridging the Unavoidable A Priori: A Framework for Comparative Causal Modeling", "abstract": "AI/ML models have rapidly gained prominence as innovations for solving previously unsolved problems and their unintended consequences from amplifying human biases. Advocates for responsible AI/ML have sought ways to draw on the richer causal models of system dynamics to better inform the development of responsible AI/ML. However, a major barrier to advancing this work is the difficulty of bringing together methods rooted in different underlying assumptions (i.e., Dana Meadow's \"the unavoidable a priori\"). This paper brings system dynamics and structural equation modeling together into a common mathematical framework that can be used to generate systems from distributions, develop methods, and compare results to inform the underlying epistemology of system dynamics for data science and AI/ML applications.", "date": "2025-11-26", "categories": []}
{"id": "2511.21675", "title": "On Evolution-Based Models for Experimentation Under Interference", "abstract": "Causal effect estimation in networked systems is central to data-driven decision making. In such settings, interventions on one unit can spill over to others, and in complex physical or social systems, the interaction pathways driving these interference structures remain largely unobserved. We argue that for identifying population-level causal effects, it is not necessary to recover the exact network structure; instead, it suffices to characterize how those interactions contribute to the evolution of outcomes. Building on this principle, we study an evolution-based approach that investigates how outcomes change across observation rounds in response to interventions, hence compensating for missing network information. Using an exposure-mapping perspective, we give an axiomatic characterization of when the empirical distribution of outcomes follows a low-dimensional recursive equation, and identify minimal structural conditions under which such evolution mappings exist. We frame this as a distributional counterpart to difference-in-differences. Rather than assuming parallel paths for individual units, it exploits parallel evolution patterns across treatment scenarios to estimate counterfactual trajectories. A key insight is that treatment randomization plays a role beyond eliminating latent confounding; it induces an implicit sampling from hidden interference channels, enabling consistent learning about heterogeneous spillover effects. We highlight causal message passing as an instantiation of this method in dense networks while extending to more general interference structures, including influencer networks where a small set of units drives most spillovers. Finally, we discuss the limits of this approach, showing that strong temporal trends or endogenous interference can undermine identification.", "date": "2025-11-26", "categories": []}
{"id": "2307.01449", "title": "A Double Machine Learning Approach to Combining Experimental and Observational Data", "abstract": "Experimental and observational studies often lack validity due to untestable assumptions. We propose a double machine learning approach to combine experimental and observational studies, allowing practitioners to test for assumption violations and estimate treatment effects consistently. Our framework proposes a falsification test for external validity and ignorability under milder assumptions. We provide consistent treatment effect estimators even when one of the assumptions is violated. However, our no-free-lunch theorem highlights the necessity of accurately identifying the violated assumption for consistent treatment effect estimation. Through comparative analyses, we show our framework's superiority over existing data fusion methods. The practical utility of our approach is further exemplified by three real-world case studies, underscoring its potential for widespread application in empirical research.", "date": "2023-07-04", "categories": []}
{"id": "2511.20189", "title": "Learning Subgroups with Maximum Treatment Effects without Causal Heuristics", "abstract": "Discovering subgroups with the maximum average treatment effect is crucial for targeted decision making in domains such as precision medicine, public policy, and education. While most prior work is formulated in the potential outcome framework, the corresponding structural causal model (SCM) for this task has been largely overlooked. In practice, two approaches dominate. The first estimates pointwise conditional treatment effects and then fits a tree on those estimates, effectively turning subgroup estimation into the harder problem of accurate pointwise estimation. The second constructs decision trees or rule sets with ad-hoc 'causal' heuristics, typically without rigorous justification for why a given heuristic may be used or whether such heuristics are necessary at all. We address these issues by studying the problem directly under the SCM framework. Under the assumption of a partition-based model, we show that optimal subgroup discovery reduces to recovering the data-generating models and hence a standard supervised learning problem (regression or classification). This allows us to adopt any partition-based methods to learn the subgroup from data. We instantiate the approach with CART, arguably one of the most widely used tree-based methods, to learn the subgroup with maximum treatment effect. Finally, on a large collection of synthetic and semi-synthetic datasets, we compare our method against a wide range of baselines and find that our approach, which avoids such causal heuristics, more accurately identifies subgroups with maximum treatment effect. Our source code is available at https://github.com/ylincen/causal-subgroup.", "date": "2025-11-25", "categories": []}
{"id": "2312.15524", "title": "The Challenge of Using LLMs to Simulate Human Behavior: A Causal Inference Perspective", "abstract": "Large Language Models (LLMs) have shown impressive potential to simulate human behavior. We identify a fundamental challenge in using them to simulate experiments: when LLM-simulated subjects are blind to the experimental design (as is standard practice with human subjects), variations in treatment systematically affect unspecified variables that should remain constant, violating the unconfoundedness assumption. Using demand estimation as a context and an actual experiment with 40 different products as a benchmark, we show this can lead to implausible results. While confounding may in principle be addressed by controlling for covariates, this can compromise ecological validity in the context of LLM simulations: controlled covariates become artificially salient in the simulated decision process. We show formally that confoundness stems from ambiguous prompting strategies. Therefore, it can be addressed by developing unambiguous prompting strategies through unblinding, i.e., revealing the experiment design in LLM simulations. Our empirical results show that this strategy consistently enhances model performance across all tested models, including both out-of-box reasoning and non-reasoning models. We also show that it is a technique that complements fine-tuning: while fine-tuning can improve simulation performance, an unambiguous prompting strategy makes the predictions robust to the inclusion of irrelevant data in the fine-tuning process.", "date": "2023-12-24", "categories": []}
{"id": "2511.17505", "title": "Causal Intervention Sequence Analysis for Fault Tracking in Radio Access Networks", "abstract": "To keep modern Radio Access Networks (RAN) running smoothly, operators need to spot the real-world triggers behind Service-Level Agreement (SLA) breaches well before customers feel them. We introduce an AI/ML pipeline that does two things most tools miss: (1) finds the likely root-cause indicators and (2) reveals the exact order in which those events unfold. We start by labeling network data: records linked to past SLA breaches are marked `abnormal', and everything else `normal'. Our model then learns the causal chain that turns normal behavior into a fault. In Monte Carlo tests the approach pinpoints the correct trigger sequence with high precision and scales to millions of data points without loss of speed. These results show that high-resolution, causally ordered insights can move fault management from reactive troubleshooting to proactive prevention.", "date": "2025-10-02", "categories": []}
{"id": "2511.17953", "title": "On Transportability for Structural Causal Bandits", "abstract": "Intelligent agents equipped with causal knowledge can optimize their action spaces to avoid unnecessary exploration. The structural causal bandit framework provides a graphical characterization for identifying actions that are unable to maximize rewards by leveraging prior knowledge of the underlying causal structure. While such knowledge enables an agent to estimate the expected rewards of certain actions based on others in online interactions, there has been little guidance on how to transfer information inferred from arbitrary combinations of datasets collected under different conditions -- observational or experimental -- and from heterogeneous environments. In this paper, we investigate the structural causal bandit with transportability, where priors from the source environments are fused to enhance learning in the deployment setting. We demonstrate that it is possible to exploit invariances across environments to consistently improve learning. The resulting bandit algorithm achieves a sub-linear regret bound with an explicit dependence on informativeness of prior data, and it may outperform standard bandit approaches that rely solely on online learning.", "date": "2025-11-22", "categories": []}
{"id": "2511.18464", "title": "Reliable Selection of Heterogeneous Treatment Effect Estimators", "abstract": "We study the problem of selecting the best heterogeneous treatment effect (HTE) estimator from a collection of candidates in settings where the treatment effect is fundamentally unobserved. We cast estimator selection as a multiple testing problem and introduce a ground-truth-free procedure based on a cross-fitted, exponentially weighted test statistic. A key component of our method is a two-way sample splitting scheme that decouples nuisance estimation from weight learning and ensures the stability required for valid inference. Leveraging a stability-based central limit theorem, we establish asymptotic familywise error rate control under mild regularity conditions. Empirically, our procedure provides reliable error control while substantially reducing false selections compared with commonly used methods across ACIC 2016, IHDP, and Twins benchmarks, demonstrating that our method is feasible and powerful even without ground-truth treatment effects.", "date": "2025-11-23", "categories": []}
{"id": "2511.19273", "title": "Scalable Bayesian Network Structure Learning Using Tsetlin Machine to Constrain the Search Space", "abstract": "The PC algorithm is a widely used method in causal inference for learning the structure of Bayesian networks. Despite its popularity, the PC algorithm suffers from significant time complexity, particularly as the size of the dataset increases, which limits its applicability in large-scale real-world problems. In this study, we propose a novel approach that utilises the Tsetlin Machine (TM) to construct Bayesian structures more efficiently. Our method leverages the most significant literals extracted from the TM and performs conditional independence (CI) tests on these selected literals instead of the full set of variables, resulting in a considerable reduction in computational time. We implemented our approach and compared it with various state-of-the-art methods. Our evaluation includes categorical datasets from the bnlearn repository, such as Munin1, Hepar2. The findings indicate that the proposed TM-based method not only reduces computational complexity but also maintains competitive accuracy in causal discovery, making it a viable alternative to traditional PC algorithm implementations by offering improved efficiency without compromising performance.", "date": "2025-11-24", "categories": []}
{"id": "2511.19284", "title": "The Unified Non-Convex Framework for Robust Causal Inference: Overcoming the Gaussian Barrier and Optimization Fragility", "abstract": "This document proposes a Unified Robust Framework that re-engineers the estimation of the Average Treatment Effect on the Overlap (ATO). It synthesizes gamma-Divergence for outlier robustness, Graduated Non-Convexity (GNC) for global optimization, and a \"Gatekeeper\" mechanism to address the impossibility of higher-order orthogonality in Gaussian regimes.", "date": "2025-11-24", "categories": []}
{"id": "2511.19404", "title": "Nonparametric Instrumental Variable Regression with Observed Covariates", "abstract": "We study the problem of nonparametric instrumental variable regression with observed covariates, which we refer to as NPIV-O. Compared with standard nonparametric instrumental variable regression (NPIV), the additional observed covariates facilitate causal identification and enables heterogeneous causal effect estimation. However, the presence of observed covariates introduces two challenges for its theoretical analysis. First, it induces a partial identity structure, which renders previous NPIV analyses - based on measures of ill-posedness, stability conditions, or link conditions - inapplicable. Second, it imposes anisotropic smoothness on the structural function. To address the first challenge, we introduce a novel Fourier measure of partial smoothing; for the second challenge, we extend the existing kernel 2SLS instrumental variable algorithm with observed covariates, termed KIV-O, to incorporate Gaussian kernel lengthscales adaptive to the anisotropic smoothness. We prove upper $L^2$-learning rates for KIV-O and the first $L^2$-minimax lower learning rates for NPIV-O. Both rates interpolate between known optimal rates of NPIV and nonparametric regression (NPR). Interestingly, we identify a gap between our upper and lower bounds, which arises from the choice of kernel lengthscales tuned to minimize a projected risk. Our theoretical analysis also applies to proximal causal inference, an emerging framework for causal effect estimation that shares the same conditional moment restriction as NPIV-O.", "date": "2025-11-24", "categories": []}
{"id": "2310.06746", "title": "A New Causal Rule Learning Approach to Interpretable Estimation of Heterogeneous Treatment Effect", "abstract": "Interpretability plays a crucial role in the application of statistical learning to estimate heterogeneous treatment effects (HTE) in complex diseases. In this study, we leverage a rule-based workflow, namely causal rule learning (CRL), to estimate and improve our understanding of HTE for atrial septal defect, addressing an overlooked question in the previous literature: what if an individual simultaneously belongs to multiple groups with different average treatment effects? The CRL process consists of three steps: rule discovery, which generates a set of causal rules with corresponding subgroup average treatment effects; rule selection, which identifies a subset of these rules to deconstruct individual-level treatment effects as a linear combination of subgroup-level effects; and rule analysis, which presents a detailed procedure for further analyzing each selected rule from multiple perspectives to identify the most promising rules for validation. Extensive simulation studies and real-world data analysis demonstrate that CRL outperforms other methods in providing interpretable estimates of HTE, especially when dealing with complex ground truth and sufficient sample sizes.", "date": "2023-10-10", "categories": []}
{"id": "2511.16204", "title": "Causal Synthetic Data Generation in Recruitment", "abstract": "The importance of Synthetic Data Generation (SDG) has increased significantly in domains where data quality is poor or access is limited due to privacy and regulatory constraints. One such domain is recruitment, where publicly available datasets are scarce due to the sensitive nature of information typically found in curricula vitae, such as gender, disability status, or age.\nThis lack of accessible, representative data presents a significant obstacle to the development of fair and transparent machine learning models, particularly ranking algorithms that require large volumes of data to effectively learn how to recommend candidates. In the absence of such data, these models are prone to poor generalisation and may fail to perform reliably in real-world scenarios.\nRecent advances in Causal Generative Models (CGMs) offer a promising solution. CGMs enable the generation of synthetic datasets that preserve the underlying causal relationships within the data, providing greater control over fairness and interpretability in the data generation process.\nIn this study, we present a specialised SDG method involving two CGMs: one modelling job offers and the other modelling curricula. Each model is structured according to a causal graph informed by domain expertise. We use these models to generate synthetic datasets and evaluate the fairness of candidate rankings under controlled scenarios that introduce specific biases.", "date": "2025-11-20", "categories": []}
{"id": "2401.06925", "title": "Foundations of Structural Causal Models with Latent Selection", "abstract": "Three distinct phenomena complicate statistical causal analysis: latent common causes, causal cycles, and latent selection. Foundational works on Structural Causal Models (SCMs), e.g., Bongers et al. (2021, Ann. Stat., 49(5): 2885-2915), treat cycles and latent variables, while an analogous account of latent selection is missing. The goal of this article is to develop a theoretical foundation for modeling latent selection with SCMs. To achieve that, we introduce a conditioning operation for SCMs: it maps an SCM with explicit selection mechanisms to one without them while preserving the causal semantics of the selected subpopulation. Graphically, in Directed Mixed Graphs we extend bidirected edge--beyond latent common cause--to also encode latent selection. We prove that the conditioning operation preserves simplicity, acyclicity, and linearity of SCMs, and interacts well with marginalization, conditioning, and interventions. These properties make those three operations valuable tools for causal modeling, reasoning, and learning after abstracting away latent details (latent common causes and selection). Examples show how this abstraction streamlines analysis and clarifies when standard tools (e.g., adjustment, causal calculus, instrumental variables) remain valid under selection bias. We hope that these results deepen the SCM-based understanding of selection bias and become part of the standard causal modeling toolbox to build more reliable causal analysis.", "date": "2024-01-12", "categories": []}
{"id": "2505.17717", "title": "A Distributionally Robust Framework for Nuisance in Causal Effect Estimation", "abstract": "Causal inference requires evaluating models on balanced distributions between treatment and control groups, while training data often exhibits imbalance due to historical decision-making policies. Most conventional statistical methods address this distribution shift through inverse probability weighting (IPW), which requires estimating propensity scores as an intermediate step. These methods face two key challenges: inaccurate propensity estimation and instability from extreme weights. We decompose the generalization error to isolate these issues--propensity ambiguity and statistical instability--and address them through an adversarial loss function. Our approach combines distributionally robust optimization for handling propensity uncertainty with weight regularization based on weighted Rademacher complexity. Experiments on synthetic and real-world datasets demonstrate consistent improvements over existing methods.", "date": "2025-05-23", "categories": []}
{"id": "2506.06039", "title": "Do-PFN: In-Context Learning for Causal Effect Estimation", "abstract": "Estimation of causal effects is critical to a range of scientific disciplines. Existing methods for this task either require interventional data, knowledge about the ground truth causal graph, or rely on assumptions such as unconfoundedness, restricting their applicability in real-world settings. In the domain of tabular machine learning, Prior-data fitted networks (PFNs) have achieved state-of-the-art predictive performance, having been pre-trained on synthetic data to solve tabular prediction problems via in-context learning. To assess whether this can be transferred to the harder problem of causal effect estimation, we pre-train PFNs on synthetic data drawn from a wide variety of causal structures, including interventions, to predict interventional outcomes given observational data. Through extensive experiments on synthetic case studies, we show that our approach allows for the accurate estimation of causal effects without knowledge of the underlying causal graph. We also perform ablation studies that elucidate Do-PFN's scalability and robustness across datasets with a variety of causal characteristics.", "date": "2025-06-06", "categories": []}
{"id": "2511.15716", "title": "MACIE: Multi-Agent Causal Intelligence Explainer for Collective Behavior Understanding", "abstract": "As Multi Agent Reinforcement Learning systems are used in safety critical applications. Understanding why agents make decisions and how they achieve collective behavior is crucial. Existing explainable AI methods struggle in multi agent settings. They fail to attribute collective outcomes to individuals, quantify emergent behaviors, or capture complex interactions. We present MACIE Multi Agent Causal Intelligence Explainer, a framework combining structural causal models, interventional counterfactuals, and Shapley values to provide comprehensive explanations. MACIE addresses three questions. First, each agent's causal contribution using interventional attribution scores. Second, system level emergent intelligence through synergy metrics separating collective effects from individual contributions. Third, actionable explanations using natural language narratives synthesizing causal insights. We evaluate MACIE across four MARL scenarios: cooperative, competitive, and mixed motive. Results show accurate outcome attribution, mean phi_i equals 5.07, standard deviation less than 0.05, detection of positive emergence in cooperative tasks, synergy index up to 0.461, and efficient computation, 0.79 seconds per dataset on CPU. MACIE uniquely combines causal rigor, emergence quantification, and multi agent support while remaining practical for real time use. This represents a step toward interpretable, trustworthy, and accountable multi agent AI.", "date": "2025-11-11", "categories": []}
{"id": "2511.16006", "title": "Synergizing Deconfounding and Temporal Generalization For Time-series Counterfactual Outcome Estimation", "abstract": "Estimating counterfactual outcomes from time-series observations is crucial for effective decision-making, e.g. when to administer a life-saving treatment, yet remains significantly challenging because (i) the counterfactual trajectory is never observed and (ii) confounders evolve with time and distort estimation at every step. To address these challenges, we propose a novel framework that synergistically integrates two complementary approaches: Sub-treatment Group Alignment (SGA) and Random Temporal Masking (RTM). Instead of the coarse practice of aligning marginal distributions of the treatments in latent space, SGA uses iterative treatment-agnostic clustering to identify fine-grained sub-treatment groups. Aligning these fine-grained groups achieves improved distributional matching, thus leading to more effective deconfounding. We theoretically demonstrate that SGA optimizes a tighter upper bound on counterfactual risk and empirically verify its deconfounding efficacy. RTM promotes temporal generalization by randomly replacing input covariates with Gaussian noises during training. This encourages the model to rely less on potentially noisy or spuriously correlated covariates at the current step and more on stable historical patterns, thereby improving its ability to generalize across time and better preserve underlying causal relationships. Our experiments demonstrate that while applying SGA and RTM individually improves counterfactual outcome estimation, their synergistic combination consistently achieves state-of-the-art performance. This success comes from their distinct yet complementary roles: RTM enhances temporal generalization and robustness across time steps, while SGA improves deconfounding at each specific time point.", "date": "2025-11-20", "categories": []}
{"id": "2511.16016", "title": "CARE: Turning LLMs Into Causal Reasoning Expert", "abstract": "Large language models (LLMs) have recently demonstrated impressive capabilities across a range of reasoning and generation tasks. However, research studies have shown that LLMs lack the ability to identify causal relationships, a fundamental cornerstone of human intelligence. We first conduct an exploratory investigation of LLMs' behavior when asked to perform a causal-discovery task and find that they mostly rely on the semantic meaning of variable names, ignoring the observation data. This is unsurprising, given that LLMs were never trained to process structural datasets. To first tackle this challenge, we prompt the LLMs with the outputs of established causal discovery algorithms designed for observational datasets. These algorithm outputs effectively serve as the sufficient statistics of the observation data. However, quite surprisingly, we find that prompting the LLMs with these sufficient statistics decreases the LLMs' performance in causal discovery. To address this current limitation, we propose CARE, a framework that enhances LLMs' causal-reasoning ability by teaching them to effectively utilize the outputs of established causal-discovery algorithms through supervised fine-tuning. Experimental results show that a finetuned Qwen2.5-1.5B model produced by CARE significantly outperforms both traditional causal-discovery algorithms and state-of-the-art LLMs with over a thousand times more parameters, demonstrating effective utilization of its own knowledge and the external algorithmic clues.", "date": "2025-11-20", "categories": []}
{"id": "2405.15871", "title": "Explaining Time Series Classification Predictions via Causal Attributions", "abstract": "Despite the excelling performance of machine learning models, understanding their decisions remains a long-standing goal. Although commonly used attribution methods from explainable AI attempt to address this issue, they typically rely on associational rather than causal relationships. In this study, within the context of time series classification, we introduce a novel model-agnostic attribution method to assess the causal effect of concepts i.e., predefined segments within a time series, on classification outcomes. Our approach compares these causal attributions with closely related associational attributions, both theoretically and empirically. To estimate counterfactual outcomes, we use state-of-the-art diffusion models backed by state space models. We demonstrate the insights gained by our approach for a diverse set of qualitatively different time series classification tasks. Although causal and associational attributions might often share some similarities, in all cases they differ in important details, underscoring the risks associated with drawing causal conclusions from associational data alone. We believe that the proposed approach is also widely applicable in other domains to shed some light on the limits of associational attributions.", "date": "2024-05-24", "categories": []}
{"id": "2406.15229", "title": "ExDAG: an MIQP Algorithm for Learning DAGs", "abstract": "There has been a growing interest in causal learning in recent years. Commonly used representations of causal structures, including Bayesian networks and structural equation models (SEM), take the form of directed acyclic graphs (DAGs). We provide a novel mixed-integer quadratic programming formulation and an associated algorithm that identifies DAGs with a low structural Hamming distance between the identified DAG and the ground truth, under identifiability assumptions. The eventual exact learning is guaranteed by the global convergence of the branch-and-bound-and-cut algorithm, which is utilized. In addition to this, integer programming techniques give us access to the dual bound, which allows for a real time assessment of the quality of solution. Previously, integer programming techniques have been shown to lead to limited scaling in the case of DAG identification due to the super exponential number of constraints, which prevent the formation of cycles. The algorithm proposed circumvents this by selectively generating only the violated constraints using the so-called \"lazy\" constraints methodology. Our empirical results show that ExDAG outperforms state-of-the-art solvers in terms of structural Hamming distance and $F_1$ score when considering Gaussian noise on medium-sized graphs.", "date": "2024-06-21", "categories": []}
{"id": "2506.20582", "title": "Causal Representation Learning with Observational Grouping for CXR Classification", "abstract": "Identifiable causal representation learning seeks to uncover the true causal relationships underlying a data generation process. In medical imaging, this presents opportunities to improve the generalisability and robustness of task-specific latent features. This work introduces the concept of grouping observations to learn identifiable representations for disease classification in chest X-rays via an end-to-end framework. Our experiments demonstrate that these causal representations improve generalisability and robustness across multiple classification tasks when grouping is used to enforce invariance w.r.t race, sex, and imaging views.", "date": "2025-06-25", "categories": []}
{"id": "2511.14922", "title": "Integrating Causal Inference with Graph Neural Networks for Alzheimer's Disease Analysis", "abstract": "Deep graph learning has advanced Alzheimer's (AD) disease classification from MRI, but most models remain correlational, confounding demographic and genetic factors with disease specific features. We present Causal-GCN, an interventional graph convolutional framework that integrates do-calculus-based back-door adjustment to identify brain regions exerting stable causal influence on AD progression. Each subject's MRI is represented as a structural connectome where nodes denote cortical and subcortical regions and edges encode anatomical connectivity. Confounders such as age, sec, and APOE4 genotype are summarized via principal components and included in the causal adjustment set. After training, interventions on individual regions are simulated by serving their incoming edges and altering node features to estimate average causal effects on disease probability. Applied to 484 subjects from the ADNI cohort, Causal-GCN achieves performance comparable to baseline GNNs while providing interpretable causal effect rankings that highlight posterior, cingulate, and insular hubs consistent with established AD neuropathology.", "date": "2025-11-18", "categories": []}
{"id": "2511.13852", "title": "Causal computations in Semi Markovian Structural Causal Models using divide and conquer", "abstract": "Recently, Bj{\\o}ru et al. proposed a novel divide-and-conquer algorithm for bounding counterfactual probabilities in structural causal models (SCMs). They assumed that the SCMs were learned from purely observational data, leading to an imprecise characterization of the marginal distributions of exogenous variables. Their method leveraged the canonical representation of structural equations to decompose a general SCM with high-cardinality exogenous variables into a set of sub-models with low-cardinality exogenous variables. These sub-models had precise marginals over the exogenous variables and therefore admitted efficient exact inference. The aggregated results were used to bound counterfactual probabilities in the original model. The approach was developed for Markovian models, where each exogenous variable affects only a single endogenous variable. In this paper, we investigate extending the methodology to \\textit{semi-Markovian} SCMs, where exogenous variables may influence multiple endogenous variables. Such models are capable of representing confounding relationships that Markovian models cannot. We illustrate the challenges of this extension using a minimal example, which motivates a set of alternative solution strategies. These strategies are evaluated both theoretically and through a computational study.", "date": "2025-11-17", "categories": []}
{"id": "2511.14001", "title": "How to Marginalize in Causal Structure Learning?", "abstract": "Bayesian networks (BNs) are a widely used class of probabilistic graphical models employed in numerous application domains. However, inferring the network's graphical structure from data remains challenging. Bayesian structure learners approach this problem by inferring a posterior distribution over the possible directed acyclic graphs underlying the BN. The inference process often requires marginalizing over probability distributions, which is typically done using dynamic programming methods that restrict the set of possible parents for each node. Instead, we present a novel method that utilizes tractable probabilistic circuits to circumvent this restriction. This method utilizes a new learning routine that trains these circuits on both the original distribution and marginal queries. The architecture of probabilistic circuits then inherently allows for fast and exact marginalization on the learned distribution. We then show empirically that utilizing our method to answer marginals allows Bayesian structure learners to improve their performance compared to current methods.", "date": "2025-11-18", "categories": []}
{"id": "2511.14133", "title": "Synthetic Survival Control: Extending Synthetic Controls for \"When-If\" Decision", "abstract": "Estimating causal effects on time-to-event outcomes from observational data is particularly challenging due to censoring, limited sample sizes, and non-random treatment assignment. The need for answering such \"when-if\" questions--how the timing of an event would change under a specified intervention--commonly arises in real-world settings with heterogeneous treatment adoption and confounding. To address these challenges, we propose Synthetic Survival Control (SSC) to estimate counterfactual hazard trajectories in a panel data setting where multiple units experience potentially different treatments over multiple periods. In such a setting, SSC estimates the counterfactual hazard trajectory for a unit of interest as a weighted combination of the observed trajectories from other units. To provide formal justification, we introduce a panel framework with a low-rank structure for causal survival analysis. Indeed, such a structure naturally arises under classical parametric survival models. Within this framework, for the causal estimand of interest, we establish identification and finite sample guarantees for SSC. We validate our approach using a multi-country clinical dataset of cancer treatment outcomes, where the staggered introduction of new therapies creates a quasi-experimental setting. Empirically, we find that access to novel treatments is associated with improved survival, as reflected by lower post-intervention hazard trajectories relative to their synthetic counterparts. Given the broad relevance of survival analysis across medicine, economics, and public policy, our framework offers a general and interpretable tool for counterfactual survival inference using observational data.", "date": "2025-11-18", "categories": []}
{"id": "2511.14206", "title": "Causal Discovery on Higher-Order Interactions", "abstract": "Causal discovery combines data with knowledge provided by experts to learn the DAG representing the causal relationships between a given set of variables. When data are scarce, bagging is used to measure our confidence in an average DAG obtained by aggregating bootstrapped DAGs. However, the aggregation step has received little attention from the specialized literature: the average DAG is constructed using only the confidence in the individual edges of the bootstrapped DAGs, thus disregarding complex higher-order edge structures. In this paper, we introduce a novel theoretical framework based on higher-order structures and describe a new DAG aggregation algorithm. We perform a simulation study, discussing the advantages and limitations of the proposed approach. Our proposal is both computationally efficient and effective, outperforming state-of-the-art solutions, especially in low sample size regimes and under high dimensionality settings.", "date": "2025-11-18", "categories": []}
{"id": "2511.14441", "title": "Skewness-Robust Causal Discovery in Location-Scale Noise Models", "abstract": "To distinguish Markov equivalent graphs in causal discovery, it is necessary to restrict the structural causal model. Crucially, we need to be able to distinguish cause $X$ from effect $Y$ in bivariate models, that is, distinguish the two graphs $X \\to Y$ and $Y \\to X$. Location-scale noise models (LSNMs), in which the effect $Y$ is modeled based on the cause $X$ as $Y = f(X) + g(X)N$, form a flexible class of models that is general and identifiable in most cases. Estimating these models for arbitrary noise terms $N$, however, is challenging. Therefore, practical estimators are typically restricted to symmetric distributions, such as the normal distribution. As we showcase in this paper, when $N$ is a skewed random variable, which is likely in real-world domains, the reliability of these approaches decreases. To approach this limitation, we propose SkewD, a likelihood-based algorithm for bivariate causal discovery under LSNMs with skewed noise distributions. SkewD extends the usual normal-distribution framework to the skew-normal setting, enabling reliable inference under symmetric and skewed noise. For parameter estimation, we employ a combination of a heuristic search and an expectation conditional maximization algorithm. We evaluate SkewD on novel synthetically generated datasets with skewed noise as well as established benchmark datasets. Throughout our experiments, SkewD exhibits a strong performance and, in comparison to prior work, remains robust under high skewness.", "date": "2025-11-18", "categories": []}
{"id": "2511.14545", "title": "DeepBlip: Estimating Conditional Average Treatment Effects Over Time", "abstract": "Structural nested mean models (SNMMs) are a principled approach to estimate the treatment effects over time. A particular strength of SNMMs is to break the joint effect of treatment sequences over time into localized, time-specific ``blip effects''. This decomposition promotes interpretability through the incremental effects and enables the efficient offline evaluation of optimal treatment policies without re-computation. However, neural frameworks for SNMMs are lacking, as their inherently sequential g-estimation scheme prevents end-to-end, gradient-based training. Here, we propose DeepBlip, the first neural framework for SNMMs, which overcomes this limitation with a novel double optimization trick to enable simultaneous learning of all blip functions. Our DeepBlip seamlessly integrates sequential neural networks like LSTMs or transformers to capture complex temporal dependencies. By design, our method correctly adjusts for time-varying confounding to produce unbiased estimates, and its Neyman-orthogonal loss function ensures robustness to nuisance model misspecification. Finally, we evaluate our DeepBlip across various clinical datasets, where it achieves state-of-the-art performance.", "date": "2025-11-18", "categories": []}
{"id": "2405.04715", "title": "Causality Pursuit from Heterogeneous Environments via Neural Adversarial Invariance Learning", "abstract": "Pursuing causality from data is a fundamental problem in scientific discovery, treatment intervention, and transfer learning. This paper introduces a novel algorithmic method for addressing nonparametric invariance and causality learning in regression models across multiple environments, where the joint distribution of response variables and covariates varies, but the conditional expectations of outcome given an unknown set of quasi-causal variables are invariant. The challenge of finding such an unknown set of quasi-causal or invariant variables is compounded by the presence of endogenous variables that have heterogeneous effects across different environments. The proposed Focused Adversarial Invariant Regularization (FAIR) framework utilizes an innovative minimax optimization approach that drives regression models toward prediction-invariant solutions through adversarial testing. Leveraging the representation power of neural networks, FAIR neural networks (FAIR-NN) are introduced for causality pursuit. It is shown that FAIR-NN can find the invariant variables and quasi-causal variables under a minimal identification condition and that the resulting procedure is adaptive to low-dimensional composition structures in a non-asymptotic analysis. Under a structural causal model, variables identified by FAIR-NN represent pragmatic causality and provably align with exact causal mechanisms under conditions of sufficient heterogeneity. Computationally, FAIR-NN employs a novel Gumbel approximation with decreased temperature and a stochastic gradient descent ascent algorithm. The procedures are demonstrated using simulated and real-data examples.", "date": "2024-05-07", "categories": []}
{"id": "2505.17708", "title": "The Third Pillar of Causal Analysis? A Measurement Perspective on Causal Representations", "abstract": "Causal reasoning and discovery, two fundamental tasks of causal analysis, often face challenges in applications due to the complexity, noisiness, and high-dimensionality of real-world data. Despite recent progress in identifying latent causal structures using causal representation learning (CRL), what makes learned representations useful for causal downstream tasks and how to evaluate them are still not well understood. In this paper, we reinterpret CRL using a measurement model framework, where the learned representations are viewed as proxy measurements of the latent causal variables. Our approach clarifies the conditions under which learned representations support downstream causal reasoning and provides a principled basis for quantitatively assessing the quality of representations using a new Test-based Measurement EXclusivity (T-MEX) score. We validate T-MEX across diverse causal inference scenarios, including numerical simulations and real-world ecological video analysis, demonstrating that the proposed framework and corresponding score effectively assess the identification of learned representations and their usefulness for causal downstream tasks.", "date": "2025-05-23", "categories": []}
{"id": "2511.12805", "title": "Practical Causal Evaluation Metrics for Biological Networks", "abstract": "Estimating causal networks from biological data is a critical step in systems biology. When evaluating the inferred network, assessing the networks based on their intervention effects is particularly important for downstream probabilistic reasoning and the identification of potential drug targets. In the context of gene regulatory network inference, biological databases are often used as reference sources. These databases typically describe relationships in a qualitative rather than quantitative manner. However, few evaluation metrics have been developed that take this qualitative nature into account. To address this, we developed a metric, the sign-augmented Structural Intervention Distance (sSID), and a weighted sSID that incorporates the net effects of the intervention. Through simulations and analyses of real transcriptomic datasets, we found that our proposed metrics could identify a different algorithm as optimal compared to conventional metrics, and the network selected by sSID had a superior performance in the classification task of clinical covariates using transcriptomic data. This suggests that sSID can distinguish networks that are structurally correct but functionally incorrect, highlighting its potential as a more biologically meaningful and practical evaluation metric.", "date": "2025-11-16", "categories": []}
{"id": "2511.13295", "title": "Causal Inference, Biomarker Discovery, Graph Neural Network, Feature Selection", "abstract": "Biomarker discovery from high-throughput transcriptomic data is crucial for advancing precision medicine. However, existing methods often neglect gene-gene regulatory relationships and lack stability across datasets, leading to conflation of spurious correlations with genuine causal effects. To address these issues, we develop a causal graph neural network (Causal-GNN) method that integrates causal inference with multi-layer graph neural networks (GNNs). The key innovation is the incorporation of causal effect estimation for identifying stable biomarkers, coupled with a GNN-based propensity scoring mechanism that leverages cross-gene regulatory networks. Experimental results demonstrate that our method achieves consistently high predictive accuracy across four distinct datasets and four independent classifiers. Moreover, it enables the identification of more stable biomarkers compared to traditional methods. Our work provides a robust, efficient, and biologically interpretable tool for biomarker discovery, demonstrating strong potential for broad application across medical disciplines.", "date": "2025-11-17", "categories": []}
{"id": "2511.13389", "title": "Uncovering Causal Drivers of Energy Efficiency for Industrial Process in Foundry via Time-Series Causal Inference", "abstract": "Improving energy efficiency in industrial foundry processes is a critical challenge, as these operations are highly energy-intensive and marked by complex interdependencies among process variables. Correlation-based analyses often fail to distinguish true causal drivers from spurious associations, limiting their usefulness for decision-making. This paper applies a time-series causal inference framework to identify the operational factors that directly affect energy efficiency in induction furnace melting. Using production data from a Danish foundry, the study integrates time-series clustering to segment melting cycles into distinct operational modes with the PCMCI+ algorithm, a state-of-the-art causal discovery method, to uncover cause-effect relationships within each mode. Across clusters, robust causal relations among energy consumption, furnace temperature, and material weight define the core drivers of efficiency, while voltage consistently influences cooling water temperature with a delayed response. Cluster-specific differences further distinguish operational regimes: efficient clusters are characterized by stable causal structures, whereas inefficient ones exhibit reinforcing feedback loops and atypical dependencies. The contributions of this study are twofold. First, it introduces an integrated clustering-causal inference pipeline as a methodological innovation for analyzing energy-intensive processes. Second, it provides actionable insights that enable foundry operators to optimize performance, reduce energy consumption, and lower emissions.", "date": "2025-11-17", "categories": []}
{"id": "2511.10776", "title": "Potential Outcome Rankings for Counterfactual Decision Making", "abstract": "Counterfactual decision-making in the face of uncertainty involves selecting the optimal action from several alternatives using causal reasoning. Decision-makers often rank expected potential outcomes (or their corresponding utility and desirability) to compare the preferences of candidate actions. In this paper, we study new counterfactual decision-making rules by introducing two new metrics: the probabilities of potential outcome ranking (PoR) and the probability of achieving the best potential outcome (PoB). PoR reveals the most probable ranking of potential outcomes for an individual, and PoB indicates the action most likely to yield the top-ranked outcome for an individual. We then establish identification theorems and derive bounds for these metrics, and present estimation methods. Finally, we perform numerical experiments to illustrate the finite-sample properties of the estimators and demonstrate their application to a real-world dataset.", "date": "2025-11-13", "categories": []}
{"id": "2511.11004", "title": "MeCaMIL: Causality-Aware Multiple Instance Learning for Fair and Interpretable Whole Slide Image Diagnosis", "abstract": "Multiple instance learning (MIL) has emerged as the dominant paradigm for whole slide image (WSI) analysis in computational pathology, achieving strong diagnostic performance through patch-level feature aggregation. However, existing MIL methods face critical limitations: (1) they rely on attention mechanisms that lack causal interpretability, and (2) they fail to integrate patient demographics (age, gender, race), leading to fairness concerns across diverse populations. These shortcomings hinder clinical translation, where algorithmic bias can exacerbate health disparities. We introduce \\textbf{MeCaMIL}, a causality-aware MIL framework that explicitly models demographic confounders through structured causal graphs. Unlike prior approaches treating demographics as auxiliary features, MeCaMIL employs principled causal inference -- leveraging do-calculus and collider structures -- to disentangle disease-relevant signals from spurious demographic correlations. Extensive evaluation on three benchmarks demonstrates state-of-the-art performance across CAMELYON16 (ACC/AUC/F1: 0.939/0.983/0.946), TCGA-Lung (0.935/0.979/0.931), and TCGA-Multi (0.977/0.993/0.970, five cancer types). Critically, MeCaMIL achieves superior fairness -- demographic disparity variance drops by over 65% relative reduction on average across attributes, with notable improvements for underserved populations. The framework generalizes to survival prediction (mean C-index: 0.653, +0.017 over best baseline across five cancer types). Ablation studies confirm causal graph structure is essential -- alternative designs yield 0.048 lower accuracy and 4.2x times worse fairness. These results establish MeCaMIL as a principled framework for fair, interpretable, and clinically actionable AI in digital pathology. Code will be released upon acceptance.", "date": "2025-11-14", "categories": []}
{"id": "2511.11564", "title": "Estimating Total Effects in Bipartite Experiments with Spillovers and Partial Eligibility", "abstract": "We study randomized experiments in bipartite systems where only a subset of treatment-side units are eligible for assignment while all units continue to interact, generating interference. We formalize eligibility-constrained bipartite experiments and define estimands aligned with full deployment: the Primary Total Treatment Effect (PTTE) on eligible units and the Secondary Total Treatment Effect (STTE) on ineligible units. Under randomization within the eligible set, we give identification conditions and develop interference-aware ensemble estimators that combine exposure mappings, generalized propensity scores, and flexible machine learning. We further introduce a projection that links treatment- and outcome-level estimands; this mapping is exact under a Linear Additive Edges condition and enables estimation on the (typically much smaller) treatment side with deterministic aggregation to outcomes. In simulations with known ground truth across realistic exposure regimes, the proposed estimators recover PTTE and STTE with low bias and variance and reduce the bias that could arise when interference is ignored. Two field experiments illustrate practical relevance: our method corrects the direction of expected interference bias for a pre-specified metric in both studies and reverses the sign and significance of the primary decision metric in one case.", "date": "2025-11-14", "categories": []}
{"id": "2501.06926", "title": "Semiparametric Double Reinforcement Learning with Applications to Long-Term Causal Inference", "abstract": "Double Reinforcement Learning (DRL) enables efficient inference for policy values in nonparametric Markov decision processes (MDPs), but existing methods face two major obstacles: (1) they require stringent intertemporal overlap conditions on state trajectories, and (2) they rely on estimating high-dimensional occupancy density ratios. Motivated by problems in long-term causal inference, we extend DRL to a semiparametric setting and develop doubly robust, automatic estimators for general linear functionals of the Q-function in infinite-horizon, time-homogeneous MDPs. By imposing structure on the Q-function, we relax the overlap conditions required by nonparametric methods and obtain efficiency gains. The second obstacle--density-ratio estimation--typically requires computationally expensive and unstable min-max optimization. To address both challenges, we introduce superefficient nonparametric estimators whose limiting variance falls below the generalized Cramer-Rao bound. These estimators treat the Q-function as a one-dimensional summary of the state-action process, reducing high-dimensional overlap requirements to a single-dimensional condition. The procedure is simple to implement: estimate and calibrate the Q-function using fitted Q-iteration, then plug the result into the target functional, thereby avoiding density-ratio estimation altogether.", "date": "2025-01-12", "categories": []}
{"id": "2509.23213", "title": "One-Shot Multi-Label Causal Discovery in High-Dimensional Event Sequences", "abstract": "Understanding causality in event sequences with thousands of sparse event types is critical in domains such as healthcare, cybersecurity, or vehicle diagnostics, yet current methods fail to scale. We present OSCAR, a one-shot causal autoregressive method that infers per-sequence Markov Boundaries using two pretrained Transformers as density estimators. This enables efficient, parallel causal discovery without costly global CI testing. On a real-world automotive dataset with 29,100 events and 474 labels, OSCAR recovers interpretable causal structures in minutes, while classical methods fail to scale, enabling practical scientific diagnostics at production scale.", "date": "2025-09-27", "categories": []}
{"id": "2511.09814", "title": "Multiple Treatments Causal Effects Estimation with Task Embeddings and Balanced Representation Learning", "abstract": "The simultaneous application of multiple treatments is increasingly common in many fields, such as healthcare and marketing. In such scenarios, it is important to estimate the single treatment effects and the interaction treatment effects that arise from treatment combinations. Previous studies have proposed using independent outcome networks with subnetworks for interactions, or combining task embedding networks that capture treatment similarity with variational autoencoders. However, these methods suffer from the lack of parameter sharing among related treatments, or the estimation of unnecessary latent variables reduces the accuracy of causal effect estimation. To address these issues, we propose a novel deep learning framework that incorporates a task embedding network and a representation learning network with the balancing penalty. The task embedding network enables parameter sharing across related treatment patterns because it encodes elements common to single effects and contributions specific to interaction effects. The representation learning network with the balancing penalty learns representations nonparametrically from observed covariates while reducing distances in representation distributions across different treatment patterns. This process mitigates selection bias and avoids model misspecification. Simulation studies demonstrate that the proposed method outperforms existing baselines, and application to real-world marketing datasets confirms the practical implications and utility of our framework.", "date": "2025-11-12", "categories": []}
{"id": "2511.10031", "title": "Temporal Latent Variable Structural Causal Model for Causal Discovery under External Interferences", "abstract": "Inferring causal relationships from observed data is an important task, yet it becomes challenging when the data is subject to various external interferences. Most of these interferences are the additional effects of external factors on observed variables. Since these external factors are often unknown, we introduce latent variables to represent these unobserved factors that affect the observed data. Specifically, to capture the causal strength and adjacency information, we propose a new temporal latent variable structural causal model, incorporating causal strength and adjacency coefficients that represent the causal relationships between variables. Considering that expert knowledge can provide information about unknown interferences in certain scenarios, we develop a method that facilitates the incorporation of prior knowledge into parameter learning based on Variational Inference, to guide the model estimation. Experimental results demonstrate the stability and accuracy of our proposed method.", "date": "2025-11-13", "categories": []}
{"id": "2511.10120", "title": "Generalizing to Unseen Disaster Events: A Causal View", "abstract": "Due to the rapid growth of social media platforms, these tools have become essential for monitoring information during ongoing disaster events. However, extracting valuable insights requires real-time processing of vast amounts of data. A major challenge in existing systems is their exposure to event-related biases, which negatively affects their ability to generalize to emerging events. While recent advancements in debiasing and causal learning offer promising solutions, they remain underexplored in the disaster event domain. In this work, we approach bias mitigation through a causal lens and propose a method to reduce event- and domain-related biases, enhancing generalization to future events. Our approach outperforms multiple baselines by up to +1.9% F1 and significantly improves a PLM-based classifier across three disaster classification tasks.", "date": "2025-11-13", "categories": []}
{"id": "2511.10291", "title": "Causal Model-Based Reinforcement Learning for Sample-Efficient IoT Channel Access", "abstract": "Despite the advantages of multi-agent reinforcement learning (MARL) for wireless use case such as medium access control (MAC), their real-world deployment in Internet of Things (IoT) is hindered by their sample inefficiency. To alleviate this challenge, one can leverage model-based reinforcement learning (MBRL) solutions, however, conventional MBRL approaches rely on black-box models that are not interpretable and cannot reason. In contrast, in this paper, a novel causal model-based MARL framework is developed by leveraging tools from causal learn- ing. In particular, the proposed model can explicitly represent causal dependencies between network variables using structural causal models (SCMs) and attention-based inference networks. Interpretable causal models are then developed to capture how MAC control messages influence observations, how transmission actions determine outcomes, and how channel observations affect rewards. Data augmentation techniques are then used to generate synthetic rollouts using the learned causal model for policy optimization via proximal policy optimization (PPO). Analytical results demonstrate exponential sample complexity gains of causal MBRL over black-box approaches. Extensive simulations demonstrate that, on average, the proposed approach can reduce environment interactions by 58%, and yield faster convergence compared to model-free baselines. The proposed approach inherently is also shown to provide interpretable scheduling decisions via attention-based causal attribution, revealing which network conditions drive the policy. The resulting combination of sample efficiency and interpretability establishes causal MBRL as a practical approach for resource-constrained wireless systems.", "date": "2025-11-13", "categories": []}
{"id": "2511.10320", "title": "PITE: Multi-Prototype Alignment for Individual Treatment Effect Estimation", "abstract": "Estimating Individual Treatment Effects (ITE) from observational data is challenging due to confounding bias. Most studies tackle this bias by balancing distributions globally, but ignore individual heterogeneity and fail to capture the local structure that represents the natural clustering among individuals, which ultimately compromises ITE estimation. While instance-level alignment methods consider heterogeneity, they similarly overlook the local structure information. To address these issues, we propose an end-to-end Multi-\\textbf{P}rototype alignment method for \\textbf{ITE} estimation (\\textbf{PITE}). PITE effectively captures local structure within groups and enforces cross-group alignment, thereby achieving robust ITE estimation. Specifically, we first define prototypes as cluster centroids based on similar individuals under the same treatment. To identify local similarity and the distribution consistency, we perform instance-to-prototype matching to assign individuals to the nearest prototype within groups, and design a multi-prototype alignment strategy to encourage the matched prototypes to be close across treatment arms in the latent space. PITE not only reduces distribution shift through fine-grained, prototype-level alignment, but also preserves the local structures of treated and control groups, which provides meaningful constraints for ITE estimation. Extensive evaluations on benchmark datasets demonstrate that PITE outperforms 13 state-of-the-art methods, achieving more accurate and robust ITE estimation.", "date": "2025-11-13", "categories": []}
{"id": "2409.19777", "title": "Automatic debiasing of neural networks via moment-constrained learning", "abstract": "Causal and nonparametric estimands in economics and biostatistics can often be viewed as the mean of a linear functional applied to an unknown outcome regression function. Naively learning the regression function and taking a sample mean of the target functional results in biased estimators, and a rich debiasing literature has developed where one additionally learns the so-called Riesz representer (RR) of the target estimand (targeted learning, double ML, automatic debiasing etc.). Learning the RR via its derived functional form can be challenging, e.g. due to extreme inverse probability weights or the need to learn conditional density functions. Such challenges have motivated recent advances in automatic debiasing (AD), where the RR is learned directly via minimization of a bespoke loss. We propose moment-constrained learning as a new RR learning approach that addresses some shortcomings in AD, constraining the predicted moments and improving the robustness of RR estimates to optimization hyperparamters. Though our approach is not tied to a particular class of learner, we illustrate it using neural networks, and evaluate on the problems of average treatment/derivative effect estimation using semi-synthetic data. Our numerical experiments show improved performance versus state of the art benchmarks.", "date": "2024-09-29", "categories": []}
{"id": "2504.02430", "title": "How Artificial Intelligence Leads to Knowledge Why: An Inquiry Inspired by Aristotle's Posterior Analytics", "abstract": "Bayesian networks and causal models provide frameworks for handling queries about external interventions and counterfactuals, enabling tasks that go beyond what probability distributions alone can address. While these formalisms are often informally described as capturing causal knowledge, there is a lack of a formal theory characterizing the type of knowledge required to predict the effects of external interventions. This work introduces the theoretical framework of causal systems to clarify Aristotle's distinction between knowledge that and knowledge why within artificial intelligence. By interpreting existing artificial intelligence technologies as causal systems, it investigates the corresponding types of knowledge. Furthermore, it argues that predicting the effects of external interventions is feasible only with knowledge why, providing a more precise understanding of the knowledge necessary for such tasks.", "date": "2025-04-03", "categories": []}
{"id": "2505.11770", "title": "Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors", "abstract": "Interpretability research now offers a variety of techniques for identifying abstract internal mechanisms in neural networks. Can such techniques be used to predict how models will behave on out-of-distribution examples? In this work, we provide a positive answer to this question. Through a diverse set of language modeling tasks--including symbol manipulation, knowledge retrieval, and instruction following--we show that the most robust features for correctness prediction are those that play a distinctive causal role in the model's behavior. Specifically, we propose two methods that leverage causal mechanisms to predict the correctness of model outputs: counterfactual simulation (checking whether key causal variables are realized) and value probing (using the values of those variables to make predictions). Both achieve high AUC-ROC in distribution and outperform methods that rely on causal-agnostic features in out-of-distribution settings, where predicting model behaviors is more crucial. Our work thus highlights a novel and significant application for internal causal analysis of language models.", "date": "2025-05-17", "categories": []}
{"id": "2511.07484", "title": "Counterfactual Forecasting of Human Behavior using Generative AI and Causal Graphs", "abstract": "This study presents a novel framework for counterfactual user behavior forecasting that combines structural causal models with transformer-based generative artificial intelligence. To model fictitious situations, the method creates causal graphs that map the connections between user interactions, adoption metrics, and product features. The framework generates realistic behavioral trajectories under counterfactual conditions by using generative models that are conditioned on causal variables. Tested on datasets from web interactions, mobile applications, and e-commerce, the methodology outperforms conventional forecasting and uplift modeling techniques. Product teams can effectively simulate and assess possible interventions prior to deployment thanks to the framework improved interpretability through causal path visualization.", "date": "2025-11-09", "categories": []}
{"id": "2310.10559", "title": "Causal Dynamic Variational Autoencoder for Counterfactual Regression in Longitudinal Data", "abstract": "Accurately estimating treatment effects over time is crucial in fields such as precision medicine, epidemiology, economics, and marketing. Many current methods for estimating treatment effects over time assume that all confounders are observed or attempt to infer unobserved ones. In contrast, our approach focuses on unobserved adjustment variables, which specifically have a causal effect on the outcome sequence. Under the assumption of unconfoundedness, we address estimating Conditional Average Treatment Effects (CATEs) while accounting for unobserved heterogeneity in response to treatment due to these unobserved adjustment variables. Our proposed Causal Dynamic Variational Autoencoder (CDVAE) is grounded in theoretical guarantees concerning the validity of latent adjustment variables and generalization bounds on CATE estimation error. Extensive evaluations on synthetic and real-world datasets show that CDVAE outperforms existing baselines. Moreover, we demonstrate that state-of-the-art models significantly improve their CATE estimates when augmented with the latent substitutes learned by CDVAE, approaching oracle-level performance without direct access to the true adjustment variables.", "date": "2023-10-16", "categories": []}
{"id": "2506.03163", "title": "Causal Discovery in Dynamic Fading Wireless Networks", "abstract": "Dynamic causal discovery in wireless networks is essential due to evolving interference, fading, and mobility, which complicate traditional static causal models. This paper addresses causal inference challenges in dynamic fading wireless environments by proposing a sequential regression-based algorithm with a novel application of the NOTEARS acyclicity constraint, enabling efficient online updates. We derive theoretical lower and upper bounds on the detection delay required to identify structural changes, explicitly quantifying their dependence on network size, noise variance, and fading severity. Monte Carlo simulations validate these theoretical results, demonstrating linear increases in detection delay with network size, quadratic growth with noise variance, and inverse-square dependence on the magnitude of structural changes. Our findings provide rigorous theoretical insights and practical guidelines for designing robust online causal inference mechanisms to maintain network reliability under nonstationary wireless conditions.", "date": "2025-05-24", "categories": []}
{"id": "2511.06790", "title": "Robust Causal Discovery under Imperfect Structural Constraints", "abstract": "Robust causal discovery from observational data under imperfect prior knowledge remains a significant and largely unresolved challenge. Existing methods typically presuppose perfect priors or can only handle specific, pre-identified error types. And their performance degrades substantially when confronted with flawed constraints of unknown location and type. This decline arises because most of them rely on inflexible and biased thresholding strategies that may conflict with the data distribution. To overcome these limitations, we propose to harmonizes knowledge and data through prior alignment and conflict resolution. First, we assess the credibility of imperfect structural constraints through a surrogate model, which then guides a sparse penalization term measuring the loss between the learned and constrained adjacency matrices. We theoretically prove that, under ideal assumption, the knowledge-driven objective aligns with the data-driven objective. Furthermore, to resolve conflicts when this assumption is violated, we introduce a multi-task learning framework optimized via multi-gradient descent, jointly minimizing both objectives. Our proposed method is robust to both linear and nonlinear settings. Extensive experiments, conducted under diverse noise conditions and structural equation model types, demonstrate the effectiveness and efficiency of our method under imperfect structural constraints.", "date": "2025-11-10", "categories": []}
{"id": "2511.07236", "title": "Does TabPFN Understand Causal Structures?", "abstract": "Causal discovery is fundamental for multiple scientific domains, yet extracting causal information from real world data remains a significant challenge. Given the recent success on real data, we investigate whether TabPFN, a transformer-based tabular foundation model pre-trained on synthetic datasets generated from structural causal models, encodes causal information in its internal representations. We develop an adapter framework using a learnable decoder and causal tokens that extract causal signals from TabPFN's frozen embeddings and decode them into adjacency matrices for causal discovery. Our evaluations demonstrate that TabPFN's embeddings contain causal information, outperforming several traditional causal discovery algorithms, with such causal information being concentrated in mid-range layers. These findings establish a new direction for interpretable and adaptable foundation models and demonstrate the potential for leveraging pre-trained tabular models for causal discovery.", "date": "2025-11-10", "categories": []}
{"id": "2307.11127", "title": "Asymptotically Unbiased Synthetic Control Methods by Moment Matching", "abstract": "Synthetic Control Methods (SCMs) have become a fundamental tool for comparative case studies. The core idea behind SCMs is to estimate treatment effects by predicting counterfactual outcomes for a treated unit using a weighted combination of observed outcomes from untreated units. The accuracy of these predictions is crucial for evaluating the treatment effect of a policy intervention. Subsequent research has therefore focused on estimating SC weights. In this study, we highlight a key endogeneity issue in existing SCMs-namely, the correlation between the outcomes of untreated units and the error term of the synthetic control, which leads to bias in both counterfactual outcome prediction and treatment effect estimation. To address this issue, we propose a novel SCM based on moment matching, assuming that the outcome distribution of the treated unit can be approximated by a weighted mixture of the distributions of untreated units. Under this assumption, we estimate SC weights by matching the moments of the treated outcomes with the weighted sum of the moments of the untreated outcomes. Our method offers three advantages: first, under the mixture model assumption, our estimator is asymptotically unbiased; second, this asymptotic unbiasedness reduces the mean squared error in counterfactual predictions; and third, our method provides full distributions of the treatment effect rather than just expected values, thereby broadening the applicability of SCMs. Finally, we present experimental results that demonstrate the effectiveness of our approach.", "date": "2023-07-20", "categories": []}
{"id": "2506.22331", "title": "Less Greedy Equivalence Search", "abstract": "Greedy Equivalence Search (GES) is a classic score-based algorithm for causal discovery from observational data. In the sample limit, it recovers the Markov equivalence class of graphs that describe the data. Still, it faces two challenges in practice: computational cost and finite-sample accuracy. In this paper, we develop Less Greedy Equivalence Search (LGES), a variant of GES that retains its theoretical guarantees while partially addressing these limitations. LGES modifies the greedy step; rather than always applying the highest-scoring insertion, it avoids edge insertions between variables for which the score implies some conditional independence. This more targeted search yields up to a \\(10\\)-fold speed-up and a substantial reduction in structural error relative to GES. Moreover, LGES can guide the search using prior knowledge, and can correct this knowledge when contradicted by data. Finally, LGES can use interventional data to refine the learned observational equivalence class. We prove that LGES recovers the true equivalence class in the sample limit, even with misspecified knowledge. Experiments demonstrate that LGES outperforms GES and other baselines in speed, accuracy, and robustness to misspecified knowledge. Our code is available at https://github.com/CausalAILab/lges.", "date": "2025-06-27", "categories": []}
{"id": "2507.02275", "title": "It's Hard to Be Normal: The Impact of Noise on Structure-agnostic Estimation", "abstract": "Structure-agnostic causal inference studies how well one can estimate a treatment effect given black-box machine learning estimates of nuisance functions (like the impact of confounders on treatment and outcomes). Here, we find that the answer depends in a surprising way on the distribution of the treatment noise. Focusing on the partially linear model of \\citet{robinson1988root}, we first show that the widely adopted double machine learning (DML) estimator is minimax rate-optimal for Gaussian treatment noise, resolving an open problem of \\citet{mackey2018orthogonal}. Meanwhile, for independent non-Gaussian treatment noise, we show that DML is always suboptimal by constructing new practical procedures with higher-order robustness to nuisance errors. These \\emph{ACE} procedures use structure-agnostic cumulant estimators to achieve $r$-th order insensitivity to nuisance errors whenever the $(r+1)$-st treatment cumulant is non-zero. We complement these core results with novel minimax guarantees for binary treatments in the partially linear model. Finally, using synthetic demand estimation experiments, we demonstrate the practical benefits of our higher-order robust estimators.", "date": "2025-07-03", "categories": []}
{"id": "2511.04790", "title": "Causal Structure and Representation Learning with Biomedical Applications", "abstract": "Massive data collection holds the promise of a better understanding of complex phenomena and, ultimately, better decisions. Representation learning has become a key driver of deep learning applications, as it allows learning latent spaces that capture important properties of the data without requiring any supervised annotations. Although representation learning has been hugely successful in predictive tasks, it can fail miserably in causal tasks including predicting the effect of a perturbation/intervention. This calls for a marriage between representation learning and causal inference. An exciting opportunity in this regard stems from the growing availability of multi-modal data (observational and perturbational, imaging-based and sequencing-based, at the single-cell level, tissue-level, and organism-level). We outline a statistical and computational framework for causal structure and representation learning motivated by fundamental biomedical questions: how to effectively use observational and perturbational data to perform causal discovery on observed causal variables; how to use multi-modal views of the system to learn causal variables; and how to design optimal perturbations.", "date": "2025-11-06", "categories": []}
{"id": "2511.05236", "title": "The Causal Round Trip: Generating Authentic Counterfactuals by Eliminating Information Loss", "abstract": "Judea Pearl's vision of Structural Causal Models (SCMs) as engines for counterfactual reasoning hinges on faithful abduction: the precise inference of latent exogenous noise. For decades, operationalizing this step for complex, non-linear mechanisms has remained a significant computational challenge. The advent of diffusion models, powerful universal function approximators, offers a promising solution. However, we argue that their standard design, optimized for perceptual generation over logical inference, introduces a fundamental flaw for this classical problem: an inherent information loss we term the Structural Reconstruction Error (SRE). To address this challenge, we formalize the principle of Causal Information Conservation (CIC) as the necessary condition for faithful abduction. We then introduce BELM-MDCM, the first diffusion-based framework engineered to be causally sound by eliminating SRE by construction through an analytically invertible mechanism. To operationalize this framework, a Targeted Modeling strategy provides structural regularization, while a Hybrid Training Objective instills a strong causal inductive bias. Rigorous experiments demonstrate that our Zero-SRE framework not only achieves state-of-the-art accuracy but, more importantly, enables the high-fidelity, individual-level counterfactuals required for deep causal inquiries. Our work provides a foundational blueprint that reconciles the power of modern generative models with the rigor of classical causal theory, establishing a new and more rigorous standard for this emerging field.", "date": "2025-11-07", "categories": []}
{"id": "2212.09844", "title": "Robust Design and Evaluation of Predictive Algorithms under Unobserved Confounding", "abstract": "Predictive algorithms inform consequential decisions in settings with selective labels: outcomes are observed only for units selected by past decision makers. This creates an identification problem under unobserved confounding -- when selected and unselected units differ in unobserved ways that affect outcomes. We propose a framework for robust design and evaluation of predictive algorithms that bounds how much outcomes may differ between selected and unselected units with the same observed characteristics. These bounds formalize common empirical strategies including proxy outcomes and instrumental variables. Our estimators work across bounding strategies and performance measures such as conditional likelihoods, mean square error, and true/false positive rates. Using administrative data from a large Australian financial institution, we show that varying confounding assumptions substantially affects credit risk predictions and fairness evaluations across income groups.", "date": "2022-12-19", "categories": []}
{"id": "2511.03831", "title": "Higher-Order Causal Structure Learning with Additive Models", "abstract": "Causal structure learning has long been the central task of inferring causal insights from data. Despite the abundance of real-world processes exhibiting higher-order mechanisms, however, an explicit treatment of interactions in causal discovery has received little attention. In this work, we focus on extending the causal additive model (CAM) to additive models with higher-order interactions. This second level of modularity we introduce to the structure learning problem is most easily represented by a directed acyclic hypergraph which extends the DAG. We introduce the necessary definitions and theoretical tools to handle the novel structure we introduce and then provide identifiability results for the hyper DAG, extending the typical Markov equivalence classes. We next provide insights into why learning the more complex hypergraph structure may actually lead to better empirical results. In particular, more restrictive assumptions like CAM correspond to easier-to-learn hyper DAGs and better finite sample complexity. We finally develop an extension of the greedy CAM algorithm which can handle the more complex hyper DAG search space and demonstrate its empirical usefulness in synthetic experiments.", "date": "2025-11-05", "categories": []}
{"id": "2511.04619", "title": "Dynamic causal discovery in Alzheimer's disease through latent pseudotime modelling", "abstract": "The application of causal discovery to diseases like Alzheimer's (AD) is limited by the static graph assumptions of most methods; such models cannot account for an evolving pathophysiology, modulated by a latent disease pseudotime. We propose to apply an existing latent variable model to real-world AD data, inferring a pseudotime that orders patients along a data-driven disease trajectory independent of chronological age, then learning how causal relationships evolve. Pseudotime outperformed age in predicting diagnosis (AUC 0.82 vs 0.59). Incorporating minimal, disease-agnostic background knowledge substantially improved graph accuracy and orientation. Our framework reveals dynamic interactions between novel (NfL, GFAP) and established AD markers, enabling practical causal discovery despite violated assumptions.", "date": "2025-11-06", "categories": []}
{"id": "2510.04263", "title": "Efficient Latent Variable Causal Discovery: Combining Score Search and Targeted Testing", "abstract": "Learning causal structure from observational data is especially challenging when latent variables or selection bias are present. The Fast Causal Inference (FCI) algorithm addresses this setting but performs exhaustive conditional independence tests across many subsets, often leading to spurious independences, missing or extra edges, and unreliable orientations. We present a family of score-guided mixed-strategy causal search algorithms that extend this framework. First, we introduce BOSS-FCI and GRaSP-FCI, variants of GFCI (Greedy Fast Causal Inference) that substitute BOSS (Best Order Score Search) or GRaSP (Greedy Relaxations of Sparsest Permutation) for FGES (Fast Greedy Equivalence Search), preserving correctness while trading off scalability and conservativeness. Second, we develop FCI Targeted-Testing (FCIT), a novel hybrid method that replaces exhaustive testing with targeted, score-informed tests guided by BOSS. FCIT guarantees well-formed PAGs and achieves higher precision and efficiency across sample sizes. Finally, we propose a lightweight heuristic, LV-Dumb (Latent Variable \"Dumb\"), which returns the PAG of the BOSS DAG (Directed Acyclic Graph). Though not strictly sound for latent confounding, LV-Dumb often matches FCIT's accuracy while running substantially faster. Simulations and real-data analyses show that BOSS-FCI and GRaSP-FCI provide robust baselines, FCIT yields the best balance of precision and reliability, and LV-Dumb offers a fast, near-equivalent alternative. Together, these methods demonstrate that targeted and score-guided strategies can dramatically improve the efficiency and correctness of latent-variable causal discovery.", "date": "2025-10-05", "categories": []}
{"id": "2511.03168", "title": "UnCLe: Towards Scalable Dynamic Causal Discovery in Non-linear Temporal Systems", "abstract": "Uncovering cause-effect relationships from observational time series is fundamental to understanding complex systems. While many methods infer static causal graphs, real-world systems often exhibit dynamic causality-where relationships evolve over time. Accurately capturing these temporal dynamics requires time-resolved causal graphs. We propose UnCLe, a novel deep learning method for scalable dynamic causal discovery. UnCLe employs a pair of Uncoupler and Recoupler networks to disentangle input time series into semantic representations and learns inter-variable dependencies via auto-regressive Dependency Matrices. It estimates dynamic causal influences by analyzing datapoint-wise prediction errors induced by temporal perturbations. Extensive experiments demonstrate that UnCLe not only outperforms state-of-the-art baselines on static causal discovery benchmarks but, more importantly, exhibits a unique capability to accurately capture and represent evolving temporal causality in both synthetic and real-world dynamic systems (e.g., human motion). UnCLe offers a promising approach for revealing the underlying, time-varying mechanisms of complex phenomena.", "date": "2025-11-05", "categories": []}
{"id": "2501.10124", "title": "Gene Regulatory Network Inference in the Presence of Selection Bias and Latent Confounders", "abstract": "Gene regulatory network inference (GRNI) aims to discover how genes causally regulate each other from gene expression data. It is well-known that statistical dependencies in observed data do not necessarily imply causation, as spurious dependencies may arise from latent confounders, such as non-coding RNAs. Numerous GRNI methods have thus been proposed to address this confounding issue. However, dependencies may also result from selection--only cells satisfying certain survival or inclusion criteria are observed--while these selection-induced spurious dependencies are frequently overlooked in gene expression data analyses. In this work, we show that such selection is ubiquitous and, when ignored or conflated with true regulations, can lead to flawed causal interpretation and misguided intervention recommendations. To address this challenge, a fundamental question arises: can we distinguish dependencies due to regulation, confounding, and crucially, selection? We show that gene perturbations offer a simple yet effective answer: selection-induced dependencies are symmetric under perturbation, while those from regulation or confounding are not. Building on this motivation, we propose GISL (Gene regulatory network Inference in the presence of Selection bias and Latent confounders), a principled algorithm that leverages perturbation data to uncover both true gene regulatory relations and non-regulatory mechanisms of selection and confounding up to the equivalence class. Experiments on synthetic and real-world gene expression data demonstrate the effectiveness of our method.", "date": "2025-01-17", "categories": []}
{"id": "2507.01700", "title": "Relational Causal Discovery with Latent Confounders", "abstract": "Estimating causal effects from real-world relational data can be challenging when the underlying causal model and potential confounders are unknown. While several causal discovery algorithms exist for learning causal models with latent confounders from data, they assume that the data is independent and identically distributed (i.i.d.) and are not well-suited for learning from relational data. Similarly, existing relational causal discovery algorithms assume causal sufficiency, which is unrealistic for many real-world datasets. To address this gap, we propose RelFCI, a sound and complete causal discovery algorithm for relational data with latent confounders. Our work builds upon the Fast Causal Inference (FCI) and Relational Causal Discovery (RCD) algorithms and it defines new graphical models, necessary to support causal discovery in relational domains. We also establish soundness and completeness guarantees for relational d-separation with latent confounders. We present experimental results demonstrating the effectiveness of RelFCI in identifying the correct causal structure in relational causal models with latent confounders.", "date": "2025-07-02", "categories": []}
{"id": "2510.04276", "title": "Scalable Causal Discovery from Recursive Nonlinear Data via Truncated Basis Function Scores and Tests", "abstract": "Learning graphical conditional independence structures from nonlinear, continuous or mixed data is a central challenge in machine learning and the sciences, and many existing methods struggle to scale to thousands of samples or hundreds of variables. We introduce two basis-expansion tools for scalable causal discovery. First, the Basis Function BIC (BF-BIC) score uses truncated additive expansions to approximate nonlinear dependencies. BF-BIC is theoretically consistent under additive models and extends to post-nonlinear (PNL) models via an invertible reparameterization. It remains robust under moderate interactions and supports mixed data through a degenerate-Gaussian embedding for discrete variables. In simulations with fully nonlinear neural causal models (NCMs), BF-BIC outperforms kernel- and constraint-based methods (e.g., KCI, RFCI) in both accuracy and runtime. Second, the Basis Function Likelihood Ratio Test (BF-LRT) provides an approximate conditional independence test that is substantially faster than kernel tests while retaining competitive accuracy. Extensive simulations and a real-data application to Canadian wildfire risk show that, when integrated into hybrid searches, BF-based methods enable interpretable and scalable causal discovery. Implementations are available in Python, R, and Java.", "date": "2025-10-05", "categories": []}
{"id": "2511.02332", "title": "Biological Regulatory Network Inference through Circular Causal Structure Learning", "abstract": "Biological networks are pivotal in deciphering the complexity and functionality of biological systems. Causal inference, which focuses on determining the directionality and strength of interactions between variables rather than merely relying on correlations, is considered a logical approach for inferring biological networks. Existing methods for causal structure inference typically assume that causal relationships between variables can be represented by directed acyclic graphs (DAGs). However, this assumption is at odds with the reality of widespread feedback loops in biological systems, making these methods unsuitable for direct use in biological network inference. In this study, we propose a new framework named SCALD (Structural CAusal model for Loop Diagram), which employs a nonlinear structure equation model and a stable feedback loop conditional constraint through continuous optimization to infer causal regulatory relationships under feedback loops. We observe that SCALD outperforms state-of-the-art methods in inferring both transcriptional regulatory networks and signaling transduction networks. SCALD has irreplaceable advantages in identifying feedback regulation. Through transcription factor (TF) perturbation data analysis, we further validate the accuracy and sensitivity of SCALD. Additionally, SCALD facilitates the discovery of previously unknown regulatory relationships, which we have subsequently confirmed through ChIP-seq data analysis. Furthermore, by utilizing SCALD, we infer the key driver genes that facilitate the transformation from colon inflammation to cancer by examining the dynamic changes within regulatory networks during the process.", "date": "2025-11-04", "categories": []}
{"id": "2511.02536", "title": "Theoretical Guarantees for Causal Discovery on Large Random Graphs", "abstract": "We investigate theoretical guarantees for the false-negative rate (FNR) -- the fraction of true causal edges whose orientation is not recovered, under single-variable random interventions and an $\\epsilon$-interventional faithfulness assumption that accommodates latent confounding. For sparse Erd\\H{o}s--R\\'enyi directed acyclic graphs, where the edge probability scales as $p_e = \\Theta(1/d)$, we show that the FNR concentrates around its mean at rate $O(\\frac{\\log d}{\\sqrt d})$, implying that large deviations above the expected error become exponentially unlikely as dimensionality increases. This concentration ensures that derived upper bounds hold with high probability in large-scale settings. Extending the analysis to generalized Barab\\'asi--Albert graphs reveals an even stronger phenomenon: when the degree exponent satisfies $\\gamma > 3$, the deviation width scales as $O(d^{\\beta - \\frac{1}{2}})$ with $\\beta = 1/(\\gamma - 1) < \\frac{1}{2}$, and hence vanishes in the limit. This demonstrates that realistic scale-free topologies intrinsically regularize causal discovery, reducing variability in orientation error. These finite-dimension results provide the first dimension-adaptive, faithfulness-robust guarantees for causal structure recovery, and challenge the intuition that high dimensionality and network heterogeneity necessarily hinder accurate discovery. Our simulation results corroborate these theoretical predictions, showing that the FNR indeed concentrates and often vanishes in practice as dimensionality grows.", "date": "2025-11-04", "categories": []}
{"id": "2410.20057", "title": "Mechanism Learning: reverse causal inference in the presence of multiple unknown confounding through causally weighted Gaussian mixture models", "abstract": "A major limitation of machine learning (ML) prediction models is that they recover associational, rather than causal, predictive relationships between variables. In high-stakes automation applications of ML this is problematic, as the model often learns spurious, non-causal associations. This paper proposes mechanism learning, a simple method which uses causally weighted Gaussian Mixture Models (CW-GMMs) to deconfound observational data such that any appropriate ML model is forced to learn predictive relationships between effects and their causes (reverse causal inference), despite the potential presence of multiple unknown and unmeasured confounding. Effect variables can be very high-dimensional, and the predictive relationship nonlinear, as is common in ML applications. This novel method is widely applicable, the only requirement is the existence of a set of mechanism variables mediating the cause (prediction target) and effect (feature data), which is independent of the (unmeasured) confounding variables. We test our method on fully synthetic, semi-synthetic and real-world datasets, demonstrating that it can discover reliable, unbiased, causal ML predictors where by contrast, the same ML predictor trained naively using classical supervised learning on the original observational data, is heavily biased by spurious associations. We provide code to implement the results in the paper, online.", "date": "2024-10-26", "categories": []}
{"id": "2505.01037", "title": "Characterization and Learning of Causal Graphs from Hard Interventions", "abstract": "A fundamental challenge in the empirical sciences involves uncovering causal structure through observation and experimentation. Causal discovery entails linking the conditional independence (CI) invariances in observational data to their corresponding graphical constraints via d-separation. In this paper, we consider a general setting where we have access to data from multiple experimental distributions resulting from hard interventions, as well as potentially from an observational distribution. By comparing different interventional distributions, we propose a set of graphical constraints that are fundamentally linked to Pearl's do-calculus within the framework of hard interventions. These graphical constraints associate each graphical structure with a set of interventional distributions that are consistent with the rules of do-calculus. We characterize the interventional equivalence class of causal graphs with latent variables and introduce a graphical representation that can be used to determine whether two causal graphs are interventionally equivalent, i.e., whether they are associated with the same family of hard interventional distributions, where the elements of the family are indistinguishable using the invariances from do-calculus. We also propose a learning algorithm to integrate multiple datasets from hard interventions, introducing new orientation rules. The learning objective is a tuple of augmented graphs which entails a set of causal graphs. We also prove the soundness of the proposed algorithm.", "date": "2025-05-02", "categories": []}
{"id": "2507.17439", "title": "Penalized Empirical Likelihood for Doubly Robust Causal Inference under Contamination in High Dimensions", "abstract": "We propose a doubly robust estimator for the average treatment effect in high dimensional\n  low sample size observational studies, where contamination and model misspecification pose\n  serious inferential challenges. The estimator combines bounded influence estimating equations\n  for outcome modeling with covariate balancing propensity scores for treatment assignment,\n  embedded within a penalized empirical likelihood framework using nonconvex regularization.\n  It satisfies the oracle property by jointly achieving consistency under partial model correct ness, selection consistency, robustness to contamination, and asymptotic normality. For uncertainty quantification, we derive a finite sample confidence interval using cumulant generating\n  functions and influence function corrections, avoiding reliance on asymptotic approximations.\n  Simulation studies and applications to gene expression datasets (Golub and Khan) demonstrate superior performance in bias, error metrics, and interval calibration, highlighting the\n  method robustness and inferential validity in HDLSS regimes. One notable aspect is that\n  even in the absence of contamination, the proposed estimator and its confidence interval\n  remain efficient compared to those of competing models.", "date": "2025-07-23", "categories": []}
{"id": "2511.00318", "title": "A Technical Exploration of Causal Inference with Hybrid LLM Synthetic Data", "abstract": "Large Language Models (LLMs) offer a flexible means to generate synthetic tabular data, yet existing approaches often fail to preserve key causal parameters such as the average treatment effect (ATE). In this technical exploration, we first demonstrate that state-of-the-art synthetic data generators, both GAN- and LLM-based, can achieve high predictive fidelity while substantially misestimating causal effects. To address this gap, we propose a hybrid generation framework that combines model-based covariate synthesis (monitored via distance-to-closest-record filtering) with separately learned propensity and outcome models, thereby ensuring that (W, A, Y) triplets retain their underlying causal structure. We further introduce a synthetic pairing strategy to mitigate positivity violations and a realistic evaluation protocol that leverages unlimited synthetic samples to benchmark traditional estimators (IPTW, AIPW, substitution) under complex covariate distributions. This work lays the groundwork for LLM-powered data pipelines that support robust causal analysis. Our code is available at https://github.com/Xyc-arch/llm-synthetic-for-causal-inference.git.", "date": "2025-10-31", "categories": []}
{"id": "2511.01185", "title": "A Comparative Study of Model Adaptation Strategies for Multi-Treatment Uplift Modeling", "abstract": "Uplift modeling has emerged as a crucial technique for individualized treatment effect estimation, particularly in fields such as marketing and healthcare. Modeling uplift effects in multi-treatment scenarios plays a key role in real-world applications. Current techniques for modeling multi-treatment uplift are typically adapted from binary-treatment works. In this paper, we investigate and categorize all current model adaptations into two types: Structure Adaptation and Feature Adaptation. Through our empirical experiments, we find that these two adaptation types cannot maintain effectiveness under various data characteristics (noisy data, mixed with observational data, etc.). To enhance estimation ability and robustness, we propose Orthogonal Function Adaptation (OFA) based on the function approximation theorem. We conduct comprehensive experiments with multiple data characteristics to study the effectiveness and robustness of all model adaptation techniques. Our experimental results demonstrate that our proposed OFA can significantly improve uplift model performance compared to other vanilla adaptation methods and exhibits the highest robustness.", "date": "2025-11-03", "categories": []}
{"id": "2511.01329", "title": "Unbiased Platform-Level Causal Estimation for Search Systems: A Competitive Isolation PSM-DID Framework", "abstract": "Evaluating platform-level interventions in search-based two-sided marketplaces is fundamentally challenged by systemic effects such as spillovers and network interference. While widely used for causal inference, the PSM (Propensity Score Matching) - DID (Difference-in-Differences) framework remains susceptible to selection bias and cross-unit interference from unaccounted spillovers. In this paper, we introduced Competitive Isolation PSM-DID, a novel causal framework that integrates propensity score matching with competitive isolation to enable platform-level effect measurement (e.g., order volume, GMV) instead of item-level metrics in search systems.\n  Our approach provides theoretically guaranteed unbiased estimation under mutual exclusion conditions, with an open dataset released to support reproducible research on marketplace interference (github.com/xxxx). Extensive experiments demonstrate significant reductions in interference effects and estimation variance compared to baseline methods. Successful deployment in a large-scale marketplace confirms the framework's practical utility for platform-level causal inference.", "date": "2025-11-03", "categories": []}
{"id": "2511.01396", "title": "Relaxing partition admissibility in Cluster-DAGs: a causal calculus with arbitrary variable clustering", "abstract": "Cluster DAGs (C-DAGs) provide an abstraction of causal graphs in which nodes represent clusters of variables, and edges encode both cluster-level causal relationships and dependencies arisen from unobserved confounding. C-DAGs define an equivalence class of acyclic causal graphs that agree on cluster-level relationships, enabling causal reasoning at a higher level of abstraction. However, when the chosen clustering induces cycles in the resulting C-DAG, the partition is deemed inadmissible under conventional C-DAG semantics. In this work, we extend the C-DAG framework to support arbitrary variable clusterings by relaxing the partition admissibility constraint, thereby allowing cyclic C-DAG representations. We extend the notions of d-separation and causal calculus to this setting, significantly broadening the scope of causal reasoning across clusters and enabling the application of C-DAGs in previously intractable scenarios. Our calculus is both sound and atomically complete with respect to the do-calculus: all valid interventional queries at the cluster level can be derived using our rules, each corresponding to a primitive do-calculus step.", "date": "2025-11-03", "categories": []}
{"id": "2511.01641", "title": "Cross-Treatment Effect Estimation for Multi-Category, Multi-Valued Causal Inference via Dynamic Neural Masking", "abstract": "Counterfactual causal inference faces significant challenges when extended to multi-category, multi-valued treatments, where complex cross-effects between heterogeneous interventions are difficult to model. Existing methodologies remain constrained to binary or single-type treatments and suffer from restrictive assumptions, limited scalability, and inadequate evaluation frameworks for complex intervention scenarios.\n  We present XTNet, a novel network architecture for multi-category, multi-valued treatment effect estimation. Our approach introduces a cross-effect estimation module with dynamic masking mechanisms to capture treatment interactions without restrictive structural assumptions. The architecture employs a decomposition strategy separating basic effects from cross-treatment interactions, enabling efficient modeling of combinatorial treatment spaces. We also propose MCMV-AUCC, a suitable evaluation metric that accounts for treatment costs and interaction effects. Extensive experiments on synthetic and real-world datasets demonstrate that XTNet consistently outperforms state-of-the-art baselines in both ranking accuracy and effect estimation quality. The results of the real-world A/B test further confirm its effectiveness.", "date": "2025-11-03", "categories": []}
{"id": "2502.20097", "title": "Qini Curve Estimation under Clustered Network Interference", "abstract": "Qini curves are a widely used tool for assessing treatment policies under allocation constraints as they visualize the incremental gain of a new treatment policy versus the cost of its implementation. Standard Qini curve estimation assumes no interference between units: that is, that treating one unit does not influence the outcome of any other unit. In many real-life applications such as public policy or marketing, however, the presence of interference is common. Ignoring interference in these scenarios can lead to systematically biased Qini curves that over- or under-estimate a treatment policy's cost-effectiveness. In this paper, we address the problem of Qini curve estimation under clustered network interference, where interfering units form independent clusters. We propose a formal description of the problem setting with an experimental study design under which we can account for clustered network interference. Within this framework, we describe three estimation strategies, each suited to different conditions, and provide guidance for selecting the most appropriate approach by highlighting the inherent bias-variance trade-offs. To complement our theoretical analysis, we introduce a marketplace simulator that replicates clustered network interference in a typical e-commerce environment, allowing us to evaluate and compare the proposed strategies in practice.", "date": "2025-02-27", "categories": []}
{"id": "2510.16026", "title": "A tutorial on discovering and quantifying the effect of latent causal sources of multimodal EHR data", "abstract": "We provide an accessible description of a peer-reviewed generalizable causal machine learning pipeline to (i) discover latent causal sources of large-scale electronic health records observations, and (ii) quantify the source causal effects on clinical outcomes. We illustrate how imperfect multimodal clinical data can be processed, decomposed into probabilistic independent latent sources, and used to train taskspecific causal models from which individual causal effects can be estimated. We summarize the findings of the two real-world applications of the approach to date as a demonstration of its versatility and utility for medical discovery at scale.", "date": "2025-10-15", "categories": []}
{"id": "2405.20482", "title": "Sparsity regularization via tree-structured environments for disentangled representations", "abstract": "Many causal systems such as biological processes in cells can only be observed indirectly via measurements, such as gene expression. Causal representation learning -- the task of correctly mapping low-level observations to latent causal variables -- could advance scientific understanding by enabling inference of latent variables such as pathway activation. In this paper, we develop methods for inferring latent variables from multiple related datasets (environments) and tasks. As a running example, we consider the task of predicting a phenotype from gene expression, where we often collect data from multiple cell types or organisms that are related in known ways. The key insight is that the mapping from latent variables driven by gene expression to the phenotype of interest changes sparsely across closely related environments. To model sparse changes, we introduce Tree-Based Regularization (TBR), an objective that minimizes both prediction error and regularizes closely related environments to learn similar predictors. We prove that under assumptions about the degree of sparse changes, TBR identifies the true latent variables up to some simple transformations. We evaluate the theory empirically with both simulations and ground-truth gene expression data. We find that TBR recovers the latent causal variables better than related methods across these settings, even under settings that violate some assumptions of the theory.", "date": "2024-05-30", "categories": []}
{"id": "2503.04852", "title": "CAUSAL3D: A Comprehensive Benchmark for Causal Learning from Visual Data", "abstract": "True intelligence hinges on the ability to uncover and leverage hidden causal relations. Despite significant progress in AI and computer vision (CV), there remains a lack of benchmarks for assessing models' abilities to infer latent causality from complex visual data. In this paper, we introduce \\textsc{\\textbf{Causal3D}}, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning. Designed within a systematic framework, Causal3D comprises 19 3D-scene datasets capturing diverse causal relations, views, and backgrounds, enabling evaluations across scenes of varying complexity. We assess multiple state-of-the-art methods, including classical causal discovery, causal representation learning, and large/vision-language models (LLMs/VLMs). Our experiments show that as causal structures grow more complex without prior knowledge, performance declines significantly, highlighting the challenges even advanced methods face in complex causal scenarios. Causal3D serves as a vital resource for advancing causal reasoning in CV and fostering trustworthy AI in critical domains.", "date": "2025-03-06", "categories": []}
{"id": "2507.02843", "title": "LLM-Driven Treatment Effect Estimation Under Inference Time Text Confounding", "abstract": "Estimating treatment effects is crucial for personalized decision-making in medicine, but this task faces unique challenges in clinical practice. At training time, models for estimating treatment effects are typically trained on well-structured medical datasets that contain detailed patient information. However, at inference time, predictions are often made using textual descriptions (e.g., descriptions with self-reported symptoms), which are incomplete representations of the original patient information. In this work, we make three contributions. (1) We show that the discrepancy between the data available during training time and inference time can lead to biased estimates of treatment effects. We formalize this issue as an inference time text confounding problem, where confounders are fully observed during training time but only partially available through text at inference time. (2) To address this problem, we propose a novel framework for estimating treatment effects that explicitly accounts for inference time text confounding. Our framework leverages large language models together with a custom doubly robust learner to mitigate biases caused by the inference time text confounding. (3) Through a series of experiments, we demonstrate the effectiveness of our framework in real-world applications.", "date": "2025-07-03", "categories": []}
{"id": "2507.08896", "title": "Predictive Causal Inference via Spatio-Temporal Modeling and Penalized Empirical Likelihood", "abstract": "This study introduces an integrated framework for predictive causal inference designed to overcome limitations inherent in conventional single model approaches. Specifically, we combine a Hidden Markov Model (HMM) for spatial health state estimation with a Multi Task and Multi Graph Convolutional Network (MTGCN) for capturing temporal outcome trajectories. The framework asymmetrically treats temporal and spatial information regarding them as endogenous variables in the outcome regression, and exogenous variables in the propensity score model, thereby expanding the standard doubly robust treatment effect estimation to jointly enhance bias correction and predictive accuracy. To demonstrate its utility, we focus on clinical domains such as cancer, dementia, and Parkinson disease, where treatment effects are challenging to observe directly. Simulation studies are conducted to emulate latent disease dynamics and evaluate the model performance under varying conditions. Overall, the proposed framework advances predictive causal inference by structurally adapting to spatiotemporal complexities common in biomedical data.", "date": "2025-07-11", "categories": []}
{"id": "2509.17784", "title": "Revealing Multimodal Causality with Large Language Models", "abstract": "Uncovering cause-and-effect mechanisms from data is fundamental to scientific progress. While large language models (LLMs) show promise for enhancing causal discovery (CD) from unstructured data, their application to the increasingly prevalent multimodal setting remains a critical challenge. Even with the advent of multimodal LLMs (MLLMs), their efficacy in multimodal CD is hindered by two primary limitations: (1) difficulty in exploring intra- and inter-modal interactions for comprehensive causal variable identification; and (2) insufficiency to handle structural ambiguities with purely observational data. To address these challenges, we propose MLLM-CD, a novel framework for multimodal causal discovery from unstructured data. It consists of three key components: (1) a novel contrastive factor discovery module to identify genuine multimodal factors based on the interactions explored from contrastive sample pairs; (2) a statistical causal structure discovery module to infer causal relationships among discovered factors; and (3) an iterative multimodal counterfactual reasoning module to refine the discovery outcomes iteratively by incorporating the world knowledge and reasoning capabilities of MLLMs. Extensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of the proposed MLLM-CD in revealing genuine factors and causal relationships among them from multimodal unstructured data.", "date": "2025-09-22", "categories": []}
{"id": "2510.23534", "title": "Direct Debiased Machine Learning via Bregman Divergence Minimization", "abstract": "We develop a direct debiased machine learning framework comprising Neyman targeted estimation and generalized Riesz regression. Our framework unifies Riesz regression for automatic debiased machine learning, covariate balancing, targeted maximum likelihood estimation (TMLE), and density-ratio estimation. In many problems involving causal effects or structural models, the parameters of interest depend on regression functions. Plugging regression functions estimated by machine learning methods into the identifying equations can yield poor performance because of first-stage bias. To reduce such bias, debiased machine learning employs Neyman orthogonal estimating equations. Debiased machine learning typically requires estimation of the Riesz representer and the regression function. For this problem, we develop a direct debiased machine learning framework with an end-to-end algorithm. We formulate estimation of the nuisance parameters, the regression function and the Riesz representer, as minimizing the discrepancy between Neyman orthogonal scores computed with known and unknown nuisance parameters, which we refer to as Neyman targeted estimation. Neyman targeted estimation includes Riesz representer estimation, and we measure discrepancies using the Bregman divergence. The Bregman divergence encompasses various loss functions as special cases, where the squared loss yields Riesz regression and the Kullback-Leibler divergence yields entropy balancing. We refer to this Riesz representer estimation as generalized Riesz regression. Neyman targeted estimation also yields TMLE as a special case for regression function estimation. Furthermore, for specific pairs of models and Riesz representer estimation methods, we can automatically obtain the covariate balancing property without explicitly solving the covariate balancing objective.", "date": "2025-10-27", "categories": []}
{"id": "2510.26342", "title": "Linear Causal Discovery with Interventional Constraints", "abstract": "Incorporating causal knowledge and mechanisms is essential for refining causal models and improving downstream tasks such as designing new treatments. In this paper, we introduce a novel concept in causal discovery, termed interventional constraints, which differs fundamentally from interventional data. While interventional data require direct perturbations of variables, interventional constraints encode high-level causal knowledge in the form of inequality constraints on causal effects. For instance, in the Sachs dataset (Sachs et al.\\ 2005), Akt has been shown to be activated by PIP3, meaning PIP3 exerts a positive causal effect on Akt. Existing causal discovery methods allow enforcing structural constraints (for example, requiring a causal path from PIP3 to Akt), but they may still produce incorrect causal conclusions such as learning that \"PIP3 inhibits Akt\". Interventional constraints bridge this gap by explicitly constraining the total causal effect between variable pairs, ensuring learned models respect known causal influences. To formalize interventional constraints, we propose a metric to quantify total causal effects for linear causal models and formulate the problem as a constrained optimization task, solved using a two-stage constrained optimization method. We evaluate our approach on real-world datasets and demonstrate that integrating interventional constraints not only improves model accuracy and ensures consistency with established findings, making models more explainable, but also facilitates the discovery of new causal relationships that would otherwise be costly to identify.", "date": "2025-10-30", "categories": []}
{"id": "2510.26700", "title": "Assessment of the conditional exchangeability assumption in causal machine learning models: a simulation study", "abstract": "Observational studies developing causal machine learning (ML) models for the prediction of individualized treatment effects (ITEs) seldom conduct empirical evaluations to assess the conditional exchangeability assumption. We aimed to evaluate the performance of these models under conditional exchangeability violations and the utility of negative control outcomes (NCOs) as a diagnostic. We conducted a simulation study to examine confounding bias in ITE estimates generated by causal forest and X-learner models under varying conditions, including the presence or absence of true heterogeneity. We simulated data to reflect real-world scenarios with differing levels of confounding, sample size, and NCO confounding structures. We then estimated and compared subgroup-level treatment effects on the primary outcome and NCOs across settings with and without unmeasured confounding. When conditional exchangeability was violated, causal forest and X-learner models failed to recover true treatment effect heterogeneity and, in some cases, falsely indicated heterogeneity when there was none. NCOs successfully identified subgroups affected by unmeasured confounding. Even when NCOs did not perfectly satisfy its ideal assumptions, it remained informative, flagging potential bias in subgroup level estimates, though not always pinpointing the subgroup with the largest confounding. Violations of conditional exchangeability substantially limit the validity of ITE estimates from causal ML models in routinely collected observational data. NCOs serve a useful empirical diagnostic tool for detecting subgroup-specific unmeasured confounding and should be incorporated into causal ML workflows to support the credibility of individualized inference.", "date": "2025-10-30", "categories": []}
{"id": "2510.26783", "title": "A Unified Theory for Causal Inference: Direct Debiased Machine Learning via Bregman-Riesz Regression", "abstract": "This note introduces a unified theory for causal inference that integrates Riesz regression, covariate balancing, density-ratio estimation (DRE), targeted maximum likelihood estimation (TMLE), and the matching estimator in average treatment effect (ATE) estimation. In ATE estimation, the balancing weights and the regression functions of the outcome play important roles, where the balancing weights are referred to as the Riesz representer, bias-correction term, and clever covariates, depending on the context. Riesz regression, covariate balancing, DRE, and the matching estimator are methods for estimating the balancing weights, where Riesz regression is essentially equivalent to DRE in the ATE context, the matching estimator is a special case of DRE, and DRE is in a dual relationship with covariate balancing. TMLE is a method for constructing regression function estimators such that the leading bias term becomes zero. Nearest Neighbor Matching is equivalent to Least Squares Density Ratio Estimation and Riesz Regression.", "date": "2025-10-30", "categories": []}
{"id": "2505.17468", "title": "Efficient Adaptive Experimentation with Noncompliance", "abstract": "We study the problem of estimating the average treatment effect (ATE) in adaptive experiments where treatment can only be encouraged -- rather than directly assigned -- via a binary instrumental variable. Building on semiparametric efficiency theory, we derive the efficiency bound for ATE estimation under arbitrary, history-dependent instrument-assignment policies, and show it is minimized by a variance-aware allocation rule that balances outcome noise and compliance variability. Leveraging this insight, we introduce AMRIV -- an Adaptive, Multiply-Robust estimator for Instrumental-Variable settings with variance-optimal assignment. AMRIV pairs (i) an online policy that adaptively approximates the optimal allocation with (ii) a sequential, influence-function-based estimator that attains the semiparametric efficiency bound while retaining multiply-robust consistency. We establish asymptotic normality, explicit convergence rates, and anytime-valid asymptotic confidence sequences that enable sequential inference. Finally, we demonstrate the practical effectiveness of our approach through empirical studies, showing that adaptive instrument assignment, when combined with the AMRIV estimator, yields improved efficiency and robustness compared to existing baselines.", "date": "2025-05-23", "categories": []}
{"id": "2510.23906", "title": "Group Interventions on Deep Networks for Causal Discovery in Subsystems", "abstract": "Causal discovery uncovers complex relationships between variables, enhancing predictions, decision-making, and insights into real-world systems, especially in nonlinear multivariate time series. However, most existing methods primarily focus on pairwise cause-effect relationships, overlooking interactions among groups of variables, i.e., subsystems and their collective causal influence. In this study, we introduce gCDMI, a novel multi-group causal discovery method that leverages group-level interventions on trained deep neural networks and employs model invariance testing to infer causal relationships. Our approach involves three key steps. First, we use deep learning to jointly model the structural relationships among groups of all time series. Second, we apply group-wise interventions to the trained model. Finally, we conduct model invariance testing to determine the presence of causal links among variable groups. We evaluate our method on simulated datasets, demonstrating its superior performance in identifying group-level causal relationships compared to existing methods. Additionally, we validate our approach on real-world datasets, including brain networks and climate ecosystems. Our results highlight that applying group-level interventions to deep learning models, combined with invariance testing, can effectively reveal complex causal structures, offering valuable insights for domains such as neuroscience and climate science.", "date": "2025-10-27", "categories": []}
{"id": "2510.25037", "title": "Graph Distance Based on Cause-Effect Estimands with Latents", "abstract": "Causal discovery aims to recover graphs that represent causal relations among given variables from observations, and new methods are constantly being proposed. Increasingly, the community raises questions about how much progress is made, because properly evaluating discovered graphs remains notoriously difficult, particularly under latent confounding. We propose a graph distance measure for acyclic directed mixed graphs (ADMGs) based on the downstream task of cause-effect estimation under unobserved confounding. Our approach uses identification via fixing and a symbolic verifier to quantify how graph differences distort cause-effect estimands for different treatment-outcome pairs. We analyze the behavior of the measure under different graph perturbations and compare it against existing distance metrics.", "date": "2025-10-28", "categories": []}
{"id": "2502.05295", "title": "GST-UNet: A Neural Framework for Spatiotemporal Causal Inference with Time-Varying Confounding", "abstract": "Estimating causal effects from spatiotemporal observational data is essential in public health, environmental science, and policy evaluation, where randomized experiments are often infeasible. Existing approaches, however, either rely on strong structural assumptions or fail to handle key challenges such as interference, spatial confounding, temporal carryover, and time-varying confounding -- where covariates are influenced by past treatments and, in turn, affect future ones. We introduce GST-UNet (G-computation Spatio-Temporal UNet), a theoretically grounded neural framework that combines a U-Net-based spatiotemporal encoder with regression-based iterative G-computation to estimate location-specific potential outcomes under complex intervention sequences. GST-UNet explicitly adjusts for time-varying confounders and captures non-linear spatial and temporal dependencies, enabling valid causal inference from a single observed trajectory in data-scarce settings. We validate its effectiveness in synthetic experiments and in a real-world analysis of wildfire smoke exposure and respiratory hospitalizations during the 2018 California Camp Fire. Together, these results position GST-UNet as a principled and ready-to-use framework for spatiotemporal causal inference, advancing reliable estimation in policy-relevant and scientific domains.", "date": "2025-02-07", "categories": []}
{"id": "2505.17637", "title": "Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach", "abstract": "Spatio-temporal prediction plays a crucial role in intelligent transportation, weather forecasting, and urban planning. While integrating multi-modal data has shown potential for enhancing prediction accuracy, key challenges persist: (i) inadequate fusion of multi-modal information, (ii) confounding factors that obscure causal relations, and (iii) high computational complexity of prediction models. To address these challenges, we propose E^2-CSTP, an Effective and Efficient Causal multi-modal Spatio-Temporal Prediction framework. E^2-CSTP leverages cross-modal attention and gating mechanisms to effectively integrate multi-modal data. Building on this, we design a dual-branch causal inference approach: the primary branch focuses on spatio-temporal prediction, while the auxiliary branch mitigates bias by modeling additional modalities and applying causal interventions to uncover true causal dependencies. To improve model efficiency, we integrate GCN with the Mamba architecture for accelerated spatio-temporal encoding. Extensive experiments on 4 real-world datasets show that E^2-CSTP significantly outperforms 9 state-of-the-art methods, achieving up to 9.66% improvements in accuracy as well as 17.37%-56.11% reductions in computational overhead.", "date": "2025-05-23", "categories": []}
{"id": "2506.02793", "title": "Doubly-Robust Estimation of Counterfactual Policy Mean Embeddings", "abstract": "Estimating the distribution of outcomes under counterfactual policies is critical for decision-making in domains such as recommendation, advertising, and healthcare. We propose and analyze a novel framework-Counterfactual Policy Mean Embedding (CPME)-that represents the entire counterfactual outcome distribution in a reproducing kernel Hilbert space (RKHS), enabling flexible and nonparametric distributional off-policy evaluation. We introduce both a plug-in estimator and a doubly robust estimator; the latter enjoys improved convergence rates by correcting for bias in both the outcome embedding and propensity models. Building on this, we develop a doubly robust kernel test statistic for hypothesis testing, which achieves asymptotic normality and thus enables computationally efficient testing and straightforward construction of confidence intervals. Our framework also supports sampling from the counterfactual distribution. Numerical simulations illustrate the practical benefits of CPME over existing methods.", "date": "2025-06-03", "categories": []}
{"id": "2506.07918", "title": "CausalPFN: Amortized Causal Effect Estimation via In-Context Learning", "abstract": "Causal effect estimation from observational data is fundamental across various applications. However, selecting an appropriate estimator from dozens of specialized methods demands substantial manual effort and domain expertise. We present CausalPFN, a single transformer that amortizes this workflow: trained once on a large library of simulated data-generating processes that satisfy ignorability, it infers causal effects for new observational datasets out of the box. CausalPFN combines ideas from Bayesian causal inference with the large-scale training protocol of prior-fitted networks (PFNs), learning to map raw observations directly to causal effects without any task-specific adjustment. Our approach achieves superior average performance on heterogeneous and average treatment effect estimation benchmarks (IHDP, Lalonde, ACIC). Moreover, it shows competitive performance for real-world policy making on uplift modeling tasks. CausalPFN provides calibrated uncertainty estimates to support reliable decision-making based on Bayesian principles. This ready-to-use model requires no further training or tuning and takes a step toward automated causal inference (https://github.com/vdblm/CausalPFN/).", "date": "2025-06-09", "categories": []}
{"id": "2509.04415", "title": "Interpretable Clustering with Adaptive Heterogeneous Causal Structure Learning in Mixed Observational Data", "abstract": "Understanding causal heterogeneity is essential for scientific discovery in domains such as biology and medicine. However, existing methods lack causal awareness, with insufficient modeling of heterogeneity, confounding, and observational constraints, leading to poor interpretability and difficulty distinguishing true causal heterogeneity from spurious associations. We propose an unsupervised framework, HCL (Interpretable Causal Mechanism-Aware Clustering with Adaptive Heterogeneous Causal Structure Learning), that jointly infers latent clusters and their associated causal structures from mixed-type observational data without requiring temporal ordering, environment labels, interventions or other prior knowledge. HCL relaxes the homogeneity and sufficiency assumptions by introducing an equivalent representation that encodes both structural heterogeneity and confounding. It further develops a bi-directional iterative strategy to alternately refine causal clustering and structure learning, along with a self-supervised regularization that balance cross-cluster universality and specificity. Together, these components enable convergence toward interpretable, heterogeneous causal patterns. Theoretically, we show identifiability of heterogeneous causal structures under mild conditions. Empirically, HCL achieves superior performance in both clustering and structure learning tasks, and recovers biologically meaningful mechanisms in real-world single-cell perturbation data, demonstrating its utility for discovering interpretable, mechanism-level causal heterogeneity.", "date": "2025-09-04", "categories": []}
{"id": "2510.23942", "title": "Decentralized Causal Discovery using Judo Calculus", "abstract": "We describe a theory and implementation of an intuitionistic decentralized framework for causal discovery using judo calculus, which is formally defined as j-stable causal inference using j-do-calculus in a topos of sheaves. In real-world applications -- from biology to medicine and social science -- causal effects depend on regime (age, country, dose, genotype, or lab protocol). Our proposed judo calculus formalizes this context dependence formally as local truth: a causal claim is proven true on a cover of regimes, not everywhere at once. The Lawvere-Tierney modal operator j chooses which regimes are relevant; j-stability means the claim holds constructively and consistently across that family. We describe an algorithmic and implementation framework for judo calculus, combining it with standard score-based, constraint-based, and gradient-based causal discovery methods. We describe experimental results on a range of domains, from synthetic to real-world datasets from biology and economics. Our experimental results show the computational efficiency gained by the decentralized nature of sheaf-theoretic causal discovery, as well as improved performance over classical causal discovery methods.", "date": "2025-10-27", "categories": []}
{"id": "2510.24433", "title": "Nearest Neighbor Matching as Least Squares Density Ratio Estimation and Riesz Regression", "abstract": "This study proves that Nearest Neighbor (NN) matching can be interpreted as an instance of Riesz regression for automatic debiased machine learning. Lin et al. (2023) shows that NN matching is an instance of density-ratio estimation with their new density-ratio estimator. Chernozhukov et al. (2024) develops Riesz regression for automatic debiased machine learning, which directly estimates the Riesz representer (or equivalently, the bias-correction term) by minimizing the mean squared error. In this study, we first prove that the density-ratio estimation method proposed in Lin et al. (2023) is essentially equivalent to Least-Squares Importance Fitting (LSIF) proposed in Kanamori et al. (2009) for direct density-ratio estimation. Furthermore, we derive Riesz regression using the LSIF framework. Based on these results, we derive NN matching from Riesz regression. This study is based on our work Kato (2025a) and Kato (2025b).", "date": "2025-10-28", "categories": []}
{"id": "2411.16315", "title": "Local Learning for Covariate Selection in Nonparametric Causal Effect Estimation with Latent Variables", "abstract": "Estimating causal effects from nonexperimental data is a fundamental problem in many fields of science. A key component of this task is selecting an appropriate set of covariates for confounding adjustment to avoid bias. Most existing methods for covariate selection often assume the absence of latent variables and rely on learning the global network structure among variables. However, identifying the global structure can be unnecessary and inefficient, especially when our primary interest lies in estimating the effect of a treatment variable on an outcome variable. To address this limitation, we propose a novel local learning approach for covariate selection in nonparametric causal effect estimation, which accounts for the presence of latent variables. Our approach leverages testable independence and dependence relationships among observed variables to identify a valid adjustment set for a target causal relationship, ensuring both soundness and completeness under standard assumptions. We validate the effectiveness of our algorithm through extensive experiments on both synthetic and real-world data.", "date": "2024-11-25", "categories": []}
{"id": "2501.07227", "title": "MECD+: Unlocking Event-Level Causal Graph Discovery for Video Reasoning", "abstract": "Video causal reasoning aims to achieve a high-level understanding of videos from a causal perspective. However, it exhibits limitations in its scope, primarily executed in a question-answering paradigm and focusing on brief video segments containing isolated events and basic causal relations, lacking comprehensive and structured causality analysis for videos with multiple interconnected events. To fill this gap, we introduce a new task and dataset, Multi-Event Causal Discovery (MECD). It aims to uncover the causal relations between events distributed chronologically across long videos. Given visual segments and textual descriptions of events, MECD identifies the causal associations between these events to derive a comprehensive and structured event-level video causal graph explaining why and how the result event occurred. To address the challenges of MECD, we devise a novel framework inspired by the Granger Causality method, incorporating an efficient mask-based event prediction model to perform an Event Granger Test. It estimates causality by comparing the predicted result event when premise events are masked versus unmasked. Furthermore, we integrate causal inference techniques such as front-door adjustment and counterfactual inference to mitigate challenges in MECD like causality confounding and illusory causality. Additionally, context chain reasoning is introduced to conduct more robust and generalized reasoning. Experiments validate the effectiveness of our framework in reasoning complete causal relations, outperforming GPT-4o and VideoChat2 by 5.77% and 2.70%, respectively. Further experiments demonstrate that causal relation graphs can also contribute to downstream video understanding tasks such as video question answering and video event prediction.", "date": "2025-01-13", "categories": []}
{"id": "2502.04262", "title": "Efficient Randomized Experiments Using Foundation Models", "abstract": "Randomized experiments are the preferred approach for evaluating the effects of interventions, but they are costly and often yield estimates with substantial uncertainty. On the other hand, in silico experiments leveraging foundation models offer a cost-effective alternative that can potentially attain higher statistical precision. However, the benefits of in silico experiments come with a significant risk: statistical inferences are not valid if the models fail to accurately predict experimental responses to interventions. In this paper, we propose a novel approach that integrates the predictions from multiple foundation models with experimental data while preserving valid statistical inference. Our estimator is consistent and asymptotically normal, with asymptotic variance no larger than the standard estimator based on experimental data alone. Importantly, these statistical properties hold even when model predictions are arbitrarily biased. Empirical results across several randomized experiments show that our estimator offers substantial precision gains, equivalent to a reduction of up to 20% in the sample size needed to match the same precision as the standard estimator based on experimental data alone.", "date": "2025-02-06", "categories": []}
{"id": "2506.23068", "title": "Curious Causality-Seeking Agents Learn Meta Causal World", "abstract": "When building a world model, a common assumption is that the environment has a single, unchanging underlying causal rule, like applying Newton's laws to every situation. In reality, what appears as a drifting causal mechanism is often the manifestation of a fixed underlying mechanism seen through a narrow observational window. This brings about a problem that, when building a world model, even subtle shifts in policy or environment states can alter the very observed causal mechanisms. In this work, we introduce the \\textbf{Meta-Causal Graph} as world models, a minimal unified representation that efficiently encodes the transformation rules governing how causal structures shift across different latent world states. A single Meta-Causal Graph is composed of multiple causal subgraphs, each triggered by meta state, which is in the latent state space. Building on this representation, we introduce a \\textbf{Causality-Seeking Agent} whose objectives are to (1) identify the meta states that trigger each subgraph, (2) discover the corresponding causal relationships by agent curiosity-driven intervention policy, and (3) iteratively refine the Meta-Causal Graph through ongoing curiosity-driven exploration and agent experiences. Experiments on both synthetic tasks and a challenging robot arm manipulation task demonstrate that our method robustly captures shifts in causal dynamics and generalizes effectively to previously unseen contexts.", "date": "2025-06-29", "categories": []}
{"id": "2508.12650", "title": "Score-informed Neural Operator for Enhancing Ordering-based Causal Discovery", "abstract": "Ordering-based approaches to causal discovery identify topological orders of causal graphs, providing scalable alternatives to combinatorial search methods. Under the Additive Noise Model (ANM) assumption, recent causal ordering methods based on score matching require an accurate estimation of the Hessian diagonal of the log-densities. In this paper, we aim to improve the approximation of the Hessian diagonal of the log-densities, thereby enhancing the performance of ordering-based causal discovery algorithms. Existing approaches that rely on Stein gradient estimators are computationally expensive and memory-intensive, while diffusion-model-based methods remain unstable due to the second-order derivatives of score models. To alleviate these problems, we propose Score-informed Neural Operator (SciNO), a probabilistic generative model in smooth function spaces designed to stably approximate the Hessian diagonal and to preserve structural information during the score modeling. Empirical results show that SciNO reduces order divergence by 42.7% on synthetic graphs and by 31.5% on real-world datasets on average compared to DiffAN, while maintaining memory efficiency and scalability. Furthermore, we propose a probabilistic control algorithm for causal reasoning with autoregressive models that integrates SciNO's probability estimates with autoregressive model priors, enabling reliable data-driven causal ordering informed by semantic information. Consequently, the proposed method enhances causal reasoning abilities of LLMs without additional fine-tuning or prompt engineering.", "date": "2025-08-18", "categories": []}
{"id": "2509.21658", "title": "Differentiable Structure Learning and Causal Discovery for General Binary Data", "abstract": "Existing methods for differentiable structure learning in discrete data typically assume that the data are generated from specific structural equation models. However, these assumptions may not align with the true data-generating process, which limits the general applicability of such methods. Furthermore, current approaches often ignore the complex dependence structure inherent in discrete data and consider only linear effects. We propose a differentiable structure learning framework that is capable of capturing arbitrary dependencies among discrete variables. We show that although general discrete models are unidentifiable from purely observational data, it is possible to characterize the complete set of compatible parameters and structures. Additionally, we establish identifiability up to Markov equivalence under mild assumptions. We formulate the learning problem as a single differentiable optimization task in the most general form, thereby avoiding the unrealistic simplifications adopted by previous methods. Empirical results demonstrate that our approach effectively captures complex relationships in discrete data.", "date": "2025-09-25", "categories": []}
{"id": "2510.21889", "title": "Bridging Prediction and Attribution: Identifying Forward and Backward Causal Influence Ranges Using Assimilative Causal Inference", "abstract": "Causal inference identifies cause-and-effect relationships between variables. While traditional approaches rely on data to reveal causal links, a recently developed method, assimilative causal inference (ACI), integrates observations with dynamical models. It utilizes Bayesian data assimilation to trace causes back from observed effects by quantifying the reduction in uncertainty. ACI advances the detection of instantaneous causal relationships and the intermittent reversal of causal roles over time. Beyond identifying causal connections, an equally important challenge is determining the associated causal influence range (CIR), indicating when causal influences emerged and for how long they persist. In this paper, ACI is employed to develop mathematically rigorous formulations of both forward and backward CIRs at each time. The forward CIR quantifies the temporal impact of a cause, while the backward CIR traces the onset of triggers for an observed effect, thus characterizing causal predictability and attribution of outcomes at each transient phase, respectively. Objective and robust metrics for both CIRs are introduced, eliminating the need for empirical thresholds. Computationally efficient approximation algorithms to compute CIRs are developed, which facilitate the use of closed-form expressions for a broad class of nonlinear dynamical systems. Numerical simulations demonstrate how this forward and backward CIR framework provides new possibilities for probing complex dynamical systems. It advances the study of bifurcation-driven and noise-induced tipping points in Earth systems, investigates the impact from resolving the interfering variables when determining the influence ranges, and elucidates atmospheric blocking mechanisms in the equatorial region. These results have direct implications for science, policy, and decision-making.", "date": "2025-10-24", "categories": []}
{"id": "2510.22050", "title": "Towards Error-Centric Intelligence II: Energy-Structured Causal Models", "abstract": "Contemporary machine learning optimizes for predictive accuracy, yet systems that achieve state of the art performance remain causally opaque: their internal representations provide no principled handle for intervention. We can retrain such models, but we cannot surgically edit specific mechanisms while holding others fixed, because learned latent variables lack causal semantics. We argue for a conceptual reorientation: intelligence is the ability to build and refine explanations, falsifiable claims about manipulable structure that specify what changes and what remains invariant under intervention. Explanations subsume prediction but demand more: causal commitments that can be independently tested and corrected at the level of mechanisms. We introduce computational explanations, mappings from observations to intervention ready causal accounts. We instantiate these explanations with Energy Structured Causal Models (ESCMs), in which mechanisms are expressed as constraints (energy functions or vector fields) rather than explicit input output maps, and interventions act by local surgery on those constraints. This shift makes internal structure manipulable at the level where explanations live: which relations must hold, which can change, and what follows when they do. We provide concrete instantiations of the structural-causal principles LAP and ICM in the ESCM context, and also argue that empirical risk minimization systematically produces fractured, entangled representations, a failure we analyze as gauge ambiguity in encoder energy pairs. Finally, we show that under mild conditions, ESCMs recover standard SCM semantics. Building on Part I's principles (LAP, ICM, CAP) and its definition of intelligence as explanation-building under criticism, this paper offers a formal language for causal reasoning in systems that aspire to understand, not merely to predict.", "date": "2025-10-24", "categories": []}
{"id": "2510.22298", "title": "MetaCaDI: A Meta-Learning Framework for Scalable Causal Discovery with Unknown Interventions", "abstract": "Uncovering the underlying causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs and involve unknown interventions. We introduce MetaCaDI, the first framework to cast the joint discovery of a causal graph and unknown interventions as a meta-learning problem. MetaCaDI is a Bayesian framework that learns a shared causal graph structure across multiple experiments and is optimized to rapidly adapt to new, few-shot intervention target prediction tasks. A key innovation is our model's analytical adaptation, which uses a closed-form solution to bypass expensive and potentially unstable gradient-based bilevel optimization. Extensive experiments on synthetic and complex gene expression data demonstrate that MetaCaDI significantly outperforms state-of-the-art methods. It excels at both causal graph recovery and identifying intervention targets from as few as 10 data instances, proving its robustness in data-scarce scenarios.", "date": "2025-10-25", "categories": []}
{"id": "2510.22567", "title": "Semi-Supervised Learning under General Causal Models", "abstract": "Semi-supervised learning (SSL) aims to train a machine learning model using both labelled and unlabelled data. While the unlabelled data have been used in various ways to improve the prediction accuracy, the reason why unlabelled data could help is not fully understood. One interesting and promising direction is to understand SSL from a causal perspective. In light of the independent causal mechanisms principle, the unlabelled data can be helpful when the label causes the features but not vice versa. However, the causal relations between the features and labels can be complex in real world applications. In this paper, we propose a SSL framework that works with general causal models in which the variables have flexible causal relations. More specifically, we explore the causal graph structures and design corresponding causal generative models which can be learned with the help of unlabelled data. The learned causal generative model can generate synthetic labelled data for training a more accurate predictive model. We verify the effectiveness of our proposed method by empirical studies on both simulated and real data.", "date": "2025-10-26", "categories": []}
{"id": "2510.22711", "title": "Identification of Causal Direction under an Arbitrary Number of Latent Confounders", "abstract": "Recovering causal structure in the presence of latent variables is an important but challenging task. While many methods have been proposed to handle it, most of them require strict and/or untestable assumptions on the causal structure. In real-world scenarios, observed variables may be affected by multiple latent variables simultaneously, which, generally speaking, cannot be handled by these methods. In this paper, we consider the linear, non-Gaussian case, and make use of the joint higher-order cumulant matrix of the observed variables constructed in a specific way. We show that, surprisingly, causal asymmetry between two observed variables can be directly seen from the rank deficiency properties of such higher-order cumulant matrices, even in the presence of an arbitrary number of latent confounders. Identifiability results are established, and the corresponding identification methods do not even involve iterative procedures. Experimental results demonstrate the effectiveness and asymptotic correctness of our proposed method.", "date": "2025-10-26", "categories": []}
{"id": "2510.23424", "title": "Causal Deep Q Network", "abstract": "Deep Q Networks (DQN) have shown remarkable success in various reinforcement learning tasks. However, their reliance on associative learning often leads to the acquisition of spurious correlations, hindering their problem-solving capabilities. In this paper, we introduce a novel approach to integrate causal principles into DQNs, leveraging the PEACE (Probabilistic Easy vAriational Causal Effect) formula for estimating causal effects. By incorporating causal reasoning during training, our proposed framework enhances the DQN's understanding of the underlying causal structure of the environment, thereby mitigating the influence of confounding factors and spurious correlations. We demonstrate that integrating DQNs with causal capabilities significantly enhances their problem-solving capabilities without compromising performance. Experimental results on standard benchmark environments showcase that our approach outperforms conventional DQNs, highlighting the effectiveness of causal reasoning in reinforcement learning. Overall, our work presents a promising avenue for advancing the capabilities of deep reinforcement learning agents through principled causal inference.", "date": "2025-10-27", "categories": []}
{"id": "2407.07612", "title": "Teaching Transformers Causal Reasoning through Axiomatic Training", "abstract": "For text-based AI systems to interact in the real world, causal reasoning is an essential skill. Since active interventions are costly, we study to what extent a system can learn causal reasoning from symbolic demonstrations of causal axioms. Specifically, we present an axiomatic training method where the system learns from multiple demonstrations of a causal axiom (or rule), rather than incorporating the axiom as an inductive bias or inferring it from data values. A key question is whether the system would learn to generalize from the axiom demonstrations to more complex scenarios. Our results, based on applying axiomatic training to learn the transitivity axiom and d-separation rule, indicate that such generalization is possible. To avoid data contamination issues, we start with a 67 million parameter transformer model and train it from scratch. On both tasks, we find that a model trained on linear causal chains (along with some noisy variations) can generalize well to complex graphs, including longer causal chains, causal chains with reversed order, and graphs with branching.To handle diverse text inputs, the same method is extended to finetune language models. Finetuning Llama-3-8B-Instruct model on our axiomatic data leads to significant gains on causal benchmarks such as Corr2Cause and CLEAR, in some cases providing state-of-the-art performance surpassing GPT-4.", "date": "2024-07-10", "categories": []}
{"id": "2502.06343", "title": "Prediction-Powered Causal Inferences", "abstract": "In many scientific experiments, the data annotating cost constraints the pace for testing novel hypotheses. Yet, modern machine learning pipelines offer a promising solution, provided their predictions yield correct conclusions. We focus on Prediction-Powered Causal Inferences (PPCI), i.e., estimating the treatment effect in an unlabeled target experiment, relying on training data with the same outcome annotated but potentially different treatment or effect modifiers. We first show that conditional calibration guarantees valid PPCI at population level. Then, we introduce a sufficient representation constraint transferring validity across experiments, which we propose to enforce in practice in Deconfounded Empirical Risk Minimization, our new model-agnostic training objective. We validate our method on synthetic and real-world scientific data, solving impossible problem instances for Empirical Risk Minimization even with standard invariance constraints. In particular, for the first time, we achieve valid causal inference on a scientific experiment with complex recording and no human annotations, fine-tuning a foundational model on our similar annotated experiment.", "date": "2025-02-10", "categories": []}
{"id": "2503.15114", "title": "DeCaFlow: A deconfounding causal generative model", "abstract": "We introduce DeCaFlow, a deconfounding causal generative model. Training once per dataset using just observational data and the underlying causal graph, DeCaFlow enables accurate causal inference on continuous variables under the presence of hidden confounders. Specifically, we extend previous results on causal estimation under hidden confounding to show that a single instance of DeCaFlow provides correct estimates for all causal queries identifiable with do-calculus, leveraging proxy variables to adjust for the causal effects when do-calculus alone is insufficient. Moreover, we show that counterfactual queries are identifiable as long as their interventional counterparts are identifiable, and thus are also correctly estimated by DeCaFlow. Our empirical results on diverse settings (including the Ecoli70 dataset, with 3 independent hidden confounders, tens of observed variables and hundreds of causal queries) show that DeCaFlow outperforms existing approaches, while demonstrating its out-of-the-box applicability to any given causal graph. An implementation can be found in https://github.com/aalmodovares/DeCaFlow", "date": "2025-03-19", "categories": []}
{"id": "2506.09891", "title": "Causal Climate Emulation with Bayesian Filtering", "abstract": "Traditional models of climate change use complex systems of coupled equations to simulate physical processes across the Earth system. These simulations are highly computationally expensive, limiting our predictions of climate change and analyses of its causes and effects. Machine learning has the potential to quickly emulate data from climate models, but current approaches are not able to incorporate physically-based causal relationships. Here, we develop an interpretable climate model emulator based on causal representation learning. We derive a novel approach including a Bayesian filter for stable long-term autoregressive emulation. We demonstrate that our emulator learns accurate climate dynamics, and we show the importance of each one of its components on a realistic synthetic dataset and data from two widely deployed climate models.", "date": "2025-06-11", "categories": []}
{"id": "2509.07036", "title": "Methodological Insights into Structural Causal Modelling and Uncertainty-Aware Forecasting for Economic Indicators", "abstract": "This paper presents a methodological approach to financial time series analysis by combining causal discovery and uncertainty-aware forecasting. As a case study, we focus on four key U.S. macroeconomic indicators -- GDP, economic growth, inflation, and unemployment -- and we apply the LPCMCI framework with Gaussian Process Distance Correlation (GPDC) to uncover dynamic causal relationships in quarterly data from 1970 to 2021. Our results reveal a robust unidirectional causal link from economic growth to GDP and highlight the limited connectivity of inflation, suggesting the influence of latent factors. Unemployment exhibits strong autoregressive dependence, motivating its use as a case study for probabilistic forecasting. Leveraging the Chronos framework, a large language model trained for time series, we perform zero-shot predictions on unemployment. This approach delivers accurate forecasts one and two quarters ahead, without requiring task-specific training. Crucially, the model's uncertainty-aware predictions yield 90\\% confidence intervals, enabling effective anomaly detection through statistically principled deviation analysis. This study demonstrates the value of combining causal structure learning with probabilistic language models to inform economic policy and enhance forecasting robustness.", "date": "2025-09-08", "categories": []}
{"id": "2510.20884", "title": "ROPES: Robotic Pose Estimation via Score-Based Causal Representation Learning", "abstract": "Causal representation learning (CRL) has emerged as a powerful unsupervised framework that (i) disentangles the latent generative factors underlying high-dimensional data, and (ii) learns the cause-and-effect interactions among the disentangled variables. Despite extensive recent advances in identifiability and some practical progress, a substantial gap remains between theory and real-world practice. This paper takes a step toward closing that gap by bringing CRL to robotics, a domain that has motivated CRL. Specifically, this paper addresses the well-defined robot pose estimation -- the recovery of position and orientation from raw images -- by introducing Robotic Pose Estimation via Score-Based CRL (ROPES). Being an unsupervised framework, ROPES embodies the essence of interventional CRL by identifying those generative factors that are actuated: images are generated by intrinsic and extrinsic latent factors (e.g., joint angles, arm/limb geometry, lighting, background, and camera configuration) and the objective is to disentangle and recover the controllable latent variables, i.e., those that can be directly manipulated (intervened upon) through actuation. Interventional CRL theory shows that variables that undergo variations via interventions can be identified. In robotics, such interventions arise naturally by commanding actuators of various joints and recording images under varied controls. Empirical evaluations in semi-synthetic manipulator experiments demonstrate that ROPES successfully disentangles latent generative factors with high fidelity with respect to the ground truth. Crucially, this is achieved by leveraging only distributional changes, without using any labeled data. The paper also includes a comparison with a baseline based on a recently proposed semi-supervised framework. This paper concludes by positioning robot pose estimation as a near-practical testbed for CRL.", "date": "2025-10-23", "categories": []}
{"id": "2510.21110", "title": "Confounding Robust Deep Reinforcement Learning: A Causal Approach", "abstract": "A key task in Artificial Intelligence is learning effective policies for controlling agents in unknown environments to optimize performance measures. Off-policy learning methods, like Q-learning, allow learners to make optimal decisions based on past experiences. This paper studies off-policy learning from biased data in complex and high-dimensional domains where \\emph{unobserved confounding} cannot be ruled out a priori. Building on the well-celebrated Deep Q-Network (DQN), we propose a novel deep reinforcement learning algorithm robust to confounding biases in observed data. Specifically, our algorithm attempts to find a safe policy for the worst-case environment compatible with the observations. We apply our method to twelve confounded Atari games, and find that it consistently dominates the standard DQN in all games where the observed input to the behavioral and target policies mismatch and unobserved confounders exist.", "date": "2025-10-24", "categories": []}
{"id": "2510.21181", "title": "Shylock: Causal Discovery in Multivariate Time Series based on Hybrid Constraints", "abstract": "Causal relationship discovery has been drawing increasing attention due to its prevalent application. Existing methods rely on human experience, statistical methods, or graphical criteria methods which are error-prone, stuck at the idealized assumption, and rely on a huge amount of data. And there is also a serious data gap in accessing Multivariate time series(MTS) in many areas, adding difficulty in finding their causal relationship. Existing methods are easy to be over-fitting on them. To fill the gap we mentioned above, in this paper, we propose Shylock, a novel method that can work well in both few-shot and normal MTS to find the causal relationship. Shylock can reduce the number of parameters exponentially by using group dilated convolution and a sharing kernel, but still learn a better representation of variables with time delay. By combing the global constraint and the local constraint, Shylock achieves information sharing among networks to help improve the accuracy. To evaluate the performance of Shylock, we also design a data generation method to generate MTS with time delay. We evaluate it on commonly used benchmarks and generated datasets. Extensive experiments show that Shylock outperforms two existing state-of-art methods on both few-shot and normal MTS. We also developed Tcausal, a library for easy use and deployed it on the EarthDataMiner platform", "date": "2025-10-24", "categories": []}
{"id": "2510.21333", "title": "CausalRec: A CausalBoost Attention Model for Sequential Recommendation", "abstract": "Recent advances in correlation-based sequential recommendation systems have demonstrated substantial success. Specifically, the attention-based model outperforms other RNN-based and Markov chains-based models by capturing both short- and long-term dependencies more effectively. However, solely focusing on item co-occurrences overlooks the underlying motivations behind user behaviors, leading to spurious correlations and potentially inaccurate recommendations. To address this limitation, we present a novel framework that integrates causal attention for sequential recommendation, CausalRec. It incorporates a causal discovery block and a CausalBooster. The causal discovery block learns the causal graph in user behavior sequences, and we provide a theory to guarantee the identifiability of the learned causal graph. The CausalBooster utilizes the discovered causal graph to refine the attention mechanism, prioritizing behaviors with causal significance. Experimental evaluations on real-world datasets indicate that CausalRec outperforms several state-of-the-art methods, with average improvements of 7.21% in Hit Rate (HR) and 8.65% in Normalized Discounted Cumulative Gain (NDCG). To the best of our knowledge, this is the first model to incorporate causality through the attention mechanism in sequential recommendation, demonstrating the value of causality in generating more accurate and reliable recommendations.", "date": "2025-10-24", "categories": []}
{"id": "2510.21427", "title": "Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems", "abstract": "Large-scale networked systems, such as traffic, power, and wireless grids, challenge reinforcement-learning agents with both scale and environment shifts. To address these challenges, we propose GSAC (Generalizable and Scalable Actor-Critic), a framework that couples causal representation learning with meta actor-critic learning to achieve both scalability and domain generalization. Each agent first learns a sparse local causal mask that provably identifies the minimal neighborhood variables influencing its dynamics, yielding exponentially tight approximately compact representations (ACRs) of state and domain factors. These ACRs bound the error of truncating value functions to $\\kappa$-hop neighborhoods, enabling efficient learning on graphs. A meta actor-critic then trains a shared policy across multiple source domains while conditioning on the compact domain factors; at test time, a few trajectories suffice to estimate the new domain factor and deploy the adapted policy. We establish finite-sample guarantees on causal recovery, actor-critic convergence, and adaptation gap, and show that GSAC adapts rapidly and significantly outperforms learning-from-scratch and conventional adaptation baselines.", "date": "2025-10-24", "categories": []}
{"id": "2510.21457", "title": "Estimating Treatment Effects in Networks using Domain Adversarial Training", "abstract": "Estimating heterogeneous treatment effects in network settings is complicated by interference, meaning that the outcome of an instance can be influenced by the treatment status of others. Existing causal machine learning approaches usually assume a known exposure mapping that summarizes how the outcome of a given instance is influenced by others' treatment, a simplification that is often unrealistic. Furthermore, the interaction between homophily -- the tendency of similar instances to connect -- and the treatment assignment mechanism can induce a network-level covariate shift that may lead to inaccurate treatment effect estimates, a phenomenon that has not yet been explicitly studied. To address these challenges, we propose HINet, a novel method that integrates graph neural networks with domain adversarial training. This combination allows estimating treatment effects under unknown exposure mappings while mitigating the impact of (network-level) covariate shift. An extensive empirical evaluation on synthetic and semi-synthetic network datasets demonstrates the effectiveness of our approach.", "date": "2025-10-24", "categories": []}
{"id": "2406.09567", "title": "Causal Post-Processing of Predictive Models", "abstract": "Organizations increasingly rely on predictive models to decide who should be targeted for interventions, such as marketing campaigns, customer retention offers, or medical treatments. Yet these models are usually built to predict outcomes (e.g., likelihood of purchase or churn), not the actual impact of an intervention. As a result, the scores (predicted values) they produce are often imperfect guides for allocating resources. Causal effects can be estimated with randomized experiments, but experiments are costly, limited in scale, and tied to specific actions. We propose causal post-processing (CPP), a family of techniques that uses limited experimental data to refine the outputs of predictive models, so they better align with causal decision making. The CPP family spans approaches that trade off flexibility against data efficiency, unifying existing methods and motivating new ones. Through simulations and an empirical study in digital advertising, we show that CPP can improve intervention decisions, particularly when predictive models capture a useful but imperfect causal signal. Our results show how organizations can combine predictive modeling with experimental evidence to make more effective and scalable intervention decisions.", "date": "2024-06-13", "categories": []}
{"id": "2506.17065", "title": "Flow based approach for Dynamic Temporal Causal models with non-Gaussian or Heteroscedastic Noises", "abstract": "Understanding causal relationships in multivariate time series is crucial in many scenarios, such as those dealing with financial or neurological data. Many such time series exhibit multiple regimes, i.e., consecutive temporal segments with a priori unknown boundaries, with each regime having its own causal structure. Inferring causal dependencies and regime shifts is critical for analyzing the underlying processes. However, causal structure learning in this setting is challenging due to (1) non-stationarity, i.e., each regime can have its own causal graph and mixing function, and (2) complex noise distributions, which may be nonGaussian or heteroscedastic. Existing causal discovery approaches cannot address these challenges, since generally assume stationarity or Gaussian noise with constant variance. Hence, we introduce FANTOM, a unified framework for causal discovery that handles non-stationary processes along with non-Gaussian and heteroscedastic noises. FANTOM simultaneously infers the number of regimes and their corresponding indices and learns each regime's Directed Acyclic Graph. It uses a Bayesian Expectation Maximization algorithm that maximizes the evidence lower bound of the data log-likelihood. On the theoretical side, we prove, under mild assumptions, that temporal heteroscedastic causal models, introduced in FANTOM's formulation, are identifiable in both stationary and non-stationary settings. In addition, extensive experiments on synthetic and real data show that FANTOM outperforms existing methods.", "date": "2025-06-20", "categories": []}
{"id": "2510.20106", "title": "Competition is the key: A Game Theoretic Causal Discovery Approach", "abstract": "Causal discovery remains a central challenge in machine learning, yet existing methods face a fundamental gap: algorithms like GES and GraN-DAG achieve strong empirical performance but lack finite-sample guarantees, while theoretically principled approaches fail to scale. We close this gap by introducing a game-theoretic reinforcement learning framework for causal discovery, where a DDQN agent directly competes against a strong baseline (GES or GraN-DAG), always warm-starting from the opponent's solution. This design yields three provable guarantees: the learned graph is never worse than the opponent, warm-starting strictly accelerates convergence, and most importantly, with high probability the algorithm selects the true best candidate graph. To the best of our knowledge, our result makes a first-of-its-kind progress in explaining such finite-sample guarantees in causal discovery: on synthetic SEMs (30 nodes), the observed error probability decays with n, tightly matching theory. On real-world benchmarks including Sachs, Asia, Alarm, Child, Hepar2, Dream, and Andes, our method consistently improves upon GES and GraN-DAG while remaining theoretically safe. Remarkably, it scales to large graphs such as Hepar2 (70 nodes), Dream (100 nodes), and Andes (220 nodes). Together, these results establish a new class of RL-based causal discovery algorithms that are simultaneously provably consistent, sample-efficient, and practically scalable, marking a decisive step toward unifying empirical performance with rigorous finite-sample theory.", "date": "2025-10-23", "categories": []}
{"id": "2510.20221", "title": "FinCARE: Financial Causal Analysis with Reasoning and Evidence", "abstract": "Portfolio managers rely on correlation-based analysis and heuristic methods that fail to capture true causal relationships driving performance. We present a hybrid framework that integrates statistical causal discovery algorithms with domain knowledge from two complementary sources: a financial knowledge graph extracted from SEC 10-K filings and large language model reasoning. Our approach systematically enhances three representative causal discovery paradigms, constraint-based (PC), score-based (GES), and continuous optimization (NOTEARS), by encoding knowledge graph constraints algorithmically and leveraging LLM conceptual reasoning for hypothesis generation. Evaluated on a synthetic financial dataset of 500 firms across 18 variables, our KG+LLM-enhanced methods demonstrate consistent improvements across all three algorithms: PC (F1: 0.622 vs. 0.459 baseline, +36%), GES (F1: 0.735 vs. 0.367, +100%), and NOTEARS (F1: 0.759 vs. 0.163, +366%). The framework enables reliable scenario analysis with mean absolute error of 0.003610 for counterfactual predictions and perfect directional accuracy for intervention effects. It also addresses critical limitations of existing methods by grounding statistical discoveries in financial domain expertise while maintaining empirical validation, providing portfolio managers with the causal foundation necessary for proactive risk management and strategic decision-making in dynamic market environments.", "date": "2025-10-23", "categories": []}
{"id": "2303.04201", "title": "DR-VIDAL -- Doubly Robust Variational Information-theoretic Deep Adversarial Learning for Counterfactual Prediction and Treatment Effect Estimation on Real World Data", "abstract": "Determining causal effects of interventions onto outcomes from real-world, observational (non-randomized) data, e.g., treatment repurposing using electronic health records, is challenging due to underlying bias. Causal deep learning has improved over traditional techniques for estimating individualized treatment effects (ITE). We present the Doubly Robust Variational Information-theoretic Deep Adversarial Learning (DR-VIDAL), a novel generative framework that combines two joint models of treatment and outcome, ensuring an unbiased ITE estimation even when one of the two is misspecified. DR-VIDAL integrates: (i) a variational autoencoder (VAE) to factorize confounders into latent variables according to causal assumptions; (ii) an information-theoretic generative adversarial network (Info-GAN) to generate counterfactuals; (iii) a doubly robust block incorporating treatment propensities for outcome predictions. On synthetic and real-world datasets (Infant Health and Development Program, Twin Birth Registry, and National Supported Work Program), DR-VIDAL achieves better performance than other non-generative and generative methods. In conclusion, DR-VIDAL uniquely fuses causal assumptions, VAE, Info-GAN, and doubly robustness into a comprehensive, performant framework. Code is available at: https://github.com/Shantanu48114860/DR-VIDAL-AMIA-22 under MIT license.", "date": "2023-03-07", "categories": []}
{"id": "2406.19195", "title": "Estimating Long-term Heterogeneous Dose-response Curve: Generalization Bound Leveraging Optimal Transport Weights", "abstract": "Long-term treatment effect estimation is a significant but challenging problem in many applications. Existing methods rely on ideal assumptions, such as no unobserved confounders or binary treatment, to estimate long-term average treatment effects. However, in numerous real-world applications, these assumptions could be violated, and average treatment effects are insufficient for personalized decision-making. In this paper, we address a more general problem of estimating long-term Heterogeneous Dose-Response Curve (HDRC) while accounting for unobserved confounders and continuous treatment. Specifically, to remove the unobserved confounders in the long-term observational data, we introduce an optimal transport weighting framework to align the long-term observational data to an auxiliary short-term experimental data. Furthermore, to accurately predict the heterogeneous effects of continuous treatment, we establish a generalization bound on counterfactual prediction error by leveraging the reweighted distribution induced by optimal transport. Finally, we develop a long-term HDRC estimator building upon the above theoretical foundations. Extensive experiments on synthetic and semi-synthetic datasets demonstrate the effectiveness of our approach.", "date": "2024-06-27", "categories": []}
{"id": "2502.18994", "title": "Long-term Causal Inference via Modeling Sequential Latent Confounding", "abstract": "Long-term causal inference is an important but challenging problem across various scientific domains. To solve the latent confounding problem in long-term observational studies, existing methods leverage short-term experimental data. Ghassami et al. propose an approach based on the Conditional Additive Equi-Confounding Bias (CAECB) assumption, which asserts that the confounding bias in the short-term outcome is equal to that in the long-term outcome, so that the long-term confounding bias and the causal effects can be identified. While effective in certain cases, this assumption is limited to scenarios where there is only one short-term outcome with the same scale as the long-term outcome. In this paper, we introduce a novel assumption that extends the CAECB assumption to accommodate temporal short-term outcomes. Our proposed assumption states a functional relationship between sequential confounding biases across temporal short-term outcomes, under which we theoretically establish the identification of long-term causal effects. Based on the identification result, we develop an estimator and conduct a theoretical analysis of its asymptotic properties. Extensive experiments validate our theoretical results and demonstrate the effectiveness of the proposed method.", "date": "2025-02-26", "categories": []}
{"id": "2506.14329", "title": "Adjustment for Confounding using Pre-Trained Representations", "abstract": "There is growing interest in extending average treatment effect (ATE) estimation to incorporate non-tabular data, such as images and text, which may act as sources of confounding. Neglecting these effects risks biased results and flawed scientific conclusions. However, incorporating non-tabular data necessitates sophisticated feature extractors, often in combination with ideas of transfer learning. In this work, we investigate how latent features from pre-trained neural networks can be leveraged to adjust for sources of confounding. We formalize conditions under which these latent features enable valid adjustment and statistical inference in ATE estimation, demonstrating results along the example of double machine learning. We discuss critical challenges inherent to latent feature learning and downstream parameter estimation arising from the high dimensionality and non-identifiability of representations. Common structural assumptions for obtaining fast convergence rates with additive or sparse linear models are shown to be unrealistic for latent features. We argue, however, that neural networks are largely insensitive to these issues. In particular, we show that neural networks can achieve fast convergence rates by adapting to intrinsic notions of sparsity and dimension of the learning problem.", "date": "2025-06-17", "categories": []}
{"id": "2510.19138", "title": "InvarGC: Invariant Granger Causality for Heterogeneous Interventional Time Series under Latent Confounding", "abstract": "Granger causality is widely used for causal structure discovery in complex systems from multivariate time series data. Traditional Granger causality tests based on linear models often fail to detect even mild non-linear causal relationships. Therefore, numerous recent studies have investigated non-linear Granger causality methods, achieving improved performance. However, these methods often rely on two key assumptions: causal sufficiency and known interventional targets. Causal sufficiency assumes the absence of latent confounders, yet their presence can introduce spurious correlations. Moreover, real-world time series data usually come from heterogeneous environments, without prior knowledge of interventions. Therefore, in practice, it is difficult to distinguish intervened environments from non-intervened ones, and even harder to identify which variables or timesteps are affected. To address these challenges, we propose Invariant Granger Causality (InvarGC), which leverages cross-environment heterogeneity to mitigate the effects of latent confounding and to distinguish intervened from non-intervened environments with edge-level granularity, thereby recovering invariant causal relations. In addition, we establish the identifiability under these conditions. Extensive experiments on both synthetic and real-world datasets demonstrate the competitive performance of our approach compared to state-of-the-art methods.", "date": "2025-10-22", "categories": []}
{"id": "2510.19517", "title": "Bi-Level Decision-Focused Causal Learning for Large-Scale Marketing Optimization: Bridging Observational and Experimental Data", "abstract": "Online Internet platforms require sophisticated marketing strategies to optimize user retention and platform revenue -- a classical resource allocation problem. Traditional solutions adopt a two-stage pipeline: machine learning (ML) for predicting individual treatment effects to marketing actions, followed by operations research (OR) optimization for decision-making. This paradigm presents two fundamental technical challenges. First, the prediction-decision misalignment: Conventional ML methods focus solely on prediction accuracy without considering downstream optimization objectives, leading to improved predictive metrics that fail to translate to better decisions. Second, the bias-variance dilemma: Observational data suffers from multiple biases (e.g., selection bias, position bias), while experimental data (e.g., randomized controlled trials), though unbiased, is typically scarce and costly -- resulting in high-variance estimates. We propose Bi-level Decision-Focused Causal Learning (Bi-DFCL) that systematically addresses these challenges. First, we develop an unbiased estimator of OR decision quality using experimental data, which guides ML model training through surrogate loss functions that bridge discrete optimization gradients. Second, we establish a bi-level optimization framework that jointly leverages observational and experimental data, solved via implicit differentiation. This novel formulation enables our unbiased OR estimator to correct learning directions from biased observational data, achieving optimal bias-variance tradeoff. Extensive evaluations on public benchmarks, industrial marketing datasets, and large-scale online A/B tests demonstrate the effectiveness of Bi-DFCL, showing statistically significant improvements over state-of-the-art. Currently, Bi-DFCL has been deployed at Meituan, one of the largest online food delivery platforms in the world.", "date": "2025-10-22", "categories": []}
{"id": "2306.07032", "title": "Mitigating Prior Errors in Causal Structure Learning: A Resilient Approach via Bayesian Networks", "abstract": "Causal structure learning (CSL), a prominent technique for encoding cause-and-effect relationships among variables, through Bayesian Networks (BNs). Although recovering causal structure solely from data is a challenge, the integration of prior knowledge, revealing partial structural truth, can markedly enhance learning quality. However, current methods based on prior knowledge exhibit limited resilience to errors in the prior, with hard constraint methods disregarding priors entirely, and soft constraints accepting priors based on a predetermined confidence level, which may require expert intervention. To address this issue, we propose a strategy resilient to edge-level prior errors for CSL, thereby minimizing human intervention. We classify prior errors into different types and provide their theoretical impact on the Structural Hamming Distance (SHD) under the presumption of sufficient data. Intriguingly, we discover and prove that the strong hazard of prior errors is associated with a unique acyclic closed structure, defined as ``quasi-circle''. Leveraging this insight, a post-hoc strategy is employed to identify the prior errors by its impact on the increment of ``quasi-circles''. Through empirical evaluation on both real and synthetic datasets, we demonstrate our strategy's robustness against prior errors. Specifically, we highlight its substantial ability to resist order-reversed errors while maintaining the majority of correct prior.", "date": "2023-06-12", "categories": []}
{"id": "2410.12539", "title": "Counterfactual Effect Decomposition in Multi-Agent Sequential Decision Making", "abstract": "We address the challenge of explaining counterfactual outcomes in multi-agent Markov decision processes. In particular, we aim to explain the total counterfactual effect of an agent's action on the outcome of a realized scenario through its influence on the environment dynamics and the agents' behavior. To achieve this, we introduce a novel causal explanation formula that decomposes the counterfactual effect by attributing to each agent and state variable a score reflecting their respective contributions to the effect. First, we show that the total counterfactual effect of an agent's action can be decomposed into two components: one measuring the effect that propagates through all subsequent agents' actions and another related to the effect that propagates through the state transitions. Building on recent advancements in causal contribution analysis, we further decompose these two effects as follows. For the former, we consider agent-specific effects -- a causal concept that quantifies the counterfactual effect of an agent's action that propagates through a subset of agents. Based on this notion, we use Shapley value to attribute the effect to individual agents. For the latter, we consider the concept of structure-preserving interventions and attribute the effect to state variables based on their \"intrinsic\" contributions. Through extensive experimentation, we demonstrate the interpretability of our approach in a Gridworld environment with LLM-assisted agents and a sepsis management simulator.", "date": "2024-10-16", "categories": []}
{"id": "2502.01106", "title": "Can We Validate Counterfactual Estimations in the Presence of General Network Interference?", "abstract": "Randomized experiments have become a cornerstone of evidence-based decision-making in contexts ranging from online platforms to public health. However, in experimental settings with network interference, a unit's treatment can influence outcomes of other units, challenging both causal effect estimation and its validation. Classic validation approaches fail as outcomes are only observable under a single treatment scenario and exhibit complex correlation patterns due to interference. To address these challenges, we introduce a framework that facilitates the use of machine learning tools for both estimation and validation in causal inference. Central to our approach is the new distribution-preserving network bootstrap, a theoretically-grounded technique that generates multiple statistically-valid subpopulations from a single experiment's data. This amplification of experimental samples enables our second contribution: a counterfactual cross-validation procedure. This procedure adapts the principles of model validation to the unique constraints of causal settings, providing a rigorous, data-driven method for selecting and evaluating estimators. We extend recent causal message-passing developments by incorporating heterogeneous unit-level characteristics and varying local interactions, ensuring reliable finite-sample performance through non-asymptotic analysis. Additionally, we develop and publicly release a comprehensive benchmark toolbox featuring diverse experimental environments, from networks of interacting AI agents to ride-sharing applications. These environments provide known ground truth values while maintaining realistic complexities, enabling systematic evaluation of causal inference methods. Extensive testing across these environments demonstrates our method's robustness to diverse forms of network interference.", "date": "2025-02-03", "categories": []}
{"id": "2508.04478", "title": "Who cuts emissions, who turns up the heat? causal machine learning estimates of energy efficiency interventions", "abstract": "Reducing domestic energy demand is central to climate mitigation and fuel poverty strategies, yet the impact of energy efficiency interventions is highly heterogeneous. Using a causal machine learning model trained on nationally representative data of the English housing stock, we estimate average and conditional treatment effects of wall insulation on gas consumption, focusing on distributional effects across energy burden subgroups. While interventions reduce gas demand on average (by as much as 19 percent), low energy burden groups achieve substantial savings, whereas those experiencing high energy burdens see little to no reduction. This pattern reflects a behaviourally-driven mechanism: households constrained by high costs-to-income ratios (e.g. more than 0.1) reallocate savings toward improved thermal comfort rather than lowering consumption. Far from wasteful, such responses represent rational adjustments in contexts of prior deprivation, with potential co-benefits for health and well-being. These findings call for a broader evaluation framework that accounts for both climate impacts and the equity implications of domestic energy policy.", "date": "2025-08-06", "categories": []}
{"id": "2510.17944", "title": "Intuitionistic $j$-Do-Calculus in Topos Causal Models", "abstract": "In this paper, we generalize Pearl's do-calculus to an Intuitionistic setting called $j$-stable causal inference inside a topos of sheaves. Our framework is an elaboration of the recently proposed framework of Topos Causal Models (TCMs), where causal interventions are defined as subobjects. We generalize the original setting of TCM using the Lawvere-Tierney topology on a topos, defined by a modal operator $j$ on the subobject classifier $\\Omega$. We introduce $j$-do-calculus, where we replace global truth with local truth defined by Kripke-Joyal semantics, and formalize causal reasoning as structure-preserving morphisms that are stable along $j$-covers. $j$-do-calculus is a sound rule system whose premises and conclusions are formulas of the internal Intuitionistic logic of the causal topos. We define $j$-stability for conditional independences and interventional claims as local truth in the internal logic of the causal topos. We give three inference rules that mirror Pearl's insertion/deletion and action/observation exchange, and we prove soundness in the Kripke-Joyal semantics. A companion paper in preparation will describe how to estimate the required entities from data and instantiate $j$-do with standard discovery procedures (e.g., score-based and constraint-based methods), and will include experimental results on how to (i) form data-driven $j$-covers (via regime/section constructions), (ii) compute chartwise conditional independences after graph surgeries, and (iii) glue them to certify the premises of the $j$-do rules in practice", "date": "2025-10-20", "categories": []}
{"id": "2510.18052", "title": "Measure-Theoretic Anti-Causal Representation Learning", "abstract": "Causal representation learning in the anti-causal setting (labels cause features rather than the reverse) presents unique challenges requiring specialized approaches. We propose Anti-Causal Invariant Abstractions (ACIA), a novel measure-theoretic framework for anti-causal representation learning. ACIA employs a two-level design, low-level representations capture how labels generate observations, while high-level representations learn stable causal patterns across environment-specific variations. ACIA addresses key limitations of existing approaches by accommodating prefect and imperfect interventions through interventional kernels, eliminating dependency on explicit causal structures, handling high-dimensional data effectively, and providing theoretical guarantees for out-of-distribution generalization. Experiments on synthetic and real-world medical datasets demonstrate that ACIA consistently outperforms state-of-the-art methods in both accuracy and invariance metrics. Furthermore, our theoretical results establish tight bounds on performance gaps between training and unseen environments, confirming the efficacy of our approach for robust anti-causal learning.", "date": "2025-10-16", "categories": []}
{"id": "2510.18310", "title": "Towards Identifiability of Hierarchical Temporal Causal Representation Learning", "abstract": "Modeling hierarchical latent dynamics behind time series data is critical for capturing temporal dependencies across multiple levels of abstraction in real-world tasks. However, existing temporal causal representation learning methods fail to capture such dynamics, as they fail to recover the joint distribution of hierarchical latent variables from \\textit{single-timestep observed variables}. Interestingly, we find that the joint distribution of hierarchical latent variables can be uniquely determined using three conditionally independent observations. Building on this insight, we propose a Causally Hierarchical Latent Dynamic (CHiLD) identification framework. Our approach first employs temporal contextual observed variables to identify the joint distribution of multi-layer latent variables. Sequentially, we exploit the natural sparsity of the hierarchical structure among latent variables to identify latent variables within each layer. Guided by the theoretical results, we develop a time series generative model grounded in variational inference. This model incorporates a contextual encoder to reconstruct multi-layer latent variables and normalize flow-based hierarchical prior networks to impose the independent noise condition of hierarchical latent dynamics. Empirical evaluations on both synthetic and real-world datasets validate our theoretical claims and demonstrate the effectiveness of CHiLD in modeling hierarchical latent dynamics.", "date": "2025-10-21", "categories": []}
{"id": "2510.18841", "title": "A Hybrid Enumeration Framework for Optimal Counterfactual Generation in Post-Acute COVID-19 Heart Failure", "abstract": "Counterfactual inference provides a mathematical framework for reasoning about hypothetical outcomes under alternative interventions, bridging causal reasoning and predictive modeling. We present a counterfactual inference framework for individualized risk estimation and intervention analysis, illustrated through a clinical application to post-acute sequelae of COVID-19 (PASC) among patients with pre-existing heart failure (HF). Using longitudinal diagnosis, laboratory, and medication data from a large health-system cohort, we integrate regularized predictive modeling with counterfactual search to identify actionable pathways to PASC-related HF hospital admissions. The framework combines exact enumeration with optimization-based methods, including the Nearest Instance Counterfactual Explanations (NICE) and Multi-Objective Counterfactuals (MOC) algorithms, to efficiently explore high-dimensional intervention spaces. Applied to more than 2700 individuals with confirmed SARS-CoV-2 infection and prior HF, the model achieved strong discriminative performance (AUROC: 0.88, 95% CI: 0.84-0.91) and generated interpretable, patient-specific counterfactuals that quantify how modifying comorbidity patterns or treatment factors could alter predicted outcomes. This work demonstrates how counterfactual reasoning can be formalized as an optimization problem over predictive functions, offering a rigorous, interpretable, and computationally efficient approach to personalized inference in complex biomedical systems.", "date": "2025-10-21", "categories": []}
{"id": "2311.06928", "title": "Attention for Causal Relationship Discovery from Biological Neural Dynamics", "abstract": "This paper explores the potential of the transformer models for learning Granger causality in networks with complex nonlinear dynamics at every node, as in neurobiological and biophysical networks. Our study primarily focuses on a proof-of-concept investigation based on simulated neural dynamics, for which the ground-truth causality is known through the underlying connectivity matrix. For transformer models trained to forecast neuronal population dynamics, we show that the cross attention module effectively captures the causal relationship among neurons, with an accuracy equal or superior to that for the most popular Granger causality analysis method. While we acknowledge that real-world neurobiology data will bring further challenges, including dynamic connectivity and unobserved variability, this research offers an encouraging preliminary glimpse into the utility of the transformer model for causal representation learning in neuroscience.", "date": "2023-11-12", "categories": []}
{"id": "2401.13408", "title": "Toward A Causal Framework for Modeling Perception", "abstract": "Perception occurs when individuals interpret the same information differently. It is a known cognitive phenomenon with implications for bias in human decision-making. Perception, however, remains understudied in machine learning (ML). This is problematic as modern decision flows, whether partially or fully automated by ML applications, always involve human experts. For instance, how might we account for cases in which two experts interpret differently the same deferred instance or explanation from a ML model? Addressing this and similar questions requires first a formulation of perception, particularly, in a manner that integrates with ML-enabled decision flows. In this work, we present a first approach to modeling perception causally. We define perception under causal reasoning using structural causal models (SCMs). Our approach formalizes individual experience as additional causal knowledge that comes with and is used by the expert decision-maker in the form of a SCM. We define two kinds of probabilistic causal perception: structural and parametrical. We showcase our framework through a series of examples of modern decision flows. We also emphasize the importance of addressing perception in fair ML, discussing relevant fairness implications and possible applications.", "date": "2024-01-24", "categories": []}
{"id": "2410.13381", "title": "Learning Counterfactual Distributions via Kernel Nearest Neighbors", "abstract": "Consider a setting with multiple units (e.g., individuals, cohorts, geographic locations) and outcomes (e.g., treatments, times, items), where the goal is to learn a multivariate distribution for each unit-outcome entry, such as the distribution of a user's weekly spend and engagement under a specific mobile app version. A common challenge is the prevalence of missing not at random data, where observations are available only for certain unit-outcome combinations and the observation availability can be correlated with the properties of distributions themselves, i.e., there is unobserved confounding. An additional challenge is that for any observed unit-outcome entry, we only have a finite number of samples from the underlying distribution. We tackle these two challenges by casting the problem into a novel distributional matrix completion framework and introduce a kernel based distributional generalization of nearest neighbors to estimate the underlying distributions. By leveraging maximum mean discrepancies and a suitable factor model on the kernel mean embeddings of the underlying distributions, we establish consistent recovery of the underlying distributions even when data is missing not at random and positivity constraints are violated. Furthermore, we demonstrate that our nearest neighbors approach is robust to heteroscedastic noise, provided we have access to two or more measurements for the observed unit-outcome entries, a robustness not present in prior works on nearest neighbors with single measurements.", "date": "2024-10-17", "categories": []}
{"id": "2505.08750", "title": "HCR-Reasoner: Synergizing Large Language Models and Theory for Human-like Causal Reasoning", "abstract": "Genuine human-like causal reasoning is fundamental for strong artificial intelligence. Humans typically identify whether an event is part of the causal chain first, and then influenced by modulatory factors such as morality, normality, and intention to make the final judgment. These two stages naturally map to the fields of 1) actual causality that provides formalisms for causal chain membership and 2) causal judgment from cognitive science that studies psychological modulators that influence causal selection. However, these two domains have largely been studied in isolation, leaving a gap for a systematic method based on LLMs. Therefore, we introduce HCR-Reasoner, a framework that systematically integrates the theory of actual causality and causal judgment into LLMs for human-like causal reasoning. It simulates humans by using actual causality formalisms to filter for structurally necessary candidate causes and causal judgment factors to determine the psychologically selected cause. For fine-grained evaluation, we introduce HCR-Bench, a challenging benchmark with 1,093 annotated instances with detailed reasoning steps. Results show HCR-Reasoner consistently and significantly improves LLMs' causal alignment with humans, and that explicitly integrating theory-guided reasoning into LLMs is highly effective for achieving faithful human-like causal reasoning.", "date": "2025-05-13", "categories": []}
{"id": "2510.16382", "title": "Humanoid-inspired Causal Representation Learning for Domain Generalization", "abstract": "This paper proposes the Humanoid-inspired Structural Causal Model (HSCM), a novel causal framework inspired by human intelligence, designed to overcome the limitations of conventional domain generalization models. Unlike approaches that rely on statistics to capture data-label dependencies and learn distortion-invariant representations, HSCM replicates the hierarchical processing and multi-level learning of human vision systems, focusing on modeling fine-grained causal mechanisms. By disentangling and reweighting key image attributes such as color, texture, and shape, HSCM enhances generalization across diverse domains, ensuring robust performance and interpretability. Leveraging the flexibility and adaptability of human intelligence, our approach enables more effective transfer and learning in dynamic, complex environments. Through both theoretical and empirical evaluations, we demonstrate that HSCM outperforms existing domain generalization models, providing a more principled method for capturing causal relationships and improving model robustness. The code is available at https://github.com/lambett/HSCM.", "date": "2025-10-18", "categories": []}
{"id": "2510.16419", "title": "A Relative Error-Based Evaluation Framework of Heterogeneous Treatment Effect Estimators", "abstract": "While significant progress has been made in heterogeneous treatment effect (HTE) estimation, the evaluation of HTE estimators remains underdeveloped. In this article, we propose a robust evaluation framework based on relative error, which quantifies performance differences between two HTE estimators. We first derive the key theoretical conditions on the nuisance parameters that are necessary to achieve a robust estimator of relative error. Building on these conditions, we introduce novel loss functions and design a neural network architecture to estimate nuisance parameters and obtain robust estimation of relative error, thereby achieving reliable evaluation of HTE estimators. We provide the large sample properties of the proposed relative error estimator. Furthermore, beyond evaluation, we propose a new learning algorithm for HTE that leverages both the previously HTE estimators and the nuisance parameters learned through our neural network architecture. Extensive experiments demonstrate that our evaluation framework supports reliable comparisons across HTE estimators, and the proposed learning algorithm for HTE exhibits desirable performance.", "date": "2025-10-18", "categories": []}
{"id": "2510.16530", "title": "Realizing LLMs' Causal Potential Requires Science-Grounded, Novel Benchmarks", "abstract": "Recent claims of strong performance by Large Language Models (LLMs) on causal discovery are undermined by a key flaw: many evaluations rely on benchmarks likely included in pretraining corpora. Thus, apparent success suggests that LLM-only methods, which ignore observational data, outperform classical statistical approaches. We challenge this narrative by asking: Do LLMs truly reason about causal structure, and how can we measure it without memorization concerns? Can they be trusted for real-world scientific discovery? We argue that realizing LLMs' potential for causal analysis requires two shifts: (P.1) developing robust evaluation protocols based on recent scientific studies to guard against dataset leakage, and (P.2) designing hybrid methods that combine LLM-derived knowledge with data-driven statistics. To address P.1, we encourage evaluating discovery methods on novel, real-world scientific studies. We outline a practical recipe for extracting causal graphs from recent publications released after an LLM's training cutoff, ensuring relevance and preventing memorization while capturing both established and novel relations. Compared to benchmarks like BNLearn, where LLMs achieve near-perfect accuracy, they perform far worse on our curated graphs, underscoring the need for statistical grounding. Supporting P.2, we show that using LLM predictions as priors for the classical PC algorithm significantly improves accuracy over both LLM-only and purely statistical methods. We call on the community to adopt science-grounded, leakage-resistant benchmarks and invest in hybrid causal discovery methods suited to real-world inquiry.", "date": "2025-10-18", "categories": []}
{"id": "2510.16656", "title": "Simulation-free Structure Learning for Stochastic Dynamics", "abstract": "Modeling dynamical systems and unraveling their underlying causal relationships is central to many domains in the natural sciences. Various physical systems, such as those arising in cell biology, are inherently high-dimensional and stochastic in nature, and admit only partial, noisy state measurements. This poses a significant challenge for addressing the problems of modeling the underlying dynamics and inferring the network structure of these systems. Existing methods are typically tailored either for structure learning or modeling dynamics at the population level, but are limited in their ability to address both problems together. In this work, we address both problems simultaneously: we present StructureFlow, a novel and principled simulation-free approach for jointly learning the structure and stochastic population dynamics of physical systems. We showcase the utility of StructureFlow for the tasks of structure learning from interventions and dynamical (trajectory) inference of conditional population dynamics. We empirically evaluate our approach on high-dimensional synthetic systems, a set of biologically plausible simulated systems, and an experimental single-cell dataset. We show that StructureFlow can learn the structure of underlying systems while simultaneously modeling their conditional population dynamics -- a key step toward the mechanistic understanding of systems behavior.", "date": "2025-10-18", "categories": []}
{"id": "2510.17318", "title": "CausalMamba: Scalable Conditional State Space Models for Neural Causal Inference", "abstract": "We introduce CausalMamba, a scalable framework that addresses fundamental limitations in fMRI-based causal inference: the ill-posed nature of inferring neural causality from hemodynamically distorted BOLD signals and the computational intractability of existing methods like Dynamic Causal Modeling (DCM). Our approach decomposes this complex inverse problem into two tractable stages: BOLD deconvolution to recover latent neural activity, followed by causal graph inference using a novel Conditional Mamba architecture. On simulated data, CausalMamba achieves 37% higher accuracy than DCM. Critically, when applied to real task fMRI data, our method recovers well-established neural pathways with 88% fidelity, whereas conventional approaches fail to identify these canonical circuits in over 99% of subjects. Furthermore, our network analysis of working memory data reveals that the brain strategically shifts its primary causal hub-recruiting executive or salience networks depending on the stimulus-a sophisticated reconfiguration that remains undetected by traditional methods. This work provides neuroscientists with a practical tool for large-scale causal inference that captures both fundamental circuit motifs and flexible network dynamics underlying cognitive function.", "date": "2025-10-20", "categories": []}
{"id": "2501.12706", "title": "REX: Causal discovery based on machine learning and explainability techniques", "abstract": "Explainable Artificial Intelligence (XAI) techniques hold significant potential for enhancing the causal discovery process, which is crucial for understanding complex systems in areas like healthcare, economics, and artificial intelligence. However, no causal discovery methods currently incorporate explainability into their models to derive the causal graphs. Thus, in this paper we explore this innovative approach, as it offers substantial potential and represents a promising new direction worth investigating. Specifically, we introduce ReX, a causal discovery method that leverages machine learning (ML) models coupled with explainability techniques, specifically Shapley values, to identify and interpret significant causal relationships among variables. Comparative evaluations on synthetic datasets comprising continuous tabular data reveal that ReX outperforms state-of-the-art causal discovery methods across diverse data generation processes, including non-linear and additive noise models. Moreover, ReX was tested on the Sachs single-cell protein-signaling dataset, achieving a precision of 0.952 and recovering key causal relationships with no incorrect edges. Taking together, these results showcase ReX's effectiveness in accurately recovering true causal structures while minimizing false positive predictions, its robustness across diverse datasets, and its applicability to real-world problems. By combining ML and explainability techniques with causal discovery, ReX bridges the gap between predictive modeling and causal inference, offering an effective tool for understanding complex causal structures.", "date": "2025-01-22", "categories": []}
{"id": "2510.14049", "title": "CausalVerse: Benchmarking Causal Representation Learning with Configurable High-Fidelity Simulations", "abstract": "Causal Representation Learning (CRL) aims to uncover the data-generating process and identify the underlying causal variables and relations, whose evaluation remains inherently challenging due to the requirement of known ground-truth causal variables and causal structure. Existing evaluations often rely on either simplistic synthetic datasets or downstream performance on real-world tasks, generally suffering a dilemma between realism and evaluative precision. In this paper, we introduce a new benchmark for CRL using high-fidelity simulated visual data that retains both realistic visual complexity and, more importantly, access to ground-truth causal generating processes. The dataset comprises around 200 thousand images and 3 million video frames across 24 sub-scenes in four domains: static image generation, dynamic physical simulations, robotic manipulations, and traffic situation analysis. These scenarios range from static to dynamic settings, simple to complex structures, and single to multi-agent interactions, offering a comprehensive testbed that hopefully bridges the gap between rigorous evaluation and real-world applicability. In addition, we provide flexible access to the underlying causal structures, allowing users to modify or configure them to align with the required assumptions in CRL, such as available domain labels, temporal dependencies, or intervention histories. Leveraging this benchmark, we evaluated representative CRL methods across diverse paradigms and offered empirical insights to assist practitioners and newcomers in choosing or extending appropriate CRL frameworks to properly address specific types of real problems that can benefit from the CRL perspective. Welcome to visit our: Project page:https://causal-verse.github.io/, Dataset:https://huggingface.co/CausalVerse.", "date": "2025-10-15", "categories": []}
{"id": "2510.15128", "title": "Towards Error Centric Intelligence I, Beyond Observational Learning", "abstract": "We argue that progress toward AGI is theory limited rather than data or scale limited. Building on the critical rationalism of Popper and Deutsch, we challenge the Platonic Representation Hypothesis. Observationally equivalent worlds can diverge under interventions, so observational adequacy alone cannot guarantee interventional competence. We begin by laying foundations, definitions of knowledge, learning, intelligence, counterfactual competence and AGI, and then analyze the limits of observational learning that motivate an error centric shift. We recast the problem as three questions about how explicit and implicit errors evolve under an agent's actions, which errors are unreachable within a fixed hypothesis space, and how conjecture and criticism expand that space. From these questions we propose Causal Mechanics, a mechanisms first program in which hypothesis space change is a first class operation and probabilistic structure is used when useful rather than presumed. We advance structural principles that make error discovery and correction tractable, including a differential Locality and Autonomy Principle for modular interventions, a gauge invariant form of Independent Causal Mechanisms for separability, and the Compositional Autonomy Principle for analogy preservation, together with actionable diagnostics. The aim is a scaffold for systems that can convert unreachable errors into reachable ones and correct them.", "date": "2025-10-16", "categories": []}
{"id": "2510.15265", "title": "Causal Time Series Modeling of Supraglacial Lake Evolution in Greenland under Distribution Shift", "abstract": "Causal modeling offers a principled foundation for uncovering stable, invariant relationships in time-series data, thereby improving robustness and generalization under distribution shifts. Yet its potential is underutilized in spatiotemporal Earth observation, where models often depend on purely correlational features that fail to transfer across heterogeneous domains. We propose RIC-TSC, a regionally-informed causal time-series classification framework that embeds lag-aware causal discovery directly into sequence modeling, enabling both predictive accuracy and scientific interpretability. Using multi-modal satellite and reanalysis data-including Sentinel-1 microwave backscatter, Sentinel-2 and Landsat-8 optical reflectance, and CARRA meteorological variables-we leverage Joint PCMCI+ (J-PCMCI+) to identify region-specific and invariant predictors of supraglacial lake evolution in Greenland. Causal graphs are estimated globally and per basin, with validated predictors and their time lags supplied to lightweight classifiers. On a balanced benchmark of 1000 manually labeled lakes from two contrasting melt seasons (2018-2019), causal models achieve up to 12.59% higher accuracy than correlation-based baselines under out-of-distribution evaluation. These results show that causal discovery is not only a means of feature selection but also a pathway to generalizable and mechanistically grounded models of dynamic Earth surface processes.", "date": "2025-10-17", "categories": []}
{"id": "2510.15458", "title": "Robust Optimization in Causal Models and G-Causal Normalizing Flows", "abstract": "In this paper, we show that interventionally robust optimization problems in causal models are continuous under the $G$-causal Wasserstein distance, but may be discontinuous under the standard Wasserstein distance. This highlights the importance of using generative models that respect the causal structure when augmenting data for such tasks. To this end, we propose a new normalizing flow architecture that satisfies a universal approximation property for causal structural models and can be efficiently trained to minimize the $G$-causal Wasserstein distance. Empirically, we demonstrate that our model outperforms standard (non-causal) generative models in data augmentation for causal regression and mean-variance portfolio optimization in causal factor models.", "date": "2025-10-17", "categories": []}
{"id": "2510.13985", "title": "Do Large Language Models Show Biases in Causal Learning? Insights from Contingency Judgment", "abstract": "Causal learning is the cognitive process of developing the capability of making causal inferences based on available information, often guided by normative principles. This process is prone to errors and biases, such as the illusion of causality, in which people perceive a causal relationship between two variables despite lacking supporting evidence. This cognitive bias has been proposed to underlie many societal problems, including social prejudice, stereotype formation, misinformation, and superstitious thinking. In this work, we examine whether large language models are prone to developing causal illusions when faced with a classic cognitive science paradigm: the contingency judgment task. To investigate this, we constructed a dataset of 1,000 null contingency scenarios (in which the available information is not sufficient to establish a causal relationship between variables) within medical contexts and prompted LLMs to evaluate the effectiveness of potential causes. Our findings show that all evaluated models systematically inferred unwarranted causal relationships, revealing a strong susceptibility to the illusion of causality. While there is ongoing debate about whether LLMs genuinely understand causality or merely reproduce causal language without true comprehension, our findings support the latter hypothesis and raise concerns about the use of language models in domains where accurate causal reasoning is essential for informed decision-making.", "date": "2025-10-15", "categories": []}
{"id": "2510.14488", "title": "From Guess2Graph: When and How Can Unreliable Experts Safely Boost Causal Discovery in Finite Samples?", "abstract": "Causal discovery algorithms often perform poorly with limited samples. While integrating expert knowledge (including from LLMs) as constraints promises to improve performance, guarantees for existing methods require perfect predictions or uncertainty estimates, making them unreliable for practical use. We propose the Guess2Graph (G2G) framework, which uses expert guesses to guide the sequence of statistical tests rather than replacing them. This maintains statistical consistency while enabling performance improvements. We develop two instantiations of G2G: PC-Guess, which augments the PC algorithm, and gPC-Guess, a learning-augmented variant designed to better leverage high-quality expert input. Theoretically, both preserve correctness regardless of expert error, with gPC-Guess provably outperforming its non-augmented counterpart in finite samples when experts are \"better than random.\" Empirically, both show monotonic improvement with expert accuracy, with gPC-Guess achieving significantly stronger gains.", "date": "2025-10-16", "categories": []}
{"id": "2510.14694", "title": "Response to Discussions of \"Causal and Counterfactual Views of Missing Data Models\"", "abstract": "We are grateful to the discussants, Levis and Kennedy [2025], Luo and Geng [2025], Wang and van der Laan [2025], and Yang and Kim [2025], for their thoughtful comments on our paper (Nabi et al., 2025). In this rejoinder, we summarize our main contributions and respond to each discussion in turn.", "date": "2025-10-16", "categories": []}
{"id": "2510.14780", "title": "Causal Discovery for Linear DAGs with Dependent Latent Variables via Higher-order Cumulants", "abstract": "This paper addresses the problem of estimating causal directed acyclic graphs in linear non-Gaussian acyclic models with latent confounders (LvLiNGAM). Existing methods assume mutually independent latent confounders or cannot properly handle models with causal relationships among observed variables.\n  We propose a novel algorithm that identifies causal DAGs in LvLiNGAM, allowing causal structures among latent variables, among observed variables, and between the two. The proposed method leverages higher-order cumulants of observed data to identify the causal structure. Extensive simulations and experiments with real-world data demonstrate the validity and practical utility of the proposed algorithm.", "date": "2025-10-16", "categories": []}
{"id": "2412.11511", "title": "Constructing Confidence Intervals for Average Treatment Effects from Multiple Datasets", "abstract": "Constructing confidence intervals (CIs) for the average treatment effect (ATE) from patient records is crucial to assess the effectiveness and safety of drugs. However, patient records typically come from different hospitals, thus raising the question of how multiple observational datasets can be effectively combined for this purpose. In our paper, we propose a new method that estimates the ATE from multiple observational datasets and provides valid CIs. Our method makes little assumptions about the observational datasets and is thus widely applicable in medical practice. The key idea of our method is that we leverage prediction-powered inferences and thereby essentially `shrink' the CIs so that we offer more precise uncertainty quantification as compared to na\\\"ive approaches. We further prove the unbiasedness of our method and the validity of our CIs. We confirm our theoretical results through various numerical experiments. Finally, we provide an extension of our method for constructing CIs from combinations of experimental and observational datasets.", "date": "2024-12-16", "categories": []}
{"id": "2508.15692", "title": "Effect Identification and Unit Categorization in the Multi-Score Regression Discontinuity Design with Application to LED Manufacturing", "abstract": "RDD (Regression discontinuity design) is a widely used framework for identifying and estimating causal effects at the cutoff of a single running variable. In practice, however, decision-making often involves multiple thresholds and criteria, especially in production systems. Standard MRD (multi-score RDD) methods address this complexity by reducing the problem to a one-dimensional design. This simplification allows existing approaches to be used to identify and estimate causal effects, but it can introduce non-compliance by misclassifying units relative to the original cutoff rules. We develop theoretical tools to detect and reduce \"fuzziness\" when estimating the cutoff effect for units that comply with individual subrules of a multi-rule system. In particular, we propose a formal definition and categorization of unit behavior types under multi-dimensional cutoff rules, extending standard classifications of compliers, alwaystakers, and nevertakers, and incorporating defiers and indecisive units. We further identify conditions under which cutoff effects for compliers can be estimated in multiple dimensions, and establish when identification remains valid after excluding nevertakers and alwaystakers. In addition, we examine how decomposing complex Boolean cutoff rules (such as AND- and OR-type rules) into simpler components affects the classification of units into behavioral types and improves estimation by making it possible to identify and remove non-compliant units more accurately. We validate our framework using both semi-synthetic simulations calibrated to production data and real-world data from opto-electronic semiconductor manufacturing. The empirical results demonstrate that our approach has practical value in refining production policies and reduces estimation variance. This underscores the usefulness of the MRD framework in manufacturing contexts.", "date": "2025-08-21", "categories": []}
{"id": "2206.12532", "title": "Causal Ordering Without Effect Estimation: A Framework for Using Proxies in Treatment Prioritization", "abstract": "Who should we prioritize for treatment when causal effects cannot be estimated? In practice, organizations often rely on predictive proxies: ads are targeted using purchase probabilities, and retention incentives are allocated using churn-risk scores. These models are not causal, but they are often used with the aim of ranking individuals by treatment effects, a task we call causal ordering. We develop a decision-focused framework to reason about this practice. We identify conditions under which proxies recover the correct effect ordering, which hold when a proxy reflects a dominant moderator of treatment effects. We show how these conditions emerge as a useful approximation in discrete choice settings, where the propensity to act without an intervention moderates persuasion. Moreover, we extend beyond this case, demonstrating that proxies capturing a non-dominant moderator can still outperform CATE estimates when they target signals that are easier to estimate precisely. Building on these insights, we introduce diagnostic tools to assess proxy usefulness in practice. Finally, we illustrate the framework in advertising, where a simple predictive proxy outperforms heterogeneous-effect estimation methods.", "date": "2022-06-25", "categories": []}
{"id": "2408.06849", "title": "Causal Agent based on Large Language Model", "abstract": "The large language model (LLM) has achieved significant success across various domains. However, the inherent complexity of causal problems and causal theory poses challenges in accurately describing them in natural language, making it difficult for LLM to comprehend and use them effectively. Causal methods are not easily conveyed through natural language, which hinders LLM's ability to apply them accurately. Additionally, causal datasets are typically tabular, while LLM excels in handling natural language data, creating a structural mismatch that impedes effective reasoning with tabular data. To address these challenges, we have equipped the LLM with causal tools within an agent framework, named the Causal Agent, enabling it to tackle causal problems. The causal agent comprises tools, memory, and reasoning modules. In the tool module, the causal agent calls Python code and uses the encapsulated causal function module to align tabular data with natural language. In the reasoning module, the causal agent performs reasoning through multiple iterations with the tools. In the memory module, the causal agent maintains a dictionary instance where the keys are unique names and the values are causal graphs. To verify the causal ability of the causal agent, we established a Causal Tabular Question Answer (CausalTQA) benchmark consisting of four levels of causal problems: variable level, edge level, causal graph level, and causal effect level. CausalTQA consists of about 1.4K for these four levels questions. Causal agent demonstrates remarkable efficacy on the four-level causal problems, with accuracy rates all above 80\\%. Through verification on the real-world dataset QRData, the causal agent is 6\\% higher than the original SOTA. For further insights and implementation details, our code is accessible via the GitHub repository https://github.com/kairong-han/causal_agent.", "date": "2024-08-13", "categories": []}
{"id": "2412.02869", "title": "Constrained Identifiability of Causal Effects", "abstract": "We study the identification of causal effects in the presence of different types of constraints (e.g., logical constraints) in addition to the causal graph. These constraints impose restrictions on the models (parameterizations) induced by the causal graph, reducing the set of models considered by the identifiability problem. We formalize the notion of constrained identifiability, which takes a set of constraints as another input to the classical definition of identifiability. We then introduce a framework for testing constrained identifiability by employing tractable Arithmetic Circuits (ACs), which enables us to accommodate constraints systematically. We show that this AC-based approach is at least as complete as existing algorithms (e.g., do-calculus) for testing classical identifiability, which only assumes the constraint of strict positivity. We use examples to demonstrate the effectiveness of this AC-based approach by showing that unidentifiable causal effects may become identifiable under different types of constraints.", "date": "2024-12-03", "categories": []}
{"id": "2510.12033", "title": "CausalTrace: A Neurosymbolic Causal Analysis Agent for Smart Manufacturing", "abstract": "Modern manufacturing environments demand not only accurate predictions but also interpretable insights to process anomalies, root causes, and potential interventions. Existing AI systems often function as isolated black boxes, lacking the seamless integration of prediction, explanation, and causal reasoning required for a unified decision-support solution. This fragmentation limits their trustworthiness and practical utility in high-stakes industrial environments. In this work, we present CausalTrace, a neurosymbolic causal analysis module integrated into the SmartPilot industrial CoPilot. CausalTrace performs data-driven causal analysis enriched by industrial ontologies and knowledge graphs, including advanced functions such as causal discovery, counterfactual reasoning, and root cause analysis (RCA). It supports real-time operator interaction and is designed to complement existing agents by offering transparent, explainable decision support. We conducted a comprehensive evaluation of CausalTrace using multiple causal assessment methods and the C3AN framework (i.e. Custom, Compact, Composite AI with Neurosymbolic Integration), which spans principles of robustness, intelligence, and trustworthiness. In an academic rocket assembly testbed, CausalTrace achieved substantial agreement with domain experts (ROUGE-1: 0.91 in ontology QA) and strong RCA performance (MAP@3: 94%, PR@2: 97%, MRR: 0.92, Jaccard: 0.92). It also attained 4.59/5 in the C3AN evaluation, demonstrating precision and reliability for live deployment.", "date": "2025-10-14", "categories": []}
{"id": "2510.12325", "title": "Causal Inspired Multi Modal Recommendation", "abstract": "Multimodal recommender systems enhance personalized recommendations in e-commerce and online advertising by integrating visual, textual, and user-item interaction data. However, existing methods often overlook two critical biases: (i) modal confounding, where latent factors (e.g., brand style or product category) simultaneously drive multiple modalities and influence user preference, leading to spurious feature-preference associations; (ii) interaction bias, where genuine user preferences are mixed with noise from exposure effects and accidental clicks. To address these challenges, we propose a Causal-inspired multimodal Recommendation framework. Specifically, we introduce a dual-channel cross-modal diffusion module to identify hidden modal confounders, utilize back-door adjustment with hierarchical matching and vector-quantized codebooks to block confounding paths, and apply front-door adjustment combined with causal topology reconstruction to build a deconfounded causal subgraph. Extensive experiments on three real-world e-commerce datasets demonstrate that our method significantly outperforms state-of-the-art baselines while maintaining strong interpretability.", "date": "2025-10-14", "categories": []}
{"id": "2510.12503", "title": "The Robustness of Differentiable Causal Discovery in Misspecified Scenarios", "abstract": "Causal discovery aims to learn causal relationships between variables from targeted data, making it a fundamental task in machine learning. However, causal discovery algorithms often rely on unverifiable causal assumptions, which are usually difficult to satisfy in real-world data, thereby limiting the broad application of causal discovery in practical scenarios. Inspired by these considerations, this work extensively benchmarks the empirical performance of various mainstream causal discovery algorithms, which assume i.i.d. data, under eight model assumption violations. Our experimental results show that differentiable causal discovery methods exhibit robustness under the metrics of Structural Hamming Distance and Structural Intervention Distance of the inferred graphs in commonly used challenging scenarios, except for scale variation. We also provide the theoretical explanations for the performance of differentiable causal discovery methods. Finally, our work aims to comprehensively benchmark the performance of recent differentiable causal discovery methods under model assumption violations, and provide the standard for reasonable evaluation of causal discovery, as well as to further promote its application in real-world scenarios.", "date": "2025-10-14", "categories": []}
{"id": "2402.03941", "title": "Discovering and Reasoning of Causality in the Hidden World with Large Language Models", "abstract": "Revealing hidden causal variables alongside the underlying causal mechanisms is essential to the development of science. Despite the progress in the past decades, existing practice in causal discovery (CD) heavily relies on high-quality measured variables, which are usually given by human experts. In fact, the lack of well-defined high-level variables behind unstructured data has been a longstanding roadblock to a broader real-world application of CD. This procedure can naturally benefit from an automated process that can suggest potential hidden variables in the system. Interestingly, Large language models (LLMs) are trained on massive observations of the world and have demonstrated great capability in processing unstructured data. To leverage the power of LLMs, we develop a new framework termed Causal representatiOn AssistanT (COAT) that incorporates the rich world knowledge of LLMs to propose useful measured variables for CD with respect to high-value target variables on their paired unstructured data. Instead of directly inferring causality with LLMs, COAT constructs feedback from intermediate CD results to LLMs to refine the proposed variables. Given the target variable and the paired unstructured data, we first develop COAT-MB that leverages the predictivity of the proposed variables to iteratively uncover the Markov Blanket of the target variable. Built upon COAT-MB, COAT-PAG further extends to uncover a more complete causal graph, i.e., Partial Ancestral Graph, by iterating over the target variables and actively seeking new high-level variables. Moreover, the reliable CD capabilities of COAT also extend the debiased causal inference to unstructured data by discovering an adjustment set. We establish theoretical guarantees for the CD results and verify their efficiency and reliability across realistic benchmarks and real-world case studies.", "date": "2024-02-06", "categories": []}
{"id": "2502.06577", "title": "The Minimal Search Space for Conditional Causal Bandits", "abstract": "Causal knowledge can be used to support decision-making problems. This has been recognized in the causal bandits literature, where a causal (multi-armed) bandit is characterized by a causal graphical model and a target variable. The arms are then interventions on the causal model, and rewards are samples of the target variable. Causal bandits were originally studied with a focus on hard interventions. We focus instead on cases where the arms are conditional interventions, which more accurately model many real-world decision-making problems by allowing the value of the intervened variable to be chosen based on the observed values of other variables. This paper presents a graphical characterization of the minimal set of nodes guaranteed to contain the optimal conditional intervention, which maximizes the expected reward. We then propose an efficient algorithm with a time complexity of $O(|V| + |E|)$ to identify this minimal set of nodes. We prove that the graphical characterization and the proposed algorithm are correct. Finally, we empirically demonstrate that our algorithm significantly prunes the search space and substantially accelerates convergence rates when integrated into standard multi-armed bandit algorithms.", "date": "2025-02-10", "categories": []}
{"id": "2510.09684", "title": "Using LLMs to Directly Guess Conditional Expectations Can Improve Efficiency in Causal Estimation", "abstract": "We propose a simple yet effective use of LLM-powered AI tools to improve causal estimation. In double machine learning, the accuracy of causal estimates of the effect of a treatment on an outcome in the presence of a high-dimensional confounder depends on the performance of estimators of conditional expectation functions. We show that predictions made by generative models trained on historical data can be used to improve the performance of these estimators relative to approaches that solely rely on adjusting for embeddings extracted from these models. We argue that the historical knowledge and reasoning capacities associated with these generative models can help overcome curse-of-dimensionality problems in causal inference problems. We consider a case study using a small dataset of online jewelry auctions, and demonstrate that inclusion of LLM-generated guesses as predictors can improve efficiency in estimation.", "date": "2025-10-09", "categories": []}
{"id": "2510.09846", "title": "CALM: A Causal Analysis Language Model for Tabular Data in Complex Systems with Local Scores, Conditional Independence Tests, and Relation Attributes", "abstract": "Causal discovery from observational data is fundamental to scientific fields like biology, where controlled experiments are often impractical. However, existing methods, including constraint-based (e.g., PC, causalMGM) and score-based approaches (e.g., NOTEARS), face significant limitations. These include an inability to resolve causal direction, restrictions to linear associations, sensitivity to violations of the faithfulness assumption, and inefficiency in searching vast hypothesis spaces. While large language models (LLMs) offer powerful reasoning capabilities, their application is hindered by a fundamental discrepancy: they are designed for text, while most causal data is tabular. To address these challenges, we introduce CALM, a novel causal analysis language model specifically designed for tabular data in complex systems. CALM leverages a Mamba-based architecture to classify causal patterns from pairwise variable relationships. It integrates a comprehensive suite of evidence, including local causal scores, conditional independence tests, and relational attributes, to capture a wide spectrum of linear, nonlinear, and conditional causal mechanisms. Trained on a diverse corpus of synthetic data (from linear, mixed, and nonlinear models) and 10 real-world biological datasets with rigorously validated causal relationships, our model ensures robustness and generalizability. Empirical evaluation demonstrates that CALM significantly outperforms existing methods in both simulation studies, achieving over 91% accuracy, and in a real-world application identifying causal factors in Hepatitis C virus progression. This work represents a significant step towards accurate and generalizable causal discovery by successfully adapting the pattern recognition capabilities of language models to the intricacies of tabular data.", "date": "2025-10-10", "categories": []}
{"id": "2510.11133", "title": "Test-Time Adaptation by Causal Trimming", "abstract": "Test-time adaptation aims to improve model robustness under distribution shifts by adapting models with access to unlabeled target samples. A primary cause of performance degradation under such shifts is the model's reliance on features that lack a direct causal relationship with the prediction target. We introduce Test-time Adaptation by Causal Trimming (TACT), a method that identifies and removes non-causal components from representations for test distributions. TACT applies data augmentations that preserve causal features while varying non-causal ones. By analyzing the changes in the representations using Principal Component Analysis, TACT identifies the highest variance directions associated with non-causal features. It trims the representations by removing their projections on the identified directions, and uses the trimmed representations for the predictions. During adaptation, TACT continuously tracks and refines these directions to get a better estimate of non-causal features. We theoretically analyze the effectiveness of this approach and empirically validate TACT on real-world out-of-distribution benchmarks. TACT consistently outperforms state-of-the-art methods by a significant margin.", "date": "2025-10-13", "categories": []}
{"id": "2407.19638", "title": "On the Reliability of Large Language Models for Causal Discovery", "abstract": "This study investigates the efficacy of Large Language Models (LLMs) in causal discovery. Using newly available open-source LLMs, OLMo and BLOOM, which provide access to their pre-training corpora, we investigate how LLMs address causal discovery through three research questions. We examine: (i) the impact of memorization for accurate causal relation prediction, (ii) the influence of incorrect causal relations in pre-training data, and (iii) the contextual nuances that influence LLMs' understanding of causal relations. Our findings indicate that while LLMs are effective in recognizing causal relations that occur frequently in pre-training data, their ability to generalize to new or rare causal relations is limited. Moreover, the presence of incorrect causal relations significantly undermines the confidence of LLMs in corresponding correct causal relations, and the contextual information critically affects the outcomes of LLMs to discern causal connections between random variables.", "date": "2024-07-29", "categories": []}
{"id": "2502.09981", "title": "Exploring Neural Granger Causality with xLSTMs: Unveiling Temporal Dependencies in Complex Data", "abstract": "Causality in time series can be challenging to determine, especially in the presence of non-linear dependencies. Granger causality helps analyze potential relationships between variables, thereby offering a method to determine whether one time series can predict-Granger cause-future values of another. Although successful, Granger causal methods still struggle with capturing long-range relations between variables. To this end, we leverage the recently successful Extended Long Short-Term Memory (xLSTM) architecture and propose Granger causal xLSTMs (GC-xLSTM). It first enforces sparsity between the time series components by using a novel dynamic loss penalty on the initial projection. Specifically, we adaptively improve the model and identify sparsity candidates. Our joint optimization procedure then ensures that the Granger causal relations are recovered robustly. Our experimental evaluation on six diverse datasets demonstrates the overall efficacy of GC-xLSTM.", "date": "2025-02-14", "categories": []}
{"id": "2505.16620", "title": "CausalDynamics: A large-scale benchmark for structural discovery of dynamical causal models", "abstract": "Causal discovery for dynamical systems poses a major challenge in fields where active interventions are infeasible. Most methods used to investigate these systems and their associated benchmarks are tailored to deterministic, low-dimensional and weakly nonlinear time-series data. To address these limitations, we present CausalDynamics, a large-scale benchmark and extensible data generation framework to advance the structural discovery of dynamical causal models. Our benchmark consists of true causal graphs derived from thousands of both linearly and nonlinearly coupled ordinary and stochastic differential equations as well as two idealized climate models. We perform a comprehensive evaluation of state-of-the-art causal discovery algorithms for graph reconstruction on systems with noisy, confounded, and lagged dynamics. CausalDynamics consists of a plug-and-play, build-your-own coupling workflow that enables the construction of a hierarchy of physical systems. We anticipate that our framework will facilitate the development of robust causal discovery algorithms that are broadly applicable across domains while addressing their unique challenges. We provide a user-friendly implementation and documentation on https://kausable.github.io/CausalDynamics.", "date": "2025-05-22", "categories": []}
{"id": "2505.21641", "title": "PrivATE: Differentially Private Confidence Intervals for Average Treatment Effects", "abstract": "The average treatment effect (ATE) is widely used to evaluate the effectiveness of drugs and other medical interventions. In safety-critical applications like medicine, reliable inferences about the ATE typically require valid uncertainty quantification, such as through confidence intervals (CIs). However, estimating treatment effects in these settings often involves sensitive data that must be kept private. In this work, we present PrivATE, a novel machine learning framework for computing CIs for the ATE under differential privacy. Specifically, we focus on deriving valid privacy-preserving CIs for the ATE from observational data. Our PrivATE framework consists of three steps: (i) estimating the differentially private ATE through output perturbation; (ii) estimating the differentially private variance in a doubly robust manner; and (iii) constructing the CIs while accounting for the uncertainty from both the estimation and privatization steps. Our PrivATE framework is model agnostic, doubly robust, and ensures valid CIs. We demonstrate the effectiveness of our framework using synthetic and real-world medical datasets. To the best of our knowledge, we are the first to derive a general, doubly robust framework for valid CIs of the ATE under ($\\varepsilon,\\delta$)-differential privacy.", "date": "2025-05-27", "categories": []}
{"id": "2510.08758", "title": "A Design-based Solution for Causal Inference with Text: Can a Language Model Be Too Large?", "abstract": "Many social science questions ask how linguistic properties causally affect an audience's attitudes and behaviors. Because text properties are often interlinked (e.g., angry reviews use profane language), we must control for possible latent confounding to isolate causal effects. Recent literature proposes adapting large language models (LLMs) to learn latent representations of text that successfully predict both treatment and the outcome. However, because the treatment is a component of the text, these deep learning methods risk learning representations that actually encode the treatment itself, inducing overlap bias. Rather than depending on post-hoc adjustments, we introduce a new experimental design that handles latent confounding, avoids the overlap issue, and unbiasedly estimates treatment effects. We apply this design in an experiment evaluating the persuasiveness of expressing humility in political communication. Methodologically, we demonstrate that LLM-based methods perform worse than even simple bag-of-words models using our real text and outcomes from our experiment. Substantively, we isolate the causal effect of expressing humility on the perceived persuasiveness of political statements, offering new insights on communication effects for social media platforms, policy makers, and social scientists.", "date": "2025-10-09", "categories": []}
{"id": "2510.08762", "title": "Spatial Deconfounder: Interference-Aware Deconfounding for Spatial Causal Inference", "abstract": "Causal inference in spatial domains faces two intertwined challenges: (1) unmeasured spatial factors, such as weather, air pollution, or mobility, that confound treatment and outcome, and (2) interference from nearby treatments that violate standard no-interference assumptions. While existing methods typically address one by assuming away the other, we show they are deeply connected: interference reveals structure in the latent confounder. Leveraging this insight, we propose the Spatial Deconfounder, a two-stage method that reconstructs a substitute confounder from local treatment vectors using a conditional variational autoencoder (CVAE) with a spatial prior, then estimates causal effects via a flexible outcome model. We show that this approach enables nonparametric identification of both direct and spillover effects under weak assumptions--without requiring multiple treatment types or a known model of the latent field. Empirically, we extend SpaCE, a benchmark suite for spatial confounding, to include treatment interference, and show that the Spatial Deconfounder consistently improves effect estimation across real-world datasets in environmental health and social science. By turning interference into a multi-cause signal, our framework bridges spatial and deconfounding literatures to advance robust causal inference in structured data.", "date": "2025-10-09", "categories": []}
{"id": "2510.09217", "title": "IRIS: An Iterative and Integrated Framework for Verifiable Causal Discovery in the Absence of Tabular Data", "abstract": "Causal discovery is fundamental to scientific research, yet traditional statistical algorithms face significant challenges, including expensive data collection, redundant computation for known relations, and unrealistic assumptions. While recent LLM-based methods excel at identifying commonly known causal relations, they fail to uncover novel relations. We introduce IRIS (Iterative Retrieval and Integrated System for Real-Time Causal Discovery), a novel framework that addresses these limitations. Starting with a set of initial variables, IRIS automatically collects relevant documents, extracts variables, and uncovers causal relations. Our hybrid causal discovery method combines statistical algorithms and LLM-based methods to discover known and novel causal relations. In addition to causal discovery on initial variables, the missing variable proposal component of IRIS identifies and incorporates missing variables to expand the causal graphs. Our approach enables real-time causal discovery from only a set of initial variables without requiring pre-existing datasets.", "date": "2025-10-10", "categories": []}
{"id": "2410.23975", "title": "Average Controlled and Average Natural Micro Direct Effects in Summary Causal Graphs", "abstract": "In this paper, we investigate the identifiability of average controlled direct effects and average natural direct effects in causal systems represented by summary causal graphs, which are abstractions of full causal graphs, often used in dynamic systems where cycles and omitted temporal information complicate causal inference. Unlike in the traditional linear setting, where direct effects are typically easier to identify and estimate, non-parametric direct effects, which are crucial for handling real-world complexities, particularly in epidemiological contexts where relationships between variables (e.g, genetic, environmental, and behavioral factors) are often non-linear, are much harder to define and identify. In particular, we give sufficient conditions for identifying average controlled micro direct effect and average natural micro direct effect from summary causal graphs in the presence of hidden confounding. Furthermore, we show that the conditions given for the average controlled micro direct effect become also necessary in the setting where there is no hidden confounding and where we are only interested in identifiability by adjustment.", "date": "2024-10-31", "categories": []}
{"id": "2510.07501", "title": "Evaluating and Learning Optimal Dynamic Treatment Regimes under Truncation by Death", "abstract": "Truncation by death, a prevalent challenge in critical care, renders traditional dynamic treatment regime (DTR) evaluation inapplicable due to ill-defined potential outcomes. We introduce a principal stratification-based method, focusing on the always-survivor value function. We derive a semiparametrically efficient, multiply robust estimator for multi-stage DTRs, demonstrating its robustness and efficiency. Empirical validation and an application to electronic health records showcase its utility for personalized treatment optimization.", "date": "2025-10-08", "categories": []}
{"id": "2510.08207", "title": "DODO: Causal Structure Learning with Budgeted Interventions", "abstract": "Artificial Intelligence has achieved remarkable advancements in recent years, yet much of its progress relies on identifying increasingly complex correlations. Enabling causality awareness in AI has the potential to enhance its performance by enabling a deeper understanding of the underlying mechanisms of the environment. In this paper, we introduce DODO, an algorithm defining how an Agent can autonomously learn the causal structure of its environment through repeated interventions. We assume a scenario where an Agent interacts with a world governed by a causal Directed Acyclic Graph (DAG), which dictates the system's dynamics but remains hidden from the Agent. The Agent's task is to accurately infer the causal DAG, even in the presence of noise. To achieve this, the Agent performs interventions, leveraging causal inference techniques to analyze the statistical significance of observed changes. Results show better performance for DODO, compared to observational approaches, in all but the most limited resource conditions. DODO is often able to reconstruct with as low as zero errors the structure of the causal graph. In the most challenging configuration, DODO outperforms the best baseline by +0.25 F1 points.", "date": "2025-10-09", "categories": []}
{"id": "2506.07844", "title": "Conditional Local Independence Testing for It\\^o processes with Applications to Dynamic Causal Discovery", "abstract": "Inferring causal relationships from dynamical systems is the central interest of many scientific inquiries. Conditional local independence, which describes whether the evolution of one process is influenced by another process given additional processes, is important for causal learning in such systems. In this paper, we propose a hypothesis test for conditional local independence in It\\^o processes. Our test is grounded in the semimartingale decomposition of the It\\^o process, with which we introduce a stochastic integral process that is a martingale under the null hypothesis. We then apply a test for the martingale property, quantifying potential deviation from local independence. The test statistics is estimated using the optimal filtering equation. We show the consistency of the estimation, thereby establishing the level and power of our test. Numerical verification and a real-world application to causal discovery in brain resting-state fMRIs are conducted.", "date": "2025-06-09", "categories": []}
{"id": "2509.25282", "title": "Toward Causal-Visual Programming: Enhancing Agentic Reasoning in Low-Code Environments", "abstract": "Large language model (LLM) agents are increasingly capable of orchestrating complex tasks in low-code environments. However, these agents often exhibit hallucinations and logical inconsistencies because their inherent reasoning mechanisms rely on probabilistic associations rather than genuine causal understanding. This paper introduces a new programming paradigm: Causal-Visual Programming (CVP), designed to address this fundamental issue by explicitly introducing causal structures into the workflow design. CVP allows users to define a simple \"world model\" for workflow modules through an intuitive low-code interface, effectively creating a Directed Acyclic Graph (DAG) that explicitly defines the causal relationships between modules. This causal graph acts as a crucial constraint during the agent's reasoning process, anchoring its decisions to a user-defined causal structure and significantly reducing logical errors and hallucinations by preventing reliance on spurious correlations. To validate the effectiveness of CVP, we designed a synthetic experiment that simulates a common real-world problem: a distribution shift between the training and test environments. Our results show that a causally anchored model maintained stable accuracy in the face of this shift, whereas a purely associative baseline model that relied on probabilistic correlations experienced a significant performance drop. The primary contributions of this study are: a formal definition of causal structures for workflow modules; the proposal and implementation of a CVP framework that anchors agent reasoning to a user-defined causal graph; and empirical evidence demonstrating the framework's effectiveness in enhancing agent robustness and reducing errors caused by causal confusion in dynamic environments. CVP offers a viable path toward building more interpretable, reliable, and trustworthy AI agents.", "date": "2025-09-29", "categories": []}
{"id": "2510.06995", "title": "Root Cause Analysis of Outliers in Unknown Cyclic Graphs", "abstract": "We study the propagation of outliers in cyclic causal graphs with linear structural equations, tracing them back to one or several \"root cause\" nodes. We show that it is possible to identify a short list of potential root causes provided that the perturbation is sufficiently strong and propagates according to the same structural equations as in the normal mode. This shortlist consists of the true root causes together with those of its parents lying on a cycle with the root cause. Notably, our method does not require prior knowledge of the causal graph.", "date": "2025-10-08", "categories": []}
{"id": "2503.01722", "title": "Learning Exposure Mapping Functions for Inferring Heterogeneous Peer Effects", "abstract": "In causal inference, interference refers to the phenomenon in which the actions of peers in a network can influence an individual's outcome. Peer effect refers to the difference in counterfactual outcomes of an individual for different levels of peer exposure, the extent to which an individual is exposed to the treatments, actions, or behaviors of peers. Estimating peer effects requires deciding how to represent peer exposure. Typically, researchers define an exposure mapping function that aggregates peer treatments and outputs peer exposure. Most existing approaches for defining exposure mapping functions assume peer exposure based on the number or fraction of treated peers. Recent studies have investigated more complex functions of peer exposure which capture that different peers can exert different degrees of influence. However, none of these works have explicitly considered the problem of automatically learning the exposure mapping function. In this work, we focus on learning this function for the purpose of estimating heterogeneous peer effects, where heterogeneity refers to the variation in counterfactual outcomes for the same peer exposure but different individual's contexts. We develop EgoNetGNN, a graph neural network (GNN)-based method, to automatically learn the appropriate exposure mapping function allowing for complex peer influence mechanisms that, in addition to peer treatments, can involve the local neighborhood structure and edge attributes. We show that GNN models that use peer exposure based on the number or fraction of treated peers or learn peer exposure naively face difficulty accounting for such influence mechanisms. Our comprehensive evaluation on synthetic and semi-synthetic network data shows that our method is more robust to different unknown underlying influence mechanisms when estimating heterogeneous peer effects when compared to state-of-the-art baselines.", "date": "2025-03-03", "categories": []}
{"id": "2401.16661", "title": "Generalization of LiNGAM that allows confounding", "abstract": "LiNGAM determines the variable order from cause to effect using additive noise models, but it faces challenges with confounding. Previous methods maintained LiNGAM's fundamental structure while trying to identify and address variables affected by confounding. As a result, these methods required significant computational resources regardless of the presence of confounding, and they did not ensure the detection of all confounding types. In contrast, this paper enhances LiNGAM by introducing LiNGAM-MMI, a method that quantifies the magnitude of confounding using KL divergence and arranges the variables to minimize its impact. This method efficiently achieves a globally optimal variable order through the shortest path problem formulation. LiNGAM-MMI processes data as efficiently as traditional LiNGAM in scenarios without confounding while effectively addressing confounding situations. Our experimental results suggest that LiNGAM-MMI more accurately determines the correct variable order, both in the presence and absence of confounding. The code is in the supplementary file in this link: https://github.com/SkyJoyTianle/ISIT2024.", "date": "2024-01-30", "categories": []}
{"id": "2411.01625", "title": "Counterfactual explainability and analysis of variance", "abstract": "Existing tools for explaining complex models and systems are associational rather than causal and do not provide mechanistic understanding. We propose a new notion called counterfactual explainability for causal attribution that is motivated by the concept of genetic heritability in twin studies. Counterfactual explainability extends methods for global sensitivity analysis (including the functional analysis of variance and Sobol's indices), which assumes independent explanatory variables, to dependent explanations by using a directed acyclic graphs to describe their causal relationship. Therefore, this explanability measure directly incorporates causal mechanisms by construction. Under a comonotonicity assumption, we discuss methods for estimating counterfactual explainability and apply them to a real dataset dataset to explain income inequality by gender, race, and educational attainment.", "date": "2024-11-03", "categories": []}
{"id": "2510.04816", "title": "On Predicting Post-Click Conversion Rate via Counterfactual Inference", "abstract": "Accurately predicting conversion rate (CVR) is essential in various recommendation domains such as online advertising systems and e-commerce. These systems utilize user interaction logs, which consist of exposures, clicks, and conversions. CVR prediction models are typically trained solely based on clicked samples, as conversions can only be determined following clicks. However, the sparsity of clicked instances necessitates the collection of a substantial amount of logs for effective model training. Recent works address this issue by devising frameworks that leverage non-clicked samples. While these frameworks aim to reduce biases caused by the discrepancy between clicked and non-clicked samples, they often rely on heuristics. Against this background, we propose a method to counterfactually generate conversion labels for non-clicked samples by using causality as a guiding principle, attempting to answer the question, \"Would the user have converted if he or she had clicked the recommended item?\" Our approach is named the Entire Space Counterfactual Inference Multi-task Model (ESCIM). We initially train a structural causal model (SCM) of user sequential behaviors and conduct a hypothetical intervention (i.e., click) on non-clicked items to infer counterfactual CVRs. We then introduce several approaches to transform predicted counterfactual CVRs into binary counterfactual conversion labels for the non-clicked samples. Finally, the generated samples are incorporated into the training process. Extensive experiments on public datasets illustrate the superiority of the proposed algorithm. Online A/B testing further empirically validates the effectiveness of our proposed algorithm in real-world scenarios. In addition, we demonstrate the improved performance of the proposed method on latent conversion data, showcasing its robustness and superior generalization capabilities.", "date": "2025-10-06", "categories": []}
{"id": "2510.05033", "title": "Causal Abstractions, Categorically Unified", "abstract": "We present a categorical framework for relating causal models that represent the same system at different levels of abstraction. We define a causal abstraction as natural transformations between appropriate Markov functors, which concisely consolidate desirable properties a causal abstraction should exhibit. Our approach unifies and generalizes previously considered causal abstractions, and we obtain categorical proofs and generalizations of existing results on causal abstractions. Using string diagrammatical tools, we can explicitly describe the graphs that serve as consistent abstractions of a low-level graph under interventions. We discuss how methods from mechanistic interpretability, such as circuit analysis and sparse autoencoders, fit within our categorical framework. We also show how applying do-calculus on a high-level graphical abstraction of an acyclic-directed mixed graph (ADMG), when unobserved confounders are present, gives valid results on the low-level graph, thus generalizing an earlier statement by Anand et al. (2023). We argue that our framework is more suitable for modeling causal abstractions compared to existing categorical frameworks. Finally, we discuss how notions such as $\\tau$-consistency and constructive $\\tau$-abstractions can be recovered with our framework.", "date": "2025-10-06", "categories": []}
{"id": "2302.08854", "title": "Post Reinforcement Learning Inference", "abstract": "We study estimation and inference using data collected by reinforcement learning (RL) algorithms. These algorithms adaptively experiment by interacting with individual units over multiple stages, updating their strategies based on past outcomes. Our goal is to evaluate a counterfactual policy after data collection and estimate structural parameters, such as dynamic treatment effects, that support credit assignment and quantify the impact of early actions on final outcomes. These parameters can often be defined as solutions to moment equations, motivating moment-based estimation methods developed for static data. In RL settings, however, data are often collected adaptively under nonstationary behavior policies. As a result, standard estimators fail to achieve asymptotic normality due to time-varying variance. We propose a weighted generalized method of moments (GMM) approach that uses adaptive weights to stabilize this variance. We characterize weighting schemes that ensure consistency and asymptotic normality of the weighted GMM estimators, enabling valid hypothesis testing and uniform confidence region construction. Key applications include dynamic treatment effect estimation and dynamic off-policy evaluation.", "date": "2023-02-17", "categories": []}
{"id": "2303.01412", "title": "The Challenges of Hyperparameter Tuning for Accurate Causal Effect Estimation", "abstract": "ML is playing an increasingly crucial role in estimating causal effects of treatments on outcomes from observational data. Many ML methods (`causal estimators') have been proposed for this task. All of these methods, as with any ML approach, require extensive hyperparameter tuning. For non-causal predictive tasks, there is a consensus on the choice of tuning metrics (e.g. mean squared error), making it simple to compare models. However, for causal inference tasks, such a consensus is yet to be reached, making any comparison of causal models difficult. On top of that, there is no ideal metric on which to tune causal estimators, so one must rely on proxies. Furthermore, the fact that model selection in causal inference involves multiple components (causal estimator, ML regressor, hyperparameters, metric), complicates the issue even further. In order to evaluate the importance of each component, we perform an extensive empirical study on their combination. Our experimental setup involves many commonly used causal estimators, regressors (`base learners' henceforth) and metrics applied to four well-known causal inference benchmark datasets. Our results show that hyperparameter tuning increased the probability of reaching state-of-the-art performance in average ($65\\% {\\rightarrow} 81\\%$) and individualised ($50\\% {\\rightarrow} 57\\%$) effect estimation with only commonly used estimators. We also show that the performance of standard metrics can be inconsistent across different scenarios. Our findings highlight the need for further research to establish whether metrics uniformly capable of state-of-the-art performance in causal model evaluation can be found.", "date": "2023-03-02", "categories": []}
{"id": "2403.14125", "title": "Learning causal graphs using variable grouping according to ancestral relationship", "abstract": "Several causal discovery algorithms have been proposed. However, when the sample size is small relative to the number of variables, the accuracy of estimating causal graphs using existing methods decreases. And some methods are not feasible when the sample size is smaller than the number of variables. To circumvent these problems, some researchers proposed causal structure learning algorithms using divide-and-conquer approaches. For learning the entire causal graph, the approaches first split variables into several subsets according to the conditional independence relationships among the variables, then apply a conventional causal discovery algorithm to each subset and merge the estimated results. Since the divide-and-conquer approach reduces the number of variables to which a causal structure learning algorithm is applied, it is expected to improve the estimation accuracy of causal graphs, especially when the sample size is small relative to the number of variables and the model is sparse. However, existing methods are either computationally expensive or do not provide sufficient accuracy when the sample size is small. This paper proposes a new algorithm for grouping variables based the ancestral relationships among the variables, under the LiNGAM assumption, where the causal relationships are linear, and the mutually independent noise are distributed as continuous non-Gaussian distributions. We call the proposed algorithm CAG. The time complexity of the ancestor finding in CAG is shown to be cubic to the number of variables. Extensive computer experiments confirm that the proposed method outperforms the original DirectLiNGAM without grouping variables and other divide-and-conquer approaches not only in estimation accuracy but also in computation time when the sample size is small relative to the number of variables and the model is sparse.", "date": "2024-03-21", "categories": []}
{"id": "2502.06398", "title": "Learning Counterfactual Outcomes Under Rank Preservation", "abstract": "Counterfactual inference aims to estimate the counterfactual outcome at the individual level given knowledge of an observed treatment and the factual outcome, with broad applications in fields such as epidemiology, econometrics, and management science. Previous methods rely on a known structural causal model (SCM) or assume the homogeneity of the exogenous variable and strict monotonicity between the outcome and exogenous variable. In this paper, we propose a principled approach for identifying and estimating the counterfactual outcome. We first introduce a simple and intuitive rank preservation assumption to identify the counterfactual outcome without relying on a known structural causal model. Building on this, we propose a novel ideal loss for theoretically unbiased learning of the counterfactual outcome and further develop a kernel-based estimator for its empirical estimation. Our theoretical analysis shows that the rank preservation assumption is not stronger than the homogeneity and strict monotonicity assumptions, and shows that the proposed ideal loss is convex, and the proposed estimator is unbiased. Extensive semi-synthetic and real-world experiments are conducted to demonstrate the effectiveness of the proposed method.", "date": "2025-02-10", "categories": []}
{"id": "2210.11003", "title": "Synthetic Blips: Generalizing Synthetic Controls for Dynamic Treatment Effects", "abstract": "We propose a generalization of the synthetic control and interventions methods to the setting with dynamic treatment effects. We consider the estimation of unit-specific treatment effects from panel data collected under a general treatment sequence. Here, each unit receives multiple treatments sequentially, according to an adaptive policy that depends on a latent, endogenously time-varying confounding state. Under a low-rank latent factor model assumption, we develop an identification strategy for any unit-specific mean outcome under any sequence of interventions. The latent factor model we propose admits linear time-varying and time-invariant dynamical systems as special cases. Our approach can be viewed as an identification strategy for structural nested mean models -- a widely used framework for dynamic treatment effects -- under a low-rank latent factor assumption on the blip effects. Unlike these models, however, it is more permissive in observational settings, thereby broadening its applicability. Our method, which we term synthetic blip effects, is a backwards induction process in which the blip effect of a treatment at each period and for a target unit is recursively expressed as a linear combination of the blip effects of a group of other units that received the designated treatment. This strategy avoids the combinatorial explosion in the number of units that would otherwise be required by a naive application of prior synthetic control and intervention methods in dynamic treatment settings. We provide estimation algorithms that are easy to implement in practice and yield estimators with desirable properties. Using unique Korean firm-level panel data, we demonstrate how the proposed framework can be used to estimate individualized dynamic treatment effects and to derive optimal treatment allocation rules in the context of financial support for exporting firms.", "date": "2022-10-20", "categories": []}
{"id": "2309.14073", "title": "Neural Network Parameter-optimization of Gaussian pmDAGs", "abstract": "Finding the parameters of a latent variable causal model is central to causal inference and causal identification. In this article, we show that existing graphical structures that are used in causal inference are not stable under marginalization of Gaussian Bayesian networks, and present a graphical structure that faithfully represent margins of Gaussian Bayesian networks. We present the first duality between parameter optimization of a latent variable model and training a feed-forward neural network in the parameter space of the assumed family of distributions. Based on this observation, we develop an algorithm for parameter optimization of these graphical structures based on a given observational distribution. Then, we provide conditions for causal effect identifiability in the Gaussian setting. We propose an meta-algorithm that checks whether a causal effect is identifiable or not. Moreover, we lay a grounding for generalizing the duality between a neural network and a causal model from the Gaussian to other distributions.", "date": "2023-09-25", "categories": []}
{"id": "2510.01562", "title": "Large-Scale Bayesian Causal Discovery with Interventional Data", "abstract": "Inferring the causal relationships among a set of variables in the form of a directed acyclic graph (DAG) is an important but notoriously challenging problem. Recently, advancements in high-throughput genomic perturbation screens have inspired development of methods that leverage interventional data to improve model identification. However, existing methods still suffer poor performance on large-scale tasks and fail to quantify uncertainty. Here, we propose Interventional Bayesian Causal Discovery (IBCD), an empirical Bayesian framework for causal discovery with interventional data. Our approach models the likelihood of the matrix of total causal effects, which can be approximated by a matrix normal distribution, rather than the full data matrix. We place a spike-and-slab horseshoe prior on the edges and separately learn data-driven weights for scale-free and Erd\\H{o}s-R\\'enyi structures from observational data, treating each edge as a latent variable to enable uncertainty-aware inference. Through extensive simulation, we show that IBCD achieves superior structure recovery compared to existing baselines. We apply IBCD to CRISPR perturbation (Perturb-seq) data on 521 genes, demonstrating that edge posterior inclusion probabilities enable identification of robust graph structures.", "date": "2025-10-02", "categories": []}
{"id": "2510.02117", "title": "DAG DECORation: Continuous Optimization for Structure Learning under Hidden Confounding", "abstract": "We study structure learning for linear Gaussian SEMs in the presence of latent confounding. Existing continuous methods excel when errors are independent, while deconfounding-first pipelines rely on pervasive factor structure or nonlinearity. We propose \\textsc{DECOR}, a single likelihood-based and fully differentiable estimator that jointly learns a DAG and a correlated noise model. Our theory gives simple sufficient conditions for global parameter identifiability: if the mixed graph is bow free and the noise covariance has a uniform eigenvalue margin, then the map from $(\\B,\\OmegaMat)$ to the observational covariance is injective, so both the directed structure and the noise are uniquely determined. The estimator alternates a smooth-acyclic graph update with a convex noise update and can include a light bow complementarity penalty or a post hoc reconciliation step. On synthetic benchmarks that vary confounding density, graph density, latent rank, and dimension with $n<p$, \\textsc{DECOR} matches or outperforms strong baselines and is especially robust when confounding is non-pervasive, while remaining competitive under pervasiveness.", "date": "2025-10-02", "categories": []}
{"id": "2510.00163", "title": "Partial Identification Approach to Counterfactual Fairness Assessment", "abstract": "The wide adoption of AI decision-making systems in critical domains such as criminal justice, loan approval, and hiring processes has heightened concerns about algorithmic fairness. As we often only have access to the output of algorithms without insights into their internal mechanisms, it was natural to examine how decisions would alter when auxiliary sensitive attributes (such as race) change. This led the research community to come up with counterfactual fairness measures, but how to evaluate the measure from available data remains a challenging task. In many practical applications, the target counterfactual measure is not identifiable, i.e., it cannot be uniquely determined from the combination of quantitative data and qualitative knowledge. This paper addresses this challenge using partial identification, which derives informative bounds over counterfactual fairness measures from observational data. We introduce a Bayesian approach to bound unknown counterfactual fairness measures with high confidence. We demonstrate our algorithm on the COMPAS dataset, examining fairness in recidivism risk scores with respect to race, age, and sex. Our results reveal a positive (spurious) effect on the COMPAS score when changing race to African-American (from all others) and a negative (direct causal) effect when transitioning from young to old age.", "date": "2025-09-30", "categories": []}
{"id": "2408.16218", "title": "Large-Scale Targeted Cause Discovery via Learning from Simulated Data", "abstract": "We propose a novel machine learning approach for inferring causal variables of a target variable from observations. Our focus is on directly inferring a set of causal factors without requiring full causal graph reconstruction, which is computationally challenging in large-scale systems. The identified causal set consists of all potential regulators of the target variable under experimental settings, enabling efficient regulation through intervention. To achieve this, we train a neural network using supervised learning on simulated data to infer causality. By employing a subsampled-ensemble inference strategy, our approach scales with linear complexity in the number of variables, efficiently scaling up to thousands of variables. Empirical results demonstrate superior performance in identifying causal relationships within large-scale gene regulatory networks, outperforming existing methods that emphasize full-graph discovery. We validate our model's generalization capability across out-of-distribution graph structures and generating mechanisms, including gene regulatory networks of E. coli and the human K562 cell line. Implementation codes are available at https://github.com/snu-mllab/Targeted-Cause-Discovery.", "date": "2024-08-29", "categories": []}
{"id": "2509.25715", "title": "MuPlon: Multi-Path Causal Optimization for Claim Verification through Controlling Confounding", "abstract": "As a critical task in data quality control, claim verification aims to curb the spread of misinformation by assessing the truthfulness of claims based on a wide range of evidence. However, traditional methods often overlook the complex interactions between evidence, leading to unreliable verification results. A straightforward solution represents the claim and evidence as a fully connected graph, which we define as the Claim-Evidence Graph (C-E Graph). Nevertheless, claim verification methods based on fully connected graphs face two primary confounding challenges, Data Noise and Data Biases. To address these challenges, we propose a novel framework, Multi-Path Causal Optimization (MuPlon). MuPlon integrates a dual causal intervention strategy, consisting of the back-door path and front-door path. In the back-door path, MuPlon dilutes noisy node interference by optimizing node probability weights, while simultaneously strengthening the connections between relevant evidence nodes. In the front-door path, MuPlon extracts highly relevant subgraphs and constructs reasoning paths, further applying counterfactual reasoning to eliminate data biases within these paths. The experimental results demonstrate that MuPlon outperforms existing methods and achieves state-of-the-art performance.", "date": "2025-09-30", "categories": []}
{"id": "2509.26275", "title": "Wasserstein Distributionally Robust Optimization Through the Lens of Structural Causal Models and Individual Fairness", "abstract": "In recent years, Wasserstein Distributionally Robust Optimization (DRO) has garnered substantial interest for its efficacy in data-driven decision-making under distributional uncertainty. However, limited research has explored the application of DRO to address individual fairness concerns, particularly when considering causal structures and sensitive attributes in learning problems. To address this gap, we first formulate the DRO problem from causality and individual fairness perspectives. We then present the DRO dual formulation as an efficient tool to convert the DRO problem into a more tractable and computationally efficient form. Next, we characterize the closed form of the approximate worst-case loss quantity as a regularizer, eliminating the max-step in the min-max DRO problem. We further estimate the regularizer in more general cases and explore the relationship between DRO and classical robust optimization. Finally, by removing the assumption of a known structural causal model, we provide finite sample error bounds when designing DRO with empirical distributions and estimated causal structures to ensure efficiency and robust learning.", "date": "2025-09-30", "categories": []}
{"id": "2208.08871", "title": "Network inference via process motifs for lagged correlation in linear stochastic processes", "abstract": "A major challenge for causal inference from time-series data is the trade-off between computational feasibility and accuracy. Motivated by process motifs for lagged covariance in an autoregressive model with slow mean-reversion, we propose to infer networks of causal relations via pairwise edge measure (PEMs) that one can easily compute from lagged correlation matrices. Motivated by contributions of process motifs to covariance and lagged variance, we formulate two PEMs that correct for confounding factors and for reverse causation. To demonstrate the performance of our PEMs, we consider network interference from simulations of linear stochastic processes, and we show that our proposed PEMs can infer networks accurately and efficiently. Specifically, for slightly autocorrelated time-series data, our approach achieves accuracies higher than or similar to Granger causality, transfer entropy, and convergent crossmapping -- but with much shorter computation time than possible with any of these methods. Our fast and accurate PEMs are easy-to-implement methods for network inference with a clear theoretical underpinning. They provide promising alternatives to current paradigms for the inference of linear models from time-series data, including Granger causality, vector-autoregression, and sparse inverse covariance estimation.", "date": "2022-08-18", "categories": []}
{"id": "2311.06012", "title": "Double Machine Learning Based Structure Identification from Temporal Data", "abstract": "Learning the causes of time-series data is a fundamental task in many applications, spanning from finance to earth sciences or bio-medical applications. Common approaches for this task are based on vector auto-regression, and they do not take into account unknown confounding between potential causes. However, in settings with many potential causes and noisy data, these approaches may be substantially biased. Furthermore, potential causes may be correlated in practical applications or even contain cycles. To address these challenges, we propose a new double machine learning based method for structure identification from temporal data (DR-SIT). We provide theoretical guarantees, showing that our method asymptotically recovers the true underlying causal structure. Our analysis extends to cases where the potential causes have cycles, and they may even be confounded. We further perform extensive experiments to showcase the superior performance of our method. Code: https://github.com/sdi1100041/TMLR_submission_DR_SIT", "date": "2023-11-10", "categories": []}
{"id": "2505.16288", "title": "No Black Boxes: Interpretable and Interactable Predictive Healthcare with Knowledge-Enhanced Agentic Causal Discovery", "abstract": "Deep learning models trained on extensive Electronic Health Records (EHR) data have achieved high accuracy in diagnosis prediction, offering the potential to assist clinicians in decision-making and treatment planning. However, these models lack two crucial features that clinicians highly value: interpretability and interactivity. The ``black-box'' nature of these models makes it difficult for clinicians to understand the reasoning behind predictions, limiting their ability to make informed decisions. Additionally, the absence of interactive mechanisms prevents clinicians from incorporating their own knowledge and experience into the decision-making process. To address these limitations, we propose II-KEA, a knowledge-enhanced agent-driven causal discovery framework that integrates personalized knowledge databases and agentic LLMs. II-KEA enhances interpretability through explicit reasoning and causal analysis, while also improving interactivity by allowing clinicians to inject their knowledge and experience through customized knowledge bases and prompts. II-KEA is evaluated on both MIMIC-III and MIMIC-IV, demonstrating superior performance along with enhanced interpretability and interactivity, as evidenced by its strong results from extensive case studies.", "date": "2025-05-22", "categories": []}
{"id": "2509.23570", "title": "Improving constraint-based discovery with robust propagation and reliable LLM priors", "abstract": "Learning causal structure from observational data is central to scientific modeling and decision-making. Constraint-based methods aim to recover conditional independence (CI) relations in a causal directed acyclic graph (DAG). Classical approaches such as PC and subsequent methods orient v-structures first and then propagate edge directions from these seeds, assuming perfect CI tests and exhaustive search of separating subsets -- assumptions often violated in practice, leading to cascading errors in the final graph. Recent work has explored using large language models (LLMs) as experts, prompting sets of nodes for edge directions, and could augment edge orientation when assumptions are not met. However, such methods implicitly assume perfect experts, which is unrealistic for hallucination-prone LLMs. We propose MosaCD, a causal discovery method that propagates edges from a high-confidence set of seeds derived from both CI tests and LLM annotations. To filter hallucinations, we introduce shuffled queries that exploit LLMs' positional bias, retaining only high-confidence seeds. We then apply a novel confidence-down propagation strategy that orients the most reliable edges first, and can be integrated with any skeleton-based discovery method. Across multiple real-world graphs, MosaCD achieves higher accuracy in final graph construction than existing constraint-based methods, largely due to the improved reliability of initial seeds and robust propagation strategies.", "date": "2025-09-28", "categories": []}
{"id": "2509.23992", "title": "Guide: Generalized-Prior and Data Encoders for DAG Estimation", "abstract": "Modern causal discovery methods face critical limitations in scalability, computational efficiency, and adaptability to mixed data types, as evidenced by benchmarks on node scalability (30, $\\le 50$, $\\ge 70$ nodes), computational energy demands, and continuous/non-continuous data handling. While traditional algorithms like PC, GES, and ICA-LiNGAM struggle with these challenges, exhibiting prohibitive energy costs for higher-order nodes and poor scalability beyond 70 nodes, we propose \\textbf{GUIDE}, a framework that integrates Large Language Model (LLM)-generated adjacency matrices with observational data through a dual-encoder architecture. GUIDE uniquely optimizes computational efficiency, reducing runtime on average by $\\approx 42%$ compared to RL-BIC and KCRL methods, while achieving an average $\\approx 117%$ improvement in accuracy over both NOTEARS and GraN-DAG individually. During training, GUIDE's reinforcement learning agent dynamically balances reward maximization (accuracy) and penalty avoidance (DAG constraints), enabling robust performance across mixed data types and scalability to $\\ge 70$ nodes -- a setting where baseline methods fail.", "date": "2025-09-28", "categories": []}
{"id": "2509.24293", "title": "ActiveCQ: Active Estimation of Causal Quantities", "abstract": "Estimating causal quantities (CQs) typically requires large datasets, which can be expensive to obtain, especially when measuring individual outcomes is costly. This challenge highlights the importance of sample-efficient active learning strategies. To address the narrow focus of prior work on the conditional average treatment effect, we formalize the broader task of Actively estimating Causal Quantities (ActiveCQ) and propose a unified framework for this general problem. Built upon the insight that many CQs are integrals of regression functions, our framework models the regression function with a Gaussian Process. For the distribution component, we explore both a baseline using explicit density estimators and a more integrated method using conditional mean embeddings in a reproducing kernel Hilbert space. This latter approach offers key advantages: it bypasses explicit density estimation, operates within the same function space as the GP, and adaptively refines the distributional model after each update. Our framework enables the principled derivation of acquisition strategies from the CQ's posterior uncertainty; we instantiate this principle with two utility functions based on information gain and total variance reduction. A range of simulated and semi-synthetic experiments demonstrate that our principled framework significantly outperforms relevant baselines, achieving substantial gains in sample efficiency across a variety of CQs.", "date": "2025-09-29", "categories": []}
{"id": "2502.20115", "title": "Multi-View Causal Discovery without Non-Gaussianity: Identifiability and Algorithms", "abstract": "Causal discovery is a difficult problem that typically relies on strong assumptions on the data-generating model, such as non-Gaussianity. In practice, many modern applications provide multiple related views of the same system, which has rarely been considered for causal discovery. Here, we leverage this multi-view structure to achieve causal discovery with weak assumptions. We propose a multi-view linear Structural Equation Model (SEM) that extends the well-known framework of non-Gaussian disturbances by alternatively leveraging correlation over views. We prove the identifiability of the model for acyclic SEMs. Subsequently, we propose several multi-view causal discovery algorithms, inspired by single-view algorithms (DirectLiNGAM, PairwiseLiNGAM, and ICA-LiNGAM). The new methods are validated through simulations and applications on neuroimaging data, where they enable the estimation of causal graphs between brain regions.", "date": "2025-02-27", "categories": []}
{"id": "2509.21607", "title": "Causal Abstraction Inference under Lossy Representations", "abstract": "The study of causal abstractions bridges two integral components of human intelligence: the ability to determine cause and effect, and the ability to interpret complex patterns into abstract concepts. Formally, causal abstraction frameworks define connections between complicated low-level causal models and simple high-level ones. One major limitation of most existing definitions is that they are not well-defined when considering lossy abstraction functions in which multiple low-level interventions can have different effects while mapping to the same high-level intervention (an assumption called the abstract invariance condition). In this paper, we introduce a new type of abstractions called projected abstractions that generalize existing definitions to accommodate lossy representations. We show how to construct a projected abstraction from the low-level model and how it translates equivalent observational, interventional, and counterfactual causal queries from low to high-level. Given that the true model is rarely available in practice we prove a new graphical criteria for identifying and estimating high-level causal queries from limited low-level data. Finally, we experimentally show the effectiveness of projected abstraction models in high-dimensional image settings.", "date": "2025-09-25", "categories": []}
{"id": "2509.21805", "title": "Towards Minimal Causal Representations for Human Multimodal Language Understanding", "abstract": "Human Multimodal Language Understanding (MLU) aims to infer human intentions by integrating related cues from heterogeneous modalities. Existing works predominantly follow a ``learning to attend\" paradigm, which maximizes mutual information between data and labels to enhance predictive performance. However, such methods are vulnerable to unintended dataset biases, causing models to conflate statistical shortcuts with genuine causal features and resulting in degraded out-of-distribution (OOD) generalization. To alleviate this issue, we introduce a Causal Multimodal Information Bottleneck (CaMIB) model that leverages causal principles rather than traditional likelihood. Concretely, we first applies the information bottleneck to filter unimodal inputs, removing task-irrelevant noise. A parameterized mask generator then disentangles the fused multimodal representation into causal and shortcut subrepresentations. To ensure global consistency of causal features, we incorporate an instrumental variable constraint, and further adopt backdoor adjustment by randomly recombining causal and shortcut features to stabilize causal estimation. Extensive experiments on multimodal sentiment analysis, humor detection, and sarcasm detection, along with OOD test sets, demonstrate the effectiveness of CaMIB. Theoretical and empirical analyses further highlight its interpretability and soundness.", "date": "2025-09-26", "categories": []}
{"id": "2509.21866", "title": "Causal-EPIG: A Prediction-Oriented Active Learning Framework for CATE Estimation", "abstract": "Estimating the Conditional Average Treatment Effect (CATE) is often constrained by the high cost of obtaining outcome measurements, making active learning essential. However, conventional active learning strategies suffer from a fundamental objective mismatch. They are designed to reduce uncertainty in model parameters or in observable factual outcomes, failing to directly target the unobservable causal quantities that are the true objects of interest. To address this misalignment, we introduce the principle of causal objective alignment, which posits that acquisition functions should target unobservable causal quantities, such as the potential outcomes and the CATE, rather than indirect proxies. We operationalize this principle through the Causal-EPIG framework, which adapts the information-theoretic criterion of Expected Predictive Information Gain (EPIG) to explicitly quantify the value of a query in terms of reducing uncertainty about unobservable causal quantities. From this unified framework, we derive two distinct strategies that embody a fundamental trade-off: a comprehensive approach that robustly models the full causal mechanisms via the joint potential outcomes, and a focused approach that directly targets the CATE estimand for maximum sample efficiency. Extensive experiments demonstrate that our strategies consistently outperform standard baselines, and crucially, reveal that the optimal strategy is context-dependent, contingent on the base estimator and data complexity. Our framework thus provides a principled guide for sample-efficient CATE estimation in practice.", "date": "2025-09-26", "categories": []}
{"id": "2407.07338", "title": "Towards Complete Causal Explanation with Expert Knowledge", "abstract": "We study the problem of restricting a Markov equivalence class of maximal ancestral graphs (MAGs) to only those MAGs that contain certain edge marks, which we refer to as expert or orientation knowledge. Such a restriction of the Markov equivalence class can be uniquely represented by a restricted essential ancestral graph. Our contributions are several-fold. First, we prove certain properties for the entire Markov equivalence class including a conjecture from Ali et al. (2009). Second, we present several new sound graphical orientation rules for adding orientation knowledge to an essential ancestral graph. We also show that some orientation rules of Zhang (2008b) are not needed for restricting the Markov equivalence class with orientation knowledge. Third, we provide an algorithm for including this orientation knowledge and show that in certain settings the output of our algorithm is a restricted essential ancestral graph. Finally, outside of the specified settings, we provide an algorithm for checking whether a graph is a restricted essential graph and discuss its runtime. This work can be seen as a generalization of Meek (1995) to settings which allow for latent confounding.", "date": "2024-07-10", "categories": []}
{"id": "2409.02604", "title": "Context-Aware Reasoning On Parametric Knowledge for Inferring Causal Variables", "abstract": "Scientific discovery catalyzes human intellectual advances, driven by the cycle of hypothesis generation, experimental design, evaluation, and assumption refinement. Central to this process is causal inference, uncovering the mechanisms behind observed phenomena. While randomized experiments provide strong inferences, they are often infeasible due to ethical or practical constraints. However, observational studies are prone to confounding or mediating biases. While crucial, identifying such backdoor paths is expensive and heavily depends on scientists' domain knowledge to generate hypotheses. We introduce a novel benchmark where the objective is to complete a partial causal graph. We design a benchmark with varying difficulty levels with over 4000 queries. We show the strong ability of LLMs to hypothesize the backdoor variables between a cause and its effect. Unlike simple knowledge memorization of fixed associations, our task requires the LLM to reason according to the context of the entire graph.", "date": "2024-09-04", "categories": []}
{"id": "2508.04495", "title": "Causal Reflection with Language Models", "abstract": "While LLMs exhibit impressive fluency and factual recall, they struggle with robust causal reasoning, often relying on spurious correlations and brittle patterns. Similarly, traditional Reinforcement Learning agents also lack causal understanding, optimizing for rewards without modeling why actions lead to outcomes. We introduce Causal Reflection, a framework that explicitly models causality as a dynamic function over state, action, time, and perturbation, enabling agents to reason about delayed and nonlinear effects. Additionally, we define a formal Reflect mechanism that identifies mismatches between predicted and observed outcomes and generates causal hypotheses to revise the agent's internal model. In this architecture, LLMs serve not as black-box reasoners, but as structured inference engines translating formal causal outputs into natural language explanations and counterfactuals. Our framework lays the theoretical groundwork for Causal Reflective agents that can adapt, self-correct, and communicate causal understanding in evolving environments.", "date": "2025-08-06", "categories": []}
{"id": "2506.18187", "title": "The Impact of Medication Non-adherence on Adverse Outcomes: Evidence from Schizophrenia Patients via Survival Analysis", "abstract": "This study quantifies the association between non-adherence to antipsychotic medications and adverse outcomes in individuals with schizophrenia. We frame the problem using survival analysis, focusing on the time to the earliest of several adverse events (early death, involuntary hospitalization, jail booking). We extend standard causal inference methods (T-learner, S-learner, nearest neighbor matching) to utilize various survival models to estimate individual and average treatment effects, where treatment corresponds to medication non-adherence. Analyses are repeated using different amounts of longitudinal information (3, 6, 9, and 12 months). Using data from Allegheny County in western Pennsylvania, we find strong evidence that non-adherence advances adverse outcomes by approximately 1 to 4 months. Ablation studies confirm that county-provided risk scores adjust for key confounders, as their removal amplifies the estimated effects. Subgroup analyses by medication formulation (injectable vs. oral) and medication type consistently show that non-adherence is associated with earlier adverse events. These findings highlight the clinical importance of adherence in delaying psychiatric crises and show that integrating survival analysis with causal inference tools can yield policy-relevant insights. We caution that although we apply causal inference, we only make associative claims and discuss assumptions needed for causal interpretation.", "date": "2025-06-22", "categories": []}
{"id": "2507.00093", "title": "$\\sigma$-Maximal Ancestral Graphs", "abstract": "Maximal Ancestral Graphs (MAGs) provide an abstract representation of Directed Acyclic Graphs (DAGs) with latent (selection) variables. These graphical objects encode information about ancestral relations and d-separations of the DAGs they represent. This abstract representation has been used amongst others to prove the soundness and completeness of the FCI algorithm for causal discovery, and to derive a do-calculus for its output. One significant inherent limitation of MAGs is that they rule out the possibility of cyclic causal relationships. In this work, we address that limitation. We introduce and study a class of graphical objects that we coin ''$\\sigma$-Maximal Ancestral Graphs'' (''$\\sigma$-MAGs''). We show how these graphs provide an abstract representation of (possibly cyclic) Directed Graphs (DGs) with latent (selection) variables, analogously to how MAGs represent DAGs. We study the properties of these objects and provide a characterization of their Markov equivalence classes.", "date": "2025-06-30", "categories": []}
{"id": "2509.19705", "title": "Causal Machine Learning for Surgical Interventions", "abstract": "Surgical decision-making is complex and requires understanding causal relationships between patient characteristics, interventions, and outcomes. In high-stakes settings like spinal fusion or scoliosis correction, accurate estimation of individualized treatment effects (ITEs) remains limited due to the reliance on traditional statistical methods that struggle with complex, heterogeneous data. In this study, we develop a multi-task meta-learning framework, X-MultiTask, for ITE estimation that models each surgical decision (e.g., anterior vs. posterior approach, surgery vs. no surgery) as a distinct task while learning shared representations across tasks. To strengthen causal validity, we incorporate the inverse probability weighting (IPW) into the training objective. We evaluate our approach on two datasets: (1) a public spinal fusion dataset (1,017 patients) to assess the effect of anterior vs. posterior approaches on complication severity; and (2) a private AIS dataset (368 patients) to analyze the impact of posterior spinal fusion (PSF) vs. non-surgical management on patient-reported outcomes (PROs). Our model achieves the highest average AUC (0.84) in the anterior group and maintains competitive performance in the posterior group (0.77). It outperforms baselines in treatment effect estimation with the lowest overall $\\epsilon_{\\text{NN-PEHE}}$ (0.2778) and $\\epsilon_{\\text{ATE}}$ (0.0763). Similarly, when predicting PROs in AIS, X-MultiTask consistently shows superior performance across all domains, with $\\epsilon_{\\text{NN-PEHE}}$ = 0.2551 and $\\epsilon_{\\text{ATE}}$ = 0.0902. By providing robust, patient-specific causal estimates, X-MultiTask offers a powerful tool to advance personalized surgical care and improve patient outcomes. The code is available at https://github.com/Wizaaard/X-MultiTask.", "date": "2025-09-24", "categories": []}
{"id": "2402.04906", "title": "Conformal Convolution and Monte Carlo Meta-learners for Predictive Inference of Individual Treatment Effects", "abstract": "Generating probabilistic forecasts of potential outcomes and individual treatment effects (ITE) is essential for risk-aware decision-making in domains such as healthcare, policy, marketing, and finance. We propose two novel methods: the conformal convolution T-learner (CCT) and the conformal Monte Carlo (CMC) meta-learner, that generate full predictive distributions of both potential outcomes and ITEs. Our approaches combine weighted conformal predictive systems with either analytic convolution of potential outcome distributions or Monte Carlo sampling, addressing covariate shift through propensity score weighting. In contrast to other approaches that allow the generation of potential outcome predictive distributions, our approaches are model agnostic, universal, and come with finite-sample guarantees of probabilistic calibration under knowledge of the propensity score. Regarding estimating the ITE distribution, we formally characterize how assumptions about potential outcomes' noise dependency impact distribution validity and establish universal consistency under independence noise assumptions. Experiments on synthetic and semi-synthetic datasets demonstrate that the proposed methods achieve probabilistically calibrated predictive distributions while maintaining narrow prediction intervals and having performant continuous ranked probability scores. Besides probabilistic forecasting performance, we observe significant efficiency gains for the CCT- and CMC meta-learners compared to other conformal approaches that produce prediction intervals for ITE with coverage guarantees.", "date": "2024-02-07", "categories": []}
{"id": "2411.08019", "title": "Language Models as Causal Effect Generators", "abstract": "In this work, we present sequence-driven structural causal models (SD-SCMs), a framework for specifying causal models with user-defined structure and language-model-defined mechanisms. We characterize how an SD-SCM enables sampling from observational, interventional, and counterfactual distributions according to the desired causal structure. We then leverage this procedure to propose a new type of benchmark for causal inference methods, generating individual-level counterfactual data to test treatment effect estimation. We create an example benchmark consisting of thousands of datasets, and test a suite of popular estimation methods for average, conditional average, and individual treatment effect estimation. We find under this benchmark that (1) causal methods outperform non-causal methods and that (2) even state-of-the-art methods struggle with individualized effect estimation, suggesting this benchmark captures some inherent difficulties in causal estimation. Apart from generating data, this same technique can underpin the auditing of language models for (un)desirable causal effects, such as misinformation or discrimination. We believe SD-SCMs can serve as a useful tool in any application that would benefit from sequential data with controllable causal structure.", "date": "2024-11-12", "categories": []}
{"id": "2509.18148", "title": "Augmenting Limited and Biased RCTs through Pseudo-Sample Matching-Based Observational Data Fusion Method", "abstract": "In the online ride-hailing pricing context, companies often conduct randomized controlled trials (RCTs) and utilize uplift models to assess the effect of discounts on customer orders, which substantially influences competitive market outcomes. However, due to the high cost of RCTs, the proportion of trial data relative to observational data is small, which only accounts for 0.65\\% of total traffic in our context, resulting in significant bias when generalizing to the broader user base. Additionally, the complexity of industrial processes reduces the quality of RCT data, which is often subject to heterogeneity from potential interference and selection bias, making it difficult to correct. Moreover, existing data fusion methods are challenging to implement effectively in complex industrial settings due to the high dimensionality of features and the strict assumptions that are hard to verify with real-world data. To address these issues, we propose an empirical data fusion method called pseudo-sample matching. By generating pseudo-samples from biased, low-quality RCT data and matching them with the most similar samples from large-scale observational data, the method expands the RCT dataset while mitigating its heterogeneity. We validated the method through simulation experiments, conducted offline and online tests using real-world data. In a week-long online experiment, we achieved a 0.41\\% improvement in profit, which is a considerable gain when scaled to industrial scenarios with hundreds of millions in revenue. In addition, we discuss the harm to model training, offline evaluation, and online economic benefits when the RCT data quality is not high, and emphasize the importance of improving RCT data quality in industrial scenarios. Further details of the simulation experiments can be found in the GitHub repository https://github.com/Kairong-Han/Pseudo-Matching.", "date": "2025-09-16", "categories": []}
{"id": "2509.18156", "title": "Event Causality Identification with Synthetic Control", "abstract": "Event causality identification (ECI), a process that extracts causal relations between events from text, is crucial for distinguishing causation from correlation. Traditional approaches to ECI have primarily utilized linguistic patterns and multi-hop relational inference, risking false causality identification due to informal usage of causality and specious graphical inference. In this paper, we adopt the Rubin Causal Model to identify event causality: given two temporally ordered events, we see the first event as the treatment and the second one as the observed outcome. Determining their causality involves manipulating the treatment and estimating the resultant change in the likelihood of the outcome. Given that it is only possible to implement manipulation conceptually in the text domain, as a work-around, we try to find a twin for the protagonist from existing corpora. This twin should have identical life experiences with the protagonist before the treatment but undergoes an intervention of treatment. However, the practical difficulty of locating such a match limits its feasibility. Addressing this issue, we use the synthetic control method to generate such a twin' from relevant historical data, leveraging text embedding synthesis and inversion techniques. This approach allows us to identify causal relations more robustly than previous methods, including GPT-4, which is demonstrated on a causality benchmark, COPES-hard.", "date": "2025-09-16", "categories": []}
{"id": "2509.19058", "title": "Towards Causal Representation Learning with Observable Sources as Auxiliaries", "abstract": "Causal representation learning seeks to recover latent factors that generate observational data through a mixing function. Needing assumptions on latent structures or relationships to achieve identifiability in general, prior works often build upon conditional independence given known auxiliary variables. However, prior frameworks limit the scope of auxiliary variables to be external to the mixing function. Yet, in some cases, system-driving latent factors can be easily observed or extracted from data, possibly facilitating identification. In this paper, we introduce a framework of observable sources being auxiliaries, serving as effective conditioning variables. Our main results show that one can identify entire latent variables up to subspace-wise transformations and permutations using volume-preserving encoders. Moreover, when multiple known auxiliary variables are available, we offer a variable-selection scheme to choose those that maximize recoverability of the latent factors given knowledge of the latent causal graph. Finally, we demonstrate the effectiveness of our framework through experiments on synthetic graph and image data, thereby extending the boundaries of current approaches.", "date": "2025-09-23", "categories": []}
{"id": "2502.12013", "title": "Unsupervised Structural-Counterfactual Generation under Domain Shift", "abstract": "Motivated by the burgeoning interest in cross-domain learning, we present a novel generative modeling challenge: generating counterfactual samples in a target domain based on factual observations from a source domain. Our approach operates within an unsupervised paradigm devoid of parallel or joint datasets, relying exclusively on distinct observational samples and causal graphs for each domain. This setting presents challenges that surpass those of conventional counterfactual generation. Central to our methodology is the disambiguation of exogenous causes into effect-intrinsic and domain-intrinsic categories. This differentiation facilitates the integration of domain-specific causal graphs into a unified joint causal graph via shared effect-intrinsic exogenous variables. We propose leveraging Neural Causal models within this joint framework to enable accurate counterfactual generation under standard identifiability assumptions. Furthermore, we introduce a novel loss function that effectively segregates effect-intrinsic from domain-intrinsic variables during model training. Given a factual observation, our framework combines the posterior distribution of effect-intrinsic variables from the source domain with the prior distribution of domain-intrinsic variables from the target domain to synthesize the desired counterfactuals, adhering to Pearl's causal hierarchy. Intriguingly, when domain shifts are restricted to alterations in causal mechanisms without accompanying covariate shifts, our training regimen parallels the resolution of a conditional optimal transport problem. Empirical evaluations on a synthetic dataset show that our framework generates counterfactuals in the target domain that very closely resemble the ground truth.", "date": "2025-02-17", "categories": []}
{"id": "2507.16370", "title": "Canonical Representations of Markovian Structural Causal Models: A Framework for Counterfactual Reasoning", "abstract": "Counterfactual reasoning aims at answering contrary-to-fact questions like ``Would have Alice recovered had she taken aspirin?'' and corresponds to the most fine-grained layer of causation. Critically, while many counterfactual statements cannot be falsified-even by randomized experiments-they underpin fundamental concepts like individual-wise fairness. Therefore, providing models to formalize and implement counterfactual beliefs remains a fundamental scientific problem. In the Markovian setting of Pearl's causal framework, we propose an alternative approach to structural causal models to represent counterfactuals compatible with a given causal graphical model. More precisely, we introduce counterfactual models, also called canonical representations of structural causal models. They enable analysts to choose a counterfactual assumption via random-process probability distributions with preassigned marginals and characterize the counterfactual equivalence class of structural causal models. Using these representations, we present a normalization procedure to disentangle the (arbitrary and unfalsifiable) counterfactual choice from the (typically testable) interventional constraints. In contrast to structural causal models, this allows to implement many counterfactual assumptions while preserving interventional knowledge, and does not require any estimation step at the individual-counterfactual layer: only to make a choice. Finally, we illustrate the specific role of counterfactuals in causality and the benefits of our approach on theoretical and numerical examples.", "date": "2025-07-22", "categories": []}
{"id": "2509.16397", "title": "GRID: Graph-based Reasoning for Intervention and Discovery in Built Environments", "abstract": "Manual HVAC fault diagnosis in commercial buildings takes 8-12 hours per incident and achieves only 60 percent diagnostic accuracy, reflecting analytics that stop at correlation instead of causation. To close this gap, we present GRID (Graph-based Reasoning for Intervention and Discovery), a three-stage causal discovery pipeline that combines constraint-based search, neural structural equation modeling, and language model priors to recover directed acyclic graphs from building sensor data. Across six benchmarks: synthetic rooms, EnergyPlus simulation, the ASHRAE Great Energy Predictor III dataset, and a live office testbed, GRID achieves F1 scores ranging from 0.65 to 1.00, with exact recovery (F1 = 1.00) in three controlled environments (Base, Hidden, Physical) and strong performance on real-world data (F1 = 0.89 on ASHRAE, 0.86 in noisy conditions). The method outperforms ten baseline approaches across all evaluation scenarios. Intervention scheduling achieves low operational impact in most scenarios (cost <= 0.026) while reducing risk metrics compared to baseline approaches. The framework integrates constraint-based methods, neural architectures, and domain-specific language model prompts to address the observational-causal gap in building analytics.", "date": "2025-09-19", "categories": []}
{"id": "2509.16463", "title": "Entropic Causal Inference: Graph Identifiability", "abstract": "Entropic causal inference is a recent framework for learning the causal graph between two variables from observational data by finding the information-theoretically simplest structural explanation of the data, i.e., the model with smallest entropy. In our work, we first extend the causal graph identifiability result in the two-variable setting under relaxed assumptions. We then show the first identifiability result using the entropic approach for learning causal graphs with more than two nodes. Our approach utilizes the property that ancestrality between a source node and its descendants can be determined using the bivariate entropic tests. We provide a sound sequential peeling algorithm for general graphs that relies on this property. We also propose a heuristic algorithm for small graphs that shows strong empirical performance. We rigorously evaluate the performance of our algorithms on synthetic data generated from a variety of models, observing improvement over prior work. Finally we test our algorithms on real-world datasets.", "date": "2025-09-19", "categories": []}
{"id": "2509.16842", "title": "DoubleGen: Debiased Generative Modeling of Counterfactuals", "abstract": "Generative models for counterfactual outcomes face two key sources of bias. Confounding bias arises when approaches fail to account for systematic differences between those who receive the intervention and those who do not. Misspecification bias arises when methods attempt to address confounding through estimation of an auxiliary model, but specify it incorrectly. We introduce DoubleGen, a doubly robust framework that modifies generative modeling training objectives to mitigate these biases. The new objectives rely on two auxiliaries -- a propensity and outcome model -- and successfully address confounding bias even if only one of them is correct. We provide finite-sample guarantees for this robustness property. We further establish conditions under which DoubleGen achieves oracle optimality -- matching the convergence rates standard approaches would enjoy if interventional data were available -- and minimax rate optimality. We illustrate DoubleGen with three examples: diffusion models, flow matching, and autoregressive language models.", "date": "2025-09-20", "categories": []}
{"id": "2509.17228", "title": "Causal Representation Learning from Multimodal Clinical Records under Non-Random Modality Missingness", "abstract": "Clinical notes contain rich patient information, such as diagnoses or medications, making them valuable for patient representation learning. Recent advances in large language models have further improved the ability to extract meaningful representations from clinical texts. However, clinical notes are often missing. For example, in our analysis of the MIMIC-IV dataset, 24.5% of patients have no available discharge summaries. In such cases, representations can be learned from other modalities such as structured data, chest X-rays, or radiology reports. Yet the availability of these modalities is influenced by clinical decision-making and varies across patients, resulting in modality missing-not-at-random (MMNAR) patterns. We propose a causal representation learning framework that leverages observed data and informative missingness in multimodal clinical records. It consists of: (1) an MMNAR-aware modality fusion component that integrates structured data, imaging, and text while conditioning on missingness patterns to capture patient health and clinician-driven assignment; (2) a modality reconstruction component with contrastive learning to ensure semantic sufficiency in representation learning; and (3) a multitask outcome prediction model with a rectifier that corrects for residual bias from specific modality observation patterns. Comprehensive evaluations across MIMIC-IV and eICU show consistent gains over the strongest baselines, achieving up to 13.8% AUC improvement for hospital readmission and 13.1% for ICU admission.", "date": "2025-09-21", "categories": []}
{"id": "2509.17380", "title": "Correlation or Causation: Analyzing the Causal Structures of LLM and LRM Reasoning Process", "abstract": "LLMs suffer from critical reasoning issues such as unfaithfulness, bias, and inconsistency, since they lack robust causal underpinnings and may rely on superficial correlations rather than genuine understanding. Successive LRMs have emerged as a promising alternative, leveraging advanced training techniques such as reinforcement learning (RL) and distillation to improve task accuracy. However, the impact of these training methods on causality remains largely unexplored. In this study, we conduct a systematic causal analysis on LLMs and LRMs, examining structural causal models (SCMs) of four key variables: problem instruction (Z), thinking process (T), reasoning steps (X), and answer (Y). Our findings reveal that RLVR-trained LRMs exhibit enhanced causal reasoning capabilities, aligning more closely with ideal causal structures, while LLMs and distilled LRMs fail to address causality-related deficiencies. Our further investigation indicates that RLVR reduces spurious correlations and strengthens genuine causal patterns, thereby mitigating unfaithfulness and bias. In addition, our inspection on the dynamics of the RLVR training process observes a high correlation between reduced spurious features and improved causal structures, where the causal relationships consistently improve in the training process. This study contributes to the understanding of causality in reasoning models, highlights the critical role of RLVR in enhancing causal reasoning, and provides insights for designing future AI systems with stronger causal foundations. We release our code and data at https://github.com/Harryking1999/CoT_Causal_Analysis.", "date": "2025-09-22", "categories": []}
{"id": "2501.11447", "title": "Decomposing Interventional Causality into Synergistic, Redundant, and Unique Components", "abstract": "We introduce a novel framework for decomposing interventional causal effects into synergistic, redundant, and unique components, building on the intuition of Partial Information Decomposition (PID) and the principle of M\\\"obius inversion. While recent work has explored a similar decomposition of an observational measure, we argue that a proper causal decomposition must be interventional in nature. We develop a mathematical approach that systematically quantifies how causal power is distributed among variables in a system, using a recently derived closed-form expression for the M\\\"obius function of the redundancy lattice. The formalism is then illustrated by decomposing the causal power in logic gates, cellular automata, chemical reaction networks, and a transformer language model. Our results reveal how the distribution of causal power can be context- and parameter-dependent. The decomposition provides new insights into complex systems by revealing how causal influences are shared and combined among multiple variables, with potential applications ranging from attribution of responsibility in legal or AI systems, to the analysis of biological networks or climate models.", "date": "2025-01-20", "categories": []}
{"id": "2509.16027", "title": "What is a good matching of probability measures? A counterfactual lens on transport maps", "abstract": "Coupling probability measures lies at the core of many problems in statistics and machine learning, from domain adaptation to transfer learning and causal inference. Yet, even when restricted to deterministic transports, such couplings are not identifiable: two atomless marginals admit infinitely many transport maps. The common recourse to optimal transport, motivated by cost minimization and cyclical monotonicity, obscures the fact that several distinct notions of multivariate monotone matchings coexist. In this work, we first carry a comparative analysis of three constructions of transport maps: cyclically monotone, quantile-preserving and triangular monotone maps. We establish necessary and sufficient conditions for their equivalence, thereby clarifying their respective structural properties. In parallel, we formulate counterfactual reasoning within the framework of structural causal models as a problem of selecting transport maps between fixed marginals, which makes explicit the role of untestable assumptions in counterfactual reasoning. Then, we are able to connect these two perspectives by identifying conditions on causal graphs and structural equations under which counterfactual maps coincide with classical statistical transports. In this way, we delineate the circumstances in which causal assumptions support the use of a specific structure of transport map. Taken together, our results aim to enrich the theoretical understanding of families of transport maps and to clarify their possible causal interpretations. We hope this work contributes to establishing new bridges between statistical transport and causal inference.", "date": "2025-09-19", "categories": []}
{"id": "2509.05775", "title": "Causal Clustering for Conditional Average Treatment Effects Estimation and Subgroup Discovery", "abstract": "Estimating heterogeneous treatment effects is critical in domains such as personalized medicine, resource allocation, and policy evaluation. A central challenge lies in identifying subpopulations that respond differently to interventions, thereby enabling more targeted and effective decision-making. While clustering methods are well-studied in unsupervised learning, their integration with causal inference remains limited. We propose a novel framework that clusters individuals based on estimated treatment effects using a learned kernel derived from causal forests, revealing latent subgroup structures. Our approach consists of two main steps. First, we estimate debiased Conditional Average Treatment Effects (CATEs) using orthogonalized learners via the Robinson decomposition, yielding a kernel matrix that encodes sample-level similarities in treatment responsiveness. Second, we apply kernelized clustering to this matrix to uncover distinct, treatment-sensitive subpopulations and compute cluster-level average CATEs. We present this kernelized clustering step as a form of regularization within the residual-on-residual regression framework. Through extensive experiments on semi-synthetic and real-world datasets, supported by ablation studies and exploratory analyses, we demonstrate the effectiveness of our method in capturing meaningful treatment effect heterogeneity.", "date": "2025-09-06", "categories": []}
{"id": "2509.12497", "title": "Prediction and Causality of functional MRI and synthetic signal using a Zero-Shot Time-Series Foundation Model", "abstract": "Time-series forecasting and causal discovery are central in neuroscience, as predicting brain activity and identifying causal relationships between neural populations and circuits can shed light on the mechanisms underlying cognition and disease. With the rise of foundation models, an open question is how they compare to traditional methods for brain signal forecasting and causality analysis, and whether they can be applied in a zero-shot setting. In this work, we evaluate a foundation model against classical methods for inferring directional interactions from spontaneous brain activity measured with functional magnetic resonance imaging (fMRI) in humans. Traditional approaches often rely on Wiener-Granger causality. We tested the forecasting ability of the foundation model in both zero-shot and fine-tuned settings, and assessed causality by comparing Granger-like estimates from the model with standard Granger causality. We validated the approach using synthetic time series generated from ground-truth causal models, including logistic map coupling and Ornstein-Uhlenbeck processes. The foundation model achieved competitive zero-shot forecasting fMRI time series (mean absolute percentage error of 0.55 in controls and 0.27 in patients). Although standard Granger causality did not show clear quantitative differences between models, the foundation model provided a more precise detection of causal interactions.\n  Overall, these findings suggest that foundation models offer versatility, strong zero-shot performance, and potential utility for forecasting and causal discovery in time-series data.", "date": "2025-09-15", "categories": []}
{"id": "2509.12387", "title": "Causal-Symbolic Meta-Learning (CSML): Inducing Causal World Models for Few-Shot Generalization", "abstract": "Modern deep learning models excel at pattern recognition but remain fundamentally limited by their reliance on spurious correlations, leading to poor generalization and a demand for massive datasets. We argue that a key ingredient for human-like intelligence-robust, sample-efficient learning-stems from an understanding of causal mechanisms. In this work, we introduce Causal-Symbolic Meta-Learning (CSML), a novel framework that learns to infer the latent causal structure of a task distribution. CSML comprises three key modules: a perception module that maps raw inputs to disentangled symbolic representations; a differentiable causal induction module that discovers the underlying causal graph governing these symbols and a graph-based reasoning module that leverages this graph to make predictions. By meta-learning a shared causal world model across a distribution of tasks, CSML can rapidly adapt to novel tasks, including those requiring reasoning about interventions and counterfactuals, from only a handful of examples. We introduce CausalWorld, a new physics-based benchmark designed to test these capabilities. Our experiments show that CSML dramatically outperforms state-of-the-art meta-learning and neuro-symbolic baselines, particularly on tasks demanding true causal inference.", "date": "2025-09-15", "categories": []}
{"id": "2405.09493", "title": "C-Learner: Constrained Learning for Causal Inference", "abstract": "Popular debiased estimation methods for causal inference -- such as augmented inverse propensity weighting and targeted maximum likelihood estimation -- enjoy desirable asymptotic properties like statistical efficiency and double robustness but they can produce unstable estimates when there is limited overlap between treatment and control, requiring additional assumptions or ad hoc adjustments in practice (e.g., truncating propensity scores). In contrast, simple plug-in estimators are stable but lack desirable asymptotic properties. We propose a novel debiasing approach that achieves the best of both worlds, producing stable plug-in estimates with desirable asymptotic properties. Our constrained learning framework solves for the best plug-in estimator under the constraint that the first-order error with respect to the plugged-in quantity is zero, and can leverage flexible model classes including neural networks and tree ensembles. In several experimental settings, including ones in which we handle text-based covariates by fine-tuning language models, our constrained learning-based estimator outperforms basic versions of one-step estimation and targeting in challenging settings with limited overlap between treatment and control, and performs similarly otherwise.", "date": "2024-05-15", "categories": []}
{"id": "2506.19010", "title": "Simulation-Based Sensitivity Analysis in Optimal Treatment Regimes and Causal Decomposition with Individualized Interventions", "abstract": "Causal decomposition analysis aims to assess the effect of modifying risk factors on reducing social disparities in outcomes. Recently, this analysis has incorporated individual characteristics when modifying risk factors by utilizing optimal treatment regimes (OTRs). Since the newly defined individualized effects rely on the no omitted confounding assumption, developing sensitivity analyses to account for potential omitted confounding is essential. Moreover, OTRs and individualized effects are primarily based on binary risk factors, and no formal approach currently exists to benchmark the strength of omitted confounding using observed covariates for binary risk factors. To address this gap, we extend a simulation-based sensitivity analysis that simulates unmeasured confounders, addressing two sources of bias emerging from deriving OTRs and estimating individualized effects. Additionally, we propose a formal bounding strategy that benchmarks the strength of omitted confounding for binary risk factors. Using the High School Longitudinal Study 2009 (HSLS:09), we demonstrate this sensitivity analysis and benchmarking method.", "date": "2025-06-23", "categories": []}
{"id": "2509.11676", "title": "An Interventional Approach to Real-Time Disaster Assessment via Causal Attribution", "abstract": "Traditional disaster analysis and modelling tools for assessing the severity of a disaster are predictive in nature. Based on the past observational data, these tools prescribe how the current input state (e.g., environmental conditions, situation reports) results in a severity assessment. However, these systems are not meant to be interventional in the causal sense, where the user can modify the current input state to simulate counterfactual \"what-if\" scenarios. In this work, we provide an alternative interventional tool that complements traditional disaster modelling tools by leveraging real-time data sources like satellite imagery, news, and social media. Our tool also helps understand the causal attribution of different factors on the estimated severity, over any given region of interest. In addition, we provide actionable recourses that would enable easier mitigation planning. Our source code is publicly available.", "date": "2025-09-15", "categories": []}
{"id": "2508.13355", "title": "Counterfactual Probabilistic Diffusion with Expert Models", "abstract": "Predicting counterfactual distributions in complex dynamical systems is essential for scientific modeling and decision-making in domains such as public health and medicine. However, existing methods often rely on point estimates or purely data-driven models, which tend to falter under data scarcity. We propose a time series diffusion-based framework that incorporates guidance from imperfect expert models by extracting high-level signals to serve as structured priors for generative modeling. Our method, ODE-Diff, bridges mechanistic and data-driven approaches, enabling more reliable and interpretable causal inference. We evaluate ODE-Diff across semi-synthetic COVID-19 simulations, synthetic pharmacological dynamics, and real-world case studies, demonstrating that it consistently outperforms strong baselines in both point prediction and distributional accuracy.", "date": "2025-08-18", "categories": []}
{"id": "2409.03962", "title": "Average Causal Effect Estimation in DAGs with Hidden Variables: Beyond Back-Door and Front-Door Criteria", "abstract": "The identification theory for causal effects in directed acyclic graphs (DAGs) with hidden variables is well established, but methods for estimating and inferring functionals that extend beyond the g-formula remain underdeveloped. Previous studies have introduced semiparametric estimators for such functionals in a broad class of DAGs with hidden variables. While these estimators exhibit desirable statistical properties such as double robustness in certain cases, they also face significant limitations. Notably, they encounter substantial computational challenges, particularly involving density estimation and numerical integration for continuous variables, and their estimates may fall outside the parameter space of the target estimand. Additionally, the asymptotic properties of these estimators is underexplored, especially when integrating flexible statistical and machine learning models for nuisance functional estimations. This paper addresses these challenges by introducing novel one-step corrected plug-in and targeted minimum loss-based estimators of causal effects for a class of hidden variable DAGs that go beyond classical back-door and front-door criteria (known as the treatment primal fixability criterion in prior literature). These estimators leverage data-adaptive machine learning algorithms to minimize modeling assumptions while ensuring key statistical properties including double robustness, efficiency, boundedness within the target parameter space, and asymptotic linearity under $L^2(P)$-rate conditions for nuisance functional estimates that yield root-n consistent causal effect estimates. To ensure our estimation methods are accessible in practice, we provide the flexCausal package in R.", "date": "2024-09-06", "categories": []}
{"id": "2509.09128", "title": "Learning What Matters: Causal Time Series Modeling for Arctic Sea Ice Prediction", "abstract": "Conventional machine learning and deep learning models typically rely on correlation-based learning, which often fails to distinguish genuine causal relationships from spurious associations, limiting their robustness, interpretability, and ability to generalize. To overcome these limitations, we introduce a causality-aware deep learning framework that integrates Multivariate Granger Causality (MVGC) and PCMCI+ for causal feature selection within a hybrid neural architecture. Leveraging 43 years (1979-2021) of Arctic Sea Ice Extent (SIE) data and associated ocean-atmospheric variables at daily and monthly resolutions, the proposed method identifies causally influential predictors, prioritizes direct causes of SIE dynamics, reduces unnecessary features, and enhances computational efficiency. Experimental results show that incorporating causal inputs leads to improved prediction accuracy and interpretability across varying lead times. While demonstrated on Arctic SIE forecasting, the framework is broadly applicable to other dynamic, high-dimensional domains, offering a scalable approach that advances both the theoretical foundations and practical performance of causality-informed predictive modeling.", "date": "2025-09-11", "categories": []}
{"id": "2509.08682", "title": "Automatic Failure Attribution and Critical Step Prediction Method for Multi-Agent Systems Based on Causal Inference", "abstract": "Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who\\&When, state-of-the-art methods achieve less than 15\\% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who\\&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2\\% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4\\%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.", "date": "2025-09-10", "categories": []}
{"id": "2407.19618", "title": "Improving the Estimation of Lifetime Effects in A/B Testing via Treatment Locality", "abstract": "Utilizing randomized experiments to evaluate the effect of short-term treatments on the short-term outcomes has been well understood and become the golden standard in industrial practice. However, as service systems become increasingly dynamical and personalized, much focus is shifting toward maximizing long-term outcomes, such as customer lifetime value, through lifetime exposure to interventions. Our goal is to assess the impact of treatment and control policies on long-term outcomes from relatively short-term observations, such as those generated by A/B testing. A key managerial observation is that many practical treatments are local, affecting only targeted states while leaving other parts of the policy unchanged. This paper rigorously investigates whether and how such locality can be exploited to improve estimation of long-term effects in Markov Decision Processes (MDPs), a fundamental model of dynamic systems. We first develop optimal inference techniques for general A/B testing in MDPs and establish corresponding efficiency bounds. We then propose methods to harness the localized structure by sharing information on the non-targeted states. Our new estimator can achieve a linear reduction with the number of test arms for a major part of the variance without sacrificing unbiasedness. It also matches a tighter variance lower bound that accounts for locality. Furthermore, we extend our framework to a broad class of differentiable estimators, which encompasses many widely used approaches in practice. We show that all such estimators can benefit from variance reduction through information sharing without increasing their bias. Together, these results provide both theoretical foundations and practical tools for conducting efficient experiments in dynamic service systems with local treatments.", "date": "2024-07-29", "categories": []}
{"id": "2509.01535", "title": "CAT: Causal Attention Tuning For Injecting Fine-grained Causal Knowledge into Large Language Models", "abstract": "Large Language Models (LLMs) have achieved remarkable success across various domains. However, a fundamental question remains: Can LLMs effectively utilize causal knowledge for prediction and generation? Through empirical studies, we find that LLMs trained directly on large-scale data often capture spurious correlations rather than true causal relationships, leading to suboptimal performance, especially in out-of-distribution (OOD) scenarios. To address this challenge, we propose Causal Attention Tuning (CAT), a novel approach that injects fine-grained causal knowledge into the attention mechanism. We propose an automated pipeline that leverages human priors to automatically generate token-level causal signals and introduce the Re-Attention mechanism to guide training, helping the model focus on causal structures while mitigating noise and biases in attention scores. Experimental results on our proposed Spurious Token Game (STG) benchmark and multiple downstream tasks demonstrate that our approach effectively leverages causal knowledge for prediction and remains robust in OOD scenarios. The CAT achieves an average improvement of 5.76% on the STG dataset and 1.56% on downstream tasks. Notably, the OOD performance of the Llama-3.1-8B model on STG_M increased from 64.5% to 90.5%, and Qwen's OOD performance on the STG_H dataset improved from 25.4% to 55.9%. Implementation details can be found at https://github.com/Kairong-Han/CAT.", "date": "2025-09-01", "categories": []}
{"id": "2509.07701", "title": "Building causation links in stochastic nonlinear systems from data", "abstract": "Causal relationships play a fundamental role in understanding the world around us. The ability to identify and understand cause-effect relationships is critical to making informed decisions, predicting outcomes, and developing effective strategies. However, deciphering causal relationships from observational data is a difficult task, as correlations alone may not provide definitive evidence of causality. In recent years, the field of machine learning (ML) has emerged as a powerful tool, offering new opportunities for uncovering hidden causal mechanisms and better understanding complex systems. In this work, we address the issue of detecting the intrinsic causal links of a large class of complex systems in the framework of the response theory in physics. We develop some theoretical ideas put forward by [1], and technically we use state-of-the-art ML techniques to build up models from data. We consider both linear stochastic and non-linear systems. Finally, we compute the asymptotic efficiency of the linear response based causal predictor in a case of large scale Markov process network of linear interactions.", "date": "2025-09-09", "categories": []}
{"id": "2507.03897", "title": "GenAI-Powered Inference", "abstract": "We introduce GenAI-Powered Inference (GPI), a statistical framework for both causal and predictive inference using unstructured data, including text and images. GPI leverages open-source Generative Artificial Intelligence (GenAI) models -- such as large language models and diffusion models -- not only to generate unstructured data at scale but also to extract low-dimensional representations that are guaranteed to capture their underlying structure. Applying machine learning to these representations, GPI enables estimation of causal and predictive effects while quantifying associated estimation uncertainty. Unlike existing approaches to representation learning, GPI does not require fine-tuning of generative models, making it computationally efficient and broadly accessible. We illustrate the versatility of the GPI framework through three applications: (1) analyzing Chinese social media censorship, (2) estimating predictive effects of candidates' facial appearance on electoral outcomes, and (3) assessing the persuasiveness of political rhetoric. An open-source software package is available for implementing GPI.", "date": "2025-07-05", "categories": []}
{"id": "2508.10479", "title": "Confounding is a Pervasive Problem in Real World Recommender Systems", "abstract": "Unobserved confounding arises when an unmeasured feature influences both the treatment and the outcome, leading to biased causal effect estimates. This issue undermines observational studies in fields like economics, medicine, ecology or epidemiology. Recommender systems leveraging fully observed data seem not to be vulnerable to this problem. However many standard practices in recommender systems result in observed features being ignored, resulting in effectively the same problem. This paper will show that numerous common practices such as feature engineering, A/B testing and modularization can in fact introduce confounding into recommendation systems and hamper their performance. Several illustrations of the phenomena are provided, supported by simulation studies with practical suggestions about how practitioners may reduce or avoid the affects of confounding in real systems.", "date": "2025-08-14", "categories": []}
{"id": "2509.02130", "title": "Online Identification of IT Systems through Active Causal Learning", "abstract": "Identifying a causal model of an IT system is fundamental to many branches of systems engineering and operation. Such a model can be used to predict the effects of control actions, optimize operations, diagnose failures, detect intrusions, etc., which is central to achieving the longstanding goal of automating network and system management tasks. Traditionally, causal models have been designed and maintained by domain experts. This, however, proves increasingly challenging with the growing complexity and dynamism of modern IT systems. In this paper, we present the first principled method for online, data-driven identification of an IT system in the form of a causal model. The method, which we call active causal learning, estimates causal functions that capture the dependencies among system variables in an iterative fashion using Gaussian process regression based on system measurements, which are collected through a rollout-based intervention policy. We prove that this method is optimal in the Bayesian sense and that it produces effective interventions. Experimental validation on a testbed shows that our method enables accurate identification of a causal system model while inducing low interference with system operations.", "date": "2025-09-02", "categories": []}
{"id": "2301.11732", "title": "Convolutional neural networks for valid and efficient causal inference", "abstract": "Convolutional neural networks (CNN) have been successful in machine learning applications. Their success relies on their ability to consider space invariant local features. We consider the use of CNN to fit nuisance models in semiparametric estimation of the average causal effect of a treatment. In this setting, nuisance models are functions of pre-treatment covariates that need to be controlled for. In an application where we want to estimate the effect of early retirement on a health outcome, we propose to use CNN to control for time-structured covariates. Thus, CNN is used when fitting nuisance models explaining the treatment and the outcome. These fits are then combined into an augmented inverse probability weighting estimator yielding efficient and uniformly valid inference. Theoretically, we contribute by providing rates of convergence for CNN equipped with the rectified linear unit activation function and compare it to an existing result for feedforward neural networks. We also show when those rates guarantee uniformly valid inference. A Monte Carlo study is provided where the performance of the proposed estimator is evaluated and compared with other strategies. Finally, we give results on a study of the effect of early retirement on hospitalization using data covering the whole Swedish population.", "date": "2023-01-27", "categories": []}
{"id": "2502.12796", "title": "Learning Counterfactually Fair Models via Improved Generation with Neural Causal Models", "abstract": "One of the main concerns while deploying machine learning models in real-world applications is fairness. Counterfactual fairness has emerged as an intuitive and natural definition of fairness. However, existing methodologies for enforcing counterfactual fairness seem to have two limitations: (i) generating counterfactual samples faithful to the underlying causal graph, and (ii) as we argue in this paper, existing regularizers are mere proxies and do not directly enforce the exact definition of counterfactual fairness. In this work, our aim is to mitigate both issues. Firstly, we propose employing Neural Causal Models (NCMs) for generating the counterfactual samples. For implementing the abduction step in NCMs, the posteriors of the exogenous variables need to be estimated given a counterfactual query, as they are not readily available. As a consequence, $\\mathcal{L}_3$ consistency with respect to the underlying causal graph cannot be guaranteed in practice due to the estimation errors involved. To mitigate this issue, we propose a novel kernel least squares loss term that enforces the $\\mathcal{L}_3$ constraints explicitly. Thus, we obtain an improved counterfactual generation suitable for the counterfactual fairness task. Secondly, we propose a new MMD-based regularizer term that explicitly enforces the counterfactual fairness conditions into the base model while training. We show an improved trade-off between counterfactual fairness and generalization over existing baselines on synthetic and benchmark datasets.", "date": "2025-02-18", "categories": []}
{"id": "2509.03548", "title": "Multilinear and Linear Programs for Partially Identifiable Queries in Quasi-Markovian Structural Causal Models", "abstract": "We investigate partially identifiable queries in a class of causal models. We focus on acyclic Structural Causal Models that are quasi-Markovian (that is, each endogenous variable is connected with at most one exogenous confounder). We look into scenarios where endogenous variables are observed (and a distribution over them is known), while exogenous variables are not fully specified. This leads to a representation that is in essence a Bayesian network where the distribution of root variables is not uniquely determined. In such circumstances, it may not be possible to precisely compute a probability value of interest. We thus study the computation of tight probability bounds, a problem that has been solved by multilinear programming in general, and by linear programming when a single confounded component is intervened upon. We present a new algorithm to simplify the construction of such programs by exploiting input probabilities over endogenous variables. For scenarios with a single intervention, we apply column generation to compute a probability bound through a sequence of auxiliary linear integer programs, thus showing that a representation with polynomial cardinality for exogenous variables is possible. Experiments show column generation techniques to be superior to existing methods.", "date": "2025-09-02", "categories": []}
{"id": "2509.04296", "title": "Using causal abstractions to accelerate decision-making in complex bandit problems", "abstract": "Although real-world decision-making problems can often be encoded as causal multi-armed bandits (CMABs) at different levels of abstraction, a general methodology exploiting the information and computational advantages of each abstraction level is missing. In this paper, we propose AT-UCB, an algorithm which efficiently exploits shared information between CMAB problem instances defined at different levels of abstraction. More specifically, AT-UCB leverages causal abstraction (CA) theory to explore within a cheap-to-simulate and coarse-grained CMAB instance, before employing the traditional upper confidence bound (UCB) algorithm on a restricted set of potentially optimal actions in the CMAB of interest, leading to significant reductions in cumulative regret when compared to the classical UCB algorithm. We illustrate the advantages of AT-UCB theoretically, through a novel upper bound on the cumulative regret, and empirically, by applying AT-UCB to epidemiological simulators with varying resolution and computational cost.", "date": "2025-09-04", "categories": []}
{"id": "2507.17686", "title": "Debiased maximum-likelihood estimators for hazard ratios under kernel-based machine-learning adjustment", "abstract": "Previous studies have shown that hazard ratios between treatment groups estimated with the Cox model are uninterpretable because the unspecified baseline hazard of the model fails to identify temporal change in the risk set composition due to treatment assignment and unobserved factors among multiple, contradictory scenarios. To alleviate this problem, especially in studies based on observational data with uncontrolled dynamic treatment and real-time measurement of many covariates, we propose abandoning the baseline hazard and using kernel-based machine learning to explicitly model the change in the risk set with or without latent variables. For this framework, we clarify the context in which hazard ratios can be causally interpreted, and then develop a method based on Neyman orthogonality to compute debiased maximum-likelihood estimators of hazard ratios, proving necessary convergence results. Numerical simulations confirm that the proposed method identifies the true hazard ratios with minimal bias. These results lay the foundation for developing a useful, alternative method for causal inference with uncontrolled, observational data in modern epidemiology.", "date": "2025-07-23", "categories": []}
{"id": "2509.03438", "title": "Non-Linear Counterfactual Aggregate Optimization", "abstract": "We consider the problem of directly optimizing a non-linear function of an outcome, where this outcome itself is the sum of many small contributions. The non-linearity of the function means that the problem is not equivalent to the maximization of the expectation of the individual contribution. By leveraging the concentration properties of the sum of individual outcomes, we derive a scalable descent algorithm that directly optimizes for our stated objective. This allows for instance to maximize the probability of successful A/B test, for which it can be wiser to target a success criterion, such as exceeding a given uplift, rather than chasing the highest expected payoff.", "date": "2025-09-03", "categories": []}
{"id": "2503.00134", "title": "Personalized Causal Graph Reasoning for LLMs: An Implementation for Dietary Recommendations", "abstract": "Large Language Models (LLMs) excel at general-purpose reasoning by leveraging broad commonsense knowledge, but they remain limited in tasks requiring personalized reasoning over multifactorial personal data. This limitation constrains their applicability in domains such as healthcare, where decisions must adapt to individual contexts. We introduce Personalized Causal Graph Reasoning, a framework that enables LLMs to reason over individual-specific causal graphs constructed from longitudinal data. Each graph encodes how user-specific factors influence targeted outcomes. In response to a query, the LLM traverses the graph to identify relevant causal pathways, rank them by estimated impact, simulate potential outcomes, and generate tailored responses. We implement this framework in the context of nutrient-oriented dietary recommendations, where variability in metabolic responses demands personalized reasoning. Using counterfactual evaluation, we assess the effectiveness of LLM-generated food suggestions for glucose control. Our method reduces postprandial glucose iAUC across three time windows compared to prior approaches. Additional LLM-as-a-judge evaluations further confirm improvements in personalization quality.", "date": "2025-02-28", "categories": []}
{"id": "2504.01551", "title": "Identifying Macro Causal Effects in a C-DMG over ADMGs", "abstract": "Causal effect identification using causal graphs is a fundamental challenge in causal inference. While extensive research has been conducted in this area, most existing methods assume the availability of fully specified directed acyclic graphs or acyclic directed mixed graphs. However, in complex domains such as medicine and epidemiology, complete causal knowledge is often unavailable, and only partial information about the system is accessible. This paper focuses on causal effect identification within partially specified causal graphs, with particular emphasis on cluster-directed mixed graphs (C-DMGs) which can represent many different acyclic directed mixed graphs (ADMGs). These graphs provide a higher-level representation of causal relationships by grouping variables into clusters, offering a more practical approach for handling complex systems. Unlike fully specified ADMGs, C-DMGs can contain cycles, which complicate their analysis and interpretation. Furthermore, their cluster-based nature introduces new challenges, as it gives rise to two distinct types of causal effects: macro causal effects and micro causal effects, each with different properties. In this work, we focus on macro causal effects, which describe the effects of entire clusters on other clusters. We establish that the do-calculus is both sound and complete for identifying these effects in C-DMGs over ADMGs when the cluster sizes are either unknown or of size greater than one. Additionally, we provide a graphical characterization of non-identifiability for macro causal effects in these graphs.", "date": "2025-04-02", "categories": []}
{"id": "2509.00538", "title": "Identifying Causal Direction via Dense Functional Classes", "abstract": "We address the problem of determining the causal direction between two univariate, continuous-valued variables, X and Y, under the assumption of no hidden confounders. In general, it is not possible to make definitive statements about causality without some assumptions on the underlying model. To distinguish between cause and effect, we propose a bivariate causal score based on the Minimum Description Length (MDL) principle, using functions that possess the density property on a compact real interval. We prove the identifiability of these causal scores under specific conditions. These conditions can be easily tested. Gaussianity of the noise in the causal model equations is not assumed, only that the noise is low. The well-studied class of cubic splines possesses the density property on a compact real interval. We propose LCUBE as an instantiation of the MDL-based causal score utilizing cubic regression splines. LCUBE is an identifiable method that is also interpretable, simple, and very fast. It has only one hyperparameter. Empirical evaluations compared to state-of-the-art methods demonstrate that LCUBE achieves superior precision in terms of AUDRC on the real-world Tuebingen cause-effect pairs dataset. It also shows superior average precision across common 10 benchmark datasets and achieves above average precision on 13 datasets.", "date": "2025-08-30", "categories": []}
{"id": "2509.00602", "title": "TranCIT: Transient Causal Interaction Toolbox", "abstract": "Quantifying transient causal interactions from non-stationary neural signals is a fundamental challenge in neuroscience. Traditional methods are often inadequate for brief neural events, and advanced, event-specific techniques have lacked accessible implementations within the Python ecosystem. Here, we introduce trancit (Transient Causal Interaction Toolbox), an open-source Python package designed to bridge this gap. TranCIT implements a comprehensive analysis pipeline, including Granger Causality, Transfer Entropy, and the more robust Structural Causal Model-based Dynamic Causal Strength (DCS) and relative Dynamic Causal Strength (rDCS) for accurately detecting event-driven causal effects. We demonstrate TranCIT's utility by successfully capturing causality in high-synchrony regimes where traditional methods fail and by identifying the known transient information flow from hippocampal CA3 to CA1 during sharp-wave ripple events in real-world data. The package offers a user-friendly, validated solution for investigating the transient causal dynamics that govern complex systems.", "date": "2025-08-30", "categories": []}
{"id": "2509.00846", "title": "Causal SHAP: Feature Attribution with Dependency Awareness through Causal Discovery", "abstract": "Explaining machine learning (ML) predictions has become crucial as ML models are increasingly deployed in high-stakes domains such as healthcare. While SHapley Additive exPlanations (SHAP) is widely used for model interpretability, it fails to differentiate between causality and correlation, often misattributing feature importance when features are highly correlated. We propose Causal SHAP, a novel framework that integrates causal relationships into feature attribution while preserving many desirable properties of SHAP. By combining the Peter-Clark (PC) algorithm for causal discovery and the Intervention Calculus when the DAG is Absent (IDA) algorithm for causal strength quantification, our approach addresses the weakness of SHAP. Specifically, Causal SHAP reduces attribution scores for features that are merely correlated with the target, as validated through experiments on both synthetic and real-world datasets. This study contributes to the field of Explainable AI (XAI) by providing a practical framework for causal-aware model explanations. Our approach is particularly valuable in domains such as healthcare, where understanding true causal relationships is critical for informed decision-making.", "date": "2025-08-31", "categories": []}
{"id": "2509.01352", "title": "Causal Sensitivity Identification using Generative Learning", "abstract": "In this work, we propose a novel generative method to identify the causal impact and apply it to prediction tasks. We conduct causal impact analysis using interventional and counterfactual perspectives. First, applying interventions, we identify features that have a causal influence on the predicted outcome, which we refer to as causally sensitive features, and second, applying counterfactuals, we evaluate how changes in the cause affect the effect. Our method exploits the Conditional Variational Autoencoder (CVAE) to identify the causal impact and serve as a generative predictor. We are able to reduce confounding bias by identifying causally sensitive features. We demonstrate the effectiveness of our method by recommending the most likely locations a user will visit next in their spatiotemporal trajectory influenced by the causal relationships among various features. Experiments on the large-scale GeoLife [Zheng et al., 2010] dataset and the benchmark Asia Bayesian network validate the ability of our method to identify causal impact and improve predictive performance.", "date": "2025-09-01", "categories": []}
{"id": "2509.01558", "title": "Direct Profit Estimation Using Uplift Modeling under Clustered Network Interference", "abstract": "Uplift modeling is a key technique for promotion optimization in recommender systems, but standard methods typically fail to account for interference, where treating one item affects the outcomes of others. This violation of the Stable Unit Treatment Value Assumption (SUTVA) leads to suboptimal policies in real-world marketplaces. Recent developments in interference-aware estimators such as Additive Inverse Propensity Weighting (AddIPW) have not found their way into the uplift modeling literature yet, and optimising policies using these estimators is not well-established. This paper proposes a practical methodology to bridge this gap. We use the AddIPW estimator as a differentiable learning objective suitable for gradient-based optimization. We demonstrate how this framework can be integrated with proven response transformation techniques to directly optimize for economic outcomes like incremental profit. Through simulations, we show that our approach significantly outperforms interference-naive methods, especially as interference effects grow. Furthermore, we find that adapting profit-centric uplift strategies within our framework can yield superior performance in identifying the highest-impact interventions, offering a practical path toward more profitable incentive personalization.", "date": "2025-09-01", "categories": []}
{"id": "2509.01621", "title": "Effects of Distributional Biases on Gradient-Based Causal Discovery in the Bivariate Categorical Case", "abstract": "Gradient-based causal discovery shows great potential for deducing causal structure from data in an efficient and scalable way. Those approaches however can be susceptible to distributional biases in the data they are trained on. We identify two such biases: Marginal Distribution Asymmetry, where differences in entropy skew causal learning toward certain factorizations, and Marginal Distribution Shift Asymmetry, where repeated interventions cause faster shifts in some variables than in others. For the bivariate categorical setup with Dirichlet priors, we illustrate how these biases can occur even in controlled synthetic data. To examine their impact on gradient-based methods, we employ two simple models that derive causal factorizations by learning marginal or conditional data distributions - a common strategy in gradient-based causal discovery. We demonstrate how these models can be susceptible to both biases. We additionally show how the biases can be controlled. An empirical evaluation of two related, existing approaches indicates that eliminating competition between possible causal factorizations can make models robust to the presented biases.", "date": "2025-09-01", "categories": []}
{"id": "2509.01887", "title": "Design of Experiment for Discovering Directed Mixed Graph", "abstract": "We study the problem of experimental design for accurately identifying the causal graph structure of a simple structural causal model (SCM), where the underlying graph may include both cycles and bidirected edges induced by latent confounders. The presence of cycles renders it impossible to recover the graph skeleton using observational data alone, while confounding can further invalidate traditional conditional independence (CI) tests in certain scenarios. To address these challenges, we establish lower bounds on both the maximum number of variables that can be intervened upon in a single experiment and the total number of experiments required to identify all directed edges and non-adjacent bidirected edges. Leveraging both CI tests and do see tests, and accounting for $d$ separation and $\\sigma$ separation, we develop two classes of algorithms, i.e., bounded and unbounded, that can recover all causal edges except for double adjacent bidirected edges. We further show that, up to logarithmic factors, the proposed algorithms are tight with respect to the derived lower bounds.", "date": "2025-09-02", "categories": []}
{"id": "2509.02535", "title": "Probabilities of Causation and Root Cause Analysis with Quasi-Markovian Models", "abstract": "Probabilities of causation provide principled ways to assess causal relationships but face computational challenges due to partial identifiability and latent confounding. This paper introduces both algorithmic simplifications, significantly reducing the computational complexity of calculating tighter bounds for these probabilities, and a novel methodological framework for Root Cause Analysis that systematically employs these causal metrics to rank entire causal paths.", "date": "2025-09-02", "categories": []}
{"id": "2508.21722", "title": "Inferring Effects of Major Events through Discontinuity Forecasting of Population Anxiety", "abstract": "Estimating community-specific mental health effects of local events is vital for public health policy. While forecasting mental health scores alone offers limited insights into the impact of events on community well-being, quasi-experimental designs like the Longitudinal Regression Discontinuity Design (LRDD) from econometrics help researchers derive more effects that are more likely to be causal from observational data. LRDDs aim to extrapolate the size of changes in an outcome (e.g. a discontinuity in running scores for anxiety) due to a time-specific event. Here, we propose adapting LRDDs beyond traditional forecasting into a statistical learning framework whereby future discontinuities (i.e. time-specific shifts) and changes in slope (i.e. linear trajectories) are estimated given a location's history of the score, dynamic covariates (other running assessments), and exogenous variables (static representations). Applying our framework to predict discontinuities in the anxiety of US counties from COVID-19 events, we found the task was difficult but more achievable as the sophistication of models was increased, with the best results coming from integrating exogenous and dynamic covariates. Our approach shows strong improvement ($r=+.46$ for discontinuity and $r = +.65$ for slope) over traditional static community representations. Discontinuity forecasting raises new possibilities for estimating the idiosyncratic effects of potential future or hypothetical events on specific communities.", "date": "2025-08-29", "categories": []}
{"id": "2505.20130", "title": "Balancing Interference and Correlation in Spatial Experimental Designs: A Causal Graph Cut Approach", "abstract": "This paper focuses on the design of spatial experiments to optimize the amount of information derived from the experimental data and enhance the accuracy of the resulting causal effect estimator. We propose a surrogate function for the mean squared error (MSE) of the estimator, which facilitates the use of classical graph cut algorithms to learn the optimal design. Our proposal offers three key advances: (1) it accommodates moderate to large spatial interference effects; (2) it adapts to different spatial covariance functions; (3) it is computationally efficient. Theoretical results and numerical experiments based on synthetic environments and a dispatch simulator that models a city-scale ridesharing market, further validate the effectiveness of our design. A python implementation of our method is available at https://github.com/Mamba413/CausalGraphCut.", "date": "2025-05-26", "categories": []}
{"id": "2508.20259", "title": "Latent Variable Modeling for Robust Causal Effect Estimation", "abstract": "Latent variable models provide a powerful framework for incorporating and inferring unobserved factors in observational data. In causal inference, they help account for hidden factors influencing treatment or outcome, thereby addressing challenges posed by missing or unmeasured covariates. This paper proposes a new framework that integrates latent variable modeling into the double machine learning (DML) paradigm to enable robust causal effect estimation in the presence of such hidden factors. We consider two scenarios: one where a latent variable affects only the outcome, and another where it may influence both treatment and outcome. To ensure tractability, we incorporate latent variables only in the second stage of DML, separating representation learning from latent inference. We demonstrate the robustness and effectiveness of our method through extensive experiments on both synthetic and real-world datasets.", "date": "2025-08-27", "categories": []}
{"id": "2508.20335", "title": "Dynamic Synthetic Controls vs. Panel-Aware Double Machine Learning for Geo-Level Marketing Impact Estimation", "abstract": "Accurately quantifying geo-level marketing lift in two-sided marketplaces is challenging: the Synthetic Control Method (SCM) often exhibits high power yet systematically under-estimates effect size, while panel-style Double Machine Learning (DML) is seldom benchmarked against SCM. We build an open, fully documented simulator that mimics a typical large-scale geo roll-out: N_unit regional markets are tracked for T_pre weeks before launch and for a further T_post-week campaign window, allowing all key parameters to be varied by the user and probe both families under five stylized stress tests: 1) curved baseline trends, 2) heterogeneous response lags, 3) treated-biased shocks, 4) a non-linear outcome link, and 5) a drifting control group trend.\n  Seven estimators are evaluated: three standard Augmented SCM (ASC) variants and four panel-DML flavors (TWFE, CRE/Mundlak, first-difference, and within-group). Across 100 replications per scenario, ASC models consistently demonstrate severe bias and near-zero coverage in challenging scenarios involving nonlinearities or external shocks. By contrast, panel-DML variants dramatically reduce this bias and restore nominal 95%-CI coverage, proving far more robust.\n  The results indicate that while ASC provides a simple baseline, it is unreliable in common, complex situations. We therefore propose a 'diagnose-first' framework where practitioners first identify the primary business challenge (e.g., nonlinear trends, response lags) and then select the specific DML model best suited for that scenario, providing a more robust and reliable blueprint for analyzing geo-experiments.", "date": "2025-08-28", "categories": []}
{"id": "2508.20371", "title": "P2C: Path to Counterfactuals", "abstract": "Machine-learning models are increasingly driving decisions in high-stakes settings, such as finance, law, and hiring, thus, highlighting the need for transparency. However, the key challenge is to balance transparency -- clarifying `why' a decision was made -- with recourse: providing actionable steps on `how' to achieve a favourable outcome from an unfavourable outcome. Counterfactual explanations reveal `why' an undesired outcome occurred and `how' to reverse it through targeted feature changes (interventions).\n  Current counterfactual approaches have limitations: 1) they often ignore causal dependencies between features, and 2) they typically assume all interventions can happen simultaneously, an unrealistic assumption in practical scenarios where actions are typically taken in a sequence. As a result, these counterfactuals are often not achievable in the real world.\n  We present P2C (Path-to-Counterfactuals), a model-agnostic framework that produces a plan (ordered sequence of actions) converting an unfavourable outcome to a causally consistent favourable outcome. P2C addresses both limitations by 1) Explicitly modelling causal relationships between features and 2) Ensuring that each intermediate state in the plan is feasible and causally valid. P2C uses the goal-directed Answer Set Programming system s(CASP) to generate the plan accounting for feature changes that happen automatically due to causal dependencies. Furthermore, P2C refines cost (effort) computation by only counting changes actively made by the user, resulting in realistic cost estimates. Finally, P2C highlights how its causal planner outperforms standard planners, which lack causal knowledge and thus can generate illegal actions.", "date": "2025-08-28", "categories": []}
{"id": "2508.19327", "title": "Quantum Entanglement as Super-Confounding: From Bell's Theorem to Robust Machine Learning", "abstract": "Bell's theorem reveals a profound conflict between quantum mechanics and local realism, a conflict we reinterpret through the modern lens of causal inference. We propose and computationally validate a framework where quantum entanglement acts as a \"super-confounding\" resource, generating correlations that violate the classical causal bounds set by Bell's inequalities. This work makes three key contributions: First, we establish a physical hierarchy of confounding (Quantum > Classical) and introduce Confounding Strength (CS) to quantify this effect. Second, we provide a circuit-based implementation of the quantum $\\mathcal{DO}$-calculus to distinguish causality from spurious correlation. Finally, we apply this calculus to a quantum machine learning problem, where causal feature selection yields a statistically significant 11.3% average absolute improvement in model robustness. Our framework bridges quantum foundations and causal AI, offering a new, practical perspective on quantum correlations.", "date": "2025-08-26", "categories": []}
{"id": "2305.10032", "title": "A Survey on Causal Discovery: Theory and Practice", "abstract": "Understanding the laws that govern a phenomenon is the core of scientific progress. This is especially true when the goal is to model the interplay between different aspects in a causal fashion. Indeed, causal inference itself is specifically designed to quantify the underlying relationships that connect a cause to its effect. Causal discovery is a branch of the broader field of causality in which causal graphs are recovered from data (whenever possible), enabling the identification and estimation of causal effects. In this paper, we explore recent advancements in causal discovery in a unified manner, provide a consistent overview of existing algorithms developed under different settings, report useful tools and data, present real-world applications to understand why and how these methods can be fruitfully exploited.", "date": "2023-05-17", "categories": []}
{"id": "2306.16902", "title": "Integrating Large Language Model for Improved Causal Discovery", "abstract": "Recovering the structure of causal graphical models from observational data is an essential yet challenging task for causal discovery in scientific scenarios. Domain-specific causal discovery usually relies on expert validation or prior analysis to improve the reliability of recovered causality, which is yet limited by the scarcity of expert resources. Recently, Large Language Models (LLM) have been used for causal analysis across various domain-specific scenarios, suggesting its potential as autonomous expert roles in guiding data-based structure learning. However, integrating LLMs into causal discovery faces challenges due to inaccuracies in LLM-based reasoning on revealing the actual causal structure. To address this challenge, we propose an error-tolerant LLM-driven causal discovery framework. The error-tolerant mechanism is designed three-fold with sufficient consideration on potential inaccuracies. In the LLM-based reasoning process, an accuracy-oriented prompting strategy restricts causal analysis to a reliable range. Next, a knowledge-to-structure transition aligns LLM-derived causal statements with structural causal interactions. In the structure learning process, the goodness-of-fit to data and adherence to LLM-derived priors are balanced to further address prior inaccuracies. Evaluation of eight real-world causal structures demonstrates the efficacy of our LLM-driven approach in improving data-based causal discovery, along with its robustness to inaccurate LLM-derived priors. Codes are available at https://github.com/tyMadara/LLM-CD.", "date": "2023-06-29", "categories": []}
{"id": "2508.17576", "title": "CausalSent: Interpretable Sentiment Classification with RieszNet", "abstract": "Despite the overwhelming performance improvements offered by recent natural language processing (NLP) models, the decisions made by these models are largely a black box. Towards closing this gap, the field of causal NLP combines causal inference literature with modern NLP models to elucidate causal effects of text features. We replicate and extend Bansal et al's work on regularizing text classifiers to adhere to estimated effects, focusing instead on model interpretability. Specifically, we focus on developing a two-headed RieszNet-based neural network architecture which achieves better treatment effect estimation accuracy. Our framework, CausalSent, accurately predicts treatment effects in semi-synthetic IMDB movie reviews, reducing MAE of effect estimates by 2-3x compared to Bansal et al's MAE on synthetic Civil Comments data. With an ensemble of validated models, we perform an observational case study on the causal effect of the word \"love\" in IMDB movie reviews, finding that the presence of the word \"love\" causes a +2.9% increase in the probability of a positive sentiment.", "date": "2025-08-25", "categories": []}
{"id": "2508.18873", "title": "MOCHA: Discovering Multi-Order Dynamic Causality in Temporal Point Processes", "abstract": "Discovering complex causal dependencies in temporal point processes (TPPs) is critical for modeling real-world event sequences. Existing methods typically rely on static or first-order causal structures, overlooking the multi-order and time-varying nature of causal relationships. In this paper, we propose MOCHA, a novel framework for discovering multi-order dynamic causality in TPPs. MOCHA characterizes multi-order influences as multi-hop causal paths over a latent time-evolving graph. To model such dynamics, we introduce a time-varying directed acyclic graph (DAG) with learnable structural weights, where acyclicity and sparsity constraints are enforced to ensure structural validity. We design an end-to-end differentiable framework that jointly models causal discovery and TPP dynamics, enabling accurate event prediction and revealing interpretable structures. Extensive experiments on real-world datasets demonstrate that MOCHA not only achieves state-of-the-art performance in event prediction, but also reveals meaningful and interpretable causal structures.", "date": "2025-08-26", "categories": []}
{"id": "2508.16910", "title": "Unbiased Reasoning for Knowledge-Intensive Tasks in Large Language Models via Conditional Front-Door Adjustment", "abstract": "Large Language Models (LLMs) have shown impressive capabilities in natural language processing but still struggle to perform well on knowledge-intensive tasks that require deep reasoning and the integration of external knowledge. Although methods such as Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) have been proposed to enhance LLMs with external knowledge, they still suffer from internal bias in LLMs, which often leads to incorrect answers. In this paper, we propose a novel causal prompting framework, Conditional Front-Door Prompting (CFD-Prompting), which enables the unbiased estimation of the causal effect between the query and the answer, conditional on external knowledge, while mitigating internal bias. By constructing counterfactual external knowledge, our framework simulates how the query behaves under varying contexts, addressing the challenge that the query is fixed and is not amenable to direct causal intervention. Compared to the standard front-door adjustment, the conditional variant operates under weaker assumptions, enhancing both robustness and generalisability of the reasoning process. Extensive experiments across multiple LLMs and benchmark datasets demonstrate that CFD-Prompting significantly outperforms existing baselines in both accuracy and robustness.", "date": "2025-08-23", "categories": []}
{"id": "2508.17136", "title": "Factor Informed Double Deep Learning For Average Treatment Effect Estimation", "abstract": "We investigate the problem of estimating the average treatment effect (ATE) under a very general setup where the covariates can be high-dimensional, highly correlated, and can have sparse nonlinear effects on the propensity and outcome models. We present the use of a Double Deep Learning strategy for estimation, which involves combining recently developed factor-augmented deep learning-based estimators, FAST-NN, for both the response functions and propensity scores to achieve our goal. By using FAST-NN, our method can select variables that contribute to propensity and outcome models in a completely nonparametric and algorithmic manner and adaptively learn low-dimensional function structures through neural networks. Our proposed novel estimator, FIDDLE (Factor Informed Double Deep Learning Estimator), estimates ATE based on the framework of augmented inverse propensity weighting AIPW with the FAST-NN-based response and propensity estimates. FIDDLE consistently estimates ATE even under model misspecification and is flexible to also allow for low-dimensional covariates. Our method achieves semiparametric efficiency under a very flexible family of propensity and outcome models. We present extensive numerical studies on synthetic and real datasets to support our theoretical guarantees and establish the advantages of our methods over other traditional choices, especially when the data dimension is large.", "date": "2025-08-23", "categories": []}
{"id": "2508.17207", "title": "Explainable Counterfactual Reasoning in Depression Medication Selection at Multi-Levels (Personalized and Population)", "abstract": "Background: This study investigates how variations in Major Depressive Disorder (MDD) symptoms, quantified by the Hamilton Rating Scale for Depression (HAM-D), causally influence the prescription of SSRIs versus SNRIs. Methods: We applied explainable counterfactual reasoning with counterfactual explanations (CFs) to assess the impact of specific symptom changes on antidepressant choice. Results: Among 17 binary classifiers, Random Forest achieved highest performance (accuracy, F1, precision, recall, ROC-AUC near 0.85). Sample-based CFs revealed both local and global feature importance of individual symptoms in medication selection. Conclusions: Counterfactual reasoning elucidates which MDD symptoms most strongly drive SSRI versus SNRI selection, enhancing interpretability of AI-based clinical decision support systems. Future work should validate these findings on more diverse cohorts and refine algorithms for clinical deployment.", "date": "2025-08-24", "categories": []}
{"id": "2508.15928", "title": "Transforming Causality: Transformer-Based Temporal Causal Discovery with Prior Knowledge Integration", "abstract": "We introduce a novel framework for temporal causal discovery and inference that addresses two key challenges: complex nonlinear dependencies and spurious correlations. Our approach employs a multi-layer Transformer-based time-series forecaster to capture long-range, nonlinear temporal relationships among variables. After training, we extract the underlying causal structure and associated time lags from the forecaster using gradient-based analysis, enabling the construction of a causal graph. To mitigate the impact of spurious causal relationships, we introduce a prior knowledge integration mechanism based on attention masking, which consistently enforces user-excluded causal links across multiple Transformer layers. Extensive experiments show that our method significantly outperforms other state-of-the-art approaches, achieving a 12.8% improvement in F1-score for causal discovery and 98.9% accuracy in estimating causal lags.", "date": "2025-08-21", "categories": []}
{"id": "2303.16148", "title": "Causal Modelling of Cryptocurrency Price Movements Using Discretisation-Aware Bayesian Networks", "abstract": "This study identifies the key factors influencing the price movements of major cryptocurrencies, Bitcoin, Binance Coin, Ethereum, Litecoin, Ripple, and Tether, using Bayesian networks (BNs). This study addresses two key challenges: modelling price movements in highly volatile cryptocurrency markets and enhancing predictive performance through discretisation-aware Bayesian Networks. It analyses both macro-financial indicators (gold, oil, MSCI, S and P 500, USDX) and social media signals (tweet volume) as potential price drivers. Moreover, since discretisation is a critical step in the effectiveness of BNs, we implement a structured procedure to build 54 BNs models by combining three discretisation methods (equal interval, equal quantile, and k-means) with several bin counts. These models are evaluated using four metrics, including balanced accuracy, F1 score, area under the ROC curve and a composite score. Results show that equal interval with two bins consistently yields the best predictive performance. We also provide deeper insights into each network's structure through inference, sensitivity, and influence strength analyses. These analyses reveal distinct price-driving patterns for each cryptocurrency, underscore the importance of coin-specific analysis, and demonstrate the value of BNs for interpretable causal modelling in volatile cryptocurrency markets.", "date": "2023-03-26", "categories": []}
{"id": "2409.05500", "title": "Scalable Time-Series Causal Discovery with Approximate Causal Ordering", "abstract": "Causal discovery in time-series data presents a significant computational challenge. Standard algorithms are often prohibitively expensive for datasets with many variables or samples. This study introduces and validates a heuristic approximation of the VarLiNGAM algorithm to address this scalability problem. The standard VarLiNGAM method relies on an iterative search, recalculating statistical dependencies after each step. Our heuristic modifies this procedure by omitting the iterative refinement. This change permits a one-time precomputation of all necessary statistical values. The algorithmic modification reduces the time complexity from $O(m^3n)$ to $O(m^2n + m^3)$ while keeping the space complexity at $O(m^2)$, where $m$ is the number of variables and $n$ is the number of samples. While an approximation, our approach retains VarLiNGAM's essential structure and empirical reliability. On large-scale financial data with up to 400 variables, our algorithm achieves a 7--13x speedup over the standard implementation and a 4.5x speedup over a GPU-accelerated version. Evaluations across medical imaging, web server monitoring, and finance demonstrate the heuristic's robustness and practical scalability. This work offers a validated balance between computational efficiency and discovery quality, making large-scale causal analysis feasible on personal computers.", "date": "2024-09-09", "categories": []}
{"id": "2501.08429", "title": "Modeling Discrimination with Causal Abstraction", "abstract": "A person is directly racially discriminated against only if her race caused her worse treatment. This implies that race is an attribute sufficiently separable from other attributes to isolate its causal role. But race is embedded in a nexus of social factors that resist isolated treatment. If race is socially constructed, in what sense can it cause worse treatment? Some propose that the perception of race, rather than race itself, causes worse treatment. Others suggest that since causal models require \\textit{modularity}, i.e. the ability to isolate causal effects, attempts to causally model discrimination are misguided.\n  This paper addresses the problem differently. We introduce a framework for reasoning about discrimination, in which race is a high-level \\textit{abstraction} of lower-level features. In this framework, race can be modeled as itself causing worse treatment. Modularity is ensured by allowing assumptions about social construction to be precisely and explicitly stated, via an alignment between race and its constituents. Such assumptions can then be subjected to normative and empirical challenges, which lead to different views of when discrimination occurs. By distinguishing constitutive and causal relations, the abstraction framework pinpoints disagreements in the current literature on modeling discrimination, while preserving a precise causal account of discrimination.", "date": "2025-01-14", "categories": []}
{"id": "2507.17792", "title": "Causal Mechanism Estimation in Multi-Sensor Systems Across Multiple Domains", "abstract": "To gain deeper insights into a complex sensor system through the lens of causality, we present common and individual causal mechanism estimation (CICME), a novel three-step approach to inferring causal mechanisms from heterogeneous data collected across multiple domains. By leveraging the principle of Causal Transfer Learning (CTL), CICME is able to reliably detect domain-invariant causal mechanisms when provided with sufficient samples. The identified common causal mechanisms are further used to guide the estimation of the remaining causal mechanisms in each domain individually. The performance of CICME is evaluated on linear Gaussian models under scenarios inspired from a manufacturing process. Building upon existing continuous optimization-based causal discovery methods, we show that CICME leverages the benefits of applying causal discovery on the pooled data and repeatedly on data from individual domains, and it even outperforms both baseline methods under certain scenarios.", "date": "2025-07-23", "categories": []}
{"id": "2508.14432", "title": "Personalized Counterfactual Framework: Generating Potential Outcomes from Wearable Data", "abstract": "Wearable sensor data offer opportunities for personalized health monitoring, yet deriving actionable insights from their complex, longitudinal data streams is challenging. This paper introduces a framework to learn personalized counterfactual models from multivariate wearable data. This enables exploring what-if scenarios to understand potential individual-specific outcomes of lifestyle choices. Our approach first augments individual datasets with data from similar patients via multi-modal similarity analysis. We then use a temporal PC (Peter-Clark) algorithm adaptation to discover predictive relationships, modeling how variables at time t-1 influence physiological changes at time t. Gradient Boosting Machines are trained on these discovered relationships to quantify individual-specific effects. These models drive a counterfactual engine projecting physiological trajectories under hypothetical interventions (e.g., activity or sleep changes). We evaluate the framework via one-step-ahead predictive validation and by assessing the plausibility and impact of interventions. Evaluation showed reasonable predictive accuracy (e.g., mean heart rate MAE 4.71 bpm) and high counterfactual plausibility (median 0.9643). Crucially, these interventions highlighted significant inter-individual variability in response to hypothetical lifestyle changes, showing the framework's potential for personalized insights. This work provides a tool to explore personalized health dynamics and generate hypotheses on individual responses to lifestyle changes.", "date": "2025-08-20", "categories": []}
{"id": "2508.13607", "title": "Bounding Causal Effects and Counterfactuals", "abstract": "Causal inference often hinges on strong assumptions - such as no unmeasured confounding or perfect compliance - that are rarely satisfied in practice. Partial identification offers a principled alternative: instead of relying on unverifiable assumptions to estimate causal effects precisely, it derives bounds that reflect the uncertainty inherent in the data. Despite its theoretical appeal, partial identification remains underutilized in applied work, in part due to the fragmented nature of existing methods and the lack of practical guidance. This thesis addresses these challenges by systematically comparing a diverse set of bounding algorithms across multiple causal scenarios. We implement, extend, and unify state-of-the-art methods - including symbolic, optimization-based, and information-theoretic approaches - within a common evaluation framework. In particular, we propose an extension of a recently introduced entropy-bounded method, making it applicable to counterfactual queries such as the Probability of Necessity and Sufficiency (PNS). Our empirical study spans thousands of randomized simulations involving both discrete and continuous data-generating processes. We assess each method in terms of bound tightness, computational efficiency, and robustness to assumption violations. To support practitioners, we distill our findings into a practical decision tree for algorithm selection and train a machine learning model to predict the best-performing method based on observable data characteristics.\n  All implementations are released as part of an open-source Python package, CausalBoundingEngine, which enables users to apply and compare bounding methods through a unified interface.", "date": "2025-08-19", "categories": []}
{"id": "2508.13967", "title": "A PC Algorithm for Max-Linear Bayesian Networks", "abstract": "Max-linear Bayesian networks (MLBNs) are a relatively recent class of structural equation models which arise when the random variables involved have heavy-tailed distributions. Unlike most directed graphical models, MLBNs are typically not faithful to d-separation and thus classical causal discovery algorithms such as the PC algorithm or greedy equivalence search can not be used to accurately recover the true graph structure. In this paper, we begin the study of constraint-based discovery algorithms for MLBNs given an oracle for testing conditional independence in the true, unknown graph. We show that if the oracle is given by the $\\ast$-separation criteria in the true graph, then the PC algorithm remains consistent despite the presence of additional CI statements implied by $\\ast$-separation. We also introduce a new causal discovery algorithm named \"PCstar\" which assumes faithfulness to $C^\\ast$-separation and is able to orient additional edges which cannot be oriented with only d- or $\\ast$-separation.", "date": "2025-08-19", "categories": []}
{"id": "2311.01301", "title": "TRIALSCOPE: A Unifying Causal Framework for Scaling Real-World Evidence Generation with Biomedical Language Models", "abstract": "The rapid digitization of real-world data presents an unprecedented opportunity to optimize healthcare delivery and accelerate biomedical discovery. However, these data are often found in unstructured forms such as clinical notes in electronic medical records (EMRs), and is typically plagued by confounders, making it challenging to generate robust real-world evidence (RWE). Therefore, we present TRIALSCOPE, a framework designed to distil RWE from population level observational data at scale. TRIALSCOPE leverages biomedical language models to structure clinical text at scale, employs advanced probabilistic modeling for denoising and imputation, and incorporates state-of-the-art causal inference techniques to address common confounders in treatment effect estimation. Extensive experiments were conducted on a large-scale dataset of over one million cancer patients from a single large healthcare network in the United States. TRIALSCOPE was shown to automatically curate high-quality structured patient data, expanding the dataset and incorporating key patient attributes only available in unstructured form. The framework reduces confounding in treatment effect estimation, generating comparable results to randomized controlled lung cancer trials. Additionally, we demonstrate simulations of unconducted clinical trials - including a pancreatic cancer trial with varying eligibility criteria - using a suite of validation tests to ensure robustness. Thorough ablation studies were conducted to better understand key components of TRIALSCOPE and establish best practices for RWE generation from EMRs. TRIALSCOPE was able to extract data cancer treatment data from EMRs, overcoming limitations of manual curation. We were also able to show that TRIALSCOPE could reproduce results of lung and pancreatic cancer clinical trials from the extracted real world data.", "date": "2023-11-02", "categories": []}
{"id": "2410.08362", "title": "Towards Optimal Environmental Policies: Policy Learning under Arbitrary Bipartite Network Interference", "abstract": "The substantial effect of air pollution on cardiovascular disease and mortality burdens is well-established. Emissions-reducing interventions on coal-fired power plants -- a major source of hazardous air pollution -- have proven to be an effective, but costly, strategy for reducing pollution-related health burdens. Targeting the power plants that achieve maximum health benefits while satisfying realistic cost constraints is challenging. The primary difficulty lies in quantifying the health benefits of intervening at particular plants. This is further complicated because interventions are applied on power plants, while health impacts occur in potentially distant communities, a setting known as bipartite network interference (BNI). In this paper, we introduce novel policy learning methods based on Q- and A-Learning to determine the optimal policy under arbitrary BNI. We derive asymptotic properties and demonstrate finite sample efficacy in simulations. We apply our novel methods to a comprehensive dataset of Medicare claims, power plant data, and pollution transport networks. Our goal is to determine the optimal strategy for installing power plant scrubbers to minimize ischemic heart disease (IHD) hospitalizations under various cost constraints. We find that annual IHD hospitalization rates could be reduced in a range from 23.37-55.30 per 10,000 person-years through optimal policies under different cost constraints.", "date": "2024-10-10", "categories": []}
{"id": "2410.14789", "title": "Differentially Private Covariate Balancing Causal Inference", "abstract": "Differential privacy is the leading mathematical framework for privacy protection, providing a probabilistic guarantee that safeguards individuals' private information when publishing statistics from a dataset. This guarantee is achieved by applying a randomized algorithm to the original data, which introduces unique challenges in data analysis by distorting inherent patterns. In particular, causal inference using observational data in privacy-sensitive contexts is challenging because it requires covariate balance between treatment groups, yet checking the true covariates is prohibited to prevent leakage of sensitive information. In this article, we present a differentially private two-stage covariate balancing weighting estimator to infer causal effects from observational data. Our algorithm produces both point and interval estimators with statistical guarantees, such as consistency and rate optimality, under a given privacy budget.", "date": "2024-10-18", "categories": []}
{"id": "2508.11741", "title": "BaMANI: Bayesian Multi-Algorithm causal Network Inference", "abstract": "Improved computational power has enabled different disciplines to predict causal relationships among modeled variables using Bayesian network inference. While many alternative algorithms have been proposed to improve the efficiency and reliability of network prediction, the predicted causal networks reflect the generative process but also bear an opaque imprint of the specific computational algorithm used. Following a ``wisdom of the crowds\" strategy, we developed an ensemble learning approach to marginalize the impact of a single algorithm on Bayesian causal network inference. To introduce the approach, we first present the theoretical foundation of this framework. Next, we present a comprehensive implementation of the framework in terms of a new software tool called BaMANI (Bayesian Multi-Algorithm causal Network Inference). Finally, we describe a BaMANI use-case from biology, particularly within human breast cancer studies.", "date": "2025-08-15", "categories": []}
{"id": "2508.12500", "title": "Root Cause Analysis of Hydrogen Bond Separation in Spatio-Temporal Molecular Dynamics using Causal Models", "abstract": "Molecular dynamics simulations (MDS) face challenges, including resource-heavy computations and the need to manually scan outputs to detect \"interesting events,\" such as the formation and persistence of hydrogen bonds between atoms of different molecules. A critical research gap lies in identifying the underlying causes of hydrogen bond formation and separation -understanding which interactions or prior events contribute to their emergence over time. With this challenge in mind, we propose leveraging spatio-temporal data analytics and machine learning models to enhance the detection of these phenomena. In this paper, our approach is inspired by causal modeling and aims to identify the root cause variables of hydrogen bond formation and separation events. Specifically, we treat the separation of hydrogen bonds as an \"intervention\" occurring and represent the causal structure of the bonding and separation events in the MDS as graphical causal models. These causal models are built using a variational autoencoder-inspired architecture that enables us to infer causal relationships across samples with diverse underlying causal graphs while leveraging shared dynamic information. We further include a step to infer the root causes of changes in the joint distribution of the causal models. By constructing causal models that capture shifts in the conditional distributions of molecular interactions during bond formation or separation, this framework provides a novel perspective on root cause analysis in molecular dynamic systems. We validate the efficacy of our model empirically on the atomic trajectories that used MDS for chiral separation, demonstrating that we can predict many steps in the future and also find the variables driving the observed changes in the system.", "date": "2025-08-17", "categories": []}
{"id": "2508.12798", "title": "A Shift in Perspective on Causality in Domain Generalization", "abstract": "The promise that causal modelling can lead to robust AI generalization has been challenged in recent work on domain generalization (DG) benchmarks. We revisit the claims of the causality and DG literature, reconciling apparent contradictions and advocating for a more nuanced theory of the role of causality in generalization. We also provide an interactive demo at https://chai-uk.github.io/ukairs25-causal-predictors/.", "date": "2025-08-18", "categories": []}
{"id": "2505.07503", "title": "Identifying Causal Direction via Variational Bayesian Compression", "abstract": "Telling apart the cause and effect between two random variables with purely observational data is a challenging problem that finds applications in various scientific disciplines. A key principle utilized in this task is the algorithmic Markov condition, which postulates that the joint distribution, when factorized according to the causal direction, yields a more succinct codelength compared to the anti-causal direction. Previous approaches approximate these codelengths by relying on simple functions or Gaussian processes (GPs) with easily evaluable complexity, compromising between model fitness and computational complexity. To address these limitations, we propose leveraging the variational Bayesian learning of neural networks as an interpretation of the codelengths. This allows the improvement of model fitness, while maintaining the succinctness of the codelengths, and the avoidance of the significant computational complexity of the GP-based approaches. Extensive experiments on both synthetic and real-world benchmarks in cause-effect identification demonstrate the effectiveness of our proposed method, showing promising performance enhancements on several datasets in comparison to most related methods.", "date": "2025-05-12", "categories": []}
{"id": "2508.10581", "title": "Technical Report: Facilitating the Adoption of Causal Inference Methods Through LLM-Empowered Co-Pilot", "abstract": "Estimating treatment effects (TE) from observational data is a critical yet complex task in many fields, from healthcare and economics to public policy. While recent advances in machine learning and causal inference have produced powerful estimation techniques, their adoption remains limited due to the need for deep expertise in causal assumptions, adjustment strategies, and model selection. In this paper, we introduce CATE-B, an open-source co-pilot system that uses large language models (LLMs) within an agentic framework to guide users through the end-to-end process of treatment effect estimation. CATE-B assists in (i) constructing a structural causal model via causal discovery and LLM-based edge orientation, (ii) identifying robust adjustment sets through a novel Minimal Uncertainty Adjustment Set criterion, and (iii) selecting appropriate regression methods tailored to the causal structure and dataset characteristics. To encourage reproducibility and evaluation, we release a suite of benchmark tasks spanning diverse domains and causal complexities. By combining causal inference with intelligent, interactive assistance, CATE-B lowers the barrier to rigorous causal analysis and lays the foundation for a new class of benchmarks in automated treatment effect estimation.", "date": "2025-08-14", "categories": []}
{"id": "2505.17917", "title": "M-learner:A Flexible And Powerful Framework To Study Heterogeneous Treatment Effect In Mediation Model", "abstract": "We propose a novel method, termed the M-learner, for estimating heterogeneous indirect and total treatment effects and identifying relevant subgroups within a mediation framework. The procedure comprises four key steps. First, we compute individual-level conditional average indirect/total treatment effect Second, we construct a distance matrix based on pairwise differences. Third, we apply tSNE to project this matrix into a low-dimensional Euclidean space, followed by K-means clustering to identify subgroup structures. Finally, we calibrate and refine the clusters using a threshold-based procedure to determine the optimal configuration. To the best of our knowledge, this is the first approach specifically designed to capture treatment effect heterogeneity in the presence of mediation. Experimental results validate the robustness and effectiveness of the proposed framework. Application to the real-world Jobs II dataset highlights the broad adaptability and potential applicability of our method.Code is available at https: //anonymous.4open.science/r/M-learner-C4BB.", "date": "2025-05-23", "categories": []}
{"id": "2508.09154", "title": "Peer Effect Estimation in the Presence of Simultaneous Feedback and Unobserved Confounders", "abstract": "Estimating peer causal effects within complex real-world networks such as social networks is challenging, primarily due to simultaneous feedback between peers and unobserved confounders. Existing methods either address unobserved confounders while ignoring the simultaneous feedback, or account for feedback but under restrictive linear assumptions, thus failing to obtain accurate peer effect estimation. In this paper, we propose DIG2RSI, a novel Deep learning framework which leverages I-G transformation (matrix operation) and 2SRI (an instrumental variable or IV technique) to address both simultaneous feedback and unobserved confounding, while accommodating complex, nonlinear and high-dimensional relationships. DIG2RSI first applies the I-G transformation to disentangle mutual peer influences and eliminate the bias due to the simultaneous feedback. To deal with unobserved confounding, we first construct valid IVs from network data. In stage 1 of 2RSI, we train a neural network on these IVs to predict peer exposure, and extract residuals as proxies for the unobserved confounders. In the stage 2, we fit a separate neural network augmented by an adversarial discriminator that incorporates these residuals as a control function and enforces the learned representation to contain no residual confounding signal. The expressive power of deep learning models in capturing complex non-linear relationships and adversarial debiasing enhances the effectiveness of DIG2RSI in eliminating bias from both feedback loops and hidden confounders. We prove consistency of our estimator under standard regularity conditions, ensuring asymptotic recovery of the true peer effect. Empirical results on two semi-synthetic benchmarks and a real-world dataset demonstrate that DIG2RSI outperforms existing approaches.", "date": "2025-08-05", "categories": []}
{"id": "2508.09447", "title": "NEXICA: Discovering Road Traffic Causality (Extended arXiv Version)", "abstract": "Road traffic congestion is a persistent problem. Focusing resources on the causes of congestion is a potentially efficient strategy for reducing slowdowns. We present NEXICA, an algorithm to discover which parts of the highway system tend to cause slowdowns on other parts of the highway. We use time series of road speeds as inputs to our causal discovery algorithm. Finding other algorithms inadequate, we develop a new approach that is novel in three ways. First, it concentrates on just the presence or absence of events in the time series, where an event indicates the temporal beginning of a traffic slowdown. Second, we develop a probabilistic model using maximum likelihood estimation to compute the probabilities of spontaneous and caused slowdowns between two locations on the highway. Third, we train a binary classifier to identify pairs of cause/effect locations trained on pairs of road locations where we are reasonably certain a priori of their causal connections, both positive and negative. We test our approach on six months of road speed data from 195 different highway speed sensors in the Los Angeles area, showing that our approach is superior to state-of-the-art baselines in both accuracy and computation speed.", "date": "2025-08-13", "categories": []}
{"id": "2508.08295", "title": "Topos Causal Models", "abstract": "We propose topos causal models (TCMs), a novel class of causal models that exploit the key properties of a topos category: they are (co)complete, meaning all (co)limits exist, they admit a subobject classifier, and allow exponential objects. The main goal of this paper is to show that these properties are central to many applications in causal inference. For example, subobject classifiers allow a categorical formulation of causal intervention, which creates sub-models. Limits and colimits allow causal diagrams of arbitrary complexity to be ``solved\", using a novel interpretation of causal approximation. Exponential objects enable reasoning about equivalence classes of operations on causal models, such as covered edge reversal and causal homotopy. Analogous to structural causal models (SCMs), TCMs are defined by a collection of functions, each defining a ``local autonomous\" causal mechanism that assemble to induce a unique global function from exogenous to endogenous variables. Since the category of TCMs is (co)complete, which we prove in this paper, every causal diagram has a ``solution\" in the form of a (co)limit: this implies that any arbitrary causal model can be ``approximated\" by some global function with respect to the morphisms going into or out of the diagram. Natural transformations are crucial in measuring the quality of approximation. In addition, we show that causal interventions are modeled by subobject classifiers: any sub-model is defined by a monic arrow into its parent model. Exponential objects permit reasoning about entire classes of causal equivalences and interventions. Finally, as TCMs form a topos, they admit an internal logic defined as a Mitchell-Benabou language with an associated Kripke-Joyal semantics. We show how to reason about causal models in TCMs using this internal logic.", "date": "2025-08-05", "categories": []}
{"id": "2508.08752", "title": "Sensitivity Analysis to Unobserved Confounding with Copula-based Normalizing Flows", "abstract": "We propose a novel method for sensitivity analysis to unobserved confounding in causal inference. The method builds on a copula-based causal graphical normalizing flow that we term $\\rho$-GNF, where $\\rho \\in [-1,+1]$ is the sensitivity parameter. The parameter represents the non-causal association between exposure and outcome due to unobserved confounding, which is modeled as a Gaussian copula. In other words, the $\\rho$-GNF enables scholars to estimate the average causal effect (ACE) as a function of $\\rho$, accounting for various confounding strengths. The output of the $\\rho$-GNF is what we term the $\\rho_{curve}$, which provides the bounds for the ACE given an interval of assumed $\\rho$ values. The $\\rho_{curve}$ also enables scholars to identify the confounding strength required to nullify the ACE. We also propose a Bayesian version of our sensitivity analysis method. Assuming a prior over the sensitivity parameter $\\rho$ enables us to derive the posterior distribution over the ACE, which enables us to derive credible intervals. Finally, leveraging on experiments from simulated and real-world data, we show the benefits of our sensitivity analysis method.", "date": "2025-08-12", "categories": []}
{"id": "2508.08883", "title": "Position: Causal Machine Learning Requires Rigorous Synthetic Experiments for Broader Adoption", "abstract": "Causal machine learning has the potential to revolutionize decision-making by combining the predictive power of machine learning algorithms with the theory of causal inference. However, these methods remain underutilized by the broader machine learning community, in part because current empirical evaluations do not permit assessment of their reliability and robustness, undermining their practical utility. Specifically, one of the principal criticisms made by the community is the extensive use of synthetic experiments. We argue, on the contrary, that synthetic experiments are essential and necessary to precisely assess and understand the capabilities of causal machine learning methods. To substantiate our position, we critically review the current evaluation practices, spotlight their shortcomings, and propose a set of principles for conducting rigorous empirical analyses with synthetic data. Adopting the proposed principles will enable comprehensive evaluations that build trust in causal machine learning methods, driving their broader adoption and impactful real-world use.", "date": "2025-08-12", "categories": []}
{"id": "2508.09059", "title": "Causal Machine Learning for Patient-Level Intraoperative Opioid Dose Prediction from Electronic Health Records", "abstract": "This paper introduces the OPIAID algorithm, a novel approach for predicting and recommending personalized opioid dosages for individual patients. The algorithm optimizes pain management while minimizing opioid related adverse events (ORADE) by employing machine learning models trained on observational electronic health records (EHR) data. It leverages a causal machine learning approach to understand the relationship between opioid dose, case specific patient and intraoperative characteristics, and pain versus ORADE outcomes. The OPIAID algorithm considers patient-specific characteristics and the influence of different opiates, enabling personalized dose recommendations. This paper outlines the algorithm's methodology and architecture, and discusses key assumptions, and approaches to evaluating its performance.", "date": "2025-08-12", "categories": []}
{"id": "2407.13313", "title": "Sortability of Time Series Data", "abstract": "Evaluating the performance of causal discovery algorithms that aim to find causal relationships between time-dependent processes remains a challenging topic. In this paper, we show that certain characteristics of datasets, such as varsortability (Reisach et al. 2021) and $R^2$-sortability (Reisach et al. 2023), also occur in datasets for autocorrelated stationary time series. We illustrate this empirically using four types of data: simulated data based on SVAR models and Erd\\H{o}s-R\\'enyi graphs, the data used in the 2019 causality-for-climate challenge (Runge et al. 2019), real-world river stream datasets, and real-world data generated by the Causal Chamber of (Gamella et al. 2024). To do this, we adapt var- and $R^2$-sortability to time series data. We also investigate the extent to which the performance of score-based causal discovery methods goes hand in hand with high sortability. Arguably, our most surprising finding is that the investigated real-world datasets exhibit high varsortability and low $R^2$-sortability indicating that scales may carry a significant amount of causal information.", "date": "2024-07-18", "categories": []}
{"id": "2508.02183", "title": "Multi-Treatment-DML: Causal Estimation for Multi-Dimensional Continuous Treatments with Monotonicity Constraints in Personal Loan Risk Optimization", "abstract": "Optimizing credit limits, interest rates, and loan terms is crucial for managing borrower risk and lifetime value (LTV) in personal loan platform. However, counterfactual estimation of these continuous, multi-dimensional treatments faces significant challenges: randomized trials are often prohibited by risk controls and long repayment cycles, forcing reliance on biased observational data. Existing causal methods primarily handle binary/discrete treatments and struggle with continuous, multi-dimensional settings. Furthermore, financial domain knowledge mandates provably monotonic treatment-outcome relationships (e.g., risk increases with credit limit).To address these gaps, we propose Multi-Treatment-DML, a novel framework leveraging Double Machine Learning (DML) to: (i) debias observational data for causal effect estimation; (ii) handle arbitrary-dimensional continuous treatments; and (iii) enforce monotonic constraints between treatments and outcomes, guaranteeing adherence to domain requirements.Extensive experiments on public benchmarks and real-world industrial datasets demonstrate the effectiveness of our approach. Furthermore, online A/B testing conducted on a realworld personal loan platform, confirms the practical superiority of Multi-Treatment-DML in real-world loan operations.", "date": "2025-08-04", "categories": []}
{"id": "2508.06742", "title": "Learning Causal Structure Distributions for Robust Planning", "abstract": "Structural causal models describe how the components of a robotic system interact. They provide both structural and functional information about the relationships that are present in the system. The structural information outlines the variables among which there is interaction. The functional information describes how such interactions work, via equations or learned models. In this paper we find that learning the functional relationships while accounting for the uncertainty about the structural information leads to more robust dynamics models which improves downstream planning, while using significantly lower computational resources. This in contrast with common model-learning methods that ignore the causal structure and fail to leverage the sparsity of interactions in robotic systems. We achieve this by estimating a causal structure distribution that is used to sample causal graphs that inform the latent-space representations in an encoder-multidecoder probabilistic model. We show that our model can be used to learn the dynamics of a robot, which together with a sampling-based planner can be used to perform new tasks in novel environments, provided an objective function for the new requirement is available. We validate our method using manipulators and mobile robots in both simulation and the real-world. Additionally, we validate the learned dynamics' adaptability and increased robustness to corrupted inputs and changes in the environment, which is highly desirable in challenging real-world robotics scenarios. Video: https://youtu.be/X6k5t7OOnNc.", "date": "2025-08-08", "categories": []}
{"id": "2508.07221", "title": "LLM-based Agents for Automated Confounder Discovery and Subgroup Analysis in Causal Inference", "abstract": "Estimating individualized treatment effects from observational data presents a persistent challenge due to unmeasured confounding and structural bias. Causal Machine Learning (causal ML) methods, such as causal trees and doubly robust estimators, provide tools for estimating conditional average treatment effects. These methods have limited effectiveness in complex real-world environments due to the presence of latent confounders or those described in unstructured formats. Moreover, reliance on domain experts for confounder identification and rule interpretation introduces high annotation cost and scalability concerns. In this work, we proposed Large Language Model-based agents for automated confounder discovery and subgroup analysis that integrate agents into the causal ML pipeline to simulate domain expertise. Our framework systematically performs subgroup identification and confounding structure discovery by leveraging the reasoning capabilities of LLM-based agents, which reduces human dependency while preserving interpretability. Experiments on real-world medical datasets show that our proposed approach enhances treatment effect estimation robustness by narrowing confidence intervals and uncovering unrecognized confounding biases. Our findings suggest that LLM-based agents offer a promising path toward scalable, trustworthy, and semantically aware causal inference.", "date": "2025-08-10", "categories": []}
{"id": "2503.01290", "title": "ACTIVA: Amortized Causal Effect Estimation via Transformer-based Variational Autoencoder", "abstract": "Predicting the distribution of outcomes under hypothetical interventions is crucial across healthcare, economics, and policy-making. However, existing methods often require restrictive assumptions, and are typically limited by the lack of amortization across problem instances. We propose ACTIVA, a transformer-based conditional variational autoencoder (VAE) architecture for amortized causal inference, which estimates interventional distributions directly from observational data without. ACTIVA learns a latent representation conditioned on observational inputs and intervention queries, enabling zero-shot inference by amortizing causal knowledge from diverse training scenarios. We provide theoretical insights showing that ACTIVA predicts interventional distributions as mixtures over observationally equivalent causal models. Empirical evaluations on synthetic and semi-synthetic datasets confirm the effectiveness of our amortized approach and highlight promising directions for future real-world applications.", "date": "2025-03-03", "categories": []}
{"id": "2401.09641", "title": "Functional Linear Non-Gaussian Acyclic Model for Causal Discovery", "abstract": "In causal discovery, non-Gaussianity has been used to characterize the complete configuration of a Linear Non-Gaussian Acyclic Model (LiNGAM), encompassing both the causal ordering of variables and their respective connection strengths. However, LiNGAM can only deal with the finite-dimensional case. To expand this concept, we extend the notion of variables to encompass vectors and even functions, leading to the Functional Linear Non-Gaussian Acyclic Model (Func-LiNGAM). Our motivation stems from the desire to identify causal relationships in brain-effective connectivity tasks involving, for example, fMRI and EEG datasets. We demonstrate why the original LiNGAM fails to handle these inherently infinite-dimensional datasets and explain the availability of functional data analysis from both empirical and theoretical perspectives. {We establish theoretical guarantees of the identifiability of the causal relationship among non-Gaussian random vectors and even random functions in infinite-dimensional Hilbert spaces.} To address the issue of sparsity in discrete time points within intrinsic infinite-dimensional functional data, we propose optimizing the coordinates of the vectors using functional principal component analysis. Experimental results on synthetic data verify the ability of the proposed framework to identify causal relationships among multivariate functions using the observed samples. For real data, we focus on analyzing the brain connectivity patterns derived from fMRI data.", "date": "2024-01-17", "categories": []}
{"id": "2508.05215", "title": "DFW: A Novel Weighting Scheme for Covariate Balancing and Treatment Effect Estimation", "abstract": "Estimating causal effects from observational data is challenging due to selection bias, which leads to imbalanced covariate distributions across treatment groups. Propensity score-based weighting methods are widely used to address this issue by reweighting samples to simulate a randomized controlled trial (RCT). However, the effectiveness of these methods heavily depends on the observed data and the accuracy of the propensity score estimator. For example, inverse propensity weighting (IPW) assigns weights based on the inverse of the propensity score, which can lead to instable weights when propensity scores have high variance-either due to data or model misspecification-ultimately degrading the ability of handling selection bias and treatment effect estimation. To overcome these limitations, we propose Deconfounding Factor Weighting (DFW), a novel propensity score-based approach that leverages the deconfounding factor-to construct stable and effective sample weights. DFW prioritizes less confounded samples while mitigating the influence of highly confounded ones, producing a pseudopopulation that better approximates a RCT. Our approach ensures bounded weights, lower variance, and improved covariate balance.While DFW is formulated for binary treatments, it naturally extends to multi-treatment settings, as the deconfounding factor is computed based on the estimated probability of the treatment actually received by each sample. Through extensive experiments on real-world benchmark and synthetic datasets, we demonstrate that DFW outperforms existing methods, including IPW and CBPS, in both covariate balancing and treatment effect estimation.", "date": "2025-08-07", "categories": []}
{"id": "2505.09660", "title": "On Measuring Intrinsic Causal Attributions in Deep Neural Networks", "abstract": "Quantifying the causal influence of input features within neural networks has become a topic of increasing interest. Existing approaches typically assess direct, indirect, and total causal effects. This work treats NNs as structural causal models (SCMs) and extends our focus to include intrinsic causal contributions (ICC). We propose an identifiable generative post-hoc framework for quantifying ICC. We also draw a relationship between ICC and Sobol' indices. Our experiments on synthetic and real-world datasets demonstrate that ICC generates more intuitive and reliable explanations compared to existing global explanation techniques.", "date": "2025-05-14", "categories": []}
{"id": "2508.04216", "title": "Causal Reward Adjustment: Mitigating Reward Hacking in External Reasoning via Backdoor Correction", "abstract": "External reasoning systems combine language models with process reward models (PRMs) to select high-quality reasoning paths for complex tasks such as mathematical problem solving. However, these systems are prone to reward hacking, where high-scoring but logically incorrect paths are assigned high scores by the PRMs, leading to incorrect answers. From a causal inference perspective, we attribute this phenomenon primarily to the presence of confounding semantic features. To address it, we propose Causal Reward Adjustment (CRA), a method that mitigates reward hacking by estimating the true reward of a reasoning path. CRA trains sparse autoencoders on the PRM's internal activations to recover interpretable features, then corrects confounding by using backdoor adjustment. Experiments on math solving datasets demonstrate that CRA mitigates reward hacking and improves final accuracy, without modifying the policy model or retraining PRM.", "date": "2025-08-06", "categories": []}
{"id": "2508.03167", "title": "Causal identification with $Y_0$", "abstract": "We present the $Y_0$ Python package, which implements causal identification algorithms that apply interventional, counterfactual, and transportability queries to data from (randomized) controlled trials, observational studies, or mixtures thereof. $Y_0$ focuses on the qualitative investigation of causation, helping researchers determine whether a causal relationship can be estimated from available data before attempting to estimate how strong that relationship is. Furthermore, $Y_0$ provides guidance on how to transform the causal query into a symbolic estimand that can be non-parametrically estimated from the available data. $Y_0$ provides a domain-specific language for representing causal queries and estimands as symbolic probabilistic expressions, tools for representing causal graphical models with unobserved confounders, such as acyclic directed mixed graphs (ADMGs), and implementations of numerous identification algorithms from the recent causal inference literature. The $Y_0$ source code can be found under the MIT License at https://github.com/y0-causal-inference/y0 and it can be installed with pip install y0.", "date": "2025-08-05", "categories": []}
{"id": "2508.01848", "title": "Causal Discovery in Multivariate Time Series through Mutual Information Featurization", "abstract": "Discovering causal relationships in complex multivariate time series is a fundamental scientific challenge. Traditional methods often falter, either by relying on restrictive linear assumptions or on conditional independence tests that become uninformative in the presence of intricate, non-linear dynamics. This paper proposes a new paradigm, shifting from statistical testing to pattern recognition. We hypothesize that a causal link creates a persistent and learnable asymmetry in the flow of information through a system's temporal graph, even when clear conditional independencies are obscured. We introduce Temporal Dependency to Causality (TD2C), a supervised learning framework that operationalizes this hypothesis. TD2C learns to recognize these complex causal signatures from a rich set of information-theoretic and statistical descriptors. Trained exclusively on a diverse collection of synthetic time series, TD2C demonstrates remarkable zero-shot generalization to unseen dynamics and established, realistic benchmarks. Our results show that TD2C achieves state-of-the-art performance, consistently outperforming established methods, particularly in high-dimensional and non-linear settings. By reframing the discovery problem, our work provides a robust and scalable new tool for uncovering causal structures in complex systems.", "date": "2025-08-03", "categories": []}
{"id": "2508.01865", "title": "Structure Maintained Representation Learning Neural Network for Causal Inference", "abstract": "Recent developments in causal inference have greatly shifted the interest from estimating the average treatment effect to the individual treatment effect. In this article, we improve the predictive accuracy of representation learning and adversarial networks in estimating individual treatment effects by introducing a structure keeper which maintains the correlation between the baseline covariates and their corresponding representations in the high dimensional space. We train a discriminator at the end of representation layers to trade off representation balance and information loss. We show that the proposed discriminator minimizes an upper bound of the treatment estimation error. We can address the tradeoff between distribution balance and information loss by considering the correlations between the learned representation space and the original covariate feature space. We conduct extensive experiments with simulated and real-world observational data to show that our proposed Structure Maintained Representation Learning (SMRL) algorithm outperforms state-of-the-art methods. We also demonstrate the algorithms on real electronic health record data from the MIMIC-III database.", "date": "2025-08-03", "categories": []}
{"id": "2508.01867", "title": "Counterfactual Reciprocal Recommender Systems for User-to-User Matching", "abstract": "Reciprocal recommender systems (RRS) in dating, gaming, and talent platforms require mutual acceptance for a match. Logged data, however, over-represents popular profiles due to past exposure policies, creating feedback loops that skew learning and fairness. We introduce Counterfactual Reciprocal Recommender Systems (CFRR), a causal framework to mitigate this bias. CFRR uses inverse propensity scored, self-normalized objectives. Experiments show CFRR improves NDCG@10 by up to 3.5% (e.g., from 0.459 to 0.475 on DBLP, from 0.299 to 0.307 on Synthetic), increases long-tail user coverage by up to 51% (from 0.504 to 0.763 on Synthetic), and reduces Gini exposure inequality by up to 24% (from 0.708 to 0.535 on Synthetic). CFRR offers a promising approach for more accurate and fair user-to-user matching.", "date": "2025-08-03", "categories": []}
{"id": "2508.02524", "title": "Causality and Interpretability for Electrical Distribution System faults", "abstract": "Causal analysis helps us understand variables that are responsible for system failures. This improves fault detection and makes system more reliable. In this work, we present a new method that combines causal inference with machine learning to classify faults in electrical distribution systems (EDS) using graph-based models. We first build causal graphs using transfer entropy (TE). Each fault case is represented as a graph, where the nodes are features such as voltage and current, and the edges demonstrate how these features influence each other. Then, the graphs are classified using machine learning and GraphSAGE where the model learns from both the node values and the structure of the graph to predict the type of fault. To make the predictions understandable, we further developed an integrated approach using GNNExplainer and Captums Integrated Gradients to highlight the nodes (features) that influences the most on the final prediction. This gives us clear insights into the possible causes of the fault. Our experiments show high accuracy: 99.44% on the EDS fault dataset, which is better than state of art models. By combining causal graphs with machine learning, our method not only predicts faults accurately but also helps understand their root causes. This makes it a strong and practical tool for improving system reliability.", "date": "2025-08-04", "categories": []}
{"id": "2211.10085", "title": "Identifying Unique Spatial-Temporal Bayesian Network without Markov Equivalence", "abstract": "Identifying vanilla Bayesian network to model spatial-temporal causality can be a critical yet challenging task. Different Markovian-equivalent directed acyclic graphs would be identified if the identifiability is not satisfied. To address this issue, Directed Cyclic Graph is proposed to drop the directed acyclic constraint. But it does not always hold, and cannot model dynamical time-series process. Then, Full Time Graph is proposed with introducing high-order time delay. Full Time Graph has no Markov equivalence class by assuming no instantaneous effects. But, it also assumes that the causality is invariant with varying time, that is not always satisfied in the spatio-temporal scenarios. Thus, in this work, a Spatial-Temporal Bayesian Network (STBN) is proposed to theoretically model the spatial-temporal causality from the perspective of information transfer. STBN explains the disappearance of network structure $X\\rightarrow Z \\rightarrow Y$ and $X\\leftarrow Z \\leftarrow Y$ by the principle of information path blocking. And finally, the uniqueness of STBN is proved. Based on this, a High-order Causal Entropy (HCE) algorithm is also proposed to uniquely identify STBN under time complexity $\\mathcal{O}(n^3\\tau_{max})$, where $n$ is the number of variables and $\\tau_{max}$ is the maximum time delay. Numerical experiments are conducted with comparison to other baseline algorithms. The results show that HCE algorithm obtains state-of-the-art identification accuracy. The code is available at https://github.com/KMY-SEU/HCE.", "date": "2022-11-18", "categories": []}
{"id": "2508.00658", "title": "Multi-Band Variable-Lag Granger Causality: A Unified Framework for Causal Time Series Inference across Frequencies", "abstract": "Understanding causal relationships in time series is fundamental to many domains, including neuroscience, economics, and behavioral science. Granger causality is one of the well-known techniques for inferring causality in time series. Typically, Granger causality frameworks have a strong fix-lag assumption between cause and effect, which is often unrealistic in complex systems. While recent work on variable-lag Granger causality (VLGC) addresses this limitation by allowing a cause to influence an effect with different time lags at each time point, it fails to account for the fact that causal interactions may vary not only in time delay but also across frequency bands. For example, in brain signals, alpha-band activity may influence another region with a shorter delay than slower delta-band oscillations. In this work, we formalize Multi-Band Variable-Lag Granger Causality (MB-VLGC) and propose a novel framework that generalizes traditional VLGC by explicitly modeling frequency-dependent causal delays. We provide a formal definition of MB-VLGC, demonstrate its theoretical soundness, and propose an efficient inference pipeline. Extensive experiments across multiple domains demonstrate that our framework significantly outperforms existing methods on both synthetic and real-world datasets, confirming its broad applicability to any type of time series data. Code and datasets are publicly available.", "date": "2025-08-01", "categories": []}
{"id": "2507.23488", "title": "Causal Reasoning in Pieces: Modular In-Context Learning for Causal Discovery", "abstract": "Causal inference remains a fundamental challenge for large language models. Recent advances in internal reasoning with large language models have sparked interest in whether state-of-the-art reasoning models can robustly perform causal discovery-a task where conventional models often suffer from severe overfitting and near-random performance under data perturbations. We study causal discovery on the Corr2Cause benchmark using the emergent OpenAI's o-series and DeepSeek-R model families and find that these reasoning-first architectures achieve significantly greater native gains than prior approaches. To capitalize on these strengths, we introduce a modular in-context pipeline inspired by the Tree-of-Thoughts and Chain-of-Thoughts methodologies, yielding nearly three-fold improvements over conventional baselines. We further probe the pipeline's impact by analyzing reasoning chain length, complexity, and conducting qualitative and quantitative comparisons between conventional and reasoning models. Our findings suggest that while advanced reasoning models represent a substantial leap forward, carefully structured in-context frameworks are essential to maximize their capabilities and offer a generalizable blueprint for causal discovery across diverse domains.", "date": "2025-07-31", "categories": []}
{"id": "2507.23495", "title": "Incorporating structural uncertainty in causal decision making", "abstract": "Practitioners making decisions based on causal effects typically ignore structural uncertainty. We analyze when this uncertainty is consequential enough to warrant methodological solutions (Bayesian model averaging over competing causal structures). Focusing on bivariate relationships ($X \\rightarrow Y$ vs. $X \\leftarrow Y$), we establish that model averaging is beneficial when: (1) structural uncertainty is moderate to high, (2) causal effects differ substantially between structures, and (3) loss functions are sufficiently sensitive to the size of the causal effect. We prove optimality results of our suggested methodological solution under regularity conditions and demonstrate through simulations that modern causal discovery methods can provide, within limits, the necessary quantification. Our framework complements existing robust causal inference approaches by addressing a distinct source of uncertainty typically overlooked in practice.", "date": "2025-07-31", "categories": []}
{"id": "2502.01575", "title": "Heterogeneous Treatment Effect in Time-to-Event Outcomes: Harnessing Censored Data with Recursively Imputed Trees", "abstract": "Tailoring treatments to individual needs is a central goal in fields such as medicine. A key step toward this goal is estimating Heterogeneous Treatment Effects (HTE) - the way treatments impact different subgroups. While crucial, HTE estimation is challenging with survival data, where time until an event (e.g., death) is key. Existing methods often assume complete observation, an assumption violated in survival data due to right-censoring, leading to bias and inefficiency. Cui et al. (2023) proposed a doubly-robust method for HTE estimation in survival data under no hidden confounders, combining a causal survival forest with an augmented inverse-censoring weighting estimator. However, we find it struggles under heavy censoring, which is common in rare-outcome problems such as Amyotrophic lateral sclerosis (ALS). Moreover, most current methods cannot handle instrumental variables, which are a crucial tool in the causal inference arsenal. We introduce Multiple Imputation for Survival Treatment Response (MISTR), a novel, general, and non-parametric method for estimating HTE in survival data. MISTR uses recursively imputed survival trees to handle censoring without directly modeling the censoring mechanism. Through extensive simulations and analysis of two real-world datasets-the AIDS Clinical Trials Group Protocol 175 and the Illinois unemployment dataset we show that MISTR outperforms prior methods under heavy censoring in the no-hidden-confounders setting, and extends to the instrumental variable setting. To our knowledge, MISTR is the first non-parametric approach for HTE estimation with unobserved confounders via instrumental variables.", "date": "2025-02-03", "categories": []}
{"id": "2507.21792", "title": "Hybrid Causal Identification and Causal Mechanism Clustering", "abstract": "Bivariate causal direction identification is a fundamental and vital problem in the causal inference field. Among binary causal methods, most methods based on additive noise only use one single causal mechanism to construct a causal model. In the real world, observations are always collected in different environments with heterogeneous causal relationships. Therefore, on observation data, this paper proposes a Mixture Conditional Variational Causal Inference model (MCVCI) to infer heterogeneous causality. Specifically, according to the identifiability of the Hybrid Additive Noise Model (HANM), MCVCI combines the superior fitting capabilities of the Gaussian mixture model and the neural network and elegantly uses the likelihoods obtained from the probabilistic bounds of the mixture conditional variational auto-encoder as causal decision criteria. Moreover, we model the casual heterogeneity into cluster numbers and propose the Mixture Conditional Variational Causal Clustering (MCVCC) method, which can reveal causal mechanism expression. Compared with state-of-the-art methods, the comprehensive best performance demonstrates the effectiveness of the methods proposed in this paper on several simulated and real data.", "date": "2025-07-29", "categories": []}
{"id": "2507.19882", "title": "Causality-aligned Prompt Learning via Diffusion-based Counterfactual Generation", "abstract": "Prompt learning has garnered attention for its efficiency over traditional model training and fine-tuning. However, existing methods, constrained by inadequate theoretical foundations, encounter difficulties in achieving causally invariant prompts, ultimately falling short of capturing robust features that generalize effectively across categories. To address these challenges, we introduce the $\\textit{\\textbf{DiCap}}$ model, a theoretically grounded $\\textbf{Di}$ffusion-based $\\textbf{C}$ounterf$\\textbf{a}$ctual $\\textbf{p}$rompt learning framework, which leverages a diffusion process to iteratively sample gradients from the marginal and conditional distributions of the causal model, guiding the generation of counterfactuals that satisfy the minimal sufficiency criterion. Grounded in rigorous theoretical derivations, this approach guarantees the identifiability of counterfactual outcomes while imposing strict bounds on estimation errors. We further employ a contrastive learning framework that leverages the generated counterfactuals, thereby enabling the refined extraction of prompts that are precisely aligned with the causal features of the data. Extensive experimental results demonstrate that our method performs excellently across tasks such as image classification, image-text retrieval, and visual question answering, with particularly strong advantages in unseen categories.", "date": "2025-07-26", "categories": []}
{"id": "2507.20349", "title": "From Observations to Causations: A GNN-based Probabilistic Prediction Framework for Causal Discovery", "abstract": "Causal discovery from observational data is challenging, especially with large datasets and complex relationships. Traditional methods often struggle with scalability and capturing global structural information. To overcome these limitations, we introduce a novel graph neural network (GNN)-based probabilistic framework that learns a probability distribution over the entire space of causal graphs, unlike methods that output a single deterministic graph. Our framework leverages a GNN that encodes both node and edge attributes into a unified graph representation, enabling the model to learn complex causal structures directly from data. The GNN model is trained on a diverse set of synthetic datasets augmented with statistical and information-theoretic measures, such as mutual information and conditional entropy, capturing both local and global data properties. We frame causal discovery as a supervised learning problem, directly predicting the entire graph structure. Our approach demonstrates superior performance, outperforming both traditional and recent non-GNN-based methods, as well as a GNN-based approach, in terms of accuracy and scalability on synthetic and real-world datasets without further training. This probabilistic framework significantly improves causal structure learning, with broad implications for decision-making and scientific discovery across various fields.", "date": "2025-07-27", "categories": []}
{"id": "2402.02672", "title": "Estimation of conditional average treatment effects on distributed confidential data", "abstract": "The estimation of conditional average treatment effects (CATEs) is an important topic in many scientific fields. CATEs can be estimated with high accuracy if data distributed across multiple parties are centralized. However, it is difficult to aggregate such data owing to confidentiality or privacy concerns. To address this issue, we propose data collaboration double machine learning, a method for estimating CATE models using privacy-preserving fusion data constructed from distributed sources, and evaluate its performance through simulations. We make three main contributions. First, our method enables estimation and testing of semi-parametric CATE models without iterative communication on distributed data, providing robustness to model mis-specification compared to parametric approaches. Second, it enables collaborative estimation across different time points and parties by accumulating a knowledge base. Third, our method performs as well as or better than existing methods in simulations using synthetic, semi-synthetic, and real-world datasets.", "date": "2024-02-05", "categories": []}
{"id": "2506.16629", "title": "Learning Causally Predictable Outcomes from Psychiatric Longitudinal Data", "abstract": "Causal inference in longitudinal biomedical data remains a central challenge, especially in psychiatry, where symptom heterogeneity and latent confounding frequently undermine classical estimators. Most existing methods for treatment effect estimation presuppose a fixed outcome variable and address confounding through observed covariate adjustment. However, the assumption of unconfoundedness may not hold for a fixed outcome in practice. To address this foundational limitation, we directly optimize the outcome definition to maximize causal identifiability. Our DEBIAS (Durable Effects with Backdoor-Invariant Aggregated Symptoms) algorithm learns non-negative, clinically interpretable weights for outcome aggregation, maximizing durable treatment effects and empirically minimizing both observed and latent confounding by leveraging the time-limited direct effects of prior treatments in psychiatric longitudinal data. The algorithm also furnishes an empirically verifiable test for outcome unconfoundedness. DEBIAS consistently outperforms state-of-the-art methods in recovering causal effects for clinically interpretable composite outcomes across comprehensive experiments in depression and schizophrenia.", "date": "2025-06-19", "categories": []}
{"id": "2506.20933", "title": "Lower Bounds on the Size of Markov Equivalence Classes", "abstract": "Causal discovery algorithms typically recover causal graphs only up to their Markov equivalence classes unless additional parametric assumptions are made. The sizes of these equivalence classes reflect the limits of what can be learned about the underlying causal graph from purely observational data. Under the assumptions of acyclicity, causal sufficiency, and a uniform model prior, Markov equivalence classes are known to be small on average. In this paper, we show that this is no longer the case when any of these assumptions is relaxed. Specifically, we prove exponentially large lower bounds for the expected size of Markov equivalence classes in three settings: sparse random directed acyclic graphs, uniformly random acyclic directed mixed graphs, and uniformly random directed cyclic graphs.", "date": "2025-06-26", "categories": []}
{"id": "2507.18903", "title": "Probably Approximately Correct Causal Discovery", "abstract": "The discovery of causal relationships is a foundational problem in artificial intelligence, statistics, epidemiology, economics, and beyond. While elegant theories exist for accurate causal discovery given infinite data, real-world applications are inherently resource-constrained. Effective methods for inferring causal relationships from observational data must perform well under finite data and time constraints, where \"performing well\" implies achieving high, though not perfect accuracy. In his seminal paper A Theory of the Learnable, Valiant highlighted the importance of resource constraints in supervised machine learning, introducing the concept of Probably Approximately Correct (PAC) learning as an alternative to exact learning. Inspired by Valiant's work, we propose the Probably Approximately Correct Causal (PACC) Discovery framework, which extends PAC learning principles to the causal field. This framework emphasizes both computational and sample efficiency for established causal methods such as propensity score techniques and instrumental variable approaches. Furthermore, we show that it can also provide theoretical guarantees for other widely used methods, such as the Self-Controlled Case Series (SCCS) method, which had previously lacked such guarantees.", "date": "2025-07-25", "categories": []}
{"id": "2507.17309", "title": "Confounded Causal Imitation Learning with Instrumental Variables", "abstract": "Imitation learning from demonstrations usually suffers from the confounding effects of unmeasured variables (i.e., unmeasured confounders) on the states and actions. If ignoring them, a biased estimation of the policy would be entailed. To break up this confounding gap, in this paper, we take the best of the strong power of instrumental variables (IV) and propose a Confounded Causal Imitation Learning (C2L) model. This model accommodates confounders that influence actions across multiple timesteps, rather than being restricted to immediate temporal dependencies. We develop a two-stage imitation learning framework for valid IV identification and policy optimization. In particular, in the first stage, we construct a testing criterion based on the defined pseudo-variable, with which we achieve identifying a valid IV for the C2L models. Such a criterion entails the sufficient and necessary identifiability conditions for IV validity. In the second stage, with the identified IV, we propose two candidate policy learning approaches: one is based on a simulator, while the other is offline. Extensive experiments verified the effectiveness of identifying the valid IV as well as learning the policy.", "date": "2025-07-23", "categories": []}
{"id": "2507.07271", "title": "Beyond the ATE: Interpretable Modelling of Treatment Effects over Dose and Time", "abstract": "The Average Treatment Effect (ATE) is a foundational metric in causal inference, widely used to assess intervention efficacy in randomized controlled trials (RCTs). However, in many applications -- particularly in healthcare -- this static summary fails to capture the nuanced dynamics of treatment effects that vary with both dose and time. We propose a framework for modelling treatment effect trajectories as smooth surfaces over dose and time, enabling the extraction of clinically actionable insights such as onset time, peak effect, and duration of benefit. To ensure interpretability, robustness, and verifiability -- key requirements in high-stakes domains -- we adapt SemanticODE, a recent framework for interpretable trajectory modelling, to the causal setting where treatment effects are never directly observed. Our approach decouples the estimation of trajectory shape from the specification of clinically relevant properties (e.g., maxima, inflection points), supporting domain-informed priors, post-hoc editing, and transparent analysis. We show that our method yields accurate, interpretable, and editable models of treatment dynamics, facilitating both rigorous causal analysis and practical decision-making.", "date": "2025-07-09", "categories": []}
{"id": "2507.15899", "title": "Structural DID with ML: Theory, Simulation, and a Roadmap for Applied Research", "abstract": "Causal inference in observational panel data has become a central concern in economics,policy analysis,and the broader social sciences.To address the core contradiction where traditional difference-in-differences (DID) struggles with high-dimensional confounding variables in observational panel data,while machine learning (ML) lacks causal structure interpretability,this paper proposes an innovative framework called S-DIDML that integrates structural identification with high-dimensional estimation.Building upon the structure of traditional DID methods,S-DIDML employs structured residual orthogonalization techniques (Neyman orthogonality+cross-fitting) to retain the group-time treatment effect (ATT) identification structure while resolving high-dimensional covariate interference issues.It designs a dynamic heterogeneity estimation module combining causal forests and semi-parametric models to capture spatiotemporal heterogeneity effects.The framework establishes a complete modular application process with standardized Stata implementation paths.The introduction of S-DIDML enriches methodological research on DID and DDML innovations, shifting causal inference from method stacking to architecture integration.This advancement enables social sciences to precisely identify policy-sensitive groups and optimize resource allocation.The framework provides replicable evaluation tools, decision optimization references,and methodological paradigms for complex intervention scenarios such as digital transformation policies and environmental regulations.", "date": "2025-07-21", "categories": []}
{"id": "2402.00849", "title": "Score-based Causal Representation Learning: Linear and General Transformations", "abstract": "This paper addresses intervention-based causal representation learning (CRL) under a general nonparametric latent causal model and an unknown transformation that maps the latent variables to the observed variables. Linear and general transformations are investigated. The paper addresses both the identifiability and achievability aspects. Identifiability refers to determining algorithm-agnostic conditions that ensure the recovery of the true latent causal variables and the underlying latent causal graph. Achievability refers to the algorithmic aspects and addresses designing algorithms that achieve identifiability guarantees. By drawing novel connections between score functions (i.e., the gradients of the logarithm of density functions) and CRL, this paper designs a score-based class of algorithms that ensures both identifiability and achievability. First, the paper focuses on linear transformations and shows that one stochastic hard intervention per node suffices to guarantee identifiability. It also provides partial identifiability guarantees for soft interventions, including identifiability up to mixing with parents for general causal models and perfect recovery of the latent graph for sufficiently nonlinear causal models. Secondly, it focuses on general transformations and demonstrates that two stochastic hard interventions per node are sufficient for identifiability. This is achieved by defining a differentiable loss function whose global optima ensure identifiability for general CRL. Notably, one does not need to know which pair of interventional environments has the same node intervened. Finally, the theoretical results are empirically validated via experiments on structured synthetic data and image data.", "date": "2024-02-01", "categories": []}
{"id": "2505.05677", "title": "Conditional Front-door Adjustment for Heterogeneous Treatment Assignment Effect Estimation Under Non-adherence", "abstract": "Estimates of heterogeneous treatment assignment effects can inform treatment decisions. Under the presence of non-adherence (e.g., patients do not adhere to their assigned treatment), both the standard backdoor adjustment (SBD) and the conditional front-door adjustment (CFD) can recover unbiased estimates of the treatment assignment effects. However, the estimation variance of these approaches may vary widely across settings, which remains underexplored in the literature. In this work, we demonstrate theoretically and empirically that CFD yields lower-variance estimates than SBD when the true effect of treatment assignment is small (i.e., assigning an intervention leads to small changes in patients' future outcome). Additionally, since CFD requires estimating multiple nuisance parameters, we introduce LobsterNet, a multi-task neural network that implements CFD with joint modeling of the nuisance parameters. Empirically, LobsterNet reduces estimation error across several semi-synthetic and real-world datasets compared to baselines. Our findings suggest CFD with shared nuisance parameter modeling can improve treatment assignment effect estimation under non-adherence.", "date": "2025-05-08", "categories": []}
{"id": "2507.14161", "title": "Complex Dynamics in Psychological Data: Mapping Individual Symptom Trajectories to Group-Level Patterns", "abstract": "This study integrates causal inference, graph analysis, temporal complexity measures, and machine learning to examine whether individual symptom trajectories can reveal meaningful diagnostic patterns. Testing on a longitudinal dataset of N=45 individuals affected by General Anxiety Disorder (GAD) and/or Major Depressive Disorder (MDD) derived from Fisher et al. 2017, we propose a novel pipeline for the analysis of the temporal dynamics of psychopathological symptoms. First, we employ the PCMCI+ algorithm with nonparametric independence test to determine the causal network of nonlinear dependencies between symptoms in individuals with different mental disorders. We found that the PCMCI+ effectively highlights the individual peculiarities of each symptom network, which could be leveraged towards personalized therapies. At the same time, aggregating the networks by diagnosis sheds light to disorder-specific causal mechanisms, in agreement with previous psychopathological literature. Then, we enrich the dataset by computing complexity-based measures (e.g. entropy, fractal dimension, recurrence) from the symptom time series, and feed it to a suitably selected machine learning algorithm to aid the diagnosis of each individual. The new dataset yields 91% accuracy in the classification of the symptom dynamics, proving to be an effective diagnostic support tool. Overall, these findings highlight how integrating causal modeling and temporal complexity can enhance diagnostic differentiation, offering a principled, data-driven foundation for both personalized assessment in clinical psychology and structural advances in psychological research.", "date": "2025-07-07", "categories": []}
{"id": "2507.14387", "title": "Incremental Causal Graph Learning for Online Cyberattack Detection in Cyber-Physical Infrastructures", "abstract": "The escalating threat of cyberattacks on real-time critical infrastructures poses serious risks to public safety, demanding detection methods that effectively capture complex system interdependencies and adapt to evolving attack patterns. Traditional real-time anomaly detection techniques often suffer from excessive false positives due to their statistical sensitivity to high data variance and class imbalance. To address these limitations, recent research has explored modeling causal relationships among system components. However, prior work mainly focuses on offline causal graph-based approaches that require static historical data and fail to generalize to real-time settings. These methods are fundamentally constrained by: (1) their inability to adapt to dynamic shifts in data distribution without retraining, and (2) the risk of catastrophic forgetting when lacking timely supervision in live systems. To overcome these challenges, we propose INCADET, a novel framework for incremental causal graph learning tailored to real-time cyberattack detection. INCADET dynamically captures evolving system behavior by incrementally updating causal graphs across streaming time windows. The framework comprises three modules: 1) Early Symptom Detection: Detects transitions in system status using divergence in edge-weight distributions across sequential causal graphs. 2) Incremental Causal Graph Learning: Leverages experience replay and edge reinforcement to continually refine causal structures while preserving prior knowledge. 3) Causal Graph Classification: Employs Graph Convolutional Networks (GCNs) to classify system status using the learned causal graphs. Extensive experiments on real-world critical infrastructure datasets demonstrate that INCADET achieves superior accuracy, robustness, and adaptability compared to both static causal and deep temporal baselines in evolving attack scenarios.", "date": "2025-07-18", "categories": []}
{"id": "2507.14528", "title": "Positive-Unlabeled Learning for Control Group Construction in Observational Causal Inference", "abstract": "In causal inference, whether through randomized controlled trials or observational studies, access to both treated and control units is essential for estimating the effect of a treatment on an outcome of interest. When treatment assignment is random, the average treatment effect (ATE) can be estimated directly by comparing outcomes between groups. In non-randomized settings, various techniques are employed to adjust for confounding and approximate the counterfactual scenario to recover an unbiased ATE. A common challenge, especially in observational studies, is the absence of units clearly labeled as controls-that is, units known not to have received the treatment. To address this, we propose positive-unlabeled (PU) learning as a framework for identifying, with high confidence, control units from a pool of unlabeled ones, using only the available treated (positive) units. We evaluate this approach using both simulated and real-world data. We construct a causal graph with diverse relationships and use it to generate synthetic data under various scenarios, assessing how reliably the method recovers control groups that allow estimates of true ATE. We also apply our approach to real-world data on optimal sowing and fertilizer treatments in sustainable agriculture. Our findings show that PU learning can successfully identify control (negative) units from unlabeled data based only on treated units and, through the resulting control group, estimate an ATE that closely approximates the true value. This work has important implications for observational causal inference, especially in fields where randomized experiments are difficult or costly. In domains such as earth, environmental, and agricultural sciences, it enables a plethora of quasi-experiments by leveraging available earth observation and climate data, particularly when treated units are available but control units are lacking.", "date": "2025-07-19", "categories": []}
{"id": "2507.14767", "title": "XplainAct: Visualization for Personalized Intervention Insights", "abstract": "Causality helps people reason about and understand complex systems, particularly through what-if analyses that explore how interventions might alter outcomes. Although existing methods embrace causal reasoning using interventions and counterfactual analysis, they primarily focus on effects at the population level. These approaches often fall short in systems characterized by significant heterogeneity, where the impact of an intervention can vary widely across subgroups. To address this challenge, we present XplainAct, a visual analytics framework that supports simulating, explaining, and reasoning interventions at the individual level within subpopulations. We demonstrate the effectiveness of XplainAct through two case studies: investigating opioid-related deaths in epidemiology and analyzing voting inclinations in the presidential election.", "date": "2025-07-19", "categories": []}
{"id": "2507.14901", "title": "Learning Nonlinear Causal Reductions to Explain Reinforcement Learning Policies", "abstract": "Why do reinforcement learning (RL) policies fail or succeed? This is a challenging question due to the complex, high-dimensional nature of agent-environment interactions. In this work, we take a causal perspective on explaining the behavior of RL policies by viewing the states, actions, and rewards as variables in a low-level causal model. We introduce random perturbations to policy actions during execution and observe their effects on the cumulative reward, learning a simplified high-level causal model that explains these relationships. To this end, we develop a nonlinear Causal Model Reduction framework that ensures approximate interventional consistency, meaning the simplified high-level model responds to interventions in a similar way as the original complex system. We prove that for a class of nonlinear causal models, there exists a unique solution that achieves exact interventional consistency, ensuring learned explanations reflect meaningful causal patterns. Experiments on both synthetic causal models and practical RL tasks-including pendulum control and robot table tennis-demonstrate that our approach can uncover important behavioral patterns, biases, and failure modes in trained RL policies.", "date": "2025-07-20", "categories": []}
{"id": "2507.15106", "title": "From Kicking to Causality: Simulating Infant Agency Detection with a Robust Intrinsic Reward", "abstract": "While human infants robustly discover their own causal efficacy, standard reinforcement learning agents remain brittle, as their reliance on correlation-based rewards fails in noisy, ecologically valid scenarios. To address this, we introduce the Causal Action Influence Score (CAIS), a novel intrinsic reward rooted in causal inference. CAIS quantifies an action's influence by measuring the 1-Wasserstein distance between the learned distribution of sensory outcomes conditional on that action, $p(h|a)$, and the baseline outcome distribution, $p(h)$. This divergence provides a robust reward that isolates the agent's causal impact from confounding environmental noise. We test our approach in a simulated infant-mobile environment where correlation-based perceptual rewards fail completely when the mobile is subjected to external forces. In stark contrast, CAIS enables the agent to filter this noise, identify its influence, and learn the correct policy. Furthermore, the high-quality predictive model learned for CAIS allows our agent, when augmented with a surprise signal, to successfully reproduce the \"extinction burst\" phenomenon. We conclude that explicitly inferring causality is a crucial mechanism for developing a robust sense of agency, offering a psychologically plausible framework for more adaptive autonomous systems.", "date": "2025-07-20", "categories": []}
{"id": "2507.15842", "title": "Identifying Conditional Causal Effects in MPDAGs", "abstract": "We consider identifying a conditional causal effect when a graph is known up to a maximally oriented partially directed acyclic graph (MPDAG). An MPDAG represents an equivalence class of graphs that is restricted by background knowledge and where all variables in the causal model are observed. We provide three results that address identification in this setting: an identification formula when the conditioning set is unaffected by treatment, a generalization of the well-known do calculus to the MPDAG setting, and an algorithm that is complete for identifying these conditional effects.", "date": "2025-07-21", "categories": []}
{"id": "2507.14126", "title": "Toward Temporal Causal Representation Learning with Tensor Decomposition", "abstract": "Temporal causal representation learning is a powerful tool for uncovering complex patterns in observational studies, which are often represented as low-dimensional time series. However, in many real-world applications, data are high-dimensional with varying input lengths and naturally take the form of irregular tensors. To analyze such data, irregular tensor decomposition is critical for extracting meaningful clusters that capture essential information. In this paper, we focus on modeling causal representation learning based on the transformed information. First, we present a novel causal formulation for a set of latent clusters. We then propose CaRTeD, a joint learning framework that integrates temporal causal representation learning with irregular tensor decomposition. Notably, our framework provides a blueprint for downstream tasks using the learned tensor factors, such as modeling latent structures and extracting causal information, and offers a more flexible regularization design to enhance tensor decomposition. Theoretically, we show that our algorithm converges to a stationary point. More importantly, our results fill the gap in theoretical guarantees for the convergence of state-of-the-art irregular tensor decomposition. Experimental results on synthetic and real-world electronic health record (EHR) datasets (MIMIC-III), with extensive benchmarks from both phenotyping and network recovery perspectives, demonstrate that our proposed method outperforms state-of-the-art techniques and enhances the explainability of causal representations.", "date": "2025-07-18", "categories": []}
{"id": "2407.17385", "title": "Formalising causal inference as prediction on a target population", "abstract": "The standard approach to causal modelling especially in social and health sciences is the potential outcomes framework due to Neyman and Rubin. In this framework, observations are thought to be drawn from a distribution over variables of interest, and the goal is to identify parameters of this distribution. Even though the stated goal is often to inform decision making on some target population, there is no straightforward way to include these target populations in the framework. Instead of modelling the relationship between the observed sample and the target population, the inductive assumptions in this framework take the form of abstract sampling and independence assumptions. In this paper, we develop a version of this framework that construes causal inference as treatment-wise predictions for finite populations where all assumptions are testable in retrospect; this means that one can not only test predictions themselves (without any fundamental problem) but also investigate sources of error when they fail. Due to close connections to the original framework, established methods can still be be analysed under the new framework.", "date": "2024-07-24", "categories": []}
{"id": "2507.12818", "title": "Self Balancing Neural Network: A Novel Method to Estimate Average Treatment Effect", "abstract": "In observational studies, confounding variables affect both treatment and outcome. Moreover, instrumental variables also influence the treatment assignment mechanism. This situation sets the study apart from a standard randomized controlled trial, where the treatment assignment is random. Due to this situation, the estimated average treatment effect becomes biased. To address this issue, a standard approach is to incorporate the estimated propensity score when estimating the average treatment effect. However, these methods incur the risk of misspecification in propensity score models. To solve this issue, a novel method called the \"Self balancing neural network\" (Sbnet), which lets the model itself obtain its pseudo propensity score from the balancing net, is proposed in this study. The proposed method estimates the average treatment effect by using the balancing net as a key part of the feedforward neural network. This formulation resolves the estimation of the average treatment effect in one step. Moreover, the multi-pseudo propensity score framework, which is estimated from the diversified balancing net and used for the estimation of the average treatment effect, is presented. Finally, the proposed methods are compared with state-of-the-art methods on three simulation setups and real-world datasets. It has been shown that the proposed self-balancing neural network shows better performance than state-of-the-art methods.", "date": "2025-07-17", "categories": []}
{"id": "2507.11381", "title": "From Observational Data to Clinical Recommendations: A Causal Framework for Estimating Patient-level Treatment Effects and Learning Policies", "abstract": "We propose a framework for building patient-specific treatment recommendation models, building on the large recent literature on learning patient-level causal models and inspired by the target trial paradigm of Hernan and Robins. We focus on safety and validity, including the crucial issue of causal identification when using observational data. We do not provide a specific model, but rather a way to integrate existing methods and know-how into a practical pipeline. We further provide a real world use-case of treatment optimization for patients with heart failure who develop acute kidney injury during hospitalization. The results suggest our pipeline can improve patient outcomes over the current treatment regime.", "date": "2025-07-15", "categories": []}
{"id": "2507.12435", "title": "Targeted Deep Architectures: A TMLE-Based Framework for Robust Causal Inference in Neural Networks", "abstract": "Modern deep neural networks are powerful predictive tools yet often lack valid inference for causal parameters, such as treatment effects or entire survival curves. While frameworks like Double Machine Learning (DML) and Targeted Maximum Likelihood Estimation (TMLE) can debias machine-learning fits, existing neural implementations either rely on \"targeted losses\" that do not guarantee solving the efficient influence function equation or computationally expensive post-hoc \"fluctuations\" for multi-parameter settings. We propose Targeted Deep Architectures (TDA), a new framework that embeds TMLE directly into the network's parameter space with no restrictions on the backbone architecture. Specifically, TDA partitions model parameters - freezing all but a small \"targeting\" subset - and iteratively updates them along a targeting gradient, derived from projecting the influence functions onto the span of the gradients of the loss with respect to weights. This procedure yields plug-in estimates that remove first-order bias and produce asymptotically valid confidence intervals. Crucially, TDA easily extends to multi-dimensional causal estimands (e.g., entire survival curves) by merging separate targeting gradients into a single universal targeting update. Theoretically, TDA inherits classical TMLE properties, including double robustness and semiparametric efficiency. Empirically, on the benchmark IHDP dataset (average treatment effects) and simulated survival data with informative censoring, TDA reduces bias and improves coverage relative to both standard neural-network estimators and prior post-hoc approaches. In doing so, TDA establishes a direct, scalable pathway toward rigorous causal inference within modern deep architectures for complex multi-parameter targets.", "date": "2025-07-16", "categories": []}
{"id": "2507.10809", "title": "Uncovering Causal Relation Shifts in Event Sequences under Out-of-Domain Interventions", "abstract": "Inferring causal relationships between event pairs in a temporal sequence is applicable in many domains such as healthcare, manufacturing, and transportation. Most existing work on causal inference primarily focuses on event types within the designated domain, without considering the impact of exogenous out-of-domain interventions. In real-world settings, these out-of-domain interventions can significantly alter causal dynamics. To address this gap, we propose a new causal framework to define average treatment effect (ATE), beyond independent and identically distributed (i.i.d.) data in classic Rubin's causal framework, to capture the causal relation shift between events of temporal process under out-of-domain intervention. We design an unbiased ATE estimator, and devise a Transformer-based neural network model to handle both long-range temporal dependencies and local patterns while integrating out-of-domain intervention information into process modeling. Extensive experiments on both simulated and real-world datasets demonstrate that our method outperforms baselines in ATE estimation and goodness-of-fit under out-of-domain-augmented point processes.", "date": "2025-07-14", "categories": []}
{"id": "2407.07290", "title": "Causal Discovery-Driven Change Point Detection in Time Series", "abstract": "Change point detection in time series aims to identify moments when the probability distribution of time series changes. It is widely applied in many areas, such as human activity sensing and medical science. In the context of multivariate time series, this typically involves examining the joint distribution of multiple variables: If the distribution of any one variable changes, the entire time series undergoes a distribution shift. However, in practical applications, we may be interested only in certain components of the time series, exploring abrupt changes in their distributions while accounting for the presence of other components. Here, assuming an underlying structural causal model that governs the time-series data generation, we address this task by proposing a two-stage non-parametric algorithm that first learns parts of the causal structure through constraint-based discovery methods, and then employs conditional relative Pearson divergence estimation to identify the change points. The conditional relative Pearson divergence quantifies the distribution difference between consecutive segments in the time series, while the causal discovery method allows a focus on the causal mechanism, facilitating access to independent and identically distributed (IID) samples. Theoretically, the typical assumption of samples being IID in conventional change point detection methods can be relaxed based on the Causal Markov Condition. Through experiments on both synthetic and real-world datasets, we validate the correctness and utility of our approach.", "date": "2024-07-10", "categories": []}
{"id": "2506.17300", "title": "Individual Causal Inference with Structural Causal Model", "abstract": "Individual causal inference (ICI) uses causal inference methods to understand and predict the effects of interventions on individuals, considering their specific characteristics / facts. It aims to estimate individual causal effect (ICE), which varies across individuals. Estimating ICE can be challenging due to the limited data available for individuals, and the fact that most causal inference methods are population-based. Structural Causal Model (SCM) is fundamentally population-based. Therefore, causal discovery (structural learning and parameter learning), association queries and intervention queries are all naturally population-based. However, exogenous variables (U) in SCM can encode individual variations and thus provide the mechanism for individualized population per specific individual characteristics / facts. Based on this, we propose ICI with SCM as a \"rung 3\" causal inference, because it involves \"imagining\" what would be the causal effect of a hypothetical intervention on an individual, given the individual's observed characteristics / facts. Specifically, we propose the indiv-operator, indiv(W), to formalize/represent the population individualization process, and the individual causal query, P(Y | indiv(W), do(X), Z), to formalize/represent ICI. We show and argue that ICI with SCM is inference on individual alternatives (possible), not individual counterfactuals (non-actual).", "date": "2025-06-17", "categories": []}
{"id": "2507.09439", "title": "Dynamic Sparse Causal-Attention Temporal Networks for Interpretable Causality Discovery in Multivariate Time Series", "abstract": "Understanding causal relationships in multivariate time series (MTS) is essential for effective decision-making in fields such as finance and marketing, where complex dependencies and lagged effects challenge conventional analytical approaches. We introduce Dynamic Sparse Causal-Attention Temporal Networks for Interpretable Causality Discovery in MTS (DyCAST-Net), a novel architecture designed to enhance causal discovery by integrating dilated temporal convolutions and dynamic sparse attention mechanisms. DyCAST-Net effectively captures multiscale temporal dependencies through dilated convolutions while leveraging an adaptive thresholding strategy in its attention mechanism to eliminate spurious connections, ensuring both accuracy and interpretability. A statistical shuffle test validation further strengthens robustness by filtering false positives and improving causal inference reliability. Extensive evaluations on financial and marketing datasets demonstrate that DyCAST-Net consistently outperforms existing models such as TCDF, GCFormer, and CausalFormer. The model provides a more precise estimation of causal delays and significantly reduces false discoveries, particularly in noisy environments. Moreover, attention heatmaps offer interpretable insights, uncovering hidden causal patterns such as the mediated effects of advertising on consumer behavior and the influence of macroeconomic indicators on financial markets. Case studies illustrate DyCAST-Net's ability to detect latent mediators and lagged causal factors, making it particularly effective in high-dimensional, dynamic settings. The model's architecture enhanced by RMSNorm stabilization and causal masking ensures scalability and adaptability across diverse application domains", "date": "2025-07-13", "categories": []}
{"id": "2507.09494", "title": "An Algorithm for Identifying Interpretable Subgroups With Elevated Treatment Effects", "abstract": "We introduce an algorithm for identifying interpretable subgroups with elevated treatment effects, given an estimate of individual or conditional average treatment effects (CATE). Subgroups are characterized by ``rule sets'' -- easy-to-understand statements of the form (Condition A AND Condition B) OR (Condition C) -- which can capture high-order interactions while retaining interpretability. Our method complements existing approaches for estimating the CATE, which often produce high dimensional and uninterpretable results, by summarizing and extracting critical information from fitted models to aid decision making, policy implementation, and scientific understanding. We propose an objective function that trades-off subgroup size and effect size, and varying the hyperparameter that controls this trade-off results in a ``frontier'' of Pareto optimal rule sets, none of which dominates the others across all criteria. Valid inference is achievable through sample splitting. We demonstrate the utility and limitations of our method using simulated and empirical examples.", "date": "2025-07-13", "categories": []}
{"id": "2502.09395", "title": "Robot Pouring: Identifying Causes of Spillage and Selecting Alternative Action Parameters Using Probabilistic Actual Causation", "abstract": "In everyday life, we perform tasks (e.g., cooking or cleaning) that involve a large variety of objects and goals. When confronted with an unexpected or unwanted outcome, we take corrective actions and try again until achieving the desired result. The reasoning performed to identify a cause of the observed outcome and to select an appropriate corrective action is a crucial aspect of human reasoning for successful task execution. Central to this reasoning is the assumption that a factor is responsible for producing the observed outcome. In this paper, we investigate the use of probabilistic actual causation to determine whether a factor is the cause of an observed undesired outcome. Furthermore, we show how the actual causation probabilities can be used to find alternative actions to change the outcome. We apply the probabilistic actual causation analysis to a robot pouring task. When spillage occurs, the analysis indicates whether a task parameter is the cause and how it should be changed to avoid spillage. The analysis requires a causal graph of the task and the corresponding conditional probability distributions. To fulfill these requirements, we perform a complete causal modeling procedure (i.e., task analysis, definition of variables, determination of the causal graph structure, and estimation of conditional probability distributions) using data from a realistic simulation of the robot pouring task, covering a large combinatorial space of task parameters. Based on the results, we discuss the implications of the variables' representation and how the alternative actions suggested by the actual causation analysis would compare to the alternative solutions proposed by a human observer. The practical use of the analysis of probabilistic actual causation to select alternative action parameters is demonstrated.", "date": "2025-02-13", "categories": []}
{"id": "2507.08177", "title": "Rethinking Spatio-Temporal Anomaly Detection: A Vision for Causality-Driven Cybersecurity", "abstract": "As cyber-physical systems grow increasingly interconnected and spatially distributed, ensuring their resilience against evolving cyberattacks has become a critical priority. Spatio-Temporal Anomaly detection plays an important role in ensuring system security and operational integrity. However, current data-driven approaches, largely driven by black-box deep learning, face challenges in interpretability, adaptability to distribution shifts, and robustness under evolving system dynamics. In this paper, we advocate for a causal learning perspective to advance anomaly detection in spatially distributed infrastructures that grounds detection in structural cause-effect relationships. We identify and formalize three key directions: causal graph profiling, multi-view fusion, and continual causal graph learning, each offering distinct advantages in uncovering dynamic cause-effect structures across time and space. Drawing on real-world insights from systems such as water treatment infrastructures, we illustrate how causal models provide early warning signals and root cause attribution, addressing the limitations of black-box detectors. Looking ahead, we outline the future research agenda centered on multi-modality, generative AI-driven, and scalable adaptive causal frameworks. Our objective is to lay a new research trajectory toward scalable, adaptive, explainable, and spatially grounded anomaly detection systems. We hope to inspire a paradigm shift in cybersecurity research, promoting causality-driven approaches to address evolving threats in interconnected infrastructures.", "date": "2025-07-10", "categories": []}
{"id": "2507.08235", "title": "InsightBuild: LLM-Powered Causal Reasoning in Smart Building Systems", "abstract": "Smart buildings generate vast streams of sensor and control data, but facility managers often lack clear explanations for anomalous energy usage. We propose InsightBuild, a two-stage framework that integrates causality analysis with a fine-tuned large language model (LLM) to provide human-readable, causal explanations of energy consumption patterns. First, a lightweight causal inference module applies Granger causality tests and structural causal discovery on building telemetry (e.g., temperature, HVAC settings, occupancy) drawn from Google Smart Buildings and Berkeley Office datasets. Next, an LLM, fine-tuned on aligned pairs of sensor-level causes and textual explanations, receives as input the detected causal relations and generates concise, actionable explanations. We evaluate InsightBuild on two real-world datasets (Google: 2017-2022; Berkeley: 2018-2020), using expert-annotated ground-truth causes for a held-out set of anomalies. Our results demonstrate that combining explicit causal discovery with LLM-based natural language generation yields clear, precise explanations that assist facility managers in diagnosing and mitigating energy inefficiencies.", "date": "2025-07-11", "categories": []}
{"id": "2410.08007", "title": "Time Can Invalidate Algorithmic Recourse", "abstract": "Algorithmic Recourse (AR) aims to provide users with actionable steps to overturn unfavourable decisions made by machine learning predictors. However, these actions often take time to implement (e.g., getting a degree can take years), and their effects may vary as the world evolves. Thus, it is natural to ask for recourse that remains valid in a dynamic environment. In this paper, we study the robustness of algorithmic recourse over time by casting the problem through the lens of causality. We demonstrate theoretically and empirically that (even robust) causal AR methods can fail over time, except in the -- unlikely -- case that the world is stationary. Even more critically, unless the world is fully deterministic, counterfactual AR cannot be solved optimally. To account for this, we propose a simple yet effective algorithm for temporal AR that explicitly accounts for time under the assumption of having access to an estimator approximating the stochastic process. Our simulations on synthetic and realistic datasets show how considering time produces more resilient solutions to potential trends in the data distribution.", "date": "2024-10-10", "categories": []}
{"id": "2507.07359", "title": "Goal-Oriented Sequential Bayesian Experimental Design for Causal Learning", "abstract": "We present GO-CBED, a goal-oriented Bayesian framework for sequential causal experimental design. Unlike conventional approaches that select interventions aimed at inferring the full causal model, GO-CBED directly maximizes the expected information gain (EIG) on user-specified causal quantities of interest, enabling more targeted and efficient experimentation. The framework is both non-myopic, optimizing over entire intervention sequences, and goal-oriented, targeting only model aspects relevant to the causal query. To address the intractability of exact EIG computation, we introduce a variational lower bound estimator, optimized jointly through a transformer-based policy network and normalizing flow-based variational posteriors. The resulting policy enables real-time decision-making via an amortized network. We demonstrate that GO-CBED consistently outperforms existing baselines across various causal reasoning and discovery tasks-including synthetic structural causal models and semi-synthetic gene regulatory networks-particularly in settings with limited experimental budgets and complex causal mechanisms. Our results highlight the benefits of aligning experimental design objectives with specific research goals and of forward-looking sequential planning.", "date": "2025-07-10", "categories": []}
{"id": "2507.07738", "title": "Efficient and Scalable Estimation of Distributional Treatment Effects with Multi-Task Neural Networks", "abstract": "We propose a novel multi-task neural network approach for estimating distributional treatment effects (DTE) in randomized experiments. While DTE provides more granular insights into the experiment outcomes over conventional methods focusing on the Average Treatment Effect (ATE), estimating it with regression adjustment methods presents significant challenges. Specifically, precision in the distribution tails suffers due to data imbalance, and computational inefficiencies arise from the need to solve numerous regression problems, particularly in large-scale datasets commonly encountered in industry. To address these limitations, our method leverages multi-task neural networks to estimate conditional outcome distributions while incorporating monotonic shape constraints and multi-threshold label learning to enhance accuracy. To demonstrate the practical effectiveness of our proposed method, we apply our method to both simulated and real-world datasets, including a randomized field experiment aimed at reducing water consumption in the US and a large-scale A/B test from a leading streaming platform in Japan. The experimental results consistently demonstrate superior performance across various datasets, establishing our method as a robust and practical solution for modern causal inference applications requiring a detailed understanding of treatment effect heterogeneity.", "date": "2025-07-10", "categories": []}
{"id": "2507.07857", "title": "Searching for actual causes: Approximate algorithms with adjustable precision", "abstract": "Causality has gained popularity in recent years. It has helped improve the performance, reliability, and interpretability of machine learning models. However, recent literature on explainable artificial intelligence (XAI) has faced criticism. The classical XAI and causality literature focuses on understanding which factors contribute to which consequences. While such knowledge is valuable for researchers and engineers, it is not what non-expert users expect as explanations. Instead, these users often await facts that cause the target consequences, i.e., actual causes. Formalizing this notion is still an open problem. Additionally, identifying actual causes is reportedly an NP-complete problem, and there are too few practical solutions to approximate formal definitions. We propose a set of algorithms to identify actual causes with a polynomial complexity and an adjustable level of precision and exhaustiveness. Our experiments indicate that the algorithms (1) identify causes for different categories of systems that are not handled by existing approaches (i.e., non-boolean, black-box, and stochastic systems), (2) can be adjusted to gain more precision and exhaustiveness with more computation time.", "date": "2025-07-10", "categories": []}
{"id": "2507.07898", "title": "Efficient Causal Discovery for Autoregressive Time Series", "abstract": "In this study, we present a novel constraint-based algorithm for causal structure learning specifically designed for nonlinear autoregressive time series. Our algorithm significantly reduces computational complexity compared to existing methods, making it more efficient and scalable to larger problems. We rigorously evaluate its performance on synthetic datasets, demonstrating that our algorithm not only outperforms current techniques, but also excels in scenarios with limited data availability. These results highlight its potential for practical applications in fields requiring efficient and accurate causal inference from nonlinear time series data.", "date": "2025-07-10", "categories": []}
{"id": "2308.03572", "title": "Efficient Transfer Learning via Causal Bounds", "abstract": "Transfer learning seeks to accelerate sequential decision-making by leveraging offline data from related agents. However, data from heterogeneous sources that differ in observed features, distributions, or unobserved confounders often render causal effects non-identifiable and bias naive estimators. We address this by forming ambiguity sets of structural causal models defined via integral constraints on their joint densities. Optimizing any causal effect over these sets leads to generally non-convex programs whose solutions tightly bound the range of possible effects under heterogeneity or confounding. To solve these programs efficiently, we develop a hit-and-run sampler that explores the entire ambiguity set and, when paired with a local optimization oracle, produces causal bound estimates that converge almost surely to the true limits. We further accommodate estimation error by relaxing the ambiguity set and exploit the Lipschitz continuity of causal effects to establish precise error propagation guarantees. These causal bounds are then embedded into bandit algorithms via arm elimination and truncated UCB indices, yielding optimal gap-dependent and minimax regret bounds. To handle estimation error, we also develop a safe algorithm for incorporating noisy causal bounds. In the contextual-bandit setting with function approximation, our method uses causal bounds to prune both the function class and the per-context action set, achieving matching upper and lower regret bounds with only logarithmic dependence on function-class complexity. Our analysis precisely characterizes when and how causal side-information accelerates online learning, and experiments on synthetic benchmarks confirm substantial regret reductions in data-scarce or confounded regimes.", "date": "2023-08-07", "categories": []}
{"id": "2504.04320", "title": "Causal Inference Isn't Special: Why It's Just Another Prediction Problem", "abstract": "Causal inference is often portrayed as fundamentally distinct from predictive modeling, with its own terminology, goals, and intellectual challenges. But at its core, causal inference is simply a structured instance of prediction under distribution shift. In both cases, we begin with labeled data from a source domain and seek to generalize to a target domain where outcomes are not observed. The key difference is that in causal inference, the labels -- potential outcomes -- are selectively observed based on treatment assignment, introducing bias that must be addressed through assumptions. This perspective reframes causal estimation as a familiar generalization problem and highlights how techniques from predictive modeling, such as reweighting and domain adaptation, apply directly to causal tasks. It also clarifies that causal assumptions are not uniquely strong -- they are simply more explicit. By viewing causal inference through the lens of prediction, we demystify its logic, connect it to familiar tools, and make it more accessible to practitioners and educators alike.", "date": "2025-04-06", "categories": []}
{"id": "2506.22675", "title": "Bayesian Invariance Modeling of Multi-Environment Data", "abstract": "Invariant prediction [Peters et al., 2016] analyzes feature/outcome data from multiple environments to identify invariant features - those with a stable predictive relationship to the outcome. Such features support generalization to new environments and help reveal causal mechanisms. Previous methods have primarily tackled this problem through hypothesis testing or regularized optimization. Here we develop Bayesian Invariant Prediction (BIP), a probabilistic model for invariant prediction. BIP encodes the indices of invariant features as a latent variable and recover them by posterior inference. Under the assumptions of Peters et al. [2016], the BIP posterior targets the true invariant features. We prove that the posterior is consistent and that greater environment heterogeneity leads to faster posterior contraction. To handle many features, we design an efficient variational approximation called VI-BIP. In simulations and real data, we find that BIP and VI-BIP are more accurate and scalable than existing methods for invariant prediction.", "date": "2025-06-27", "categories": []}
{"id": "2507.06650", "title": "Deep Disentangled Representation Network for Treatment Effect Estimation", "abstract": "Estimating individual-level treatment effect from observational data is a fundamental problem in causal inference and has attracted increasing attention in the fields of education, healthcare, and public policy.In this work, we concentrate on the study of disentangled representation methods that have shown promising outcomes by decomposing observed covariates into instrumental, confounding, and adjustment factors. However, most of the previous work has primarily revolved around generative models or hard decomposition methods for covariates, which often struggle to guarantee the attainment of precisely disentangled factors. In order to effectively model different causal relationships, we propose a novel treatment effect estimation algorithm that incorporates a mixture of experts with multi-head attention and a linear orthogonal regularizer to softly decompose the pre-treatment variables, and simultaneously eliminates selection bias via importance sampling re-weighting techniques. We conduct extensive experiments on both public semi-synthetic and real-world production datasets. The experimental results clearly demonstrate that our algorithm outperforms the state-of-the-art methods focused on individual treatment effects.", "date": "2025-07-09", "categories": []}
{"id": "2507.06888", "title": "Horizontal and Vertical Federated Causal Structure Learning via Higher-order Cumulants", "abstract": "Federated causal discovery aims to uncover the causal relationships between entities while protecting data privacy, which has significant importance and numerous applications in real-world scenarios. Existing federated causal structure learning methods primarily focus on horizontal federated settings. However, in practical situations, different clients may not necessarily contain data on the same variables. In a single client, the incomplete set of variables can easily lead to spurious causal relationships, thereby affecting the information transmitted to other clients. To address this issue, we comprehensively consider causal structure learning methods under both horizontal and vertical federated settings. We provide the identification theories and methods for learning causal structure in the horizontal and vertical federal setting via higher-order cumulants. Specifically, we first aggregate higher-order cumulant information from all participating clients to construct global cumulant estimates. These global estimates are then used for recursive source identification, ultimately yielding a global causal strength matrix. Our approach not only enables the reconstruction of causal graphs but also facilitates the estimation of causal strength coefficients. Our algorithm demonstrates superior performance in experiments conducted on both synthetic data and real-world data.", "date": "2025-07-09", "categories": []}
{"id": "2504.08773", "title": "Counterfactual Inference under Thompson Sampling", "abstract": "Recommender systems exemplify sequential decision-making under uncertainty, strategically deciding what content to serve to users, to optimise a range of potential objectives. To balance the explore-exploit trade-off successfully, Thompson sampling provides a natural and widespread paradigm to probabilistically select which action to take. Questions of causal and counterfactual inference, which underpin use-cases like offline evaluation, are not straightforward to answer in these contexts. Specifically, whilst most existing estimators rely on action propensities, these are not readily available under Thompson sampling procedures.\n  We derive exact and efficiently computable expressions for action propensities under a variety of parameter and outcome distributions, enabling the use of off-policy estimators in Thompson sampling scenarios. This opens up a range of practical use-cases where counterfactual inference is crucial, including unbiased offline evaluation of recommender systems, as well as general applications of causal inference in online advertising, personalisation, and beyond.", "date": "2025-04-03", "categories": []}
{"id": "2507.05333", "title": "Causal Foundation Models: Disentangling Physics from Instrument Properties", "abstract": "Foundation models for structured time series data must contend with a fundamental challenge: observations often conflate the true underlying physical phenomena with systematic distortions introduced by measurement instruments. This entanglement limits model generalization, especially in heterogeneous or multi-instrument settings. We present a causally-motivated foundation model that explicitly disentangles physical and instrumental factors using a dual-encoder architecture trained with structured contrastive learning. Leveraging naturally occurring observational triplets (i.e., where the same target is measured under varying conditions, and distinct targets are measured under shared conditions) our model learns separate latent representations for the underlying physical signal and instrument effects. Evaluated on simulated astronomical time series designed to resemble the complexity of variable stars observed by missions like NASA's Transiting Exoplanet Survey Satellite (TESS), our method significantly outperforms traditional single-latent space foundation models on downstream prediction tasks, particularly in low-data regimes. These results demonstrate that our model supports key capabilities of foundation models, including few-shot generalization and efficient adaptation, and highlight the importance of encoding causal structure into representation learning for structured data.", "date": "2025-07-07", "categories": []}
{"id": "2507.05510", "title": "Heterogeneous Causal Learning for Optimizing Aggregated Functions in User Growth", "abstract": "User growth is a major strategy for consumer internet companies. To optimize costly marketing campaigns and maximize user engagement, we propose a novel treatment effect optimization methodology to enhance user growth marketing. By leveraging deep learning, our algorithm learns from past experiments to optimize user selection and reward allocation, maximizing campaign impact while minimizing costs. Unlike traditional prediction methods, our model directly models uplifts in key business metrics. Further, our deep learning model can jointly optimize parameters for an aggregated loss function using softmax gating. Our approach surpasses traditional methods by directly targeting desired business metrics and demonstrates superior algorithmic flexibility in handling complex business constraints. Comprehensive evaluations, including comparisons with state-of-the-art techniques such as R-learner and Causal Forest, validate the effectiveness of our model. We experimentally demonstrate that our proposed constrained and direct optimization algorithms significantly outperform state-of-the-art methods by over $20\\%$, proving their cost-efficiency and real-world impact. The versatile methods can be applied to various product scenarios, including optimal treatment allocation. Its effectiveness has also been validated through successful worldwide production deployments.", "date": "2025-07-07", "categories": []}
{"id": "2507.05511", "title": "Deep Learning of Continuous and Structured Policies for Aggregated Heterogeneous Treatment Effects", "abstract": "As estimation of Heterogeneous Treatment Effect (HTE) is increasingly adopted across a wide range of scientific and industrial applications, the treatment action space can naturally expand, from a binary treatment variable to a structured treatment policy. This policy may include several policy factors such as a continuous treatment intensity variable, or discrete treatment assignments. From first principles, we derive the formulation for incorporating multiple treatment policy variables into the functional forms of individual and average treatment effects. Building on this, we develop a methodology to directly rank subjects using aggregated HTE functions. In particular, we construct a Neural-Augmented Naive Bayes layer within a deep learning framework to incorporate an arbitrary number of factors that satisfies the Naive Bayes assumption. The factored layer is then applied with continuous treatment variables, treatment assignment, and direct ranking of aggregated treatment effect functions. Together, these algorithms build towards a generic framework for deep learning of heterogeneous treatment policies, and we show their power to improve performance with public datasets.", "date": "2025-07-07", "categories": []}
{"id": "2507.06213", "title": "Identifiability in Causal Abstractions: A Hierarchy of Criteria", "abstract": "Identifying the effect of a treatment from observational data typically requires assuming a fully specified causal diagram. However, such diagrams are rarely known in practice, especially in complex or high-dimensional settings. To overcome this limitation, recent works have explored the use of causal abstractions-simplified representations that retain partial causal information. In this paper, we consider causal abstractions formalized as collections of causal diagrams, and focus on the identifiability of causal queries within such collections. We introduce and formalize several identifiability criteria under this setting. Our main contribution is to organize these criteria into a structured hierarchy, highlighting their relationships. This hierarchical view enables a clearer understanding of what can be identified under varying levels of causal knowledge. We illustrate our framework through examples from the literature and provide tools to reason about identifiability when full causal knowledge is unavailable.", "date": "2025-07-08", "categories": []}
{"id": "2409.18581", "title": "Deep Autoregressive Models as Causal Inference Engines", "abstract": "Existing causal inference (CI) models are often restricted to data with low-dimensional confounders and singleton actions. We propose an autoregressive (AR) CI framework capable of handling complex confounders and sequential actions commonly found in modern applications. Our approach accomplishes this using {\\em sequencification}, which transforms data from an underlying causal diagram into a sequence of tokens. Sequencification not only accommodates training with data generated from a large class of DAGs, but also extends existing CI capabilities to estimate multiple causal quantities using a {\\em single} model. We can directly compute probabilities from interventional distributions, simplifying inference and improving outcome prediction accuracy. We demonstrate that an AR model adapted for CI is efficient and effective in various complex applications such as navigating mazes, playing chess endgames, and evaluating the impact of certain keywords on paper acceptance rates, where we consider causal queries beyond standard reinforcement learning-type questions.", "date": "2024-09-27", "categories": []}
{"id": "2412.07446", "title": "A Causal World Model Underlying Next Token Prediction: Exploring GPT in a Controlled Environment", "abstract": "Are generative pre-trained transformer (GPT) models, trained only to predict the next token, implicitly learning a world model from which sequences are generated one token at a time? We address this question by deriving a causal interpretation of the attention mechanism in GPT and presenting a causal world model that arises from this interpretation. Furthermore, we propose that GPT models, at inference time, can be utilized for zero-shot causal structure learning for input sequences, and introduce a corresponding confidence score. Empirical tests were conducted in controlled environments using the setups of the Othello and Chess strategy games. A GPT, pre-trained on real-world games played with the intention of winning, was tested on out-of-distribution synthetic data consisting of sequences of random legal moves. We find that the GPT model is likely to generate legal next moves for out-of-distribution sequences for which a causal structure is encoded in the attention mechanism with high confidence. In cases where it generates illegal moves, it also fails to capture a causal structure.", "date": "2024-12-10", "categories": []}
{"id": "2507.02928", "title": "Mitigating Hidden Confounding by Progressive Confounder Imputation via Large Language Models", "abstract": "Hidden confounding remains a central challenge in estimating treatment effects from observational data, as unobserved variables can lead to biased causal estimates. While recent work has explored the use of large language models (LLMs) for causal inference, most approaches still rely on the unconfoundedness assumption. In this paper, we make the first attempt to mitigate hidden confounding using LLMs. We propose ProCI (Progressive Confounder Imputation), a framework that elicits the semantic and world knowledge of LLMs to iteratively generate, impute, and validate hidden confounders. ProCI leverages two key capabilities of LLMs: their strong semantic reasoning ability, which enables the discovery of plausible confounders from both structured and unstructured inputs, and their embedded world knowledge, which supports counterfactual reasoning under latent confounding. To improve robustness, ProCI adopts a distributional reasoning strategy instead of direct value imputation to prevent the collapsed outputs. Extensive experiments demonstrate that ProCI uncovers meaningful confounders and significantly improves treatment effect estimation across various datasets and LLMs.", "date": "2025-06-26", "categories": []}
{"id": "2507.03271", "title": "LILI clustering algorithm: Limit Inferior Leaf Interval Integrated into Causal Forest for Causal Interference", "abstract": "Causal forest methods are powerful tools in causal inference. Similar to traditional random forest in machine learning, causal forest independently considers each causal tree. However, this independence consideration increases the likelihood that classification errors in one tree are repeated in others, potentially leading to significant bias in causal e ect estimation. In this paper, we propose a novel approach that establishes connections between causal trees through the Limit Inferior Leaf Interval (LILI) clustering algorithm. LILIs are constructed based on the leaves of all causal trees, emphasizing the similarity of dataset confounders. When two instances with di erent treatments are grouped into the same leaf across a su cient number of causal trees, they are treated as counterfactual outcomes of each other. Through this clustering mechanism, LILI clustering reduces bias present in traditional causal tree methods and enhances the prediction accuracy for the average treatment e ect (ATE). By integrating LILIs into a causal forest, we develop an e cient causal inference method. Moreover, we explore several key properties of LILI by relating it to the concepts of limit inferior and limit superior in the set theory. Theoretical analysis rigorously proves the convergence of the estimated ATE using LILI clustering. Empirically, extensive comparative experiments demonstrate the superior performance of LILI clustering.", "date": "2025-07-04", "categories": []}
{"id": "2507.03528", "title": "Generating Synthetic Relational Tabular Data via Structural Causal Models", "abstract": "Synthetic tabular data generation has received increasing attention in recent years, particularly with the emergence of foundation models for tabular data. The breakthrough success of TabPFN (Hollmann et al.,2025), which leverages vast quantities of synthetic tabular datasets derived from structural causal models (SCMs), demonstrates the critical role synthetic data plays in developing powerful tabular foundation models. However, most real-world tabular data exists in relational formats spanning multiple interconnected tables - a structure not adequately addressed by current generation methods. In this work, we extend the SCM-based approach by developing a novel framework that generates realistic synthetic relational tabular data including causal relationships across tables. Our experiments confirm that this framework is able to construct relational datasets with complex inter-table dependencies mimicking real-world scenarios.", "date": "2025-07-04", "categories": []}
{"id": "2507.04332", "title": "Consistent Labeling Across Group Assignments: Variance Reduction in Conditional Average Treatment Effect Estimation", "abstract": "Numerous algorithms have been developed for Conditional Average Treatment Effect (CATE) estimation. In this paper, we first highlight a common issue where many algorithms exhibit inconsistent learning behavior for the same instance across different group assignments. We introduce a metric to quantify and visualize this inconsistency. Next, we present a theoretical analysis showing that this inconsistency indeed contributes to higher test errors and cannot be resolved through conventional machine learning techniques. To address this problem, we propose a general method called \\textbf{Consistent Labeling Across Group Assignments} (CLAGA), which eliminates the inconsistency and is applicable to any existing CATE estimation algorithm. Experiments on both synthetic and real-world datasets demonstrate significant performance improvements with CLAGA.", "date": "2025-07-06", "categories": []}
{"id": "2507.05088", "title": "How Rules Represent Causal Knowledge: Causal Modeling with Abductive Logic Programs", "abstract": "Pearl observes that causal knowledge enables predicting the effects of interventions, such as actions, whereas descriptive knowledge only permits drawing conclusions from observation. This paper extends Pearl's approach to causality and interventions to the setting of stratified abductive logic programs. It shows how stable models of such programs can be given a causal interpretation by building on philosophical foundations and recent work by Bochman and Eelink et al. In particular, it provides a translation of abductive logic programs into causal systems, thereby clarifying the informal causal reading of logic program rules and supporting principled reasoning about external actions. The main result establishes that the stable model semantics for stratified programs conforms to key philosophical principles of causation, such as causal sufficiency, natural necessity, and irrelevance of unobserved effects. This justifies the use of stratified abductive logic programs as a framework for causal modeling and for predicting the effects of interventions", "date": "2025-07-07", "categories": []}
{"id": "2505.20697", "title": "Generating Hypotheses of Dynamic Causal Graphs in Neuroscience: Leveraging Generative Factor Models of Observed Time Series", "abstract": "The field of hypothesis generation promises to reduce costs in neuroscience by narrowing the range of interventional studies needed to study various phenomena. Existing machine learning methods can generate scientific hypotheses from complex datasets, but many approaches assume causal relationships are static over time, limiting their applicability to systems with dynamic, state-dependent behavior, such as the brain. While some techniques attempt dynamic causal discovery through factor models, they often restrict relationships to linear patterns or impose other simplifying assumptions. We propose a novel method that models dynamic graphs as a conditionally weighted superposition of static graphs, where each static graph can capture nonlinear relationships. This approach enables the detection of complex, time-varying interactions between variables beyond linear limitations. Our method improves f1-scores of predicted dynamic causal patterns by roughly 22-28% on average over baselines in some of our experiments, with some improvements reaching well over 60%. A case study on real brain data demonstrates our method's ability to uncover relationships linked to specific behavioral states, offering valuable insights into neural dynamics.", "date": "2025-05-27", "categories": []}
{"id": "2409.07215", "title": "Is merging worth it? Securely evaluating the information gain for causal dataset acquisition", "abstract": "Merging datasets across institutions is a lengthy and costly procedure, especially when it involves private information. Data hosts may therefore want to prospectively gauge which datasets are most beneficial to merge with, without revealing sensitive information. For causal estimation this is particularly challenging as the value of a merge depends not only on reduction in epistemic uncertainty but also on improvement in overlap. To address this challenge, we introduce the first cryptographically secure information-theoretic approach for quantifying the value of a merge in the context of heterogeneous treatment effect estimation. We do this by evaluating the Expected Information Gain (EIG) using multi-party computation to ensure that no raw data is revealed. We further demonstrate that our approach can be combined with differential privacy (DP) to meet arbitrary privacy requirements whilst preserving more accurate computation compared to DP alone. To the best of our knowledge, this work presents the first privacy-preserving method for dataset acquisition tailored to causal estimation. We demonstrate the effectiveness and reliability of our method on a range of simulated and realistic benchmarks. Code is publicly available: https://github.com/LucileTerminassian/causal_prospective_merge.", "date": "2024-09-11", "categories": []}
{"id": "2212.08712", "title": "Causal Temporal Reasoning for Markov Decision Processes", "abstract": "We introduce $\\textit{PCFTL (Probabilistic CounterFactual Temporal Logic)}$, a new probabilistic temporal logic for the verification of Markov Decision Processes (MDP). PCFTL is the first to include operators for causal reasoning, allowing us to express interventional and counterfactual queries. Given a path formula $\\phi$, an interventional property is concerned with the satisfaction probability of $\\phi$ if we apply a particular change $I$ to the MDP (e.g., switching to a different policy); a counterfactual allows us to compute, given an observed MDP path $\\tau$, what the outcome of $\\phi$ would have been had we applied $I$ in the past. For its ability to reason about \\textit{what-if} scenarios involving different configurations of the MDP, our approach represents a departure from existing probabilistic temporal logics that can only reason about a fixed system configuration. From a syntactic viewpoint, we introduce a generalized counterfactual operator that subsumes both interventional and counterfactual probabilities as well as the traditional probabilistic operator found in e.g., PCTL. From a semantics viewpoint, our logic is interpreted over a structural causal model translation of the MDP, which gives us a representation amenable to counterfactual reasoning. We evaluate PCFTL in the context of safe reinforcement learning using a benchmark of grid-world models.", "date": "2022-12-16", "categories": []}
{"id": "2411.14003", "title": "Generative Intervention Models for Causal Perturbation Modeling", "abstract": "We consider the problem of predicting perturbation effects via causal models. In many applications, it is a priori unknown which mechanisms of a system are modified by an external perturbation, even though the features of the perturbation are available. For example, in genomics, some properties of a drug may be known, but not their causal effects on the regulatory pathways of cells. We propose a generative intervention model (GIM) that learns to map these perturbation features to distributions over atomic interventions in a jointly-estimated causal model. Contrary to prior approaches, this enables us to predict the distribution shifts of unseen perturbation features while gaining insights about their mechanistic effects in the underlying data-generating process. On synthetic data and scRNA-seq drug perturbation data, GIMs achieve robust out-of-distribution predictions on par with unstructured approaches, while effectively inferring the underlying perturbation mechanisms, often better than other causal inference methods.", "date": "2024-11-21", "categories": []}
{"id": "2505.12147", "title": "Causal Machine Learning in IoT-based Engineering Problems: A Tool Comparison in the Case of Household Energy Consumption", "abstract": "The rapid increase in computing power and the ability to store Big Data in the infrastructure has enabled predictions in a large variety of domains by Machine Learning. However, in many cases, existing Machine Learning tools are considered insufficient or incorrect since they exploit only probabilistic dependencies rather than inference logic. Causal Machine Learning methods seem to close this gap. In this paper, two prevalent tools based on Causal Machine Learning methods are compared, as well as their mathematical underpinning background. The operation of the tools is demonstrated by examining their response to 18 queries, based on the IDEAL Household Energy Dataset, published by the University of Edinburgh. First, it was important to evaluate the causal relations assumption that allowed the use of this approach; this was based on the preexisting scientific knowledge of the domain and was implemented by use of the in-built validation tools. Results were encouraging and may easily be extended to other domains.", "date": "2025-05-17", "categories": []}
{"id": "2507.00389", "title": "Causal Prompting for Implicit Sentiment Analysis with Large Language Models", "abstract": "Implicit Sentiment Analysis (ISA) aims to infer sentiment that is implied rather than explicitly stated, requiring models to perform deeper reasoning over subtle contextual cues. While recent prompting-based methods using Large Language Models (LLMs) have shown promise in ISA, they often rely on majority voting over chain-of-thought (CoT) reasoning paths without evaluating their causal validity, making them susceptible to internal biases and spurious correlations. To address this challenge, we propose CAPITAL, a causal prompting framework that incorporates front-door adjustment into CoT reasoning. CAPITAL decomposes the overall causal effect into two components: the influence of the input prompt on the reasoning chains, and the impact of those chains on the final output. These components are estimated using encoder-based clustering and the NWGM approximation, with a contrastive learning objective used to better align the encoder's representation with the LLM's reasoning space. Experiments on benchmark ISA datasets with three LLMs demonstrate that CAPITAL consistently outperforms strong prompting baselines in both accuracy and robustness, particularly under adversarial conditions. This work offers a principled approach to integrating causal inference into LLM prompting and highlights its benefits for bias-aware sentiment reasoning. The source code and case study are available at: https://github.com/whZ62/CAPITAL.", "date": "2025-07-01", "categories": []}
{"id": "2507.00440", "title": "A Recipe for Causal Graph Regression: Confounding Effects Revisited", "abstract": "Through recognizing causal subgraphs, causal graph learning (CGL) has risen to be a promising approach for improving the generalizability of graph neural networks under out-of-distribution (OOD) scenarios. However, the empirical successes of CGL techniques are mostly exemplified in classification settings, while regression tasks, a more challenging setting in graph learning, are overlooked. We thus devote this work to tackling causal graph regression (CGR); to this end we reshape the processing of confounding effects in existing CGL studies, which mainly deal with classification. Specifically, we reflect on the predictive power of confounders in graph-level regression, and generalize classification-specific causal intervention techniques to regression through a lens of contrastive learning. Extensive experiments on graph OOD benchmarks validate the efficacy of our proposals for CGR. The model implementation and the code are provided on https://github.com/causal-graph/CGR.", "date": "2025-07-01", "categories": []}
{"id": "2406.10481", "title": "DCILP: A Distributed Approach for Large-Scale Causal Structure Learning", "abstract": "Causal learning tackles the computationally demanding task of estimating causal graphs. This paper introduces a new divide-and-conquer approach for causal graph learning, called DCILP. In the divide phase, the Markov blanket MB($X_i$) of each variable $X_i$ is identified, and causal learning subproblems associated with each MB($X_i$) are independently addressed in parallel. This approach benefits from a more favorable ratio between the number of data samples and the number of variables considered. In counterpart, it can be adversely affected by the presence of hidden confounders, as variables external to MB($X_i$) might influence those within it. The reconciliation of the local causal graphs generated during the divide phase is a challenging combinatorial optimization problem, especially in large-scale applications. The main novelty of DCILP is an original formulation of this reconciliation as an integer linear programming (ILP) problem, which can be delegated and efficiently handled by an ILP solver. Through experiments on medium to large scale graphs, and comparisons with state-of-the-art methods, DCILP demonstrates significant improvements in terms of computational complexity, while preserving the learning accuracy on real-world problem and suffering at most a slight loss of accuracy on synthetic problems.", "date": "2024-06-15", "categories": []}
{"id": "2506.04194", "title": "What Makes Treatment Effects Identifiable? Characterizations and Estimators Beyond Unconfoundedness", "abstract": "Most of the widely used estimators of the average treatment effect (ATE) in causal inference rely on the assumptions of unconfoundedness and overlap. Unconfoundedness requires that the observed covariates account for all correlations between the outcome and treatment. Overlap requires the existence of randomness in treatment decisions for all individuals. Nevertheless, many types of studies frequently violate unconfoundedness or overlap, for instance, observational studies with deterministic treatment decisions - popularly known as Regression Discontinuity designs - violate overlap.\n  In this paper, we initiate the study of general conditions that enable the identification of the average treatment effect, extending beyond unconfoundedness and overlap. In particular, following the paradigm of statistical learning theory, we provide an interpretable condition that is sufficient and necessary for the identification of ATE. Moreover, this condition also characterizes the identification of the average treatment effect on the treated (ATT) and can be used to characterize other treatment effects as well. To illustrate the utility of our condition, we present several well-studied scenarios where our condition is satisfied and, hence, we prove that ATE can be identified in regimes that prior works could not capture. For example, under mild assumptions on the data distributions, this holds for the models proposed by Tan (2006) and Rosenbaum (2002), and the Regression Discontinuity design model introduced by Thistlethwaite and Campbell (1960). For each of these scenarios, we also show that, under natural additional assumptions, ATE can be estimated from finite samples.\n  We believe these findings open new avenues for bridging learning-theoretic insights and causal inference methodologies, particularly in observational studies with complex treatment mechanisms.", "date": "2025-06-04", "categories": []}
{"id": "2506.17718", "title": "Learning Time-Aware Causal Representation for Model Generalization in Evolving Domains", "abstract": "Endowing deep models with the ability to generalize in dynamic scenarios is of vital significance for real-world deployment, given the continuous and complex changes in data distribution. Recently, evolving domain generalization (EDG) has emerged to address distribution shifts over time, aiming to capture evolving patterns for improved model generalization. However, existing EDG methods may suffer from spurious correlations by modeling only the dependence between data and targets across domains, creating a shortcut between task-irrelevant factors and the target, which hinders generalization. To this end, we design a time-aware structural causal model (SCM) that incorporates dynamic causal factors and the causal mechanism drifts, and propose \\textbf{S}tatic-D\\textbf{YN}amic \\textbf{C}ausal Representation Learning (\\textbf{SYNC}), an approach that effectively learns time-aware causal representations. Specifically, it integrates specially designed information-theoretic objectives into a sequential VAE framework which captures evolving patterns, and produces the desired representations by preserving intra-class compactness of causal factors both across and within domains. Moreover, we theoretically show that our method can yield the optimal causal predictor for each time domain. Results on both synthetic and real-world datasets exhibit that SYNC can achieve superior temporal generalization performance.", "date": "2025-06-21", "categories": []}
{"id": "2506.23374", "title": "When Additive Noise Meets Unobserved Mediators: Bivariate Denoising Diffusion for Causal Discovery", "abstract": "Distinguishing cause and effect from bivariate observational data is a foundational problem in many disciplines, but challenging without additional assumptions. Additive noise models (ANMs) are widely used to enable sample-efficient bivariate causal discovery. However, conventional ANM-based methods fail when unobserved mediators corrupt the causal relationship between variables. This paper makes three key contributions: first, we rigorously characterize why standard ANM approaches break down in the presence of unmeasured mediators. Second, we demonstrate that prior solutions for hidden mediation are brittle in finite sample settings, limiting their practical utility. To address these gaps, we propose Bivariate Denoising Diffusion (BiDD) for causal discovery, a method designed to handle latent noise introduced by unmeasured mediators. Unlike prior methods that infer directionality through mean squared error loss comparisons, our approach introduces a novel independence test statistic: during the noising and denoising processes for each variable, we condition on the other variable as input and evaluate the independence of the predicted noise relative to this input. We prove asymptotic consistency of BiDD under the ANM, and conjecture that it performs well under hidden mediation. Experiments on synthetic and real-world data demonstrate consistent performance, outperforming existing methods in mediator-corrupted settings while maintaining strong performance in mediator-free settings.", "date": "2025-06-29", "categories": []}
{"id": "2407.07596", "title": "Learning treatment effects while treating those in need", "abstract": "Many social programs attempt to allocate scarce resources to people with the greatest need. Indeed, public services increasingly use algorithmic risk assessments motivated by this goal. However, targeting the highest-need recipients often conflicts with attempting to evaluate the causal effect of the program as a whole, as the best evaluations would be obtained by randomizing the allocation. We propose a framework to design randomized allocation rules which optimally balance targeting high-need individuals with learning treatment effects, presenting policymakers with a Pareto frontier between the two goals. We give sample complexity guarantees for the policy learning problem and provide a computationally efficient strategy to implement it. We then collaborate with the human services department of Allegheny County, Pennsylvania to evaluate our methods on data from real service delivery settings. Optimized policies can substantially mitigate the tradeoff between learning and targeting. For example, it is often possible to obtain 90% of the optimal utility in targeting high-need individuals while ensuring that the average treatment effect can be estimated with less than 2 times the samples that a randomized controlled trial would require. Mechanisms for targeting public services often focus on measuring need as accurately as possible. However, our results suggest that algorithmic systems in public services can be most impactful if they incorporate program evaluation as an explicit goal alongside targeting.", "date": "2024-07-10", "categories": []}
{"id": "2408.11240", "title": "Asymmetric Graph Error Control with Low Complexity in Causal Bandits", "abstract": "In this paper, the causal bandit problem is investigated, with the objective of maximizing the long-term reward by selecting an optimal sequence of interventions on nodes in an unknown causal graph. It is assumed that both the causal topology and the distribution of interventions are unknown. First, based on the difference between the two types of graph identification errors (false positives and negatives), a causal graph learning method is proposed. Numerical results suggest that this method has a much lower sample complexity relative to the prior art by learning sub-graphs. However, we note that a sample complexity analysis for the new algorithm has not been undertaken, as of yet. Under the assumption of minimum-mean squared error weight estimation, a new uncertainty bound tailored to the causal bandit problem is derived. This uncertainty bound drives an upper confidence bound-based intervention selection to optimize the reward. Further, we consider a particular instance of non-stationary bandits wherein both the causal topology and interventional distributions can change. Our solution is the design of a sub-graph change detection mechanism that requires a modest number of samples. Numerical results compare the new methodology to existing schemes and show a substantial performance improvement in stationary and non-stationary settings. Averaged over 100 randomly generated causal bandits, the proposed scheme takes significantly fewer samples to learn the causal structure and achieves a reward gain of 85% compared to existing approaches.", "date": "2024-08-20", "categories": []}
{"id": "2409.14593", "title": "Testing Causal Models with Hidden Variables in Polynomial Delay via Conditional Independencies", "abstract": "Testing a hypothesized causal model against observational data is a key prerequisite for many causal inference tasks. A natural approach is to test whether the conditional independence relations (CIs) assumed in the model hold in the data. While a model can assume exponentially many CIs (with respect to the number of variables), testing all of them is both impractical and unnecessary. Causal graphs, which encode these CIs in polynomial space, give rise to local Markov properties that enable model testing with a significantly smaller subset of CIs. Model testing based on local properties requires an algorithm to list the relevant CIs. However, existing algorithms for realistic settings with hidden variables and non-parametric distributions can take exponential time to produce even a single CI constraint. In this paper, we introduce the c-component local Markov property (C-LMP) for causal graphs with hidden variables. Since C-LMP can still invoke an exponential number of CIs, we develop a polynomial delay algorithm to list these CIs in poly-time intervals. To our knowledge, this is the first algorithm that enables poly-delay testing of CIs in causal graphs with hidden variables against arbitrary data distributions. Experiments on real-world and synthetic data demonstrate the practicality of our algorithm.", "date": "2024-09-22", "categories": []}
{"id": "2506.21154", "title": "Transformer-Based Spatial-Temporal Counterfactual Outcomes Estimation", "abstract": "The real world naturally has dimensions of time and space. Therefore, estimating the counterfactual outcomes with spatial-temporal attributes is a crucial problem. However, previous methods are based on classical statistical models, which still have limitations in performance and generalization. This paper proposes a novel framework for estimating counterfactual outcomes with spatial-temporal attributes using the Transformer, exhibiting stronger estimation ability. Under mild assumptions, the proposed estimator within this framework is consistent and asymptotically normal. To validate the effectiveness of our approach, we conduct simulation experiments and real data experiments. Simulation experiments show that our estimator has a stronger estimation capability than baseline methods. Real data experiments provide a valuable conclusion to the causal effect of conflicts on forest loss in Colombia. The source code is available at https://github.com/lihe-maxsize/DeppSTCI_Release_Version-master.", "date": "2025-06-26", "categories": []}
{"id": "2506.20169", "title": "Causal discovery in deterministic discrete LTI-DAE systems", "abstract": "Discovering pure causes or driver variables in deterministic LTI systems is of vital importance in the data-driven reconstruction of causal networks. A recent work by Kathari and Tangirala, proposed in 2022, formulated the causal discovery method as a constraint identification problem. The constraints are identified using a dynamic iterative PCA (DIPCA)-based approach for dynamical systems corrupted with Gaussian measurement errors. The DIPCA-based method works efficiently for dynamical systems devoid of any algebraic relations. However, several dynamical systems operate under feedback control and/or are coupled with conservation laws, leading to differential-algebraic (DAE) or mixed causal systems. In this work, a method, namely the partition of variables (PoV), for causal discovery in LTI-DAE systems is proposed. This method is superior to the method that was presented by Kathari and Tangirala (2022), as PoV also works for pure dynamical systems, which are devoid of algebraic equations. The proposed method identifies the causal drivers up to a minimal subset. PoV deploys DIPCA to first determine the number of algebraic relations ($n_a$), the number of dynamical relations ($n_d$) and the constraint matrix. Subsequently, the subsets are identified through an admissible partitioning of the constraint matrix by finding the condition number of it. Case studies are presented to demonstrate the effectiveness of the proposed method.", "date": "2025-06-25", "categories": []}
{"id": "2405.08498", "title": "Learning Decision Policies with Instrumental Variables through Double Machine Learning", "abstract": "A common issue in learning decision-making policies in data-rich settings is spurious correlations in the offline dataset, which can be caused by hidden confounders. Instrumental variable (IV) regression, which utilises a key unconfounded variable known as the instrument, is a standard technique for learning causal relationships between confounded action, outcome, and context variables. Most recent IV regression algorithms use a two-stage approach, where a deep neural network (DNN) estimator learnt in the first stage is directly plugged into the second stage, in which another DNN is used to estimate the causal effect. Naively plugging the estimator can cause heavy bias in the second stage, especially when regularisation bias is present in the first stage estimator. We propose DML-IV, a non-linear IV regression method that reduces the bias in two-stage IV regressions and effectively learns high-performing policies. We derive a novel learning objective to reduce bias and design the DML-IV algorithm following the double/debiased machine learning (DML) framework. The learnt DML-IV estimator has strong convergence rate and $O(N^{-1/2})$ suboptimality guarantees that match those when the dataset is unconfounded. DML-IV outperforms state-of-the-art IV regression methods on IV regression benchmarks and learns high-performing policies in the presence of instruments.", "date": "2024-05-14", "categories": []}
{"id": "2406.19573", "title": "On Counterfactual Interventions in Vector Autoregressive Models", "abstract": "Counterfactual reasoning allows us to explore hypothetical scenarios in order to explain the impacts of our decisions. However, addressing such inquires is impossible without establishing the appropriate mathematical framework. In this work, we introduce the problem of counterfactual reasoning in the context of vector autoregressive (VAR) processes. We also formulate the inference of a causal model as a joint regression task where for inference we use both data with and without interventions. After learning the model, we exploit linearity of the VAR model to make exact predictions about the effects of counterfactual interventions. Furthermore, we quantify the total causal effects of past counterfactual interventions. The source code for this project is freely available at https://github.com/KurtButler/counterfactual_interventions.", "date": "2024-06-27", "categories": []}
{"id": "2410.23499", "title": "Tangent Space Causal Inference: Leveraging Vector Fields for Causal Discovery in Dynamical Systems", "abstract": "Causal discovery with time series data remains a challenging yet increasingly important task across many scientific domains. Convergent cross mapping (CCM) and related methods have been proposed to study time series that are generated by dynamical systems, where traditional approaches like Granger causality are unreliable. However, CCM often yields inaccurate results depending upon the quality of the data. We propose the Tangent Space Causal Inference (TSCI) method for detecting causalities in dynamical systems. TSCI works by considering vector fields as explicit representations of the systems' dynamics and checks for the degree of synchronization between the learned vector fields. The TSCI approach is model-agnostic and can be used as a drop-in replacement for CCM and its generalizations. We first present a basic version of the TSCI algorithm, which is shown to be more effective than the basic CCM algorithm with very little additional computation. We additionally present augmented versions of TSCI that leverage the expressive power of latent variable models and deep learning. We validate our theory on standard systems, and we demonstrate improved causal inference performance across a number of benchmark tasks.", "date": "2024-10-30", "categories": []}
{"id": "2411.10154", "title": "Continuous Bayesian Model Selection for Multivariate Causal Discovery", "abstract": "Current causal discovery approaches require restrictive model assumptions in the absence of interventional data to ensure structure identifiability. These assumptions often do not hold in real-world applications leading to a loss of guarantees and poor performance in practice. Recent work has shown that, in the bivariate case, Bayesian model selection can greatly improve performance by exchanging restrictive modelling for more flexible assumptions, at the cost of a small probability of making an error. Our work shows that this approach is useful in the important multivariate case as well. We propose a scalable algorithm leveraging a continuous relaxation of the discrete model selection problem. Specifically, we employ the Causal Gaussian Process Conditional Density Estimator (CGP-CDE) as a Bayesian non-parametric model, using its hyperparameters to construct an adjacency matrix. This matrix is then optimised using the marginal likelihood and an acyclicity regulariser, giving the maximum a posteriori causal graph. We demonstrate the competitiveness of our approach, showing it is advantageous to perform multivariate causal discovery without infeasible assumptions using Bayesian model selection.", "date": "2024-11-15", "categories": []}
{"id": "2506.02200", "title": "Learning Treatment Representations for Downstream Instrumental Variable Regression", "abstract": "Traditional instrumental variable (IV) estimators face a fundamental constraint: they can only accommodate as many endogenous treatment variables as available instruments. This limitation becomes particularly challenging in settings where the treatment is presented in a high-dimensional and unstructured manner (e.g. descriptions of patient treatment pathways in a hospital). In such settings, researchers typically resort to applying unsupervised dimension reduction techniques to learn a low-dimensional treatment representation prior to implementing IV regression analysis. We show that such methods can suffer from substantial omitted variable bias due to implicit regularization in the representation learning step. We propose a novel approach to construct treatment representations by explicitly incorporating instrumental variables during the representation learning process. Our approach provides a framework for handling high-dimensional endogenous variables with limited instruments. We demonstrate both theoretically and empirically that fitting IV models on these instrument-informed representations ensures identification of directions that optimize outcome prediction. Our experiments show that our proposed methodology improves upon the conventional two-stage approaches that perform dimension reduction without incorporating instrument information.", "date": "2025-06-02", "categories": []}
{"id": "2506.14950", "title": "Double Machine Learning for Conditional Moment Restrictions: IV Regression, Proximal Causal Learning and Beyond", "abstract": "Solving conditional moment restrictions (CMRs) is a key problem considered in statistics, causal inference, and econometrics, where the aim is to solve for a function of interest that satisfies some conditional moment equalities. Specifically, many techniques for causal inference, such as instrumental variable (IV) regression and proximal causal learning (PCL), are CMR problems. Most CMR estimators use a two-stage approach, where the first-stage estimation is directly plugged into the second stage to estimate the function of interest. However, naively plugging in the first-stage estimator can cause heavy bias in the second stage. This is particularly the case for recently proposed CMR estimators that use deep neural network (DNN) estimators for both stages, where regularisation and overfitting bias is present. We propose DML-CMR, a two-stage CMR estimator that provides an unbiased estimate with fast convergence rate guarantees. We derive a novel learning objective to reduce bias and develop the DML-CMR algorithm following the double/debiased machine learning (DML) framework. We show that our DML-CMR estimator can achieve the minimax optimal convergence rate of $O(N^{-1/2})$ under parameterisation and mild regularity conditions, where $N$ is the sample size. We apply DML-CMR to a range of problems using DNN estimators, including IV regression and proximal causal learning on real-world datasets, demonstrating state-of-the-art performance against existing CMR estimators and algorithms tailored to those problems.", "date": "2025-06-17", "categories": []}
{"id": "2506.19049", "title": "Which Company Adjustment Matter? Insights from Uplift Modeling on Financial Health", "abstract": "Uplift modeling has achieved significant success in various fields, particularly in online marketing. It is a method that primarily utilizes machine learning and deep learning to estimate individual treatment effects. This paper we apply uplift modeling to analyze the effect of company adjustment on their financial status, and we treat these adjustment as treatments or interventions in this study. Although there have been extensive studies and application regarding binary treatments, multiple treatments, and continuous treatments, company adjustment are often more complex than these scenarios, as they constitute a series of multiple time-dependent actions. The effect estimation of company adjustment needs to take into account not only individual treatment traits but also the temporal order of this series of treatments. This study collects a real-world data set about company financial statements and reported behavior in Luxembourg for the experiments. First, we use two meta-learners and three other well-known uplift models to analyze different company adjustment by simplifying the adjustment as binary treatments. Furthermore, we propose a new uplift modeling framework (MTDnet) to address the time-dependent nature of these adjustment, and the experimental result shows the necessity of considering the timing of these adjustment.", "date": "2025-06-23", "categories": []}
{"id": "2506.19459", "title": "Tagged for Direction: Pinning Down Causal Edge Directions with Precision", "abstract": "Not every causal relation between variables is equal, and this can be leveraged for the task of causal discovery. Recent research shows that pairs of variables with particular type assignments induce a preference on the causal direction of other pairs of variables with the same type. Although useful, this assignment of a specific type to a variable can be tricky in practice. We propose a tag-based causal discovery approach where multiple tags are assigned to each variable in a causal graph. Existing causal discovery approaches are first applied to direct some edges, which are then used to determine edge relations between tags. Then, these edge relations are used to direct the undirected edges. Doing so improves upon purely type-based relations, where the assumption of type consistency lacks robustness and flexibility due to being restricted to single types for each variable. Our experimental evaluations show that this boosts causal discovery and that these high-level tag relations fit common knowledge.", "date": "2025-06-24", "categories": []}
{"id": "2506.19650", "title": "Identifying Macro Causal Effects in C-DMGs over DMGs", "abstract": "The do-calculus is a sound and complete tool for identifying causal effects in acyclic directed mixed graphs (ADMGs) induced by structural causal models (SCMs). However, in many real-world applications, especially in high-dimensional setting, constructing a fully specified ADMG is often infeasible. This limitation has led to growing interest in partially specified causal representations, particularly through cluster-directed mixed graphs (C-DMGs), which group variables into clusters and offer a more abstract yet practical view of causal dependencies. While these representations can include cycles, recent work has shown that the do-calculus remains sound and complete for identifying macro-level causal effects in C-DMGs over ADMGs under the assumption that all clusters size are greater than 1. Nevertheless, real-world systems often exhibit cyclic causal dynamics at the structural level. To account for this, input-output structural causal models (ioSCMs) have been introduced as a generalization of SCMs that allow for cycles. ioSCMs induce another type of graph structure known as a directed mixed graph (DMG). Analogous to the ADMG setting, one can define C-DMGs over DMGs as high-level representations of causal relations among clusters of variables. In this paper, we prove that, unlike in the ADMG setting, the do-calculus is unconditionally sound and complete for identifying macro causal effects in C-DMGs over DMGs. Furthermore, we show that the graphical criteria for non-identifiability of macro causal effects previously established C-DMGs over ADMGs naturally extends to a subset of C-DMGs over DMGs.", "date": "2025-06-24", "categories": []}
{"id": "2506.17826", "title": "Actionable Interpretability via Causal Hypergraphs: Unravelling Batch Size Effects in Deep Learning", "abstract": "While the impact of batch size on generalisation is well studied in vision tasks, its causal mechanisms remain underexplored in graph and text domains. We introduce a hypergraph-based causal framework, HGCNet, that leverages deep structural causal models (DSCMs) to uncover how batch size influences generalisation via gradient noise, minima sharpness, and model complexity. Unlike prior approaches based on static pairwise dependencies, HGCNet employs hypergraphs to capture higher-order interactions across training dynamics. Using do-calculus, we quantify direct and mediated effects of batch size interventions, providing interpretable, causally grounded insights into optimisation. Experiments on citation networks, biomedical text, and e-commerce reviews show that HGCNet outperforms strong baselines including GCN, GAT, PI-GNN, BERT, and RoBERTa. Our analysis reveals that smaller batch sizes causally enhance generalisation through increased stochasticity and flatter minima, offering actionable interpretability to guide training strategies in deep learning. This work positions interpretability as a driver of principled architectural and optimisation choices beyond post hoc analysis.", "date": "2025-06-21", "categories": []}
{"id": "2506.18285", "title": "Learning Causal Graphs at Scale: A Foundation Model Approach", "abstract": "Due to its human-interpretability and invariance properties, Directed Acyclic Graph (DAG) has been a foundational tool across various areas of AI research, leading to significant advancements. However, DAG learning remains highly challenging, due to its super-exponential growth in computational cost and identifiability issues, particularly in small-sample regimes. To address these two challenges, in this work we leverage the recent success of linear transformers and develop a foundation model approach for discovering multiple order-consistent DAGs across tasks. In particular, we propose Attention-DAG (ADAG), a novel attention-mechanism-based architecture for learning multiple linear Structural Equation Models (SEMs). ADAG learns the mapping from observed data to both graph structure and parameters via a nonlinear attention-based kernel, enabling efficient multi-task estimation of the underlying linear SEMs. By formulating the learning process across multiple tasks as a continuous optimization problem, the pre-trained ADAG model captures the common structural properties as a shared low-dimensional prior, thereby reducing the ill-posedness of downstream DAG learning tasks in small-sample regimes. We evaluate our proposed approach on benchmark synthetic datasets and find that ADAG achieves substantial improvements in both DAG learning accuracy and zero-shot inference efficiency. To the best of our knowledge, this is the first practical approach for pre-training a foundation model specifically designed for DAG learning, representing a step toward more efficient and generalizable down-stream applications in causal discovery.", "date": "2025-06-23", "categories": []}
{"id": "2311.16984", "title": "FedECA: Federated External Control Arms for Causal Inference with Time-To-Event Data in Distributed Settings", "abstract": "External control arms can inform early clinical development of experimental drugs and provide efficacy evidence for regulatory approval. However, accessing sufficient real-world or historical clinical trials data is challenging. Indeed, regulations protecting patients' rights by strictly controlling data processing make pooling data from multiple sources in a central server often difficult. To address these limitations, we develop a method that leverages federated learning to enable inverse probability of treatment weighting for time-to-event outcomes on separate cohorts without needing to pool data. To showcase its potential, we apply it in different settings of increasing complexity, culminating with a real-world use-case in which our method is used to compare the treatment effect of two approved chemotherapy regimens using data from three separate cohorts of patients with metastatic pancreatic cancer. By sharing our code, we hope it will foster the creation of federated research networks and thus accelerate drug development.", "date": "2023-11-28", "categories": []}
{"id": "2504.12345", "title": "Reimagining Urban Science: Scaling Causal Inference with Large Language Models", "abstract": "Urban causal research is essential for understanding the complex, dynamic processes that shape cities and for informing evidence-based policies. However, current practices are often constrained by inefficient and biased hypothesis formulation, challenges in integrating multimodal data, and fragile experimental methodologies. Imagine a system that automatically estimates the causal impact of congestion pricing on commute times by income group or measures how new green spaces affect asthma rates across neighborhoods using satellite imagery and health reports, and then generates comprehensive, policy-ready outputs, including causal estimates, subgroup analyses, and actionable recommendations. In this Perspective, we propose UrbanCIA, an LLM-driven conceptual framework composed of four distinct modular agents responsible for hypothesis generation, data engineering, experiment design and execution, and results interpretation with policy insights. We begin by examining the current landscape of urban causal research through a structured taxonomy of research topics, data sources, and methodological approaches, revealing systemic limitations across the workflow. Next, we introduce the design principles and technological roadmap for the four modules in the proposed framework. We also propose evaluation criteria to assess the rigor and transparency of these AI-augmented processes. Finally, we reflect on the broader implications for human-AI collaboration, equity, and accountability. We call for a new research agenda that embraces LLM-driven tools as catalysts for more scalable, reproducible, and inclusive urban research.", "date": "2025-04-15", "categories": []}
{"id": "2505.00310", "title": "Statistical Learning for Heterogeneous Treatment Effects: Pretraining, Prognosis, and Prediction", "abstract": "Robust estimation of heterogeneous treatment effects is a fundamental challenge for optimal decision-making in domains ranging from personalized medicine to educational policy. In recent years, predictive machine learning has emerged as a valuable toolbox for causal estimation, enabling more flexible effect estimation. However, accurately estimating conditional average treatment effects (CATE) remains a major challenge, particularly in the presence of many covariates. In this article, we propose pretraining strategies that leverage a phenomenon in real-world applications: factors that are prognostic of the outcome are frequently also predictive of treatment effect heterogeneity. In medicine, for example, components of the same biological signaling pathways frequently influence both baseline risk and treatment response. Specifically, we demonstrate our approach within the R-learner framework, which estimates the CATE by solving individual prediction problems based on a residualized loss. We use this structure to incorporate side information and develop models that can exploit synergies between risk prediction and causal effect estimation. In settings where these synergies are present, this cross-task learning enables more accurate signal detection, yields lower estimation error, reduced false discovery rates, and higher power for detecting heterogeneity.", "date": "2025-05-01", "categories": []}
{"id": "2506.15708", "title": "Refined Causal Graph Structure Learning via Curvature for Brain Disease Classification", "abstract": "Graph neural networks (GNNs) have been developed to model the relationship between regions of interest (ROIs) in brains and have shown significant improvement in detecting brain diseases. However, most of these frameworks do not consider the intrinsic relationship of causality factor between brain ROIs, which is arguably more essential to observe cause and effect interaction between signals rather than typical correlation values. We propose a novel framework called CGB (Causal Graphs for Brains) for brain disease classification/detection, which models refined brain networks based on the causal discovery method, transfer entropy, and geometric curvature strategy. CGB unveils causal relationships between ROIs that bring vital information to enhance brain disease classification performance. Furthermore, CGB also performs a graph rewiring through a geometric curvature strategy to refine the generated causal graph to become more expressive and reduce potential information bottlenecks when GNNs model it. Our extensive experiments show that CGB outperforms state-of-the-art methods in classification tasks on brain disease datasets, as measured by average F1 scores.", "date": "2025-05-30", "categories": []}
{"id": "2506.15758", "title": "Linear-Time Primitives for Algorithm Development in Graphical Causal Inference", "abstract": "We introduce CIfly, a framework for efficient algorithmic primitives in graphical causal inference that isolates reachability as a reusable core operation. It builds on the insight that many causal reasoning tasks can be reduced to reachability in purpose-built state-space graphs that can be constructed on the fly during traversal. We formalize a rule table schema for specifying such algorithms and prove they run in linear time. We establish CIfly as a more efficient alternative to the common primitives moralization and latent projection, which we show are computationally equivalent to Boolean matrix multiplication. Our open-source Rust implementation parses rule table text files and runs the specified CIfly algorithms providing high-performance execution accessible from Python and R. We demonstrate CIfly's utility by re-implementing a range of established causal inference tasks within the framework and by developing new algorithms for instrumental variables. These contributions position CIfly as a flexible and scalable backbone for graphical causal inference, guiding algorithm development and enabling easy and efficient deployment.", "date": "2025-06-18", "categories": []}
{"id": "2410.11759", "title": "LoSAM: Local Search in Additive Noise Models with Mixed Mechanisms and General Noise for Global Causal Discovery", "abstract": "Inferring causal relationships from observational data is crucial when experiments are costly or infeasible. Additive noise models (ANMs) enable unique directed acyclic graph (DAG) identification, but existing sample-efficient ANM methods often rely on restrictive assumptions on the data generating process, limiting their applicability to real-world settings. We propose local search in additive noise models, LoSAM, a topological ordering method for learning a unique DAG in ANMs with mixed causal mechanisms and general noise distributions. We introduce new causal substructures and criteria for identifying roots and leaves, enabling efficient top-down learning. We prove asymptotic consistency and polynomial runtime, ensuring scalability and sample efficiency. We test LoSAM on synthetic and real-world data, demonstrating state-of-the-art performance across all mixed mechanism settings.", "date": "2024-10-15", "categories": []}
{"id": "2503.13912", "title": "KANITE: Kolmogorov-Arnold Networks for ITE estimation", "abstract": "We introduce KANITE, a framework leveraging Kolmogorov-Arnold Networks (KANs) for Individual Treatment Effect (ITE) estimation under multiple treatments setting in causal inference. By utilizing KAN's unique abilities to learn univariate activation functions as opposed to learning linear weights by Multi-Layer Perceptrons (MLPs), we improve the estimates of ITEs. The KANITE framework comprises two key architectures: 1.Integral Probability Metric (IPM) architecture: This employs an IPM loss in a specialized manner to effectively align towards ITE estimation across multiple treatments. 2. Entropy Balancing (EB) architecture: This uses weights for samples that are learned by optimizing entropy subject to balancing the covariates across treatment groups. Extensive evaluations on benchmark datasets demonstrate that KANITE outperforms state-of-the-art algorithms in both $\\epsilon_{\\text{PEHE}}$ and $\\epsilon_{\\text{ATE}}$ metrics. Our experiments highlight the advantages of KANITE in achieving improved causal estimates, emphasizing the potential of KANs to advance causal inference methodologies across diverse application areas.", "date": "2025-03-18", "categories": []}
{"id": "2506.14862", "title": "Identifiability by common backdoor in summary causal graphs of time series", "abstract": "The identifiability problem for interventions aims at assessing whether the total effect of some given interventions can be written with a do-free formula, and thus be computed from observational data only. We study this problem, considering multiple interventions and multiple effects, in the context of time series when only abstractions of the true causal graph in the form of summary causal graphs are available. We focus in this study on identifiability by a common backdoor set, and establish, for time series with and without consistency throughout time, conditions under which such a set exists. We also provide algorithms of limited complexity to decide whether the problem is identifiable or not.", "date": "2025-06-17", "categories": []}
{"id": "2306.11908", "title": "Generalized Random Forests using Fixed-Point Trees", "abstract": "We propose a computationally efficient alternative to generalized random forests (GRFs) for estimating heterogeneous effects in large dimensions. While GRFs rely on a gradient-based splitting criterion, which in large dimensions is computationally expensive and unstable, our method introduces a fixed-point approximation that eliminates the need for Jacobian estimation. This gradient-free approach preserves GRF's theoretical guarantees of consistency and asymptotic normality while significantly improving computational efficiency. We demonstrate that our method achieves a speedup of multiple times over standard GRFs without compromising statistical accuracy. Experiments on both simulated and real-world data validate our approach. Our findings suggest that the proposed method is a scalable alternative for localized effect estimation in machine learning and causal inference applications", "date": "2023-06-20", "categories": []}
{"id": "2506.13803", "title": "Causality in the human niche: lessons for machine learning", "abstract": "Humans interpret the world around them in terms of cause and effect and communicate their understanding of the world to each other in causal terms. These causal aspects of human cognition are thought to underlie humans' ability to generalize and learn efficiently in new domains, an area where current machine learning systems are weak. Building human-like causal competency into machine learning systems may facilitate the construction of effective and interpretable AI. Indeed, the machine learning community has been importing ideas on causality formalized by the Structural Causal Model (SCM) framework, which provides a rigorous formal language for many aspects of causality and has led to significant advances. However, the SCM framework fails to capture some salient aspects of human causal cognition and has likewise not yet led to advances in machine learning in certain critical areas where humans excel. We contend that the problem of causality in the ``human niche'' -- for a social, autonomous, and goal-driven agent sensing and acting in the world in which humans live -- is quite different from the kind of causality captured by SCMs. For example, everyday objects come in similar types that have similar causal properties, and so humans readily generalize knowledge of one type of object (cups) to another related type (bowls) by drawing causal analogies between objects with similar properties, but such analogies are at best awkward to express in SCMs. We explore how such causal capabilities are adaptive in, and motivated by, the human niche. By better appreciating properties of human causal cognition and, crucially, how those properties are adaptive in the niche in which humans live, we hope that future work at the intersection of machine learning and causality will leverage more human-like inductive biases to create more capable, controllable, and interpretable systems.", "date": "2025-06-13", "categories": []}
{"id": "2506.14051", "title": "Estimation of Treatment Effects in Extreme and Unobserved Data", "abstract": "Causal effect estimation seeks to determine the impact of an intervention from observational data. However, the existing causal inference literature primarily addresses treatment effects on frequently occurring events. But what if we are interested in estimating the effects of a policy intervention whose benefits, while potentially important, can only be observed and measured in rare yet impactful events, such as extreme climate events? The standard causal inference methodology is not designed for this type of inference since the events of interest may be scarce in the observed data and some degree of extrapolation is necessary. Extreme Value Theory (EVT) provides methodologies for analyzing statistical phenomena in such extreme regimes. We introduce a novel framework for assessing treatment effects in extreme data to capture the causal effect at the occurrence of rare events of interest. In particular, we employ the theory of multivariate regular variation to model extremities. We develop a consistent estimator for extreme treatment effects and present a rigorous non-asymptotic analysis of its performance. We illustrate the performance of our estimator using both synthetic and semi-synthetic data.", "date": "2025-06-16", "categories": []}
{"id": "2506.14534", "title": "Complete Characterization for Adjustment in Summary Causal Graphs of Time Series", "abstract": "The identifiability problem for interventions aims at assessing whether the total causal effect can be written with a do-free formula, and thus be estimated from observational data only. We study this problem, considering multiple interventions, in the context of time series when only an abstraction of the true causal graph, in the form of a summary causal graph, is available. We propose in particular both necessary and sufficient conditions for the adjustment criterion, which we show is complete in this setting, and provide a pseudo-linear algorithm to decide whether the query is identifiable or not.", "date": "2025-06-17", "categories": []}
{"id": "2301.00314", "title": "Causal Deep Learning", "abstract": "We derive a set of causal deep neural networks whose architectures are a consequence of tensor (multilinear) factor analysis, a framework that facilitates causal inference. Forward causal questions are addressed with a neural network architecture composed of causal capsules and a tensor transformer. Causal capsules compute a set of invariant causal factor representations, whose interactions are governed by a tensor transformation. Inverse causal questions are addressed with a neural network that implements the multilinear projection algorithm. The architecture reverses the order of operations of a forward neural network and estimates the causes of effects. As an alternative to aggressive bottleneck dimension reduction or regularized regression that may camouflage an inherently underdetermined inverse problem, we prescribe modeling different aspects of the mechanism of data formation with piecewise tensor models whose multilinear projections produce multiple candidate solutions. Our forward and inverse questions may be addressed with shallow architectures, but for computationally scalable solutions, we derive a set of deep neural networks by taking advantage of block algebra. An interleaved kernel hierarchy results in doubly non-linear tensor factor models. The causal neural networks that are a consequence of tensor factor analysis are data agnostic, but are illustrated with facial images. Sequential, parallel and asynchronous parallel computation strategies are described.", "date": "2023-01-01", "categories": []}
{"id": "2311.03630", "title": "CATE Estimation With Potential Outcome Imputation From Local Regression", "abstract": "One of the most significant challenges in Conditional Average Treatment Effect (CATE) estimation is the statistical discrepancy between distinct treatment groups. To address this issue, we propose a model-agnostic data augmentation method for CATE estimation. First, we derive regret bounds for general data augmentation methods suggesting that a small imputation error may be necessary for accurate CATE estimation. Inspired by this idea, we propose a contrastive learning approach that reliably imputes missing potential outcomes for a selected subset of individuals formed using a similarity measure. We augment the original dataset with these reliable imputations to reduce the discrepancy between different treatment groups while inducing minimal imputation error. The augmented dataset can subsequently be employed to train standard CATE estimation models. We provide both theoretical guarantees and extensive numerical studies demonstrating the effectiveness of our approach in improving the accuracy and robustness of numerous CATE estimation models.", "date": "2023-11-07", "categories": []}
{"id": "2405.15673", "title": "Consistency of Neural Causal Partial Identification", "abstract": "Recent progress in Neural Causal Models (NCMs) showcased how identification and partial identification of causal effects can be automatically carried out via training of neural generative models that respect the constraints encoded in a given causal graph [Xia et al. 2022, Balazadeh et al. 2022]. However, formal consistency of these methods has only been proven for the case of discrete variables or only for linear causal models. In this work, we prove the consistency of partial identification via NCMs in a general setting with both continuous and categorical variables. Further, our results highlight the impact of the design of the underlying neural network architecture in terms of depth and connectivity as well as the importance of applying Lipschitz regularization in the training phase. In particular, we provide a counterexample showing that without Lipschitz regularization this method may not be asymptotically consistent. Our results are enabled by new results on the approximability of Structural Causal Models (SCMs) via neural generative models, together with an analysis of the sample complexity of the resulting architectures and how that translates into an error in the constrained optimization problem that defines the partial identification bounds.", "date": "2024-05-24", "categories": []}
{"id": "2411.05331", "title": "Discovering Latent Causal Graphs from Spatiotemporal Data", "abstract": "Many important phenomena in scientific fields like climate, neuroscience, and epidemiology are naturally represented as spatiotemporal gridded data with complex interactions. Inferring causal relationships from these data is a challenging problem compounded by the high dimensionality of such data and the correlations between spatially proximate points. We present SPACY (SPAtiotemporal Causal discoverY), a novel framework based on variational inference, designed to model latent time series and their causal relationships from spatiotemporal data. SPACY alleviates the high-dimensional challenge by discovering causal structures in the latent space. To aggregate spatially proximate, correlated grid points, we use spatial factors, parametrized by spatial kernel functions, to map observational time series to latent representations. Theoretically, we generalize the problem to a continuous spatial domain and establish identifiability when the observations arise from a nonlinear, invertible function of the product of latent series and spatial factors. Using this approach, we avoid assumptions that are often unverifiable, including those about instantaneous effects or sufficient variability. Empirically, SPACY outperforms state-of-the-art baselines on synthetic data, even in challenging settings where existing methods struggle, while remaining scalable for large grids. SPACY also identifies key known phenomena from real-world climate data. An implementation of SPACY is available at https://github.com/Rose-STL-Lab/SPACY/", "date": "2024-11-08", "categories": []}
{"id": "2411.09635", "title": "Counterfactual Uncertainty Quantification of Factual Estimand of Efficacy from Before-and-After Treatment Repeated Measures Randomized Controlled Trials", "abstract": "This article quantifies the uncertainty reduction achievable for \\textit{counterfactual} estimand, and cautions against potential bias when the estimand uses Digital Twins. Posed by Neyman (1923a) who showed unbiased \\textit{point estimation} from designed \\textit{factual} experiments is possible, \\textit{counterfactual} uncertainty quantification (CUQ) remained an open challenge for about one hundred years. The $Rx: C$ \\textit{counterfactual} efficacy we focus on is the ideal estimand for comparing treatment $Rx$ with control $C$, the expected outcome differential if each patient received \\textit{both} $Rx$ and $C$. Enabled by our new statistical modeling principle called ETZ, we show CUQ is achievable in Randomized Controlled Trials (RCTs) with \\textit{Before-and-After} Repeated Measures, common in many therapeutic areas. The CUQ we are able to achieve typically has lower variability than factual UQ.\n  We caution against using predictors with measurement error, which violates regression assumptions and can cause \\textit{attenuation} bias in estimating treatment effects. For traditional medicine and population-averaged targeted therapy, counterfactual point estimation remains unbiased. However, in both Real Human and Digital Twin approaches, estimating effects in \\emph{subgroups} may suffer attenuation bias.", "date": "2024-11-14", "categories": []}
{"id": "2412.01953", "title": "The Landscape of Causal Discovery Data: Grounding Causal Discovery in Real-World Applications", "abstract": "Causal discovery aims to automatically uncover causal relationships from data, a capability with significant potential across many scientific disciplines. However, its real-world applications remain limited. Current methods often rely on unrealistic assumptions and are evaluated only on simple synthetic toy datasets, often with inadequate evaluation metrics. In this paper, we substantiate these claims by performing a systematic review of the recent causal discovery literature. We present applications in biology, neuroscience, and Earth sciences - fields where causal discovery holds promise for addressing key challenges. We highlight available simulated and real-world datasets from these domains and discuss common assumption violations that have spurred the development of new methods. Our goal is to encourage the community to adopt better evaluation practices by utilizing realistic datasets and more adequate metrics.", "date": "2024-12-02", "categories": []}
{"id": "2506.12030", "title": "Impact, Causation and Prediction of Socio-Academic and Economic Factors in Exam-centric Student Evaluation Measures using Machine Learning and Causal Analysis", "abstract": "Understanding socio-academic and economic factors influencing students' performance is crucial for effective educational interventions. This study employs several machine learning techniques and causal analysis to predict and elucidate the impacts of these factors on academic performance. We constructed a hypothetical causal graph and collected data from 1,050 student profiles. Following meticulous data cleaning and visualization, we analyze linear relationships through correlation and variable plots, and perform causal analysis on the hypothetical graph. Regression and classification models are applied for prediction, and unsupervised causality analysis using PC, GES, ICA-LiNGAM, and GRASP algorithms is conducted. Our regression analysis shows that Ridge Regression achieve a Mean Absolute Error (MAE) of 0.12 and a Mean Squared Error (MSE) of 0.024, indicating robustness, while classification models like Random Forest achieve nearly perfect F1-scores. The causal analysis shows significant direct and indirect effects of factors such as class attendance, study hours, and group study on CGPA. These insights are validated through unsupervised causality analysis. By integrating the best regression model into a web application, we are developing a practical tool for students and educators to enhance academic outcomes based on empirical evidence.", "date": "2025-05-22", "categories": []}
{"id": "2506.12226", "title": "Learning Causality for Modern Machine Learning", "abstract": "In the past decades, machine learning with Empirical Risk Minimization (ERM) has demonstrated great capability in learning and exploiting the statistical patterns from data, or even surpassing humans. Despite the success, ERM avoids the modeling of causality the way of understanding and handling changes, which is fundamental to human intelligence. When deploying models beyond the training environment, distribution shifts are everywhere. For example, an autopilot system often needs to deal with new weather conditions that have not been seen during training, An Al-aided drug discovery system needs to predict the biochemical properties of molecules with respect to new viruses such as COVID-19. It renders the problem of Out-of-Distribution (OOD) generalization challenging to conventional machine learning.\n  In this thesis, we investigate how to incorporate and realize the causality for broader tasks in modern machine learning. In particular, we exploit the invariance implied by the principle of independent causal mechanisms (ICM), that is, the causal mechanisms generating the effects from causes do not inform or influence each other. Therefore, the conditional distribution between the target variable given its causes is invariant under distribution shifts. With the causal invariance principle, we first instantiate it to graphs -- a general data structure ubiquitous in many real-world industry and scientific applications, such as financial networks and molecules. Then, we shall see how learning the causality benefits many of the desirable properties of modern machine learning, in terms of (i) OOD generalization capability; (ii) interpretability; and (iii) robustness to adversarial attacks.\n  Realizing the causality in machine learning, on the other hand, raises a dilemma for optimization in conventional machine learning, as it often contradicts the objective of ERM...", "date": "2025-06-13", "categories": []}
{"id": "2506.12304", "title": "Conditional Average Treatment Effect Estimation Under Hidden Confounders", "abstract": "One of the major challenges in estimating conditional potential outcomes and conditional average treatment effects (CATE) is the presence of hidden confounders. Since testing for hidden confounders cannot be accomplished only with observational data, conditional unconfoundedness is commonly assumed in the literature of CATE estimation. Nevertheless, under this assumption, CATE estimation can be significantly biased due to the effects of unobserved confounders. In this work, we consider the case where in addition to a potentially large observational dataset, a small dataset from a randomized controlled trial (RCT) is available. Notably, we make no assumptions on the existence of any covariate information for the RCT dataset, we only require the outcomes to be observed. We propose a CATE estimation method based on a pseudo-confounder generator and a CATE model that aligns the learned potential outcomes from the observational data with those observed from the RCT. Our method is applicable to many practical scenarios of interest, particularly those where privacy is a concern (e.g., medical applications). Extensive numerical experiments are provided demonstrating the effectiveness of our approach for both synthetic and real-world datasets.", "date": "2025-06-14", "categories": []}
{"id": "2506.12439", "title": "Interpretable Causal Representation Learning for Biological Data in the Pathway Space", "abstract": "Predicting the impact of genomic and drug perturbations in cellular function is crucial for understanding gene functions and drug effects, ultimately leading to improved therapies. To this end, Causal Representation Learning (CRL) constitutes one of the most promising approaches, as it aims to identify the latent factors that causally govern biological systems, thus facilitating the prediction of the effect of unseen perturbations. Yet, current CRL methods fail in reconciling their principled latent representations with known biological processes, leading to models that are not interpretable. To address this major issue, we present SENA-discrepancy-VAE, a model based on the recently proposed CRL method discrepancy-VAE, that produces representations where each latent factor can be interpreted as the (linear) combination of the activity of a (learned) set of biological processes. To this extent, we present an encoder, SENA-{\\delta}, that efficiently compute and map biological processes' activity levels to the latent causal factors. We show that SENA-discrepancy-VAE achieves predictive performances on unseen combinations of interventions that are comparable with its original, non-interpretable counterpart, while inferring causal latent factors that are biologically meaningful.", "date": "2025-06-14", "categories": []}
{"id": "2506.11756", "title": "Causal Effect Identification in Heterogeneous Environments from Higher-Order Moments", "abstract": "We investigate the estimation of the causal effect of a treatment variable on an outcome in the presence of a latent confounder. We first show that the causal effect is identifiable under certain conditions when data is available from multiple environments, provided that the target causal effect remains invariant across these environments. Secondly, we propose a moment-based algorithm for estimating the causal effect as long as only a single parameter of the data-generating mechanism varies across environments -- whether it be the exogenous noise distribution or the causal relationship between two variables. Conversely, we prove that identifiability is lost if both exogenous noise distributions of both the latent and treatment variables vary across environments. Finally, we propose a procedure to identify which parameter of the data-generating mechanism has varied across the environments and evaluate the performance of our proposed methods through experiments on synthetic data.", "date": "2025-06-13", "categories": []}
{"id": "2506.11912", "title": "Breaking Habits: On the Role of the Advantage Function in Learning Causal State Representations", "abstract": "Recent work has shown that reinforcement learning agents can develop policies that exploit spurious correlations between rewards and observations. This phenomenon, known as policy confounding, arises because the agent's policy influences both past and future observation variables, creating a feedback loop that can hinder the agent's ability to generalize beyond its usual trajectories. In this paper, we show that the advantage function, commonly used in policy gradient methods, not only reduces the variance of gradient estimates but also mitigates the effects of policy confounding. By adjusting action values relative to the state representation, the advantage function downweights state-action pairs that are more likely under the current policy, breaking spurious correlations and encouraging the agent to focus on causal factors. We provide both analytical and empirical evidence demonstrating that training with the advantage function leads to improved out-of-trajectory performance.", "date": "2025-06-13", "categories": []}
{"id": "2408.14620", "title": "General targeted machine learning for modern causal mediation analysis", "abstract": "Causal mediation analyses investigate the mechanisms through which causes exert their effects, and are therefore central to scientific progress. The literature on the non-parametric definition and identification of mediational effects in rigourous causal models has grown significantly in recent years, and there has been important progress to address challenges in the interpretation and identification of such effects. Despite great progress in the causal inference front, statistical methodology for non-parametric estimation has lagged behind, with few or no methods available for tackling non-parametric estimation in the presence of multiple, continuous, or high-dimensional mediators. In this paper we show that the identification formulas for six popular non-parametric approaches to mediation analysis proposed in recent years can be recovered from just two statistical estimands. We leverage this finding to propose an all-purpose one-step estimation algorithm that can be coupled with machine learning in any mediation study that uses any of these six definitions of mediation. The estimators have desirable properties, such as $\\sqrt{n}$-convergence and asymptotic normality. Estimating the first-order correction for the one-step estimator requires estimation of complex density ratios on the potentially high-dimensional mediators, a challenge that is solved using recent advancements in so-called Riesz learning. We illustrate the properties of our methods in a simulation study and illustrate its use on real data to estimate the extent to which pain management practices mediate the total effect of having a chronic pain disorder on opioid use disorder.", "date": "2024-08-26", "categories": []}
{"id": "2411.03021", "title": "Testing Generalizability in Causal Inference", "abstract": "Ensuring robust model performance in diverse real-world scenarios requires addressing generalizability across domains with covariate shifts. However, no formal procedure exists for statistically evaluating generalizability in machine learning algorithms. Existing predictive metrics like mean squared error (MSE) help to quantify the relative performance between models, but do not directly answer whether a model can or cannot generalize. To address this gap in the domain of causal inference, we propose a systematic framework for statistically evaluating the generalizability of high-dimensional causal inference models. Our approach uses the frugal parameterization to flexibly simulate from fully and semi-synthetic causal benchmarks, offering a comprehensive evaluation for both mean and distributional regression methods. Grounded in real-world data, our method ensures more realistic evaluations, which is often missing in current work relying on simplified datasets. Furthermore, using simulations and statistical testing, our framework is robust and avoids over-reliance on conventional metrics, providing statistical safeguards for decision making.", "date": "2024-11-05", "categories": []}
{"id": "2501.18990", "title": "Permutation-Based Rank Test in the Presence of Discretization and Application in Causal Discovery with Mixed Data", "abstract": "Recent advances have shown that statistical tests for the rank of cross-covariance matrices play an important role in causal discovery. These rank tests include partial correlation tests as special cases and provide further graphical information about latent variables. Existing rank tests typically assume that all the continuous variables can be perfectly measured, and yet, in practice many variables can only be measured after discretization. For example, in psychometric studies, the continuous level of certain personality dimensions of a person can only be measured after being discretized into order-preserving options such as disagree, neutral, and agree. Motivated by this, we propose Mixed data Permutation-based Rank Test (MPRT), which properly controls the statistical errors even when some or all variables are discretized. Theoretically, we establish the exchangeability and estimate the asymptotic null distribution by permutations; as a consequence, MPRT can effectively control the Type I error in the presence of discretization while previous methods cannot. Empirically, our method is validated by extensive experiments on synthetic data and real-world data to demonstrate its effectiveness as well as applicability in causal discovery.", "date": "2025-01-31", "categories": []}
{"id": "2506.10378", "title": "Discovering Hierarchical Latent Capabilities of Language Models via Causal Representation Learning", "abstract": "Faithful evaluation of language model capabilities is crucial for deriving actionable insights that can inform model development. However, rigorous causal evaluations in this domain face significant methodological challenges, including complex confounding effects and prohibitive computational costs associated with extensive retraining. To tackle these challenges, we propose a causal representation learning framework wherein observed benchmark performance is modeled as a linear transformation of a few latent capability factors. Crucially, these latent factors are identified as causally interrelated after appropriately controlling for the base model as a common confounder. Applying this approach to a comprehensive dataset encompassing over 1500 models evaluated across six benchmarks from the Open LLM Leaderboard, we identify a concise three-node linear causal structure that reliably explains the observed performance variations. Further interpretation of this causal structure provides substantial scientific insights beyond simple numerical rankings: specifically, we reveal a clear causal direction starting from general problem-solving capabilities, advancing through instruction-following proficiency, and culminating in mathematical reasoning ability. Our results underscore the essential role of carefully controlling base model variations during evaluation, a step critical to accurately uncovering the underlying causal relationships among latent model capabilities.", "date": "2025-06-12", "categories": []}
{"id": "2312.04540", "title": "Sim-to-Real Causal Transfer: A Metric Learning Approach to Causally-Aware Interaction Representations", "abstract": "Modeling spatial-temporal interactions among neighboring agents is at the heart of multi-agent problems such as motion forecasting and crowd navigation. Despite notable progress, it remains unclear to which extent modern representations can capture the causal relationships behind agent interactions. In this work, we take an in-depth look at the causal awareness of these representations, from computational formalism to real-world practice. First, we cast doubt on the notion of non-causal robustness studied in the recent CausalAgents benchmark. We show that recent representations are already partially resilient to perturbations of non-causal agents, and yet modeling indirect causal effects involving mediator agents remains challenging. To address this challenge, we introduce a metric learning approach that regularizes latent representations with causal annotations. Our controlled experiments show that this approach not only leads to higher degrees of causal awareness but also yields stronger out-of-distribution robustness. To further operationalize it in practice, we propose a sim-to-real causal transfer method via cross-domain multi-task learning. Experiments on pedestrian datasets show that our method can substantially boost generalization, even in the absence of real-world causal annotations. We hope our work provides a new perspective on the challenges and pathways towards causally-aware representations of multi-agent interactions. Our code is available at https://github.com/vita-epfl/CausalSim2Real.", "date": "2023-12-07", "categories": []}
{"id": "2408.05860", "title": "Root Cause Attribution of Delivery Risks via Causal Discovery with Reinforcement Learning", "abstract": "This paper presents a novel approach to root cause attribution of delivery risks within supply chains by integrating causal discovery with reinforcement learning. As supply chains become increasingly complex, traditional methods of root cause analysis struggle to capture the intricate interrelationships between various factors, often leading to spurious correlations and suboptimal decision-making. Our approach addresses these challenges by leveraging causal discovery to identify the true causal relationships between operational variables, and reinforcement learning to iteratively refine the causal graph. This method enables the accurate identification of key drivers of late deliveries, such as shipping mode and delivery status, and provides actionable insights for optimizing supply chain performance. We apply our approach to a real-world supply chain dataset, demonstrating its effectiveness in uncovering the underlying causes of delivery delays and offering strategies for mitigating these risks. The findings have significant implications for improving operational efficiency, customer satisfaction, and overall profitability within supply chains.", "date": "2024-08-11", "categories": []}
{"id": "2410.00535", "title": "The Causal Information Bottleneck and Optimal Causal Variable Abstractions", "abstract": "To effectively study complex causal systems, it is often useful to construct abstractions of parts of the system by discarding irrelevant details while preserving key features. The Information Bottleneck (IB) method is a widely used approach to construct variable abstractions by compressing random variables while retaining predictive power over a target variable. Traditional methods like IB are purely statistical and ignore underlying causal structures, making them ill-suited for causal tasks. We propose the Causal Information Bottleneck (CIB), a causal extension of the IB, which compresses a set of chosen variables while maintaining causal control over a target variable. This method produces abstractions of (sets of) variables which are causally interpretable, give us insight about the interactions between the abstracted variables and the target variable, and can be used when reasoning about interventions. We present experimental results demonstrating that the learned abstractions accurately capture causal relations as intended.", "date": "2024-10-01", "categories": []}
{"id": "2506.09286", "title": "Causal Graph Recovery in Neuroimaging through Answer Set Programming", "abstract": "Learning graphical causal structures from time series data presents significant challenges, especially when the measurement frequency does not match the causal timescale of the system. This often leads to a set of equally possible underlying causal graphs due to information loss from sub-sampling (i.e., not observing all possible states of the system throughout time). Our research addresses this challenge by incorporating the effects of sub-sampling in the derivation of causal graphs, resulting in more accurate and intuitive outcomes. We use a constraint optimization approach, specifically answer set programming (ASP), to find the optimal set of answers. ASP not only identifies the most probable underlying graph, but also provides an equivalence class of possible graphs for expert selection. In addition, using ASP allows us to leverage graph theory to further prune the set of possible solutions, yielding a smaller, more accurate answer set significantly faster than traditional approaches. We validate our approach on both simulated data and empirical structural brain connectivity, and demonstrate its superiority over established methods in these experiments. We further show how our method can be used as a meta-approach on top of established methods to obtain, on average, 12% improvement in F1 score. In addition, we achieved state of the art results in terms of precision and recall of reconstructing causal graph from sub-sampled time series data. Finally, our method shows robustness to varying degrees of sub-sampling on realistic simulations, whereas other methods perform worse for higher rates of sub-sampling.", "date": "2025-06-10", "categories": []}
{"id": "2506.09433", "title": "Mitigating Spurious Correlations in LLMs via Causality-Aware Post-Training", "abstract": "While large language models (LLMs) have demonstrated remarkable capabilities in language modeling, recent studies reveal that they often fail on out-of-distribution (OOD) samples due to spurious correlations acquired during pre-training. Here, we aim to mitigate such spurious correlations through causality-aware post-training (CAPT). By decomposing a biased prediction into two unbiased steps, known as \\textit{event estimation} and \\textit{event intervention}, we reduce LLMs' pre-training biases without incurring additional fine-tuning biases, thus enhancing the model's generalization ability. Experiments on the formal causal inference benchmark CLadder and the logical reasoning dataset PrOntoQA show that 3B-scale language models fine-tuned with CAPT can outperform both traditional SFT and larger LLMs on in-distribution (ID) and OOD tasks using only 100 ID fine-tuning samples, demonstrating the effectiveness and sample efficiency of CAPT.", "date": "2025-06-11", "categories": []}
{"id": "2502.05122", "title": "Distinguishing Cause from Effect with Causal Velocity Models", "abstract": "Bivariate structural causal models (SCM) are often used to infer causal direction by examining their goodness-of-fit under restricted model classes. In this paper, we describe a parametrization of bivariate SCMs in terms of a causal velocity by viewing the cause variable as time in a dynamical system. The velocity implicitly defines counterfactual curves via the solution of initial value problems where the observation specifies the initial condition. Using tools from measure transport, we obtain a unique correspondence between SCMs and the score function of the generated distribution via its causal velocity. Based on this, we derive an objective function that directly regresses the velocity against the score function, the latter of which can be estimated non-parametrically from observational data. We use this to develop a method for bivariate causal discovery that extends beyond known model classes such as additive or location scale noise, and that requires no assumptions on the noise distributions. When the score is estimated well, the objective is also useful for detecting model non-identifiability and misspecification. We present positive results in simulation and benchmark experiments where many existing methods fail, and perform ablation studies to examine the method's sensitivity to accurate score estimation.", "date": "2025-02-07", "categories": []}
{"id": "2503.04556", "title": "Compositional Causal Reasoning Evaluation in Language Models", "abstract": "Causal reasoning and compositional reasoning are two core aspirations in AI. Measuring the extent of these behaviors requires principled evaluation methods. We explore a unified perspective that considers both behaviors simultaneously, termed compositional causal reasoning (CCR): the ability to infer how causal measures compose and, equivalently, how causal quantities propagate through graphs. We instantiate a framework for the systematic evaluation of CCR for the average treatment effect and the probability of necessity and sufficiency. As proof of concept, we demonstrate CCR evaluation for language models in the LLama, Phi, and GPT families. On a math word problem, our framework revealed a range of taxonomically distinct error patterns. CCR errors increased with the complexity of causal paths for all models except o1.", "date": "2025-03-06", "categories": []}
{"id": "2506.08747", "title": "A Sample Efficient Conditional Independence Test in the Presence of Discretization", "abstract": "In many real-world scenarios, interested variables are often represented as discretized values due to measurement limitations. Applying Conditional Independence (CI) tests directly to such discretized data, however, can lead to incorrect conclusions. To address this, recent advancements have sought to infer the correct CI relationship between the latent variables through binarizing observed data. However, this process inevitably results in a loss of information, which degrades the test's performance. Motivated by this, this paper introduces a sample-efficient CI test that does not rely on the binarization process. We find that the independence relationships of latent continuous variables can be established by addressing an over-identifying restriction problem with Generalized Method of Moments (GMM). Based on this insight, we derive an appropriate test statistic and establish its asymptotic distribution correctly reflecting CI by leveraging nodewise regression. Theoretical findings and Empirical results across various datasets demonstrate that the superiority and effectiveness of our proposed test. Our code implementation is provided in https://github.com/boyangaaaaa/DCT", "date": "2025-06-10", "categories": []}
{"id": "2506.08771", "title": "Paths to Causality: Finding Informative Subgraphs Within Knowledge Graphs for Knowledge-Based Causal Discovery", "abstract": "Inferring causal relationships between variable pairs is crucial for understanding multivariate interactions in complex systems. Knowledge-based causal discovery -- which involves inferring causal relationships by reasoning over the metadata of variables (e.g., names or textual context) -- offers a compelling alternative to traditional methods that rely on observational data. However, existing methods using Large Language Models (LLMs) often produce unstable and inconsistent results, compromising their reliability for causal inference. To address this, we introduce a novel approach that integrates Knowledge Graphs (KGs) with LLMs to enhance knowledge-based causal discovery. Our approach identifies informative metapath-based subgraphs within KGs and further refines the selection of these subgraphs using Learning-to-Rank-based models. The top-ranked subgraphs are then incorporated into zero-shot prompts, improving the effectiveness of LLMs in inferring the causal relationship. Extensive experiments on biomedical and open-domain datasets demonstrate that our method outperforms most baselines by up to 44.4 points in F1 scores, evaluated across diverse LLMs and KGs. Our code and datasets are available on GitHub: https://github.com/susantiyuni/path-to-causality", "date": "2025-06-10", "categories": []}
{"id": "2309.15793", "title": "Targeting relative risk heterogeneity with causal forests", "abstract": "The identification of heterogeneous treatment effects (HTE) across subgroups is of significant interest in clinical trial analysis. Several state-of-the-art HTE estimation methods, including causal forests, apply recursive partitioning for non-parametric identification of relevant covariates and interactions. However, the partitioning criterion is typically based on differences in absolute risk. This can dilute statistical power by masking variation in the relative risk, which is often a more appropriate quantity of clinical interest. In this work, we propose and implement a methodology for modifying causal forests to target relative risk, using a novel node-splitting procedure based on exhaustive generalized linear model comparison. We present results from simulated data that suggest relative risk causal forests can capture otherwise undetected sources of heterogeneity. We implement our method on real-world trial data to explore HTEs for liraglutide in patients with type 2 diabetes.", "date": "2023-09-26", "categories": []}
{"id": "2402.14264", "title": "Structure-agnostic Optimality of Doubly Robust Learning for Treatment Effect Estimation", "abstract": "Average treatment effect estimation is the most central problem in causal inference with application to numerous disciplines. While many estimation strategies have been proposed in the literature, the statistical optimality of these methods has still remained an open area of investigation, especially in regimes where these methods do not achieve parametric rates. In this paper, we adopt the recently introduced structure-agnostic framework of statistical lower bounds, which poses no structural properties on the nuisance functions other than access to black-box estimators that achieve some statistical estimation rate. This framework is particularly appealing when one is only willing to consider estimation strategies that use non-parametric regression and classification oracles as black-box sub-processes. Within this framework, we prove the statistical optimality of the celebrated and widely used doubly robust estimators for both the Average Treatment Effect (ATE) and the Average Treatment Effect on the Treated (ATT), as well as weighted variants of the former, which arise in policy evaluation.", "date": "2024-02-22", "categories": []}
{"id": "2412.07184", "title": "Automatic Doubly Robust Forests", "abstract": "This paper proposes the automatic Doubly Robust Random Forest (DRRF) algorithm for estimating the conditional expectation of a moment functional in the presence of high-dimensional nuisance functions. DRRF extends the automatic debiasing framework based on the Riesz representer to the conditional setting and enables nonparametric, forest-based estimation (Athey et al., 2019; Oprescu et al., 2019). In contrast to existing methods, DRRF does not require prior knowledge of the form of the debiasing term or impose restrictive parametric or semi-parametric assumptions on the target quantity. Additionally, it is computationally efficient in making predictions at multiple query points. We establish consistency and asymptotic normality results for the DRRF estimator under general assumptions, allowing for the construction of valid confidence intervals. Through extensive simulations in heterogeneous treatment effect (HTE) estimation, we demonstrate the superior performance of DRRF over benchmark approaches in terms of estimation accuracy, robustness, and computational efficiency.", "date": "2024-12-10", "categories": []}
{"id": "2412.17717", "title": "Fast Causal Discovery by Approximate Kernel-based Generalized Score Functions with Linear Computational Complexity", "abstract": "Score-based causal discovery methods can effectively identify causal relationships by evaluating candidate graphs and selecting the one with the highest score. One popular class of scores is kernel-based generalized score functions, which can adapt to a wide range of scenarios and work well in practice because they circumvent assumptions about causal mechanisms and data distributions. Despite these advantages, kernel-based generalized score functions pose serious computational challenges in time and space, with a time complexity of $\\mathcal{O}(n^3)$ and a memory complexity of $\\mathcal{O}(n^2)$, where $n$ is the sample size. In this paper, we propose an approximate kernel-based generalized score function with $\\mathcal{O}(n)$ time and space complexities by using low-rank technique and designing a set of rules to handle the complex composite matrix operations required to calculate the score, as well as developing sampling algorithms for different data types to benefit the handling of diverse data types efficiently. Our extensive causal discovery experiments on both synthetic and real-world data demonstrate that compared to the state-of-the-art method, our method can not only significantly reduce computational costs, but also achieve comparable accuracy, especially for large datasets.", "date": "2024-12-23", "categories": []}
{"id": "2502.01267", "title": "Counterfactual Situation Testing: From Single to Multidimensional Discrimination", "abstract": "We present counterfactual situation testing (CST), a causal data mining framework for detecting individual discrimination in a dataset of classifier decisions. CST answers the question ``what would have been the model outcome had the individual, or complainant, been of a different protected status?'' It extends the legally-grounded situation testing (ST) of Thanh et al. (2011) by operationalizing the notion of \"fairness given the difference\" via counterfactual reasoning. ST finds for each complainant similar protected and non-protected instances in the dataset; constructs, respectively, a control and test group; and compares the groups such that a difference in model outcomes implies a potential case of individual discrimination. CST, instead, avoids this idealized comparison by establishing the test group on the complainant's generated counterfactual, which reflects how the protected attribute when changed influences other seemingly neutral attributes of the complainant. Under CST we test for discrimination for each complainant by comparing similar individuals within the control and test group but dissimilar individuals across these groups. We consider single (e.g.,~gender) and multidimensional (e.g.,~gender and race) discrimination testing. For multidimensional discrimination we study multiple and intersectional discrimination and, as feared by legal scholars, find evidence that the former fails to account for the latter kind. Using a k-nearest neighbor implementation, we showcase CST on synthetic and real data. Experimental results show that CST uncovers a higher number of cases than ST, even when the model is counterfactually fair. CST, in fact, extends counterfactual fairness (CF) of Kusner et al. (2017) by equipping CF with confidence intervals, which we report for all experiments.", "date": "2025-02-03", "categories": []}
{"id": "2505.09706", "title": "Forests for Differences: Robust Causal Inference Beyond Parametric DiD", "abstract": "This paper introduces the Difference-in-Differences Bayesian Causal Forest (DiD-BCF), a novel non-parametric model addressing key challenges in DiD estimation, such as staggered adoption and heterogeneous treatment effects. DiD-BCF provides a unified framework for estimating Average (ATE), Group-Average (GATE), and Conditional Average Treatment Effects (CATE). A core innovation, its Parallel Trends Assumption (PTA)-based reparameterization, enhances estimation accuracy and stability in complex panel data settings. Extensive simulations demonstrate DiD-BCF's superior performance over established benchmarks, particularly under non-linearity, selection biases, and effect heterogeneity. Applied to U.S. minimum wage policy, the model uncovers significant conditional treatment effect heterogeneity related to county population, insights obscured by traditional methods. DiD-BCF offers a robust and versatile tool for more nuanced causal inference in modern DiD applications.", "date": "2025-05-14", "categories": []}
{"id": "2506.05675", "title": "Zero-Shot Event Causality Identification via Multi-source Evidence Fuzzy Aggregation with Large Language Models", "abstract": "Event Causality Identification (ECI) aims to detect causal relationships between events in textual contexts. Existing ECI models predominantly rely on supervised methodologies, suffering from dependence on large-scale annotated data. Although Large Language Models (LLMs) enable zero-shot ECI, they are prone to causal hallucination-erroneously establishing spurious causal links. To address these challenges, we propose MEFA, a novel zero-shot framework based on Multi-source Evidence Fuzzy Aggregation. First, we decompose causality reasoning into three main tasks (temporality determination, necessity analysis, and sufficiency verification) complemented by three auxiliary tasks. Second, leveraging meticulously designed prompts, we guide LLMs to generate uncertain responses and deterministic outputs. Finally, we quantify LLM's responses of sub-tasks and employ fuzzy aggregation to integrate these evidence for causality scoring and causality determination. Extensive experiments on three benchmarks demonstrate that MEFA outperforms second-best unsupervised baselines by 6.2% in F1-score and 9.3% in precision, while significantly reducing hallucination-induced errors. In-depth analysis verify the effectiveness of task decomposition and the superiority of fuzzy aggregation.", "date": "2025-06-06", "categories": []}
{"id": "2506.07049", "title": "FairPFN: A Tabular Foundation Model for Causal Fairness", "abstract": "Machine learning (ML) systems are utilized in critical sectors, such as healthcare, law enforcement, and finance. However, these systems are often trained on historical data that contains demographic biases, leading to ML decisions that perpetuate or exacerbate existing social inequalities. Causal fairness provides a transparent, human-in-the-loop framework to mitigate algorithmic discrimination, aligning closely with legal doctrines of direct and indirect discrimination. However, current causal fairness frameworks hold a key limitation in that they assume prior knowledge of the correct causal model, restricting their applicability in complex fairness scenarios where causal models are unknown or difficult to identify. To bridge this gap, we propose FairPFN, a tabular foundation model pre-trained on synthetic causal fairness data to identify and mitigate the causal effects of protected attributes in its predictions. FairPFN's key contribution is that it requires no knowledge of the causal model and still demonstrates strong performance in identifying and removing protected causal effects across a diverse set of hand-crafted and real-world scenarios relative to robust baseline methods. FairPFN paves the way for promising future research, making causal fairness more accessible to a wider variety of complex fairness problems.", "date": "2025-06-08", "categories": []}
{"id": "2506.07140", "title": "Quantile-Optimal Policy Learning under Unmeasured Confounding", "abstract": "We study quantile-optimal policy learning where the goal is to find a policy whose reward distribution has the largest $\\alpha$-quantile for some $\\alpha \\in (0, 1)$. We focus on the offline setting whose generating process involves unobserved confounders. Such a problem suffers from three main challenges: (i) nonlinearity of the quantile objective as a functional of the reward distribution, (ii) unobserved confounding issue, and (iii) insufficient coverage of the offline dataset. To address these challenges, we propose a suite of causal-assisted policy learning methods that provably enjoy strong theoretical guarantees under mild conditions. In particular, to address (i) and (ii), using causal inference tools such as instrumental variables and negative controls, we propose to estimate the quantile objectives by solving nonlinear functional integral equations. Then we adopt a minimax estimation approach with nonparametric models to solve these integral equations, and propose to construct conservative policy estimates that address (iii). The final policy is the one that maximizes these pessimistic estimates. In addition, we propose a novel regularized policy learning method that is more amenable to computation. Finally, we prove that the policies learned by these methods are $\\tilde{\\mathscr{O}}(n^{-1/2})$ quantile-optimal under a mild coverage assumption on the offline dataset. Here, $\\tilde{\\mathscr{O}}(\\cdot)$ omits poly-logarithmic factors. To the best of our knowledge, we propose the first sample-efficient policy learning algorithms for estimating the quantile-optimal policy when there exist unmeasured confounding.", "date": "2025-06-08", "categories": []}
{"id": "2506.07517", "title": "Addressing Correlated Latent Exogenous Variables in Debiased Recommender Systems", "abstract": "Recommendation systems (RS) aim to provide personalized content, but they face a challenge in unbiased learning due to selection bias, where users only interact with items they prefer. This bias leads to a distorted representation of user preferences, which hinders the accuracy and fairness of recommendations. To address the issue, various methods such as error imputation based, inverse propensity scoring, and doubly robust techniques have been developed. Despite the progress, from the structural causal model perspective, previous debiasing methods in RS assume the independence of the exogenous variables. In this paper, we release this assumption and propose a learning algorithm based on likelihood maximization to learn a prediction model. We first discuss the correlation and difference between unmeasured confounding and our scenario, then we propose a unified method that effectively handles latent exogenous variables. Specifically, our method models the data generation process with latent exogenous variables under mild normality assumptions. We then develop a Monte Carlo algorithm to numerically estimate the likelihood function. Extensive experiments on synthetic datasets and three real-world datasets demonstrate the effectiveness of our proposed method. The code is at https://github.com/WallaceSUI/kdd25-background-variable.", "date": "2025-06-09", "categories": []}
{"id": "2506.07760", "title": "Quickest Causal Change Point Detection by Adaptive Intervention", "abstract": "We propose an algorithm for change point monitoring in linear causal models that accounts for interventions. Through a special centralization technique, we can concentrate the changes arising from causal propagation across nodes into a single dimension. Additionally, by selecting appropriate intervention nodes based on Kullback-Leibler divergence, we can amplify the change magnitude. We also present an algorithm for selecting the intervention values, which aids in the identification of the most effective intervention nodes. Two monitoring methods are proposed, each with an adaptive intervention policy to make a balance between exploration and exploitation. We theoretically demonstrate the first-order optimality of the proposed methods and validate their properties using simulation datasets and two real-world case studies.", "date": "2025-06-09", "categories": []}
{"id": "2404.04399", "title": "Longitudinal Targeted Minimum Loss-based Estimation with Temporal-Difference Heterogeneous Transformer", "abstract": "We propose Deep Longitudinal Targeted Minimum Loss-based Estimation (Deep LTMLE), a novel approach to estimate the counterfactual mean of outcome under dynamic treatment policies in longitudinal problem settings. Our approach utilizes a transformer architecture with heterogeneous type embedding trained using temporal-difference learning. After obtaining an initial estimate using the transformer, following the targeted minimum loss-based likelihood estimation (TMLE) framework, we statistically corrected for the bias commonly associated with machine learning algorithms. Furthermore, our method also facilitates statistical inference by enabling the provision of 95% confidence intervals grounded in asymptotic statistical theory. Simulation results demonstrate our method's superior performance over existing approaches, particularly in complex, long time-horizon scenarios. It remains effective in small-sample, short-duration contexts, matching the performance of asymptotically efficient estimators. To demonstrate our method in practice, we applied our method to estimate counterfactual mean outcomes for standard versus intensive blood pressure management strategies in a real-world cardiovascular epidemiology cohort study.", "date": "2024-04-05", "categories": []}
{"id": "2506.05202", "title": "Causal Effect Identification in lvLiNGAM from Higher-Order Cumulants", "abstract": "This paper investigates causal effect identification in latent variable Linear Non-Gaussian Acyclic Models (lvLiNGAM) using higher-order cumulants, addressing two prominent setups that are challenging in the presence of latent confounding: (1) a single proxy variable that may causally influence the treatment and (2) underspecified instrumental variable cases where fewer instruments exist than treatments. We prove that causal effects are identifiable with a single proxy or instrument and provide corresponding estimation methods. Experimental results demonstrate the accuracy and robustness of our approaches compared to existing methods, advancing the theoretical and practical understanding of causal inference in linear systems with latent confounders.", "date": "2025-06-05", "categories": []}
{"id": "2506.05445", "title": "Causal Policy Learning in Reinforcement Learning: Backdoor-Adjusted Soft Actor-Critic", "abstract": "Hidden confounders that influence both states and actions can bias policy learning in reinforcement learning (RL), leading to suboptimal or non-generalizable behavior. Most RL algorithms ignore this issue, learning policies from observational trajectories based solely on statistical associations rather than causal effects. We propose DoSAC (Do-Calculus Soft Actor-Critic with Backdoor Adjustment), a principled extension of the SAC algorithm that corrects for hidden confounding via causal intervention estimation. DoSAC estimates the interventional policy $\\pi(a | \\mathrm{do}(s))$ using the backdoor criterion, without requiring access to true confounders or causal labels. To achieve this, we introduce a learnable Backdoor Reconstructor that infers pseudo-past variables (previous state and action) from the current state to enable backdoor adjustment from observational data. This module is integrated into a soft actor-critic framework to compute both the interventional policy and its entropy. Empirical results on continuous control benchmarks show that DoSAC outperforms baselines under confounded settings, with improved robustness, generalization, and policy reliability.", "date": "2025-06-05", "categories": []}
{"id": "2410.14812", "title": "Isolated Causal Effects of Natural Language", "abstract": "As language technologies become widespread, it is important to understand how changes in language affect reader perceptions and behaviors. These relationships may be formalized as the isolated causal effect of some focal language-encoded intervention (e.g., factual inaccuracies) on an external outcome (e.g., readers' beliefs). In this paper, we introduce a formal estimation framework for isolated causal effects of language. We show that a core challenge of estimating isolated effects is the need to approximate all non-focal language outside of the intervention. Drawing on the principle of omitted variable bias, we provide measures for evaluating the quality of both non-focal language approximations and isolated effect estimates themselves. We find that poor approximation of non-focal language can lead to bias in the corresponding isolated effect estimates due to omission of relevant variables, and we show how to assess the sensitivity of effect estimates to such bias along the two key axes of fidelity and overlap. In experiments on semi-synthetic and real-world data, we validate the ability of our framework to correctly recover isolated effects and demonstrate the utility of our proposed measures.", "date": "2024-10-18", "categories": []}
{"id": "2506.04945", "title": "Learning Joint Interventional Effects from Single-Variable Interventions in Additive Models", "abstract": "Estimating causal effects of joint interventions on multiple variables is crucial in many domains, but obtaining data from such simultaneous interventions can be challenging. Our study explores how to learn joint interventional effects using only observational data and single-variable interventions. We present an identifiability result for this problem, showing that for a class of nonlinear additive outcome mechanisms, joint effects can be inferred without access to joint interventional data. We propose a practical estimator that decomposes the causal effect into confounded and unconfounded contributions for each intervention variable. Experiments on synthetic data demonstrate that our method achieves performance comparable to models trained directly on joint interventional data, outperforming a purely observational estimator.", "date": "2025-06-05", "categories": []}
{"id": "2506.05120", "title": "Nonlinear Causal Discovery for Grouped Data", "abstract": "Inferring cause-effect relationships from observational data has gained significant attention in recent years, but most methods are limited to scalar random variables. In many important domains, including neuroscience, psychology, social science, and industrial manufacturing, the causal units of interest are groups of variables rather than individual scalar measurements. Motivated by these applications, we extend nonlinear additive noise models to handle random vectors, establishing a two-step approach for causal graph learning: First, infer the causal order among random vectors. Second, perform model selection to identify the best graph consistent with this order. We introduce effective and novel solutions for both steps in the vector case, demonstrating strong performance in simulations. Finally, we apply our method to real-world assembly line data with partial knowledge of causal ordering among variable groups.", "date": "2025-06-05", "categories": []}
{"id": "2410.00075", "title": "Optimizing Treatment Allocation in the Presence of Interference", "abstract": "In Influence Maximization (IM), the objective is to -- given a budget -- select the optimal set of entities in a network to target with a treatment so as to maximize the total effect. For instance, in marketing, the objective is to target the set of customers that maximizes the total response rate, resulting from both direct treatment effects on targeted customers and indirect, spillover, effects that follow from targeting these customers. Recently, new methods to estimate treatment effects in the presence of network interference have been proposed. However, the issue of how to leverage these models to make better treatment allocation decisions has been largely overlooked. Traditionally, in Uplift Modeling (UM), entities are ranked according to estimated treatment effect, and the top entities are allocated treatment. Since, in a network context, entities influence each other, the UM ranking approach will be suboptimal. The problem of finding the optimal treatment allocation in a network setting is \\textcolor{red}{NP-hard,} and generally has to be solved heuristically. To fill the gap between IM and UM, we propose OTAPI: Optimizing Treatment Allocation in the Presence of Interference to find solutions to the IM problem using treatment effect estimates. OTAPI consists of two steps. First, a causal estimator is trained to predict treatment effects in a network setting. Second, this estimator is leveraged to identify an optimal treatment allocation by integrating it into classic IM algorithms. We demonstrate that this novel method outperforms classic IM and UM approaches on both synthetic and semi-synthetic datasets.", "date": "2024-09-30", "categories": []}
{"id": "2506.00335", "title": "Recover Experimental Data with Selection Bias using Counterfactual Logic", "abstract": "Selection bias, arising from the systematic inclusion or exclusion of certain samples, poses a significant challenge to the validity of causal inference. While Bareinboim et al. introduced methods for recovering unbiased observational and interventional distributions from biased data using partial external information, the complexity of the backdoor adjustment and the method's strong reliance on observational data limit its applicability in many practical settings. In this paper, we formally discover the recoverability of $P(Y^*_{x^*})$ under selection bias with experimental data. By explicitly constructing counterfactual worlds via Structural Causal Models (SCMs), we analyze how selection mechanisms in the observational world propagate to the counterfactual domain. We derive a complete set of graphical and theoretical criteria to determine that the experimental distribution remain unaffected by selection bias. Furthermore, we propose principled methods for leveraging partially unbiased observational data to recover $P(Y^*_{x^*})$ from biased experimental datasets. Simulation studies replicating realistic research scenarios demonstrate the practical utility of our approach, offering concrete guidance for mitigating selection bias in applied causal inference.", "date": "2025-05-31", "categories": []}
{"id": "2506.03915", "title": "Causal Explanations Over Time: Articulated Reasoning for Interactive Environments", "abstract": "Structural Causal Explanations (SCEs) can be used to automatically generate explanations in natural language to questions about given data that are grounded in a (possibly learned) causal model. Unfortunately they work for small data only. In turn they are not attractive to offer reasons for events, e.g., tracking causal changes over multiple time steps, or a behavioral component that involves feedback loops through actions of an agent. To this end, we generalize SCEs to a (recursive) formulation of explanation trees to capture the temporal interactions between reasons. We show the benefits of this more general SCE algorithm on synthetic time-series data and a 2D grid game, and further compare it to the base SCE and other existing methods for causal explanations.", "date": "2025-06-04", "categories": []}
{"id": "2501.05844", "title": "\"Cause\" is Mechanistic Narrative within Scientific Domains: An Ordinary Language Philosophical Critique of \"Causal Machine Learning\"", "abstract": "Causal Learning has emerged as a major theme of research in statistics and machine learning in recent years, promising specific computational techniques to apply to datasets that reveal the true nature of cause and effect in a number of important domains. In this paper we consider the epistemology of recognizing true cause and effect phenomena. We apply the Ordinary Language method of engaging on the customary use of the word 'cause' to investigate valid semantics of reasoning about cause and effect. We recognize that the grammars of cause and effect are fundamentally distinct in form across scientific domains, yet they maintain a consistent and central function. This function can best be described as the mechanism underlying fundamental forces of influence as considered prominent in the respective scientific domain. We demarcate 1) physics and engineering as domains wherein mathematical models are sufficient to comprehensively describe causality, 2) biology as introducing challenges of emergence while providing opportunities for showing consistent mechanisms across scale, and 3) the social sciences as introducing grander difficulties for establishing models of low prediction error but providing, through Hermeneutics, the potential for findings that are still instrumentally useful to individuals. We posit that definitive causal claims regarding a given phenomenon (writ large) can only come through an agglomeration of consistent evidence across multiple domains. This presents important methodological questions as far as harmonizing between language games and emergence across scales. Given the role of epistemic hubris in the contemporary crisis of credibility in the sciences, exercising greater caution as far as communicating precision as to the real degree of certainty certain evidence provides for rich collections of open problems in optimizing integration of different findings.", "date": "2025-01-10", "categories": []}
{"id": "2503.05024", "title": "Kernel-based estimators for functional causal effects", "abstract": "We propose causal effect estimators based on empirical Fr\\'{e}chet means and operator-valued kernels, tailored to functional data spaces. These methods address the challenges of high-dimensionality, sequential ordering, and model complexity while preserving robustness to treatment misspecification. Using structural assumptions, we obtain compact representations of potential outcomes, enabling scalable estimation of causal effects over time and across covariates. We provide both theoretical, regarding the consistency of functional causal effects, as well as empirical comparison of a range of proposed causal effect estimators.\n  Applications to binary treatment settings with functional outcomes illustrate the framework's utility in biomedical monitoring, where outcomes exhibit complex temporal dynamics. Our estimators accommodate scenarios with registered covariates and outcomes, aligning them to the Fr\\'{e}chet means, as well as cases requiring higher-order representations to capture intricate covariate-outcome interactions. These advancements extend causal inference to dynamic and non-linear domains, offering new tools for understanding complex treatment effects in functional data settings.", "date": "2025-03-06", "categories": []}
{"id": "2506.02946", "title": "Abstract Counterfactuals for Language Model Agents", "abstract": "Counterfactual inference is a powerful tool for analysing and evaluating autonomous agents, but its application to language model (LM) agents remains challenging. Existing work on counterfactuals in LMs has primarily focused on token-level counterfactuals, which are often inadequate for LM agents due to their open-ended action spaces. Unlike traditional agents with fixed, clearly defined action spaces, the actions of LM agents are often implicit in the strings they output, making their action spaces difficult to define and interpret. Furthermore, the meanings of individual tokens can shift depending on the context, adding complexity to token-level reasoning and sometimes leading to biased or meaningless counterfactuals. We introduce \\emph{Abstract Counterfactuals}, a framework that emphasises high-level characteristics of actions and interactions within an environment, enabling counterfactual reasoning tailored to user-relevant features. Our experiments demonstrate that the approach produces consistent and meaningful counterfactuals while minimising the undesired side effects of token-level methods. We conduct experiments on text-based games and counterfactual text generation, while considering both token-level and latent-space interventions.", "date": "2025-06-03", "categories": []}
{"id": "2506.03068", "title": "Causal Explainability of Machine Learning in Heart Failure Prediction from Electronic Health Records", "abstract": "The importance of clinical variables in the prognosis of the disease is explained using statistical correlation or machine learning (ML). However, the predictive importance of these variables may not represent their causal relationships with diseases. This paper uses clinical variables from a heart failure (HF) patient cohort to investigate the causal explainability of important variables obtained in statistical and ML contexts. Due to inherent regression modeling, popular causal discovery methods strictly assume that the cause and effect variables are numerical and continuous. This paper proposes a new computational framework to enable causal structure discovery (CSD) and score the causal strength of mixed-type (categorical, numerical, binary) clinical variables for binary disease outcomes. In HF classification, we investigate the association between the importance rank order of three feature types: correlated features, features important for ML predictions, and causal features. Our results demonstrate that CSD modeling for nonlinear causal relationships is more meaningful than its linear counterparts. Feature importance obtained from nonlinear classifiers (e.g., gradient-boosting trees) strongly correlates with the causal strength of variables without differentiating cause and effect variables. Correlated variables can be causal for HF, but they are rarely identified as effect variables. These results can be used to add the causal explanation of variables important for ML-based prediction modeling.", "date": "2025-06-03", "categories": []}
{"id": "2506.03149", "title": "Causal Estimation of Tokenisation Bias", "abstract": "Modern language models are typically trained over subword sequences, but ultimately define probabilities over character-strings. Ideally, the choice of the tokeniser -- which maps character-strings to subwords -- should not affect the probability assigned to the underlying character-string; in practice, it does. We define this mismatch as tokenisation bias. In this work, we quantify one particular type of tokenisation bias: the effect of including or not a subword (e.g., $\\langle hello \\rangle$) in a tokeniser's vocabulary on the probability a trained model assigns to the corresponding characters (i.e., \\textit{``hello''}). Estimating this effect is challenging because each model is trained with only one tokeniser. We address this by framing tokenisation bias as a causal effect and estimating it using the regression discontinuity design. Specifically, we exploit the fact that tokenisation algorithms rank subwords and add the first $K$ to a tokeniser's vocabulary, where $K$ is an arbitrary cutoff point. As such, we can estimate a causal effect by comparing similar subwords around this cutoff. Experimentally, we find that tokenisation consistently affects models' outputs across scales, vocabularies, and tokenisers. Notably, a subword's presence in a small model's vocabulary may increase its characters' probability by up to 17 times, highlighting tokenisation as a key design choice in language modelling.", "date": "2025-06-03", "categories": []}
{"id": "2412.13667", "title": "Exploring Multi-Modal Data with Tool-Augmented LLM Agents for Precise Causal Discovery", "abstract": "Causal discovery is an imperative foundation for decision-making across domains, such as smart health, AI for drug discovery and AIOps. Traditional statistical causal discovery methods, while well-established, predominantly rely on observational data and often overlook the semantic cues inherent in cause-and-effect relationships. The advent of Large Language Models (LLMs) has ushered in an affordable way of leveraging the semantic cues for knowledge-driven causal discovery, but the development of LLMs for causal discovery lags behind other areas, particularly in the exploration of multi-modal data. To bridge the gap, we introduce MATMCD, a multi-agent system powered by tool-augmented LLMs. MATMCD has two key agents: a Data Augmentation agent that retrieves and processes modality-augmented data, and a Causal Constraint agent that integrates multi-modal data for knowledge-driven reasoning. The proposed design of the inner-workings ensures successful cooperation of the agents. Our empirical study across seven datasets suggests the significant potential of multi-modality enhanced causal discovery.", "date": "2024-12-18", "categories": []}
{"id": "2502.00407", "title": "Causal Abstraction Learning based on the Semantic Embedding Principle", "abstract": "Structural causal models (SCMs) allow us to investigate complex systems at multiple levels of resolution. The causal abstraction (CA) framework formalizes the mapping between high- and low-level SCMs. We address CA learning in a challenging and realistic setting, where SCMs are inaccessible, interventional data is unavailable, and sample data is misaligned. A key principle of our framework is semantic embedding, formalized as the high-level distribution lying on a subspace of the low-level one. This principle naturally links linear CA to the geometry of the Stiefel manifold. We present a category-theoretic approach to SCMs that enables the learning of a CA by finding a morphism between the low- and high-level probability measures, adhering to the semantic embedding principle. Consequently, we formulate a general CA learning problem. As an application, we solve the latter problem for linear CA; considering Gaussian measures and the Kullback-Leibler divergence as an objective. Given the nonconvexity of the learning task, we develop three algorithms building upon existing paradigms for Riemannian optimization. We demonstrate that the proposed methods succeed on both synthetic and real-world brain data with different degrees of prior information about the structure of CA.", "date": "2025-02-01", "categories": []}
{"id": "2502.06231", "title": "Falsification of Unconfoundedness by Testing Independence of Causal Mechanisms", "abstract": "A major challenge in estimating treatment effects in observational studies is the reliance on untestable conditions such as the assumption of no unmeasured confounding. In this work, we propose an algorithm that can falsify the assumption of no unmeasured confounding in a setting with observational data from multiple heterogeneous sources, which we refer to as environments. Our proposed falsification strategy leverages a key observation that unmeasured confounding can cause observed causal mechanisms to appear dependent. Building on this observation, we develop a novel two-stage procedure that detects these dependencies with high statistical power while controlling false positives. The algorithm does not require access to randomized data and, in contrast to other falsification approaches, functions even under transportability violations when the environment has a direct effect on the outcome of interest. To showcase the practical relevance of our approach, we show that our method is able to efficiently detect confounding on both simulated and semi-synthetic data.", "date": "2025-02-10", "categories": []}
{"id": "2506.00152", "title": "Aligning Language Models with Observational Data: Opportunities and Risks from a Causal Perspective", "abstract": "Large language models are being widely used across industries to generate content that contributes directly to key performance metrics, such as conversion rates. Pretrained models, however, often fall short when it comes to aligning with human preferences or optimizing for business objectives. As a result, fine-tuning with good-quality labeled data is essential to guide models to generate content that achieves better results. Controlled experiments, like A/B tests, can provide such data, but they are often expensive and come with significant engineering and logistical challenges. Meanwhile, companies have access to a vast amount of historical (observational) data that remains underutilized. In this work, we study the challenges and opportunities of fine-tuning LLMs using observational data. We show that while observational outcomes can provide valuable supervision, directly fine-tuning models on such data can lead them to learn spurious correlations. We present empirical evidence of this issue using various real-world datasets and propose DeconfoundLM, a method that explicitly removes the effect of known confounders from reward signals. Using simulation experiments, we demonstrate that DeconfoundLM improves the recovery of causal relationships and mitigates failure modes found in fine-tuning methods that ignore or naively incorporate confounding variables. Our findings highlight that while observational data presents risks, with the right causal corrections, it can be a powerful source of signal for LLM alignment. Please refer to the project page for code and related resources.", "date": "2025-05-30", "categories": []}
{"id": "2506.00402", "title": "Causal Structure Discovery for Error Diagnostics of Children's ASR", "abstract": "Children's automatic speech recognition (ASR) often underperforms compared to that of adults due to a confluence of interdependent factors: physiological (e.g., smaller vocal tracts), cognitive (e.g., underdeveloped pronunciation), and extrinsic (e.g., vocabulary limitations, background noise). Existing analysis methods examine the impact of these factors in isolation, neglecting interdependencies-such as age affecting ASR accuracy both directly and indirectly via pronunciation skills. In this paper, we introduce a causal structure discovery to unravel these interdependent relationships among physiology, cognition, extrinsic factors, and ASR errors. Then, we employ causal quantification to measure each factor's impact on children's ASR. We extend the analysis to fine-tuned models to identify which factors are mitigated by fine-tuning and which remain largely unaffected. Experiments on Whisper and Wav2Vec2.0 demonstrate the generalizability of our findings across different ASR systems.", "date": "2025-05-31", "categories": []}
{"id": "2506.00844", "title": "LLM Cannot Discover Causality, and Should Be Restricted to Non-Decisional Support in Causal Discovery", "abstract": "This paper critically re-evaluates LLMs' role in causal discovery and argues against their direct involvement in determining causal relationships. We demonstrate that LLMs' autoregressive, correlation-driven modeling inherently lacks the theoretical grounding for causal reasoning and introduces unreliability when used as priors in causal discovery algorithms. Through empirical studies, we expose the limitations of existing LLM-based methods and reveal that deliberate prompt engineering (e.g., injecting ground-truth knowledge) could overstate their performance, helping to explain the consistently favorable results reported in much of the current literature. Based on these findings, we strictly confined LLMs' role to a non-decisional auxiliary capacity: LLMs should not participate in determining the existence or directionality of causal relationships, but can assist the search process for causal graphs (e.g., LLM-based heuristic search). Experiments across various settings confirm that, by strictly isolating LLMs from causal decision-making, LLM-guided heuristic search can accelerate the convergence and outperform both traditional and LLM-based methods in causal structure learning. We conclude with a call for the community to shift focus from naively applying LLMs to developing specialized models and training method that respect the core principles of causal discovery.", "date": "2025-06-01", "categories": []}
{"id": "2506.01488", "title": "Argument-Centric Causal Intervention Method for Mitigating Bias in Cross-Document Event Coreference Resolution", "abstract": "Cross-document Event Coreference Resolution (CD-ECR) is a fundamental task in natural language processing (NLP) that seeks to determine whether event mentions across multiple documents refer to the same real-world occurrence. However, current CD-ECR approaches predominantly rely on trigger features within input mention pairs, which induce spurious correlations between surface-level lexical features and coreference relationships, impairing the overall performance of the models. To address this issue, we propose a novel cross-document event coreference resolution method based on Argument-Centric Causal Intervention (ACCI). Specifically, we construct a structural causal graph to uncover confounding dependencies between lexical triggers and coreference labels, and introduce backdoor-adjusted interventions to isolate the true causal effect of argument semantics. To further mitigate spurious correlations, ACCI integrates a counterfactual reasoning module that quantifies the causal influence of trigger word perturbations, and an argument-aware enhancement module to promote greater sensitivity to semantically grounded information. In contrast to prior methods that depend on costly data augmentation or heuristic-based filtering, ACCI enables effective debiasing in a unified end-to-end framework without altering the underlying training procedure. Extensive experiments demonstrate that ACCI achieves CoNLL F1 of 88.4% on ECB+ and 85.2% on GVC, achieving state-of-the-art performance. The implementation and materials are available at https://github.com/era211/ACCI.", "date": "2025-06-02", "categories": []}
{"id": "2506.01533", "title": "A Diffusion-Based Method for Learning the Multi-Outcome Distribution of Medical Treatments", "abstract": "In medicine, treatments often influence multiple, interdependent outcomes, such as primary endpoints, complications, adverse events, or other secondary endpoints. Hence, to make optimal treatment decisions, clinicians are interested in learning the distribution of multi-dimensional treatment outcomes. However, the vast majority of machine learning methods for predicting treatment effects focus on single-outcome settings, despite the fact that medical data often include multiple, interdependent outcomes. To address this limitation, we propose a novel diffusion-based method called DIME to learn the joint distribution of multiple outcomes of medical treatments. We addresses three challenges relevant in medical practice: (i)it is tailored to learn the joint interventional distribution of multiple medical outcomes, which enables reliable decision-making with uncertainty quantification rather than relying solely on point estimates; (ii)it explicitly captures the dependence structure between outcomes; (iii)it can handle outcomes of mixed type, including binary, categorical, and continuous variables. In DIME, we take into account the fundamental problem of causal inference through causal masking. For training, our method decomposes the joint distribution into a series of conditional distributions with a customized conditional masking to account for the dependence structure across outcomes. For inference, our method auto-regressively generates predictions. This allows our method to move beyond point estimates of causal quantities and thus learn the joint interventional distribution. To the best of our knowledge, DIME is the first neural method tailored to learn the joint, multi-outcome distribution of medical treatments. Across various experiments, we demonstrate that our method effectively learns the joint distribution and captures shared information among multiple outcomes.", "date": "2025-06-02", "categories": []}
{"id": "2408.13448", "title": "Reinforcement Learning for Causal Discovery without Acyclicity Constraints", "abstract": "Recently, reinforcement learning (RL) has proved a promising alternative for conventional local heuristics in score-based approaches to learning directed acyclic causal graphs (DAGs) from observational data. However, the intricate acyclicity constraint still challenges the efficient exploration of the vast space of DAGs in existing methods. In this study, we introduce ALIAS (reinforced dAg Learning wIthout Acyclicity conStraints), a novel approach to causal discovery powered by the RL machinery. Our method features an efficient policy for generating DAGs in just a single step with an optimal quadratic complexity, fueled by a novel parametrization of DAGs that directly translates a continuous space to the space of all DAGs, bypassing the need for explicitly enforcing acyclicity constraints. This approach enables us to navigate the search space more effectively by utilizing policy gradient methods and established scoring functions. In addition, we provide compelling empirical evidence for the strong performance of ALIAS in comparison with state-of-the-arts in causal discovery over increasingly difficult experiment conditions on both synthetic and real datasets.", "date": "2024-08-24", "categories": []}
{"id": "2410.03380", "title": "Identifying biological perturbation targets through causal differential networks", "abstract": "Identifying variables responsible for changes to a biological system enables applications in drug target discovery and cell engineering. Given a pair of observational and interventional datasets, the goal is to isolate the subset of observed variables that were the targets of the intervention. Directly applying causal discovery algorithms is challenging: the data may contain thousands of variables with as few as tens of samples per intervention, and biological systems do not adhere to classical causality assumptions. We propose a causality-inspired approach to address this practical setting. First, we infer noisy causal graphs from the observational and interventional data. Then, we learn to map the differences between these graphs, along with additional statistical features, to sets of variables that were intervened upon. Both modules are jointly trained in a supervised framework, on simulated and real data that reflect the nature of biological interventions. This approach consistently outperforms baselines for perturbation modeling on seven single-cell transcriptomics datasets. We also demonstrate significant improvements over current causal discovery methods for predicting soft and hard intervention targets across a variety of synthetic data.", "date": "2024-10-04", "categories": []}
{"id": "2501.02333", "title": "On The Causal Network Of Face-selective Regions In Human Brain During Movie Watching", "abstract": "Understanding the causal interactions in some brain tasks, such as face processing, remains a challenging and ambiguous process for researchers. In this study, we address this issue by employing a novel causal discovery method -Directed Acyclic Graphs via M-matrices for Acyclicity (DAGMA)- to investigate the causal structure of the brain's face-selective network and gain deeper insights into its mechanism. Using fMRI data of natural movie stimuli, we extract causal network of face-selective regions and analyze how frames containing faces influence this network. Specifically, our findings reveal that the presence of faces in the stimuli, causally affects the number of identified connections within the network. Additionally, our results highlight the crucial role of subcortical regions in satisfying causal sufficiency, emphasizing it's importance in causal studies of brain. This study provides a new perspective on understanding the causal architecture of the face-selective network of the brain, motivating further research on neural causality.", "date": "2025-01-04", "categories": []}
{"id": "2503.11718", "title": "The Relativity of Causal Knowledge", "abstract": "Recent advances in artificial intelligence reveal the limits of purely predictive systems and call for a shift toward causal and collaborative reasoning. Drawing inspiration from the revolution of Grothendieck in mathematics, we introduce the relativity of causal knowledge, which posits structural causal models (SCMs) are inherently imperfect, subjective representations embedded within networks of relationships. By leveraging category theory, we arrange SCMs into a functor category and show that their observational and interventional probability measures naturally form convex structures. This result allows us to encode non-intervened SCMs with convex spaces of probability measures. Next, using sheaf theory, we construct the network sheaf and cosheaf of causal knowledge. These structures enable the transfer of causal knowledge across the network while incorporating interventional consistency and the perspective of the subjects, ultimately leading to the formal, mathematical definition of relative causal knowledge.", "date": "2025-03-13", "categories": []}
{"id": "2505.23944", "title": "Retrieval Augmented Generation based Large Language Models for Causality Mining", "abstract": "Causality detection and mining are important tasks in information retrieval due to their enormous use in information extraction, and knowledge graph construction. To solve these tasks, in existing literature there exist several solutions -- both unsupervised and supervised. However, the unsupervised methods suffer from poor performance and they often require significant human intervention for causal rule selection, leading to poor generalization across different domains. On the other hand, supervised methods suffer from the lack of large training datasets. Recently, large language models (LLMs) with effective prompt engineering are found to be effective to overcome the issue of unavailability of large training dataset. Yet, in existing literature, there does not exist comprehensive works on causality detection and mining using LLM prompting. In this paper, we present several retrieval-augmented generation (RAG) based dynamic prompting schemes to enhance LLM performance in causality detection and extraction tasks. Extensive experiments over three datasets and five LLMs validate the superiority of our proposed RAG-based dynamic prompting over other static prompting schemes.", "date": "2025-05-29", "categories": []}
{"id": "2505.24296", "title": "Data Fusion for Partial Identification of Causal Effects", "abstract": "Data fusion techniques integrate information from heterogeneous data sources to improve learning, generalization, and decision making across data sciences. In causal inference, these methods leverage rich observational data to improve causal effect estimation, while maintaining the trustworthiness of randomized controlled trials. Existing approaches often relax the strong no unobserved confounding assumption by instead assuming exchangeability of counterfactual outcomes across data sources. However, when both assumptions simultaneously fail - a common scenario in practice - current methods cannot identify or estimate causal effects. We address this limitation by proposing a novel partial identification framework that enables researchers to answer key questions such as: Is the causal effect positive or negative? and How severe must assumption violations be to overturn this conclusion? Our approach introduces interpretable sensitivity parameters that quantify assumption violations and derives corresponding causal effect bounds. We develop doubly robust estimators for these bounds and operationalize breakdown frontier analysis to understand how causal conclusions change as assumption violations increase. We apply our framework to the Project STAR study, which investigates the effect of classroom size on students' third-grade standardized test performance. Our analysis reveals that the Project STAR results are robust to simultaneous violations of key assumptions, both on average and across various subgroups of interest. This strengthens confidence in the study's conclusions despite potential unmeasured biases in the data.", "date": "2025-05-30", "categories": []}
{"id": "2505.24710", "title": "Causal-aware Large Language Models: Enhancing Decision-Making Through Learning, Adapting and Acting", "abstract": "Large language models (LLMs) have shown great potential in decision-making due to the vast amount of knowledge stored within the models. However, these pre-trained models are prone to lack reasoning abilities and are difficult to adapt to new environments, further hindering their application to complex real-world tasks. To address these challenges, inspired by the human cognitive process, we propose Causal-aware LLMs, which integrate the structural causal model (SCM) into the decision-making process to model, update, and utilize structured knowledge of the environment in a ``learning-adapting-acting\" paradigm. Specifically, in the learning stage, we first utilize an LLM to extract the environment-specific causal entities and their causal relations to initialize a structured causal model of the environment. Subsequently,in the adapting stage, we update the structured causal model through external feedback about the environment, via an idea of causal intervention. Finally, in the acting stage, Causal-aware LLMs exploit structured causal knowledge for more efficient policy-making through the reinforcement learning agent. The above processes are performed iteratively to learn causal knowledge, ultimately enabling the causal-aware LLMs to achieve a more accurate understanding of the environment and make more efficient decisions. Experimental results across 22 diverse tasks within the open-world game ``Crafter\" validate the effectiveness of our proposed method.", "date": "2025-05-30", "categories": []}
{"id": "2210.15709", "title": "Improvement-Focused Causal Recourse (ICR)", "abstract": "Algorithmic recourse recommendations, such as Karimi et al.'s (2021) causal recourse (CR), inform stakeholders of how to act to revert unfavourable decisions. However, some actions lead to acceptance (i.e., revert the model's decision) but do not lead to improvement (i.e., may not revert the underlying real-world state). To recommend such actions is to recommend fooling the predictor. We introduce a novel method, Improvement-Focused Causal Recourse (ICR), which involves a conceptual shift: Firstly, we require ICR recommendations to guide towards improvement. Secondly, we do not tailor the recommendations to be accepted by a specific predictor. Instead, we leverage causal knowledge to design decision systems that predict accurately pre- and post-recourse. As a result, improvement guarantees translate into acceptance guarantees. We demonstrate that given correct causal knowledge, ICR, in contrast to existing approaches, guides towards both acceptance and improvement.", "date": "2022-10-27", "categories": []}
{"id": "2501.13890", "title": "Federated Granger Causality Learning for Interdependent Clients with State Space Representation", "abstract": "Advanced sensors and IoT devices have improved the monitoring and control of complex industrial enterprises. They have also created an interdependent fabric of geographically distributed process operations (clients) across these enterprises. Granger causality is an effective approach to detect and quantify interdependencies by examining how one client's state affects others over time. Understanding these interdependencies captures how localized events, such as faults and disruptions, can propagate throughout the system, possibly causing widespread operational impacts. However, the large volume and complexity of industrial data pose challenges in modeling these interdependencies. This paper develops a federated approach to learning Granger causality. We utilize a linear state space system framework that leverages low-dimensional state estimates to analyze interdependencies. This addresses bandwidth limitations and the computational burden commonly associated with centralized data processing. We propose augmenting the client models with the Granger causality information learned by the server through a Machine Learning (ML) function. We examine the co-dependence between the augmented client and server models and reformulate the framework as a standalone ML algorithm providing conditions for its sublinear and linear convergence rates. We also study the convergence of the framework to a centralized oracle model. Moreover, we include a differential privacy analysis to ensure data security while preserving causal insights. Using synthetic data, we conduct comprehensive experiments to demonstrate the robustness of our approach to perturbations in causality, the scalability to the size of communication, number of clients, and the dimensions of raw data. We also evaluate the performance on two real-world industrial control system datasets by reporting the volume of data saved by decentralization.", "date": "2025-01-23", "categories": []}
{"id": "2308.00957", "title": "Improving the Variance of Differentially Private Randomized Experiments through Clustering", "abstract": "Estimating causal effects from randomized experiments is only possible if participants are willing to disclose their potentially sensitive responses. Differential privacy, a widely used framework for ensuring an algorithms privacy guarantees, can encourage participants to share their responses without the risk of de-anonymization. However, many mechanisms achieve differential privacy by adding noise to the original dataset, which reduces the precision of causal effect estimation. This introduces a fundamental trade-off between privacy and variance when performing causal analyses on differentially private data. In this work, we propose a new differentially private mechanism, \"Cluster-DP\", which leverages a given cluster structure in the data to improve the privacy-variance trade-off. While our results apply to any clustering, we demonstrate that selecting higher-quality clusters, according to a quality metric we introduce, can decrease the variance penalty without compromising privacy guarantees. Finally, we evaluate the theoretical and empirical performance of our Cluster-DP algorithm on both real and simulated data, comparing it to common baselines, including two special cases of our algorithm: its unclustered version and a uniform-prior version.", "date": "2023-08-02", "categories": []}
{"id": "2501.19345", "title": "PUATE: Efficient Average Treatment Effect Estimation from Treated (Positive) and Unlabeled Units", "abstract": "The estimation of average treatment effects (ATEs), defined as the difference in expected outcomes between treatment and control groups, is a central topic in causal inference. This study develops semiparametric efficient estimators for ATE in a setting where only a treatment group and an unlabeled group, consisting of units whose treatment status is unknown, are observed. This scenario constitutes a variant of learning from positive and unlabeled data (PU learning) and can be viewed as a special case of ATE estimation with missing data. For this setting, we derive the semiparametric efficiency bounds, which characterize the lowest achievable asymptotic variance for regular estimators. We then construct semiparametric efficient ATE estimators that attain these bounds. Our results contribute to the literature on causal inference with missing data and weakly supervised learning.", "date": "2025-01-31", "categories": []}
{"id": "2504.13904", "title": "Generative Framework for Personalized Persuasion: Inferring Causal, Counterfactual, and Latent Knowledge", "abstract": "We hypothesize that optimal system responses emerge from adaptive strategies grounded in causal and counterfactual knowledge. Counterfactual inference allows us to create hypothetical scenarios to examine the effects of alternative system responses. We enhance this process through causal discovery, which identifies the strategies informed by the underlying causal structure that govern system behaviors. Moreover, we consider the psychological constructs and unobservable noises that might be influencing user-system interactions as latent factors. We show that these factors can be effectively estimated. We employ causal discovery to identify strategy-level causal relationships among user and system utterances, guiding the generation of personalized counterfactual dialogues. We model the user utterance strategies as causal factors, enabling system strategies to be treated as counterfactual actions. Furthermore, we optimize policies for selecting system responses based on counterfactual data. Our results using a real-world dataset on social good demonstrate significant improvements in persuasive system outcomes, with increased cumulative rewards validating the efficacy of causal discovery in guiding personalized counterfactual inference and optimizing dialogue policies for a persuasive dialogue system.", "date": "2025-04-08", "categories": []}
{"id": "2505.14310", "title": "Taming Recommendation Bias with Causal Intervention on Evolving Personal Popularity", "abstract": "Popularity bias occurs when popular items are recommended far more frequently than they should be, negatively impacting both user experience and recommendation accuracy. Existing debiasing methods mitigate popularity bias often uniformly across all users and only partially consider the time evolution of users or items. However, users have different levels of preference for item popularity, and this preference is evolving over time. To address these issues, we propose a novel method called CausalEPP (Causal Intervention on Evolving Personal Popularity) for taming recommendation bias, which accounts for the evolving personal popularity of users. Specifically, we first introduce a metric called {Evolving Personal Popularity} to quantify each user's preference for popular items. Then, we design a causal graph that integrates evolving personal popularity into the conformity effect, and apply deconfounded training to mitigate the popularity bias of the causal graph. During inference, we consider the evolution consistency between users and items to achieve a better recommendation. Empirical studies demonstrate that CausalEPP outperforms baseline methods in reducing popularity bias while improving recommendation accuracy.", "date": "2025-05-20", "categories": []}
{"id": "2505.20540", "title": "Causality and \"In-the-Wild\" Video-Based Person Re-ID: A Survey", "abstract": "Video-based person re-identification (Re-ID) remains brittle in real-world deployments despite impressive benchmark performance. Most existing models rely on superficial correlations such as clothing, background, or lighting that fail to generalize across domains, viewpoints, and temporal variations. This survey examines the emerging role of causal reasoning as a principled alternative to traditional correlation-based approaches in video-based Re-ID. We provide a structured and critical analysis of methods that leverage structural causal models, interventions, and counterfactual reasoning to isolate identity-specific features from confounding factors. The survey is organized around a novel taxonomy of causal Re-ID methods that spans generative disentanglement, domain-invariant modeling, and causal transformers. We review current evaluation metrics and introduce causal-specific robustness measures. In addition, we assess practical challenges of scalability, fairness, interpretability, and privacy that must be addressed for real-world adoption. Finally, we identify open problems and outline future research directions that integrate causal modeling with efficient architectures and self-supervised learning. This survey aims to establish a coherent foundation for causal video-based person Re-ID and to catalyze the next phase of research in this rapidly evolving domain.", "date": "2025-05-26", "categories": []}
{"id": "2306.07566", "title": "Learning with Selectively Labeled Data from Multiple Decision-makers", "abstract": "We study the problem of classification with selectively labeled data, whose distribution may differ from the full population due to historical decision-making. We exploit the fact that in many applications historical decisions were made by multiple decision-makers, each with different decision rules. We analyze this setup under a principled instrumental variable (IV) framework and rigorously study the identification of classification risk. We establish conditions for the exact identification of classification risk and derive tight partial identification bounds when exact identification fails. We further propose a unified cost-sensitive learning (UCL) approach to learn classifiers robust to selection bias in both identification settings. Finally, we theoretically and numerically validate the efficacy of our proposed method.", "date": "2023-06-13", "categories": []}
{"id": "2410.10502", "title": "A Practical Approach to Causal Inference over Time", "abstract": "In this paper, we focus on estimating the causal effect of an intervention over time on a dynamical system. To that end, we formally define causal interventions and their effects over time on discrete-time stochastic processes (DSPs). Then, we show under which conditions the equilibrium states of a DSP, both before and after a causal intervention, can be captured by a structural causal model (SCM). With such an equivalence at hand, we provide an explicit mapping from vector autoregressive models (VARs), broadly applied in econometrics, to linear, but potentially cyclic and/or affected by unmeasured confounders, SCMs. The resulting causal VAR framework allows us to perform causal inference over time from observational time series data. Our experiments on synthetic and real-world datasets show that the proposed framework achieves strong performance in terms of observational forecasting while enabling accurate estimation of the causal effect of interventions on dynamical systems. We demonstrate, through a case study, the potential practical questions that can be addressed using the proposed causal VAR framework.", "date": "2024-10-14", "categories": []}
{"id": "2505.20536", "title": "Covariate-Adjusted Deep Causal Learning for Heterogeneous Panel Data Models", "abstract": "This paper studies the task of estimating heterogeneous treatment effects in causal panel data models, in the presence of covariate effects. We propose a novel Covariate-Adjusted Deep Causal Learning (CoDEAL) for panel data models, that employs flexible model structures and powerful neural network architectures to cohesively deal with the underlying heterogeneity and nonlinearity of both panel units and covariate effects. The proposed CoDEAL integrates nonlinear covariate effect components (parameterized by a feed-forward neural network) with nonlinear factor structures (modeled by a multi-output autoencoder) to form a heterogeneous causal panel model. The nonlinear covariate component offers a flexible framework for capturing the complex influences of covariates on outcomes. The nonlinear factor analysis enables CoDEAL to effectively capture both cross-sectional and temporal dependencies inherent in the data panel. This latent structural information is subsequently integrated into a customized matrix completion algorithm, thereby facilitating more accurate imputation of missing counterfactual outcomes. Moreover, the use of a multi-output autoencoder explicitly accounts for heterogeneity across units and enhances the model interpretability of the latent factors. We establish theoretical guarantees on the convergence of the estimated counterfactuals, and demonstrate the compelling performance of the proposed method using extensive simulation studies and a real data application.", "date": "2025-05-26", "categories": []}
{"id": "2407.14058", "title": "Towards the Causal Complete Cause of Multi-Modal Representation Learning", "abstract": "Multi-Modal Learning (MML) aims to learn effective representations across modalities for accurate predictions. Existing methods typically focus on modality consistency and specificity to learn effective representations. However, from a causal perspective, they may lead to representations that contain insufficient and unnecessary information. To address this, we propose that effective MML representations should be causally sufficient and necessary. Considering practical issues like spurious correlations and modality conflicts, we relax the exogeneity and monotonicity assumptions prevalent in prior works and explore the concepts specific to MML, i.e., Causal Complete Cause $C^3$. We begin by defining $C^3$, which quantifies the probability of representations being causally sufficient and necessary. We then discuss the identifiability of $C^3$ and introduce an instrumental variable to support identifying $C^3$ with non-exogeneity and non-monotonicity. Building on this, we conduct the $C^3$ measurement, i.e., \\(C^3\\) risk. We propose a twin network to estimate it through (i) the real-world branch: utilizing the instrumental variable for sufficiency, and (ii) the hypothetical-world branch: applying gradient-based counterfactual modeling for necessity. Theoretical analyses confirm its reliability. Based on these results, we propose $C^3$ Regularization, a plug-and-play method that enforces the causal completeness of the learned representations by minimizing $C^3$ risk. Extensive experiments demonstrate its effectiveness.", "date": "2024-07-19", "categories": []}
{"id": "2504.11076", "title": "Using Time Structure to Estimate Causal Effects", "abstract": "There exist several approaches for estimating causal effects in time series when latent confounding is present. Many of these approaches rely on additional auxiliary observed variables or time series such as instruments, negative controls or time series that satisfy the front- or backdoor criterion in certain graphs. In this paper, we present a novel approach for estimating direct (and via Wright's path rule total) causal effects in a time series setup which does not rely on additional auxiliary observed variables or time series. This approach assumes that the underlying time series is a Structural Vector Autoregressive (SVAR) process and estimates direct causal effects by solving certain linear equation systems made up of different covariances and model parameters. We state sufficient graphical criteria in terms of the so-called full time graph under which these linear equations systems are uniquely solvable and under which their solutions contain the to-be-identified direct causal effects as components. We also state sufficient lag-based criteria under which the previously mentioned graphical conditions are satisfied and, thus, under which direct causal effects are identifiable. Several numerical experiments underline the correctness and applicability of our results.", "date": "2025-04-15", "categories": []}
{"id": "2505.19355", "title": "Estimating Online Influence Needs Causal Modeling! Counterfactual Analysis of Social Media Engagement", "abstract": "Understanding true influence in social media requires distinguishing correlation from causation--particularly when analyzing misinformation spread. While existing approaches focus on exposure metrics and network structures, they often fail to capture the causal mechanisms by which external temporal signals trigger engagement. We introduce a novel joint treatment-outcome framework that leverages existing sequential models to simultaneously adapt to both policy timing and engagement effects. Our approach adapts causal inference techniques from healthcare to estimate Average Treatment Effects (ATE) within the sequential nature of social media interactions, tackling challenges from external confounding signals. Through our experiments on real-world misinformation and disinformation datasets, we show that our models outperform existing benchmarks by 15--22% in predicting engagement across diverse counterfactual scenarios, including exposure adjustment, timing shifts, and varied intervention durations. Case studies on 492 social media users show our causal effect measure aligns strongly with the gold standard in influence estimation, the expert-based empirical influence.", "date": "2025-05-25", "categories": []}
{"id": "2409.15219", "title": "MotifDisco: Motif Causal Discovery For Time Series Motifs", "abstract": "Many time series, particularly health data streams, can be best understood as a sequence of phenomenon or events, which we call \\textit{motifs}. A time series motif is a short trace segment which may implicitly capture an underlying phenomenon within the time series. Specifically, we focus on glucose traces collected from continuous glucose monitors (CGMs), which inherently contain motifs representing underlying human behaviors such as eating and exercise. The ability to identify and quantify \\textit{causal} relationships amongst motifs can provide a mechanism to better understand and represent these patterns, useful for improving deep learning and generative models and for advanced technology development (e.g., personalized coaching and artificial insulin delivery systems). However, no previous work has developed causal discovery methods for time series motifs. Therefore, in this paper we develop MotifDisco (\\textbf{motif} \\textbf{disco}very of causality), a novel causal discovery framework to learn causal relations amongst motifs from time series traces. We formalize a notion of \\textit{Motif Causality (MC)}, inspired from Granger Causality and Transfer Entropy, and develop a Graph Neural Network-based framework that learns causality between motifs by solving an unsupervised link prediction problem. We integrate MC with three model use cases of forecasting, anomaly detection and clustering, to showcase the use of MC as a building block for downstream tasks. Finally, we evaluate our framework on different health data streams and find that Motif Causality provides a significant performance improvement in all use cases.", "date": "2024-09-23", "categories": []}
{"id": "2505.16209", "title": "MedCFVQA: A Causal Approach to Mitigate Modality Preference Bias in Medical Visual Question Answering", "abstract": "Medical Visual Question Answering (MedVQA) is crucial for enhancing the efficiency of clinical diagnosis by providing accurate and timely responses to clinicians' inquiries regarding medical images. Existing MedVQA models suffered from modality preference bias, where predictions are heavily dominated by one modality while overlooking the other (in MedVQA, usually questions dominate the answer but images are overlooked), thereby failing to learn multimodal knowledge. To overcome the modality preference bias, we proposed a Medical CounterFactual VQA (MedCFVQA) model, which trains with bias and leverages causal graphs to eliminate the modality preference bias during inference. Existing MedVQA datasets exhibit substantial prior dependencies between questions and answers, which results in acceptable performance even if the model significantly suffers from the modality preference bias. To address this issue, we reconstructed new datasets by leveraging existing MedVQA datasets and Changed their P3rior dependencies (CP) between questions and their answers in the training and test set. Extensive experiments demonstrate that MedCFVQA significantly outperforms its non-causal counterpart on both SLAKE, RadVQA and SLAKE-CP, RadVQA-CP datasets.", "date": "2025-05-22", "categories": []}
{"id": "2505.17283", "title": "Deconfounded Warm-Start Thompson Sampling with Applications to Precision Medicine", "abstract": "Randomized clinical trials often require large patient cohorts before drawing definitive conclusions, yet abundant observational data from parallel studies remains underutilized due to confounding and hidden biases. To bridge this gap, we propose Deconfounded Warm-Start Thompson Sampling (DWTS), a practical approach that leverages a Doubly Debiased LASSO (DDL) procedure to identify a sparse set of reliable measured covariates and combines them with key hidden covariates to form a reduced context. By initializing Thompson Sampling (LinTS) priors with DDL-estimated means and variances on these measured features -- while keeping uninformative priors on hidden features -- DWTS effectively harnesses confounded observational data to kick-start adaptive clinical trials. Evaluated on both a purely synthetic environment and a virtual environment created using real cardiovascular risk dataset, DWTS consistently achieves lower cumulative regret than standard LinTS, showing how offline causal insights from observational data can improve trial efficiency and support more personalized treatment decisions.", "date": "2025-05-22", "categories": []}
{"id": "2503.08245", "title": "ExMAG: Learning of Maximally Ancestral Graphs", "abstract": "In mixed graphs, there are both directed and undirected edges. An extension of acyclicity to this mixed-graph setting is known as maximally ancestral graphs. This extension is of considerable interest in causal learning in the presence of confounders. There, directed edges represent a clear direction of causality, while undirected edges represent confounding. We propose a score-based branch-and-cut algorithm for learning maximally ancestral graphs. The algorithm produces more accurate results than state-of-the-art methods, while being faster to run on small and medium-sized synthetic instances.", "date": "2025-03-11", "categories": []}
{"id": "2505.13092", "title": "Treatment Effect Estimation for Optimal Decision-Making", "abstract": "Decision-making across various fields, such as medicine, heavily relies on conditional average treatment effects (CATEs). Practitioners commonly make decisions by checking whether the estimated CATE is positive, even though the decision-making performance of modern CATE estimators is poorly understood from a theoretical perspective. In this paper, we study optimal decision-making based on two-stage CATE estimators (e.g., DR-learner), which are considered state-of-the-art and widely used in practice. We prove that, while such estimators may be optimal for estimating CATE, they can be suboptimal when used for decision-making. Intuitively, this occurs because such estimators prioritize CATE accuracy in regions far away from the decision boundary, which is ultimately irrelevant to decision-making. As a remedy, we propose a novel two-stage learning objective that retargets the CATE to balance CATE estimation error and decision performance. We then propose a neural method that optimizes an adaptively-smoothed approximation of our learning objective. Finally, we confirm the effectiveness of our method both empirically and theoretically. In sum, our work is the first to show how two-stage CATE estimators can be adapted for optimal decision-making.", "date": "2025-05-19", "categories": []}
{"id": "2505.15866", "title": "Multi-omic Causal Discovery using Genotypes and Gene Expression", "abstract": "Causal discovery in multi-omic datasets is crucial for understanding the bigger picture of gene regulatory mechanisms, but remains challenging due to high dimensionality, differentiation of direct from indirect relationships, and hidden confounders. We introduce GENESIS (GEne Network inference from Expression SIgnals and SNPs), a constraint-based algorithm that leverages the natural causal precedence of genotypes to infer ancestral relationships in transcriptomic data. Unlike traditional causal discovery methods that start with a fully connected graph, GENESIS initialises an empty ancestrality matrix and iteratively populates it with direct, indirect or non-causal relationships using a series of provably sound marginal and conditional independence tests. By integrating genotypes as fixed causal anchors, GENESIS provides a principled ``head start'' to classical causal discovery algorithms, restricting the search space to biologically plausible edges. We test GENESIS on synthetic and real-world genomic datasets. This framework offers a powerful avenue for uncovering causal pathways in complex traits, with promising applications to functional genomics, drug discovery, and precision medicine.", "date": "2025-05-21", "categories": []}
{"id": "2505.16982", "title": "Beyond Correlation: Towards Causal Large Language Model Agents in Biomedicine", "abstract": "Large Language Models (LLMs) show promise in biomedicine but lack true causal understanding, relying instead on correlations. This paper envisions causal LLM agents that integrate multimodal data (text, images, genomics, etc.) and perform intervention-based reasoning to infer cause-and-effect. Addressing this requires overcoming key challenges: designing safe, controllable agentic frameworks; developing rigorous benchmarks for causal evaluation; integrating heterogeneous data sources; and synergistically combining LLMs with structured knowledge (KGs) and formal causal inference tools. Such agents could unlock transformative opportunities, including accelerating drug discovery through automated hypothesis generation and simulation, enabling personalized medicine through patient-specific causal models. This research agenda aims to foster interdisciplinary efforts, bridging causal concepts and foundation models to develop reliable AI partners for biomedical progress.", "date": "2025-05-22", "categories": []}
{"id": "2501.15549", "title": "Optimal Transport on Categorical Data for Counterfactuals using Compositional Data and Dirichlet Transport", "abstract": "Recently, optimal transport-based approaches have gained attention for deriving counterfactuals, e.g., to quantify algorithmic discrimination. However, in the general multivariate setting, these methods are often opaque and difficult to interpret. To address this, alternative methodologies have been proposed, using causal graphs combined with iterative quantile regressions (Ple\\v{c}ko and Meinshausen (2020)) or sequential transport (Fernandes Machado et al. (2025)) to examine fairness at the individual level, often referred to as ``counterfactual fairness.'' Despite these advancements, transporting categorical variables remains a significant challenge in practical applications with real datasets. In this paper, we propose a novel approach to address this issue. Our method involves (1) converting categorical variables into compositional data and (2) transporting these compositions within the probabilistic simplex of $\\mathbb{R}^d$. We demonstrate the applicability and effectiveness of this approach through an illustration on real-world data, and discuss limitations.", "date": "2025-01-26", "categories": []}
{"id": "2505.09847", "title": "Causal Predictive Optimization and Generation for Business AI", "abstract": "The sales process involves sales functions converting leads or opportunities to customers and selling more products to existing customers. The optimization of the sales process thus is key to success of any B2B business. In this work, we introduce a principled approach to sales optimization and business AI, namely the Causal Predictive Optimization and Generation, which includes three layers: 1) prediction layer with causal ML 2) optimization layer with constraint optimization and contextual bandit 3) serving layer with Generative AI and feedback-loop for system enhancement. We detail the implementation and deployment of the system in LinkedIn, showcasing significant wins over legacy systems and sharing learning and insight broadly applicable to this field.", "date": "2025-05-14", "categories": []}
{"id": "2505.14828", "title": "Deep Koopman operator framework for causal discovery in nonlinear dynamical systems", "abstract": "We use a deep Koopman operator-theoretic formalism to develop a novel causal discovery algorithm, Kausal. Causal discovery aims to identify cause-effect mechanisms for better scientific understanding, explainable decision-making, and more accurate modeling. Standard statistical frameworks, such as Granger causality, lack the ability to quantify causal relationships in nonlinear dynamics due to the presence of complex feedback mechanisms, timescale mixing, and nonstationarity. This presents a challenge in studying many real-world systems, such as the Earth's climate. Meanwhile, Koopman operator methods have emerged as a promising tool for approximating nonlinear dynamics in a linear space of observables. In Kausal, we propose to leverage this powerful idea for causal analysis where optimal observables are inferred using deep learning. Causal estimates are then evaluated in a reproducing kernel Hilbert space, and defined as the distance between the marginal dynamics of the effect and the joint dynamics of the cause-effect observables. Our numerical experiments demonstrate Kausal's superior ability in discovering and characterizing causal signals compared to existing approaches of prescribed observables. Lastly, we extend our analysis to observations of El Ni\\~no-Southern Oscillation highlighting our algorithm's applicability to real-world phenomena. Our code is available at https://github.com/juannat7/kausal.", "date": "2025-05-20", "categories": []}
{"id": "2505.15246", "title": "Mitigating Spurious Correlations with Causal Logit Perturbation", "abstract": "Deep learning has seen widespread success in various domains such as science, industry, and society. However, it is acknowledged that certain approaches suffer from non-robustness, relying on spurious correlations for predictions. Addressing these limitations is of paramount importance, necessitating the development of methods that can disentangle spurious correlations. {This study attempts to implement causal models via logit perturbations and introduces a novel Causal Logit Perturbation (CLP) framework to train classifiers with generated causal logit perturbations for individual samples, thereby mitigating the spurious associations between non-causal attributes (i.e., image backgrounds) and classes.} {Our framework employs a} perturbation network to generate sample-wise logit perturbations using a series of training characteristics of samples as inputs. The whole framework is optimized by an online meta-learning-based learning algorithm and leverages human causal knowledge by augmenting metadata in both counterfactual and factual manners. Empirical evaluations on four typical biased learning scenarios, including long-tail learning, noisy label learning, generalized long-tail learning, and subpopulation shift learning, demonstrate that CLP consistently achieves state-of-the-art performance. Moreover, visualization results support the effectiveness of the generated causal perturbations in redirecting model attention towards causal image attributes and dismantling spurious associations.", "date": "2025-05-21", "categories": []}
{"id": "2410.11348", "title": "RATE: Causal Explainability of Reward Models with Imperfect Counterfactuals", "abstract": "Reward models are widely used as proxies for human preferences when aligning or evaluating LLMs. However, reward models are black boxes, and it is often unclear what, exactly, they are actually rewarding. In this paper we develop Rewrite-based Attribute Treatment Estimator (RATE) as an effective method for measuring the sensitivity of a reward model to high-level attributes of responses, such as sentiment, helpfulness, or complexity. Importantly, RATE measures the causal effect of an attribute on the reward. RATE uses LLMs to rewrite responses to produce imperfect counterfactuals examples that can be used to measure causal effects. A key challenge is that these rewrites are imperfect in a manner that can induce substantial bias in the estimated sensitivity of the reward model to the attribute. The core idea of RATE is to adjust for this imperfect-rewrite effect by rewriting twice. We establish the validity of the RATE procedure and show empirically that it is an effective estimator.", "date": "2024-10-15", "categories": []}
{"id": "2412.09232", "title": "Uplift modeling with continuous treatments: A predict-then-optimize approach", "abstract": "The goal of uplift modeling is to recommend actions that optimize specific outcomes by determining which entities should receive treatment. One common approach involves two steps: first, an inference step that estimates conditional average treatment effects (CATEs), and second, an optimization step that ranks entities based on their CATE values and assigns treatment to the top k within a given budget. While uplift modeling typically focuses on binary treatments, many real-world applications are characterized by continuous-valued treatments, i.e., a treatment dose. This paper presents a predict-then-optimize framework to allow for continuous treatments in uplift modeling. First, in the inference step, conditional average dose responses (CADRs) are estimated from data using causal machine learning techniques. Second, in the optimization step, we frame the assignment task of continuous treatments as a dose-allocation problem and solve it using integer linear programming (ILP). This approach allows decision-makers to efficiently and effectively allocate treatment doses while balancing resource availability, with the possibility of adding extra constraints like fairness considerations or adapting the objective function to take into account instance-dependent costs and benefits to maximize utility. The experiments compare several CADR estimators and illustrate the trade-offs between policy value and fairness, as well as the impact of an adapted objective function. This showcases the framework's advantages and flexibility across diverse applications in healthcare, lending, and human resource management. All code is available on github.com/SimonDeVos/UMCT.", "date": "2024-12-12", "categories": []}
{"id": "2505.14396", "title": "Causal Cartographer: From Mapping to Reasoning Over Counterfactual Worlds", "abstract": "Causal world models are systems that can answer counterfactual questions about an environment of interest, i.e. predict how it would have evolved if an arbitrary subset of events had been realized differently. It requires understanding the underlying causes behind chains of events and conducting causal inference for arbitrary unseen distributions. So far, this task eludes foundation models, notably large language models (LLMs), which do not have demonstrated causal reasoning capabilities beyond the memorization of existing causal relationships. Furthermore, evaluating counterfactuals in real-world applications is challenging since only the factual world is observed, limiting evaluation to synthetic datasets. We address these problems by explicitly extracting and modeling causal relationships and propose the Causal Cartographer framework. First, we introduce a graph retrieval-augmented generation agent tasked to retrieve causal relationships from data. This approach allows us to construct a large network of real-world causal relationships that can serve as a repository of causal knowledge and build real-world counterfactuals. In addition, we create a counterfactual reasoning agent constrained by causal relationships to perform reliable step-by-step causal inference. We show that our approach can extract causal knowledge and improve the robustness of LLMs for causal reasoning tasks while reducing inference costs and spurious correlations.", "date": "2025-05-20", "categories": []}
{"id": "2308.15395", "title": "The CausalBench challenge: A machine learning contest for gene network inference from single-cell perturbation data", "abstract": "In drug discovery, mapping interactions between genes within cellular systems is a crucial early step. Such maps are not only foundational for understanding the molecular mechanisms underlying disease biology but also pivotal for formulating hypotheses about potential targets for new medicines. Recognizing the need to elevate the construction of these gene-gene interaction networks, especially from large-scale, real-world datasets of perturbed single cells, the CausalBench Challenge was initiated. This challenge aimed to inspire the machine learning community to enhance state-of-the-art methods, emphasizing better utilization of expansive genetic perturbation data. Using the framework provided by the CausalBench benchmark, participants were tasked with refining the current methodologies or proposing new ones. This report provides an analysis and summary of the methods submitted during the challenge to give a partial image of the state of the art at the time of the challenge. Notably, the winning solutions significantly improved performance compared to previous baselines, establishing a new state of the art for this critical task in biology and medicine.", "date": "2023-08-29", "categories": []}
{"id": "2405.18314", "title": "Deriving Causal Order from Single-Variable Interventions: Guarantees & Algorithm", "abstract": "Targeted and uniform interventions to a system are crucial for unveiling causal relationships. While several methods have been developed to leverage interventional data for causal structure learning, their practical application in real-world scenarios often remains challenging. Recent benchmark studies have highlighted these difficulties, even when large numbers of single-variable intervention samples are available. In this work, we demonstrate, both theoretically and empirically, that such datasets contain a wealth of causal information that can be effectively extracted under realistic assumptions about the data distribution. More specifically, we introduce a novel variant of interventional faithfulness, which relies on comparisons between the marginal distributions of each variable across observational and interventional settings, and we introduce a score on causal orders. Under this assumption, we are able to prove strong theoretical guarantees on the optimum of our score that also hold for large-scale settings. To empirically verify our theory, we introduce Intersort, an algorithm designed to infer the causal order from datasets containing large numbers of single-variable interventions by approximately optimizing our score. Intersort outperforms baselines (GIES, DCDI, PC and EASE) on almost all simulated data settings replicating common benchmarks in the field. Our proposed novel approach to modeling interventional datasets thus offers a promising avenue for advancing causal inference, highlighting significant potential for further enhancements under realistic assumptions.", "date": "2024-05-28", "categories": []}
{"id": "2406.05805", "title": "Towards identifiability of micro total effects in summary causal graphs with latent confounding: extension of the front-door criterion", "abstract": "Conducting experiments to estimate total effects can be challenging due to cost, ethical concerns, or practical limitations. As an alternative, researchers often rely on causal graphs to determine whether these effects can be identified from observational data. Identifying total effects in fully specified causal graphs has received considerable attention, with Pearl's front-door criterion enabling the identification of total effects in the presence of latent confounding even when no variable set is sufficient for adjustment. However, specifying a complete causal graph is challenging in many domains. Extending these identifiability results to partially specified graphs is crucial, particularly in dynamic systems where causal relationships evolve over time. This paper addresses the challenge of identifying total effects using a specific and well-known partially specified graph in dynamic systems called a summary causal graph, which does not specify the temporal lag between causal relations and can contain cycles. In particular, this paper presents sufficient graphical conditions for identifying total effects from observational data, even in the presence of cycles and latent confounding, and when no variable set is sufficient for adjustment.", "date": "2024-06-09", "categories": []}
{"id": "2412.02878", "title": "Modeling and Discovering Direct Causes for Predictive Models", "abstract": "We introduce a causal modeling framework that captures the input-output behavior of predictive models (e.g., machine learning models). The framework enables us to identify features that directly cause the predictions, which has broad implications for data collection and model evaluation. We then present sound and complete algorithms for discovering direct causes (from data) under some assumptions. Furthermore, we propose a novel independence rule that can be integrated with the algorithms to accelerate the discovery process, as we demonstrate both theoretically and empirically.", "date": "2024-12-03", "categories": []}
{"id": "2501.08067", "title": "Optimal Policy Adaptation under Covariate Shift", "abstract": "Transfer learning of prediction models has been extensively studied, while the corresponding policy learning approaches are rarely discussed. In this paper, we propose principled approaches for learning the optimal policy in the target domain by leveraging two datasets: one with full information from the source domain and the other from the target domain with only covariates. First, under the setting of covariate shift, we formulate the problem from a perspective of causality and present the identifiability assumptions for the reward induced by a given policy. Then, we derive the efficient influence function and the semiparametric efficiency bound for the reward. Based on this, we construct a doubly robust and semiparametric efficient estimator for the reward and then learn the optimal policy by optimizing the estimated reward. Moreover, we theoretically analyze the bias and the generalization error bound for the learned policy. Extensive experiments demonstrate that the approach not only estimates the reward more accurately but also yields a policy that closely approximates the theoretically optimal policy.", "date": "2025-01-14", "categories": []}
{"id": "2503.17290", "title": "Calibration Strategies for Robust Causal Estimation: Theoretical and Empirical Insights on Propensity Score-Based Estimators", "abstract": "The partitioning of data for estimation and calibration critically impacts the performance of propensity score based estimators like inverse probability weighting (IPW) and double/debiased machine learning (DML) frameworks. We extend recent advances in calibration techniques for propensity score estimation, improving the robustness of propensity scores in challenging settings such as limited overlap, small sample sizes, or unbalanced data. Our contributions are twofold: First, we provide a theoretical analysis of the properties of calibrated estimators in the context of DML. To this end, we refine existing calibration frameworks for propensity score models, with a particular emphasis on the role of sample-splitting schemes in ensuring valid causal inference. Second, through extensive simulations, we show that calibration reduces variance of inverse-based propensity score estimators while also mitigating bias in IPW, even in small-sample regimes. Notably, calibration improves stability for flexible learners (e.g., gradient boosting) while preserving the doubly robust properties of DML. A key insight is that, even when methods perform well without calibration, incorporating a calibration step does not degrade performance, provided that an appropriate sample-splitting approach is chosen.", "date": "2025-03-21", "categories": []}
{"id": "2505.12094", "title": "Attribution Projection Calculus: A Novel Framework for Causal Inference in Bayesian Networks", "abstract": "This paper introduces Attribution Projection Calculus (AP-Calculus), a novel mathematical framework for determining causal relationships in structured Bayesian networks. We investigate a specific network architecture with source nodes connected to destination nodes through intermediate nodes, where each input maps to a single label with maximum marginal probability. We prove that for each label, exactly one intermediate node acts as a deconfounder while others serve as confounders, enabling optimal attribution of features to their corresponding labels. The framework formalizes the dual nature of intermediate nodes as both confounders and deconfounders depending on the context, and establishes separation functions that maximize distinctions between intermediate representations. We demonstrate that the proposed network architecture is optimal for causal inference compared to alternative structures, including those based on Pearl's causal framework. AP-Calculus provides a comprehensive mathematical foundation for analyzing feature-label attributions, managing spurious correlations, quantifying information gain, ensuring fairness, and evaluating uncertainty in prediction models, including large language models. Theoretical verification shows that AP-Calculus not only extends but can also subsume traditional do-calculus for many practical applications, offering a more direct approach to causal inference in supervised learning contexts.", "date": "2025-05-17", "categories": []}
{"id": "2505.12868", "title": "Causality-Inspired Robustness for Nonlinear Models via Representation Learning", "abstract": "Distributional robustness is a central goal of prediction algorithms due to the prevalent distribution shifts in real-world data. The prediction model aims to minimize the worst-case risk among a class of distributions, a.k.a., an uncertainty set. Causality provides a modeling framework with a rigorous robustness guarantee in the above sense, where the uncertainty set is data-driven rather than pre-specified as in traditional distributional robustness optimization. However, current causality-inspired robustness methods possess finite-radius robustness guarantees only in the linear settings, where the causal relationships among the covariates and the response are linear. In this work, we propose a nonlinear method under a causal framework by incorporating recent developments in identifiable representation learning and establish a distributional robustness guarantee. To our best knowledge, this is the first causality-inspired robustness method with such a finite-radius robustness guarantee in nonlinear settings. Empirical validation of the theoretical findings is conducted on both synthetic data and real-world single-cell data, also illustrating that finite-radius robustness is crucial.", "date": "2025-05-19", "categories": []}
{"id": "2505.13072", "title": "Orthogonal Survival Learners for Estimating Heterogeneous Treatment Effects from Time-to-Event Data", "abstract": "Estimating heterogeneous treatment effects (HTEs) is crucial for personalized decision-making. However, this task is challenging in survival analysis, which includes time-to-event data with censored outcomes (e.g., due to study dropout). In this paper, we propose a toolbox of novel orthogonal survival learners to estimate HTEs from time-to-event data under censoring. Our learners have three main advantages: (i) we show that learners from our toolbox are guaranteed to be orthogonal and thus come with favorable theoretical properties; (ii) our toolbox allows for incorporating a custom weighting function, which can lead to robustness against different types of low overlap, and (iii) our learners are model-agnostic (i.e., they can be combined with arbitrary machine learning models). We instantiate the learners from our toolbox using several weighting functions and, as a result, propose various neural orthogonal survival learners. Some of these coincide with existing survival learners (including survival versions of the DR- and R-learner), while others are novel and further robust w.r.t. low overlap regimes specific to the survival setting (i.e., survival overlap and censoring overlap). We then empirically verify the effectiveness of our learners for HTE estimation in different low-overlap regimes through numerical experiments. In sum, we provide practitioners with a large toolbox of learners that can be used for randomized and observational studies with censored time-to-event data.", "date": "2025-05-19", "categories": []}
{"id": "2505.13324", "title": "From What Ifs to Insights: Counterfactuals in Causal Inference vs. Explainable AI", "abstract": "Counterfactuals play a pivotal role in the two distinct data science fields of causal inference (CI) and explainable artificial intelligence (XAI). While the core idea behind counterfactuals remains the same in both fields--the examination of what would have happened under different circumstances--there are key differences in how they are used and interpreted. We introduce a formal definition that encompasses the multi-faceted concept of the counterfactual in CI and XAI. We then discuss how counterfactuals are used, evaluated, generated, and operationalized in CI vs. XAI, highlighting conceptual and practical differences. By comparing and contrasting the two, we hope to identify opportunities for cross-fertilization across CI and XAI.", "date": "2025-05-19", "categories": []}
{"id": "2401.14512", "title": "Who Are We Missing? A Principled Approach to Characterizing the Underrepresented Population", "abstract": "Randomized controlled trials (RCTs) serve as the cornerstone for understanding causal effects, yet extending inferences to target populations presents challenges due to effect heterogeneity and underrepresentation. Our paper addresses the critical issue of identifying and characterizing underrepresented subgroups in RCTs, proposing a novel framework for refining target populations to improve generalizability. We introduce an optimization-based approach, Rashomon Set of Optimal Trees (ROOT), to characterize underrepresented groups. ROOT optimizes the target subpopulation distribution by minimizing the variance of the target average treatment effect estimate, ensuring more precise treatment effect estimations. Notably, ROOT generates interpretable characteristics of the underrepresented population, aiding researchers in effective communication. Our approach demonstrates improved precision and interpretability compared to alternatives, as illustrated with synthetic data experiments. We apply our methodology to extend inferences from the Starting Treatment with Agonist Replacement Therapies (START) trial -- investigating the effectiveness of medication for opioid use disorder -- to the real-world population represented by the Treatment Episode Dataset: Admissions (TEDS-A). By refining target populations using ROOT, our framework offers a systematic approach to enhance decision-making accuracy and inform future trials in diverse populations.", "date": "2024-01-25", "categories": []}
{"id": "2505.11014", "title": "A Cautionary Tale on Integrating Studies with Disparate Outcome Measures for Causal Inference", "abstract": "Data integration approaches are increasingly used to enhance the efficiency and generalizability of studies. However, a key limitation of these methods is the assumption that outcome measures are identical across datasets -- an assumption that often does not hold in practice. Consider the following opioid use disorder (OUD) studies: the XBOT trial and the POAT study, both evaluating the effect of medications for OUD on withdrawal symptom severity (not the primary outcome of either trial). While XBOT measures withdrawal severity using the subjective opiate withdrawal scale, POAT uses the clinical opiate withdrawal scale. We analyze this realistic yet challenging setting where outcome measures differ across studies and where neither study records both types of outcomes. Our paper studies whether and when integrating studies with disparate outcome measures leads to efficiency gains. We introduce three sets of assumptions -- with varying degrees of strength -- linking both outcome measures. Our theoretical and empirical results highlight a cautionary tale: integration can improve asymptotic efficiency only under the strongest assumption linking the outcomes. However, misspecification of this assumption leads to bias. In contrast, a milder assumption may yield finite-sample efficiency gains, yet these benefits diminish as sample size increases. We illustrate these trade-offs via a case study integrating the XBOT and POAT datasets to estimate the comparative effect of two medications for opioid use disorder on withdrawal symptoms. By systematically varying the assumptions linking the SOW and COW scales, we show potential efficiency gains and the risks of bias. Our findings emphasize the need for careful assumption selection when fusing datasets with differing outcome measures, offering guidance for researchers navigating this common challenge in modern data integration.", "date": "2025-05-16", "categories": []}
{"id": "2505.11444", "title": "A Generative Framework for Causal Estimation via Importance-Weighted Diffusion Distillation", "abstract": "Estimating individualized treatment effects from observational data is a central challenge in causal inference, largely due to covariate imbalance and confounding bias from non-randomized treatment assignment. While inverse probability weighting (IPW) is a well-established solution to this problem, its integration into modern deep learning frameworks remains limited. In this work, we propose Importance-Weighted Diffusion Distillation (IWDD), a novel generative framework that combines the pretraining of diffusion models with importance-weighted score distillation to enable accurate and fast causal estimation-including potential outcome prediction and treatment effect estimation. We demonstrate how IPW can be naturally incorporated into the distillation of pretrained diffusion models, and further introduce a randomization-based adjustment that eliminates the need to compute IPW explicitly-thereby simplifying computation and, more importantly, provably reducing the variance of gradient estimates. Empirical results show that IWDD achieves state-of-the-art out-of-sample prediction performance, with the highest win rates compared to other baselines, significantly improving causal estimation and supporting the development of individualized treatment strategies. We will release our PyTorch code for reproducibility and future research.", "date": "2025-05-16", "categories": []}
{"id": "2502.11331", "title": "Transfer Learning of CATE with Kernel Ridge Regression", "abstract": "The proliferation of data has sparked significant interest in leveraging findings from one study to estimate treatment effects in a different target population without direct outcome observations. However, the transfer learning process is frequently hindered by substantial covariate shift and limited overlap between (i) the source and target populations, as well as (ii) the treatment and control groups within the source. We propose a novel method for overlap-adaptive transfer learning of conditional average treatment effect (CATE) using kernel ridge regression (KRR). Our approach involves partitioning the labeled source data into two subsets. The first one is used to train candidate CATE models based on regression adjustment and pseudo-outcomes. An optimal model is then selected using the second subset and unlabeled target data, employing another pseudo-outcome-based strategy. We provide a theoretical justification for our method through sharp non-asymptotic MSE bounds, highlighting its adaptivity to both weak overlaps and the complexity of CATE function. Extensive numerical studies confirm that our method achieves superior finite-sample efficiency and adaptability. We conclude by demonstrating the effectiveness of our approach using a 401(k) eligibility dataset.", "date": "2025-02-17", "categories": []}
{"id": "2505.09113", "title": "Sequential Treatment Effect Estimation with Unmeasured Confounders", "abstract": "This paper studies the cumulative causal effects of sequential treatments in the presence of unmeasured confounders. It is a critical issue in sequential decision-making scenarios where treatment decisions and outcomes dynamically evolve over time. Advanced causal methods apply transformer as a backbone to model such time sequences, which shows superiority in capturing long time dependence and periodic patterns via attention mechanism. However, even they control the observed confounding, these estimators still suffer from unmeasured confounders, which influence both treatment assignments and outcomes. How to adjust the latent confounding bias in sequential treatment effect estimation remains an open challenge. Therefore, we propose a novel Decomposing Sequential Instrumental Variable framework for CounterFactual Regression (DSIV-CFR), relying on a common negative control assumption. Specifically, an instrumental variable (IV) is a special negative control exposure, while the previous outcome serves as a negative control outcome. This allows us to recover the IVs latent in observation variables and estimate sequential treatment effects via a generalized moment condition. We conducted experiments on 4 datasets and achieved significant performance in one- and multi-step prediction, supported by which we can identify optimal treatments for dynamic systems.", "date": "2025-05-14", "categories": []}
{"id": "2505.09412", "title": "Counterfactual Strategies for Markov Decision Processes", "abstract": "Counterfactuals are widely used in AI to explain how minimal changes to a model's input can lead to a different output. However, established methods for computing counterfactuals typically focus on one-step decision-making, and are not directly applicable to sequential decision-making tasks. This paper fills this gap by introducing counterfactual strategies for Markov Decision Processes (MDPs). During MDP execution, a strategy decides which of the enabled actions (with known probabilistic effects) to execute next. Given an initial strategy that reaches an undesired outcome with a probability above some limit, we identify minimal changes to the initial strategy to reduce that probability below the limit. We encode such counterfactual strategies as solutions to non-linear optimization problems, and further extend our encoding to synthesize diverse counterfactual strategies. We evaluate our approach on four real-world datasets and demonstrate its practical viability in sophisticated sequential decision-making tasks.", "date": "2025-05-14", "categories": []}
{"id": "2505.08343", "title": "An Identifiable Cost-Aware Causal Decision-Making Framework Using Counterfactual Reasoning", "abstract": "Decision making under abnormal conditions is a critical process that involves evaluating the current state and determining the optimal action to restore the system to a normal state at an acceptable cost. However, in such scenarios, existing decision-making frameworks highly rely on reinforcement learning or root cause analysis, resulting in them frequently neglecting the cost of the actions or failing to incorporate causal mechanisms adequately. By relaxing the existing causal decision framework to solve the necessary cause, we propose a minimum-cost causal decision (MiCCD) framework via counterfactual reasoning to address the above challenges. Emphasis is placed on making counterfactual reasoning processes identifiable in the presence of a large amount of mixed anomaly data, as well as finding the optimal intervention state in a continuous decision space. Specifically, it formulates a surrogate model based on causal graphs, using abnormal pattern clustering labels as supervisory signals. This enables the approximation of the structural causal model among the variables and lays a foundation for identifiable counterfactual reasoning. With the causal structure approximated, we then established an optimization model based on counterfactual estimation. The Sequential Least Squares Programming (SLSQP) algorithm is further employed to optimize intervention strategies while taking costs into account. Experimental evaluations on both synthetic and real-world datasets reveal that MiCCD outperforms conventional methods across multiple metrics, including F1-score, cost efficiency, and ranking quality(nDCG@k values), thus validating its efficacy and broad applicability.", "date": "2025-05-13", "categories": []}
{"id": "2505.08378", "title": "Learning Treatment Allocations with Risk Control Under Partial Identifiability", "abstract": "Learning beneficial treatment allocations for a patient population is an important problem in precision medicine. Many treatments come with adverse side effects that are not commensurable with their potential benefits. Patients who do not receive benefits after such treatments are thereby subjected to unnecessary harm. This is a `treatment risk' that we aim to control when learning beneficial allocations. The constrained learning problem is challenged by the fact that the treatment risk is not in general identifiable using either randomized trial or observational data. We propose a certifiable learning method that controls the treatment risk with finite samples in the partially identified setting. The method is illustrated using both simulated and real data.", "date": "2025-05-13", "categories": []}
{"id": "2505.08453", "title": "Parameter Estimation using Reinforcement Learning Causal Curiosity: Limits and Challenges", "abstract": "Causal understanding is important in many disciplines of science and engineering, where we seek to understand how different factors in the system causally affect an experiment or situation and pave a pathway towards creating effective or optimising existing models. Examples of use cases are autonomous exploration and modelling of unknown environments or assessing key variables in optimising large complex systems. In this paper, we analyse a Reinforcement Learning approach called Causal Curiosity, which aims to estimate as accurately and efficiently as possible, without directly measuring them, the value of factors that causally determine the dynamics of a system. Whilst the idea presents a pathway forward, measurement accuracy is the foundation of methodology effectiveness. Focusing on the current causal curiosity's robotic manipulator, we present for the first time a measurement accuracy analysis of the future potentials and current limitations of this technique and an analysis of its sensitivity and confounding factor disentanglement capability - crucial for causal analysis. As a result of our work, we promote proposals for an improved and efficient design of Causal Curiosity methods to be applied to real-world complex scenarios.", "date": "2025-05-13", "categories": []}
{"id": "2402.01454", "title": "Integrating Large Language Models in Causal Discovery: A Statistical Causal Approach", "abstract": "In practical statistical causal discovery (SCD), embedding domain expert knowledge as constraints into the algorithm is important for reasonable causal models reflecting the broad knowledge of domain experts, despite the challenges in the systematic acquisition of background knowledge. To overcome these challenges, this paper proposes a novel method for causal inference, in which SCD and knowledge-based causal inference (KBCI) with a large language model (LLM) are synthesized through ``statistical causal prompting (SCP)'' for LLMs and prior knowledge augmentation for SCD. The experiments in this work have revealed that the results of LLM-KBCI and SCD augmented with LLM-KBCI approach the ground truths, more than the SCD result without prior knowledge. These experiments have also revealed that the SCD result can be further improved if the LLM undergoes SCP. Furthermore, with an unpublished real-world dataset, we have demonstrated that the background knowledge provided by the LLM can improve the SCD on this dataset, even if this dataset has never been included in the training data of the LLM. For future practical application of this proposed method across important domains such as healthcare, we also thoroughly discuss the limitations, risks of critical errors, expected improvement of techniques around LLMs, and realistic integration of expert checks of the results into this automatic process, with SCP simulations under various conditions both in successful and failure scenarios. The careful and appropriate application of the proposed approach in this work, with improvement and customization for each domain, can thus address challenges such as dataset biases and limitations, illustrating the potential of LLMs to improve data-driven causal inference across diverse scientific domains.\n  The code used in this work is publicly available at: www.github.com/mas-takayama/LLM-and-SCD", "date": "2024-02-02", "categories": []}
{"id": "2412.19507", "title": "Hybrid Local Causal Discovery", "abstract": "Local causal discovery aims to learn and distinguish the direct causes and effects of a target variable from observed data. Existing constraint-based local causal discovery methods use AND or OR rules in constructing the local causal skeleton, but using either rule alone is prone to produce cascading errors in the learned local causal skeleton, and thus impacting the inference of local causal relationships. On the other hand, directly applying score-based global causal discovery methods to local causal discovery may randomly return incorrect results due to the existence of local equivalence classes. To address the above issues, we propose a Hybrid Local Causal Discovery algorithm, called HLCD. Specifically, HLCD initially utilizes a constraint-based approach combined with the OR rule to obtain a candidate skeleton and then employs a score-based method to eliminate redundant portions in the candidate skeleton. Furthermore, during the local causal orientation phase, HLCD distinguishes between V-structures and equivalence classes by comparing the local structure scores between the two, thereby avoiding orientation interference caused by local equivalence classes. We conducted extensive experiments with seven state-of-the-art competitors on 14 benchmark Bayesian network datasets, and the experimental results demonstrate that HLCD significantly outperforms existing local causal discovery algorithms.", "date": "2024-12-27", "categories": []}
{"id": "2505.06367", "title": "CAST: Time-Varying Treatment Effects with Application to Chemotherapy and Radiotherapy on Head and Neck Squamous Cell Carcinoma", "abstract": "Causal machine learning (CML) enables individualized estimation of treatment effects, offering critical advantages over traditional correlation-based methods. However, existing approaches for medical survival data with censoring such as causal survival forests estimate effects at fixed time points, limiting their ability to capture dynamic changes over time. We introduce Causal Analysis for Survival Trajectories (CAST), a novel framework that models treatment effects as continuous functions of time following treatment. By combining parametric and non-parametric methods, CAST overcomes the limitations of discrete time-point analysis to estimate continuous effect trajectories. Using the RADCURE dataset [1] of 2,651 patients with head and neck squamous cell carcinoma (HNSCC) as a clinically relevant example, CAST models how chemotherapy and radiotherapy effects evolve over time at the population and individual levels. By capturing the temporal dynamics of treatment response, CAST reveals how treatment effects rise, peak, and decline over the follow-up period, helping clinicians determine when and for whom treatment benefits are maximized. This framework advances the application of CML to personalized care in HNSCC and other life-threatening medical conditions. Source code/data available at: https://github.com/CAST-FW/HNSCC", "date": "2025-05-09", "categories": []}
{"id": "2505.06542", "title": "dcFCI: Robust Causal Discovery Under Latent Confounding, Unfaithfulness, and Mixed Data", "abstract": "Causal discovery is central to inferring causal relationships from observational data. In the presence of latent confounding, algorithms such as Fast Causal Inference (FCI) learn a Partial Ancestral Graph (PAG) representing the true model's Markov Equivalence Class. However, their correctness critically depends on empirical faithfulness, the assumption that observed (in)dependencies perfectly reflect those of the underlying causal model, which often fails in practice due to limited sample sizes. To address this, we introduce the first nonparametric score to assess a PAG's compatibility with observed data, even with mixed variable types. This score is both necessary and sufficient to characterize structural uncertainty and distinguish between distinct PAGs. We then propose data-compatible FCI (dcFCI), the first hybrid causal discovery algorithm to jointly address latent confounding, empirical unfaithfulness, and mixed data types. dcFCI integrates our score into an (Anytime)FCI-guided search that systematically explores, ranks, and validates candidate PAGs. Experiments on synthetic and real-world scenarios demonstrate that dcFCI significantly outperforms state-of-the-art methods, often recovering the true PAG even in small and heterogeneous datasets. Examining top-ranked PAGs further provides valuable insights into structural uncertainty, supporting more robust and informed causal reasoning and decision-making.", "date": "2025-05-10", "categories": []}
{"id": "2505.06856", "title": "Beyond Patterns: Harnessing Causal Logic for Autonomous Driving Trajectory Prediction", "abstract": "Accurate trajectory prediction has long been a major challenge for autonomous driving (AD). Traditional data-driven models predominantly rely on statistical correlations, often overlooking the causal relationships that govern traffic behavior. In this paper, we introduce a novel trajectory prediction framework that leverages causal inference to enhance predictive robustness, generalization, and accuracy. By decomposing the environment into spatial and temporal components, our approach identifies and mitigates spurious correlations, uncovering genuine causal relationships. We also employ a progressive fusion strategy to integrate multimodal information, simulating human-like reasoning processes and enabling real-time inference. Evaluations on five real-world datasets--ApolloScape, nuScenes, NGSIM, HighD, and MoCAD--demonstrate our model's superiority over existing state-of-the-art (SOTA) methods, with improvements in key metrics such as RMSE and FDE. Our findings highlight the potential of causal reasoning to transform trajectory prediction, paving the way for robust AD systems.", "date": "2025-05-11", "categories": []}
{"id": "2505.06949", "title": "Causal knowledge graph analysis identifies adverse drug effects", "abstract": "Knowledge graphs and structural causal models have each proven valuable for organizing biomedical knowledge and estimating causal effects, but remain largely disconnected: knowledge graphs encode qualitative relationships focusing on facts and deductive reasoning without formal probabilistic semantics, while causal models lack integration with background knowledge in knowledge graphs and have no access to the deductive reasoning capabilities that knowledge graphs provide. To bridge this gap, we introduce a novel formulation of Causal Knowledge Graphs (CKGs) which extend knowledge graphs with formal causal semantics, preserving their deductive capabilities while enabling principled causal inference. CKGs support deconfounding via explicitly marked causal edges and facilitate hypothesis formulation aligned with both encoded and entailed background knowledge. We constructed a Drug-Disease CKG (DD-CKG) integrating disease progression pathways, drug indications, side-effects, and hierarchical disease classification to enable automated large-scale mediation analysis. Applied to UK Biobank and MIMIC-IV cohorts, we tested whether drugs mediate effects between indications and downstream disease progression, adjusting for confounders inferred from the DD-CKG. Our approach successfully reproduced known adverse drug reactions with high precision while identifying previously undocumented significant candidate adverse effects. Further validation through side effect similarity analysis demonstrated that combining our predicted drug effects with established databases significantly improves the prediction of shared drug indications, supporting the clinical relevance of our novel findings. These results demonstrate that our methodology provides a generalizable, knowledge-driven framework for scalable causal inference.", "date": "2025-05-11", "categories": []}
{"id": "2301.04709", "title": "Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability", "abstract": "Causal abstraction provides a theoretical foundation for mechanistic interpretability, the field concerned with providing intelligible algorithms that are faithful simplifications of the known, but opaque low-level details of black box AI models. Our contributions are (1) generalizing the theory of causal abstraction from mechanism replacement (i.e., hard and soft interventions) to arbitrary mechanism transformation (i.e., functionals from old mechanisms to new mechanisms), (2) providing a flexible, yet precise formalization for the core concepts of polysemantic neurons, the linear representation hypothesis, modular features, and graded faithfulness, and (3) unifying a variety of mechanistic interpretability methods in the common language of causal abstraction, namely, activation and path patching, causal mediation analysis, causal scrubbing, causal tracing, circuit analysis, concept erasure, sparse autoencoders, differential binary masking, distributed alignment search, and steering.", "date": "2023-01-11", "categories": []}
{"id": "2504.03122", "title": "From Observation to Orientation: an Adaptive Integer Programming Approach to Intervention Design", "abstract": "Using both observational and experimental data, a causal discovery process can identify the causal relationships between variables. A unique adaptive intervention design paradigm is presented in this work, where causal directed acyclic graphs (DAGs) are for effectively recovered with practical budgetary considerations. In order to choose treatments that optimize information gain under these considerations, an iterative integer programming (IP) approach is proposed, which drastically reduces the number of experiments required. Simulations over a broad range of graph sizes and edge densities are used to assess the effectiveness of the suggested approach. Results show that the proposed adaptive IP approach achieves full causal graph recovery with fewer intervention iterations and variable manipulations than random intervention baselines, and it is also flexible enough to accommodate a variety of practical constraints.", "date": "2025-04-04", "categories": []}
{"id": "2505.05192", "title": "Long-Term Individual Causal Effect Estimation via Identifiable Latent Representation Learning", "abstract": "Estimating long-term causal effects by combining long-term observational and short-term experimental data is a crucial but challenging problem in many real-world scenarios. In existing methods, several ideal assumptions, e.g. latent unconfoundedness assumption or additive equi-confounding bias assumption, are proposed to address the latent confounder problem raised by the observational data. However, in real-world applications, these assumptions are typically violated which limits their practical effectiveness. In this paper, we tackle the problem of estimating the long-term individual causal effects without the aforementioned assumptions. Specifically, we propose to utilize the natural heterogeneity of data, such as data from multiple sources, to identify latent confounders, thereby significantly avoiding reliance on idealized assumptions. Practically, we devise a latent representation learning-based estimator of long-term causal effects. Theoretically, we establish the identifiability of latent confounders, with which we further achieve long-term effect identification. Extensive experimental studies, conducted on multiple synthetic and semi-synthetic datasets, demonstrate the effectiveness of our proposed method.", "date": "2025-05-08", "categories": []}
{"id": "2505.04200", "title": "Estimating Causal Effects in Networks with Cluster-Based Bandits", "abstract": "  The gold standard for estimating causal effects is randomized controlled\ntrial (RCT) or A/B testing where a random group of individuals from a\npopulation of interest are given treatment and the outcome is compared to a\nrandom group of individuals from the same population. However, A/B testing is\nchallenging in the presence of interference, commonly occurring in social\nnetworks, where individuals can impact each others outcome. Moreover, A/B\ntesting can incur a high performance loss when one of the treatment arms has a\npoor performance and the test continues to treat individuals with it.\nTherefore, it is important to design a strategy that can adapt over time and\nefficiently learn the total treatment effect in the network. We introduce two\ncluster-based multi-armed bandit (MAB) algorithms to gradually estimate the\ntotal treatment effect in a network while maximizing the expected reward by\nmaking a tradeoff between exploration and exploitation. We compare the\nperformance of our MAB algorithms with a vanilla MAB algorithm that ignores\nclusters and the corresponding RCT methods on semi-synthetic data with\nsimulated interference. The vanilla MAB algorithm shows higher reward-action\nratio at the cost of higher treatment effect error due to undesired spillover.\nThe cluster-based MAB algorithms show higher reward-action ratio compared to\ntheir corresponding RCT methods without sacrificing much accuracy in treatment\neffect estimation.\n", "date": "2025-05-07", "categories": []}
{"id": "2505.04971", "title": "Moments of Causal Effects", "abstract": "  The moments of random variables are fundamental statistical measures for\ncharacterizing the shape of a probability distribution, encompassing metrics\nsuch as mean, variance, skewness, and kurtosis. Additionally, the product\nmoments, including covariance and correlation, reveal the relationships between\nmultiple random variables. On the other hand, the primary focus of causal\ninference is the evaluation of causal effects, which are defined as the\ndifference between two potential outcomes. While traditional causal effect\nassessment focuses on the average causal effect, this work provides\ndefinitions, identification theorems, and bounds for moments and product\nmoments of causal effects to analyze their distribution and relationships. We\nconduct experiments to illustrate the estimation of the moments of causal\neffects from finite samples and demonstrate their practical application using a\nreal-world medical dataset.\n", "date": "2025-05-08", "categories": []}
{"id": "2505.04983", "title": "Decomposition of Probabilities of Causation with Two Mediators", "abstract": "  Mediation analysis for probabilities of causation (PoC) provides a\nfundamental framework for evaluating the necessity and sufficiency of treatment\nin provoking an event through different causal pathways. One of the primary\nobjectives of causal mediation analysis is to decompose the total effect into\npath-specific components. In this study, we investigate the path-specific\nprobability of necessity and sufficiency (PNS) to decompose the total PNS into\npath-specific components along distinct causal pathways between treatment and\noutcome, incorporating two mediators. We define the path-specific PNS for\ndecomposition and provide an identification theorem. Furthermore, we conduct\nnumerical experiments to assess the properties of the proposed estimators from\nfinite samples and demonstrate their practical application using a real-world\neducational dataset.\n", "date": "2025-05-08", "categories": []}
{"id": "2505.05242", "title": "Enhancing Treatment Effect Estimation via Active Learning: A\n  Counterfactual Covering Perspective", "abstract": "  Although numerous complex algorithms for treatment effect estimation have\nbeen developed in recent years, their effectiveness remains limited when\nhandling insufficiently labeled training sets due to the high cost of labeling\nthe effect after treatment, e.g., expensive tumor imaging or biopsy procedures\nneeded to evaluate treatment effects. Therefore, it becomes essential to\nactively incorporate more high-quality labeled data, all while adhering to a\nconstrained labeling budget. To enable data-efficient treatment effect\nestimation, we formalize the problem through rigorous theoretical analysis\nwithin the active learning context, where the derived key measures --\n\\textit{factual} and \\textit{counterfactual covering radius} determine the risk\nupper bound. To reduce the bound, we propose a greedy radius reduction\nalgorithm, which excels under an idealized, balanced data distribution. To\ngeneralize to more realistic data distributions, we further propose FCCM, which\ntransforms the optimization objective into the \\textit{Factual} and\n\\textit{Counterfactual Coverage Maximization} to ensure effective radius\nreduction during data acquisition. Furthermore, benchmarking FCCM against other\nbaselines demonstrates its superiority across both fully synthetic and\nsemi-synthetic datasets.\n", "date": "2025-05-08", "categories": []}
{"id": "2310.18286", "title": "Optimal Transport for Treatment Effect Estimation", "abstract": "  Estimating conditional average treatment effect from observational data is\nhighly challenging due to the existence of treatment selection bias. Prevalent\nmethods mitigate this issue by aligning distributions of different treatment\ngroups in the latent space. However, there are two critical problems that these\nmethods fail to address: (1) mini-batch sampling effects (MSE), which causes\nmisalignment in non-ideal mini-batches with outcome imbalance and outliers; (2)\nunobserved confounder effects (UCE), which results in inaccurate discrepancy\ncalculation due to the neglect of unobserved confounders. To tackle these\nproblems, we propose a principled approach named Entire Space CounterFactual\nRegression (ESCFR), which is a new take on optimal transport in the context of\ncausality. Specifically, based on the framework of stochastic optimal\ntransport, we propose a relaxed mass-preserving regularizer to address the MSE\nissue and design a proximal factual outcome regularizer to handle the UCE\nissue. Extensive experiments demonstrate that our proposed ESCFR can\nsuccessfully tackle the treatment selection bias and achieve significantly\nbetter performance than state-of-the-art methods.\n", "date": "2023-10-27", "categories": []}
{"id": "2311.03382", "title": "Causal Structure Representation Learning of Confounders in Latent Space\n  for Recommendation", "abstract": "  Inferring user preferences from the historical feedback of users is a\nvaluable problem in recommender systems. Conventional approaches often rely on\nthe assumption that user preferences in the feedback data are equivalent to the\nreal user preferences without additional noise, which simplifies the problem\nmodeling. However, there are various confounders during user-item interactions,\nsuch as weather and even the recommendation system itself. Therefore,\nneglecting the influence of confounders will result in inaccurate user\npreferences and suboptimal performance of the model. Furthermore, the\nunobservability of confounders poses a challenge in further addressing the\nproblem. To address these issues, we refine the problem and propose a more\nrational solution. Specifically, we consider the influence of confounders,\ndisentangle them from user preferences in the latent space, and employ causal\ngraphs to model their interdependencies without specific labels. By cleverly\ncombining local and global causal graphs, we capture the user-specificity of\nconfounders on user preferences. We theoretically demonstrate the\nidentifiability of the obtained causal graph. Finally, we propose our model\nbased on Variational Autoencoders, named Causal Structure representation\nlearning of Confounders in latent space (CSC). We conducted extensive\nexperiments on one synthetic dataset and five real-world datasets,\ndemonstrating the superiority of our model. Furthermore, we demonstrate that\nthe learned causal representations of confounders are controllable, potentially\noffering users fine-grained control over the objectives of their recommendation\nlists with the learned causal graphs.\n", "date": "2023-11-02", "categories": []}
{"id": "2409.05459", "title": "Beyond Flatland: A Geometric Take on Matching Methods for Treatment\n  Effect Estimation", "abstract": "  Matching is a popular approach in causal inference to estimate treatment\neffects by pairing treated and control units that are most similar in terms of\ntheir covariate information. However, classic matching methods completely\nignore the geometry of the data manifold, which is crucial to define a\nmeaningful distance for matching, and struggle when covariates are noisy and\nhigh-dimensional. In this work, we propose GeoMatching, a matching method to\nestimate treatment effects that takes into account the intrinsic data geometry\ninduced by existing causal mechanisms among the confounding variables. First,\nwe learn a low-dimensional, latent Riemannian manifold that accounts for\nuncertainty and geometry of the original input data. Second, we estimate\ntreatment effects via matching in the latent space based on the learned latent\nRiemannian metric. We provide theoretical insights and empirical results in\nsynthetic and real-world scenarios, demonstrating that GeoMatching yields more\neffective treatment effect estimators, even as we increase input\ndimensionality, in the presence of outliers, or in semi-supervised scenarios.\n", "date": "2024-09-09", "categories": []}
{"id": "2505.03655", "title": "Counterfactual Inference for Eliminating Sentiment Bias in Recommender\n  Systems", "abstract": "  Recommender Systems (RSs) aim to provide personalized recommendations for\nusers. A newly discovered bias, known as sentiment bias, uncovers a common\nphenomenon within Review-based RSs (RRSs): the recommendation accuracy of users\nor items with negative reviews deteriorates compared with users or items with\npositive reviews. Critical users and niche items are disadvantaged by such\nunfair recommendations. We study this problem from the perspective of\ncounterfactual inference with two stages. At the model training stage, we build\na causal graph and model how sentiment influences the final rating score.\nDuring the inference stage, we decouple the direct and indirect effects to\nmitigate the impact of sentiment bias and remove the indirect effect using\ncounterfactual inference. We have conducted extensive experiments, and the\nresults validate that our model can achieve comparable performance on rating\nprediction for better recommendations and effective mitigation of sentiment\nbias. To the best of our knowledge, this is the first work to employ\ncounterfactual inference on sentiment bias mitigation in RSs.\n", "date": "2025-05-06", "categories": []}
{"id": "2405.01463", "title": "Dynamic Local Average Treatment Effects", "abstract": "  We consider Dynamic Treatment Regimes (DTRs) with One Sided Noncompliance\nthat arise in applications such as digital recommendations and adaptive medical\ntrials. These are settings where decision makers encourage individuals to take\ntreatments over time, but adapt encouragements based on previous\nencouragements, treatments, states, and outcomes. Importantly, individuals may\nnot comply with encouragements based on unobserved confounders. For settings\nwith binary treatments and encouragements, we provide nonparametric\nidentification, estimation, and inference for Dynamic Local Average Treatment\nEffects (LATEs), which are expected values of multiple time period treatment\neffect contrasts for the respective complier subpopulations. Under One Sided\nNoncompliance and sequential extensions of the assumptions in Imbens and\nAngrist (1994), we show that one can identify Dynamic LATEs that correspond to\ntreating at single time steps. In Staggered Adoption settings, we show that the\nassumptions are sufficient to identify Dynamic LATEs for treating in multiple\ntime periods. Moreover, this result extends to any setting where the effect of\na treatment in one period is uncorrelated with the compliance event in a\nsubsequent period.\n", "date": "2024-05-02", "categories": []}
{"id": "2505.01785", "title": "TV-SurvCaus: Dynamic Representation Balancing for Causal Survival\n  Analysis", "abstract": "  Estimating the causal effect of time-varying treatments on survival outcomes\nis a challenging task in many domains, particularly in medicine where treatment\nprotocols adapt over time. While recent advances in representation learning\nhave improved causal inference for static treatments, extending these methods\nto dynamic treatment regimes with survival outcomes remains under-explored. In\nthis paper, we introduce TV-SurvCaus, a novel framework that extends\nrepresentation balancing techniques to the time-varying treatment setting for\nsurvival analysis. We provide theoretical guarantees through (1) a generalized\nbound for time-varying precision in estimation of heterogeneous effects, (2)\nvariance control via sequential balancing weights, (3) consistency results for\ndynamic treatment regimes, (4) convergence rates for representation learning\nwith temporal dependencies, and (5) a formal bound on the bias due to\ntreatment-confounder feedback. Our neural architecture incorporates sequence\nmodeling to handle temporal dependencies while balancing time-dependent\nrepresentations. Through extensive experiments on both synthetic and real-world\ndatasets, we demonstrate that TV-SurvCaus outperforms existing methods in\nestimating individualized treatment effects with time-varying covariates and\ntreatments. Our framework advances the field of causal inference by enabling\nmore accurate estimation of treatment effects in dynamic, longitudinal settings\nwith survival outcomes.\n", "date": "2025-05-03", "categories": []}
{"id": "2505.01995", "title": "Extended Fiducial Inference for Individual Treatment Effects via Deep\n  Neural Networks", "abstract": "  Individual treatment effect estimation has gained significant attention in\nrecent data science literature. This work introduces the Double Neural Network\n(Double-NN) method to address this problem within the framework of extended\nfiducial inference (EFI). In the proposed method, deep neural networks are used\nto model the treatment and control effect functions, while an additional neural\nnetwork is employed to estimate their parameters. The universal approximation\ncapability of deep neural networks ensures the broad applicability of this\nmethod. Numerical results highlight the superior performance of the proposed\nDouble-NN method compared to the conformal quantile regression (CQR) method in\nindividual treatment effect estimation. From the perspective of statistical\ninference, this work advances the theory and methodology for statistical\ninference of large models. Specifically, it is theoretically proven that the\nproposed method permits the model size to increase with the sample size $n$ at\na rate of $O(n^{\\zeta})$ for some $0 \\leq \\zeta<1$, while still maintaining\nproper quantification of uncertainty in the model parameters. This result marks\na significant improvement compared to the range $0\\leq \\zeta < \\frac{1}{2}$\nrequired by the classical central limit theorem. Furthermore, this work\nprovides a rigorous framework for quantifying the uncertainty of deep neural\nnetworks under the neural scaling law, representing a substantial contribution\nto the statistical understanding of large-scale neural network models.\n", "date": "2025-05-04", "categories": []}
{"id": "2505.02074", "title": "Learning Local Causal World Models with State Space Models and Attention", "abstract": "  World modelling, i.e. building a representation of the rules that govern the\nworld so as to predict its evolution, is an essential ability for any agent\ninteracting with the physical world. Despite their impressive performance, many\nsolutions fail to learn a causal representation of the environment they are\ntrying to model, which would be necessary to gain a deep enough understanding\nof the world to perform complex tasks. With this work, we aim to broaden the\nresearch in the intersection of causality theory and neural world modelling by\nassessing the potential for causal discovery of the State Space Model (SSM)\narchitecture, which has been shown to have several advantages over the\nwidespread Transformer. We show empirically that, compared to an equivalent\nTransformer, a SSM can model the dynamics of a simple environment and learn a\ncausal model at the same time with equivalent or better performance, thus\npaving the way for further experiments that lean into the strength of SSMs and\nfurther enhance them with causal awareness.\n", "date": "2025-05-04", "categories": []}
{"id": "2505.02212", "title": "Exogenous Isomorphism for Counterfactual Identifiability", "abstract": "  This paper investigates $\\sim_{\\mathcal{L}_3}$-identifiability, a form of\ncomplete counterfactual identifiability within the Pearl Causal Hierarchy (PCH)\nframework, ensuring that all Structural Causal Models (SCMs) satisfying the\ngiven assumptions provide consistent answers to all causal questions. To\nsimplify this problem, we introduce exogenous isomorphism and propose\n$\\sim_{\\mathrm{EI}}$-identifiability, reflecting the strength of model\nidentifiability required for $\\sim_{\\mathcal{L}_3}$-identifiability. We explore\nsufficient assumptions for achieving $\\sim_{\\mathrm{EI}}$-identifiability in\ntwo special classes of SCMs: Bijective SCMs (BSCMs), based on counterfactual\ntransport, and Triangular Monotonic SCMs (TM-SCMs), which extend\n$\\sim_{\\mathcal{L}_2}$-identifiability. Our results unify and generalize\nexisting theories, providing theoretical guarantees for practical applications.\nFinally, we leverage neural TM-SCMs to address the consistency problem in\ncounterfactual reasoning, with experiments validating both the effectiveness of\nour method and the correctness of the theory.\n", "date": "2025-05-04", "categories": []}
{"id": "2505.02238", "title": "Federated Causal Inference in Healthcare: Methods, Challenges, and\n  Applications", "abstract": "  Federated causal inference enables multi-site treatment effect estimation\nwithout sharing individual-level data, offering a privacy-preserving solution\nfor real-world evidence generation. However, data heterogeneity across sites,\nmanifested in differences in covariate, treatment, and outcome, poses\nsignificant challenges for unbiased and efficient estimation. In this paper, we\npresent a comprehensive review and theoretical analysis of federated causal\neffect estimation across both binary/continuous and time-to-event outcomes. We\nclassify existing methods into weight-based strategies and optimization-based\nframeworks and further discuss extensions including personalized models,\npeer-to-peer communication, and model decomposition. For time-to-event\noutcomes, we examine federated Cox and Aalen-Johansen models, deriving\nasymptotic bias and variance under heterogeneity. Our analysis reveals that\nFedProx-style regularization achieves near-optimal bias-variance trade-offs\ncompared to naive averaging and meta-analysis. We review related software tools\nand conclude by outlining opportunities, challenges, and future directions for\nscalable, fair, and trustworthy federated causal inference in distributed\nhealthcare systems.\n", "date": "2025-05-04", "categories": []}
{"id": "2505.00930", "title": "Robust Root Cause Diagnosis using In-Distribution Interventions", "abstract": "  Diagnosing the root cause of an anomaly in a complex interconnected system is\na pressing problem in today's cloud services and industrial operations. We\npropose In-Distribution Interventions (IDI), a novel algorithm that predicts\nroot cause as nodes that meet two criteria: 1) **Anomaly:** root cause nodes\nshould take on anomalous values; 2) **Fix:** had the root cause nodes assumed\nusual values, the target node would not have been anomalous. Prior methods of\nassessing the fix condition rely on counterfactuals inferred from a Structural\nCausal Model (SCM) trained on historical data. But since anomalies are rare and\nfall outside the training distribution, the fitted SCMs yield unreliable\ncounterfactual estimates. IDI overcomes this by relying on interventional\nestimates obtained by solely probing the fitted SCM at in-distribution inputs.\nWe present a theoretical analysis comparing and bounding the errors in\nassessing the fix condition using interventional and counterfactual estimates.\nWe then conduct experiments by systematically varying the SCM's complexity to\ndemonstrate the cases where IDI's interventional approach outperforms the\ncounterfactual approach and vice versa. Experiments on both synthetic and\nPetShop RCD benchmark datasets demonstrate that \\our\\ consistently identifies\ntrue root causes more accurately and robustly than nine existing\nstate-of-the-art RCD baselines. Code is released at\nhttps://github.com/nlokeshiisc/IDI_release.\n", "date": "2025-05-02", "categories": []}
{"id": "2311.06840", "title": "Omitted Labels Induce Nontransitive Paradoxes in Causality", "abstract": "  We explore \"omitted label contexts,\" in which training data is limited to a\nsubset of the possible labels. This setting is standard among specialized human\nexperts or specific, focused studies. By studying Simpson's paradox, we observe\nthat ``correct'' adjustments sometimes require non-exchangeable treatment and\ncontrol groups. A generalization of Simpson's paradox leads us to study\nnetworks of conclusions drawn from different contexts, within which a paradox\nof nontransitivity arises. We prove that the space of possible nontransitive\nstructures in these networks exactly corresponds to structures that form from\naggregating ranked-choice votes.\n", "date": "2023-11-12", "categories": []}
{"id": "2406.01933", "title": "Orthogonal Causal Calibration", "abstract": "  Estimates of heterogeneous treatment effects such as conditional average\ntreatment effects (CATEs) and conditional quantile treatment effects (CQTEs)\nplay an important role in real-world decision making. Given this importance,\none should ensure these estimates are calibrated. While there is a rich\nliterature on calibrating estimators of non-causal parameters, very few methods\nhave been derived for calibrating estimators of causal parameters, or more\ngenerally estimators of quantities involving nuisance parameters. In this work,\nwe develop general algorithms for reducing the task of causal calibration to\nthat of calibrating a standard (non-causal) predictive model.\n  Throughout, we study a notion of calibration defined with respect to an\narbitrary, nuisance-dependent loss $\\ell$, under which we say an estimator\n$\\theta$ is calibrated if its predictions cannot be changed on any level set to\ndecrease loss. For losses $\\ell$ satisfying a condition called universal\northogonality, we present a simple algorithm that transforms partially-observed\ndata into generalized pseudo-outcomes and applies any off-the-shelf calibration\nprocedure. For losses $\\ell$ satisfying a weaker assumption called conditional\northogonality, we provide a similar sample splitting algorithm the performs\nempirical risk minimization over an appropriately defined class of functions.\nConvergence of both algorithms follows from a generic, two term upper bound of\nthe calibration error of any model. We demonstrate the practical applicability\nof our results in experiments on both observational and synthetic data. Our\nresults are exceedingly general, showing that essentially any existing\ncalibration algorithm can be used in causal settings, with additional loss only\narising from errors in nuisance estimation.\n", "date": "2024-06-04", "categories": []}
{"id": "2505.00229", "title": "Inference for max-linear Bayesian networks with noise", "abstract": "  Max-Linear Bayesian Networks (MLBNs) provide a powerful framework for causal\ninference in extreme-value settings; we consider MLBNs with noise parameters\nwith a given topology in terms of the max-plus algebra by taking its logarithm.\nThen, we show that an estimator of a parameter for each edge in a directed\nacyclic graph (DAG) is distributed normally. We end this paper with\ncomputational experiments with the expectation and maximization (EM) algorithm\nand quadratic optimization.\n", "date": "2025-05-01", "categories": []}
{"id": "2505.00304", "title": "Reinforcement Learning with Continuous Actions Under Unmeasured\n  Confounding", "abstract": "  This paper addresses the challenge of offline policy learning in\nreinforcement learning with continuous action spaces when unmeasured\nconfounders are present. While most existing research focuses on policy\nevaluation within partially observable Markov decision processes (POMDPs) and\nassumes discrete action spaces, we advance this field by establishing a novel\nidentification result to enable the nonparametric estimation of policy value\nfor a given target policy under an infinite-horizon framework. Leveraging this\nidentification, we develop a minimax estimator and introduce a\npolicy-gradient-based algorithm to identify the in-class optimal policy that\nmaximizes the estimated policy value. Furthermore, we provide theoretical\nresults regarding the consistency, finite-sample error bound, and regret bound\nof the resulting optimal policy. Extensive simulations and a real-world\napplication using the German Family Panel data demonstrate the effectiveness of\nour proposed methodology.\n", "date": "2025-05-01", "categories": []}
{"id": "2504.21261", "title": "Multi-Domain Causal Discovery in Bijective Causal Models", "abstract": "  We consider the problem of causal discovery (a.k.a., causal structure\nlearning) in a multi-domain setting. We assume that the causal functions are\ninvariant across the domains, while the distribution of the exogenous noise may\nvary. Under causal sufficiency (i.e., no confounders exist), we show that the\ncausal diagram can be discovered under less restrictive functional assumptions\ncompared to previous work. What enables causal discovery in this setting is\nbijective generation mechanisms (BGM), which ensures that the functional\nrelation between the exogenous noise $E$ and the endogenous variable $Y$ is\nbijective and differentiable in both directions at every level of the cause\nvariable $X = x$. BGM generalizes a variety of models including additive noise\nmodel, LiNGAM, post-nonlinear model, and location-scale noise model. Further,\nwe derive a statistical test to find the parents set of the target variable.\nExperiments on various synthetic and real-world datasets validate our\ntheoretical findings.\n", "date": "2025-04-30", "categories": []}
{"id": "2409.08295", "title": "Higher order definition of causality by optimally conditioned transfer\n  entropy", "abstract": "  The description of the dynamics of complex systems, in particular the capture\nof the interaction structure and causal relationships between elements of the\nsystem, is one of the central questions of interdisciplinary research. While\nthe characterization of pairwise causal interactions is a relatively ripe field\nwith established theoretical concepts and the current focus is on technical\nissues of their efficient estimation, it turns out that the standard concepts\nsuch as Granger causality or transfer entropy may not faithfully reflect\npossible synergies or interactions of higher orders, phenomena highly relevant\nfor many real-world complex systems. In this paper, we propose a generalization\nand refinement of the information-theoretic approach to causal inference,\nenabling the description of truly multivariate, rather than multiple pairwise,\ncausal interactions, and moving thus from causal networks to causal\nhypernetworks. In particular, while keeping the ability to control for\nmediating variables or common causes, in case of purely synergetic interactions\nsuch as the exclusive disjunction, it ascribes the causal role to the\nmultivariate causal set but \\emph{not} to individual inputs, distinguishing it\nthus from the case of e.g. two additive univariate causes. We demonstrate this\nconcept by application to illustrative theoretical examples as well as a\nbiophysically realistic simulation of biological neuronal dynamics recently\nreported to employ synergetic computations.\n", "date": "2024-08-30", "categories": []}
{"id": "2502.12181", "title": "3D ReX: Causal Explanations in 3D Neuroimaging Classification", "abstract": "  Explainability remains a significant problem for AI models in medical\nimaging, making it challenging for clinicians to trust AI-driven predictions.\nWe introduce 3D ReX, the first causality-based post-hoc explainability tool for\n3D models. 3D ReX uses the theory of actual causality to generate\nresponsibility maps which highlight the regions most crucial to the model's\ndecision. We test 3D ReX on a stroke detection model, providing insight into\nthe spatial distribution of features relevant to stroke.\n", "date": "2025-02-14", "categories": []}
{"id": "2504.20172", "title": "Causal Identification in Time Series Models", "abstract": "  In this paper, we analyze the applicability of the Causal Identification\nalgorithm to causal time series graphs with latent confounders. Since these\ngraphs extend over infinitely many time steps, deciding whether causal effects\nacross arbitrary time intervals are identifiable appears to require computation\non graph segments of unbounded size. Even for deciding the identifiability of\nintervention effects on variables that are close in time, no bound is known on\nhow many time steps in the past need to be considered. We give a first bound of\nthis kind that only depends on the number of variables per time step and the\nmaximum time lag of any direct or latent causal effect. More generally, we show\nthat applying the Causal Identification algorithm to a constant-size segment of\nthe time series graph is sufficient to decide identifiability of causal\neffects, even across unbounded time intervals.\n", "date": "2025-04-28", "categories": []}
{"id": "2504.20471", "title": "The Estimation of Continual Causal Effect for Dataset Shifting Streams", "abstract": "  Causal effect estimation has been widely used in marketing optimization. The\nframework of an uplift model followed by a constrained optimization algorithm\nis popular in practice. To enhance performance in the online environment, the\nframework needs to be improved to address the complexities caused by temporal\ndataset shift. This paper focuses on capturing the dataset shift from user\nbehavior and domain distribution changing over time. We propose an Incremental\nCausal Effect with Proxy Knowledge Distillation (ICE-PKD) framework to tackle\nthis challenge. The ICE-PKD framework includes two components: (i) a\nmulti-treatment uplift network that eliminates confounding bias using\ncounterfactual regression; (ii) an incremental training strategy that adapts to\nthe temporal dataset shift by updating with the latest data and protects\ngeneralization via replay-based knowledge distillation. We also revisit the\nuplift modeling metrics and introduce a novel metric for more precise online\nevaluation in multiple treatment scenarios. Extensive experiments on both\nsimulated and online datasets show that the proposed framework achieves better\nperformance. The ICE-PKD framework has been deployed in the marketing system of\nHuaxiaozhu, a ride-hailing platform in China.\n", "date": "2025-04-29", "categories": []}
{"id": "2504.20579", "title": "Representation Learning Preserving Ignorability and Covariate Matching\n  for Treatment Effects", "abstract": "  Estimating treatment effects from observational data is challenging due to\ntwo main reasons: (a) hidden confounding, and (b) covariate mismatch (control\nand treatment groups not having identical distributions). Long lines of works\nexist that address only either of these issues. To address the former,\nconventional techniques that require detailed knowledge in the form of causal\ngraphs have been proposed. For the latter, covariate matching and importance\nweighting methods have been used. Recently, there has been progress in\ncombining testable independencies with partial side information for tackling\nhidden confounding. A common framework to address both hidden confounding and\nselection bias is missing. We propose neural architectures that aim to learn a\nrepresentation of pre-treatment covariates that is a valid adjustment and also\nsatisfies covariate matching constraints. We combine two different neural\narchitectures: one based on gradient matching across domains created by\nsubsampling a suitable anchor variable that assumes causal side information,\nfollowed by the other, a covariate matching transformation. We prove that\napproximately invariant representations yield approximate valid adjustment sets\nwhich would enable an interval around the true causal effect. In contrast to\nusual sensitivity analysis, where an unknown nuisance parameter is varied, we\nhave a testable approximation yielding a bound on the effect estimate. We also\noutperform various baselines with respect to ATE and PEHE errors on causal\nbenchmarks that include IHDP, Jobs, Cattaneo, and an image-based Crowd\nManagement dataset.\n", "date": "2025-04-29", "categories": []}
{"id": "2310.16945", "title": "Causal Q-Aggregation for CATE Model Selection", "abstract": "  Accurate estimation of conditional average treatment effects (CATE) is at the\ncore of personalized decision making. While there is a plethora of models for\nCATE estimation, model selection is a nontrivial task, due to the fundamental\nproblem of causal inference. Recent empirical work provides evidence in favor\nof proxy loss metrics with double robust properties and in favor of model\nensembling. However, theoretical understanding is lacking. Direct application\nof prior theoretical work leads to suboptimal oracle model selection rates due\nto the non-convexity of the model selection problem. We provide regret rates\nfor the major existing CATE ensembling approaches and propose a new CATE model\nensembling approach based on Q-aggregation using the doubly robust loss. Our\nmain result shows that causal Q-aggregation achieves statistically optimal\noracle model selection regret rates of $\\frac{\\log(M)}{n}$ (with $M$ models and\n$n$ samples), with the addition of higher-order estimation error terms related\nto products of errors in the nuisance functions. Crucially, our regret rate\ndoes not require that any of the candidate CATE models be close to the truth.\nWe validate our new method on many semi-synthetic datasets and also provide\nextensions of our work to CATE model selection with instrumental variables and\nunobserved confounding.\n", "date": "2023-10-25", "categories": []}
{"id": "2408.03425", "title": "Sequential Conditional Transport on Probabilistic Graphs for\n  Interpretable Counterfactual Fairness", "abstract": "  In this paper, we link two existing approaches to derive counterfactuals:\nadaptations based on a causal graph, and optimal transport. We extend \"Knothe's\nrearrangement\" and \"triangular transport\" to probabilistic graphical models,\nand use this counterfactual approach, referred to as sequential transport, to\ndiscuss fairness at the individual level. After establishing the theoretical\nfoundations of the proposed method, we demonstrate its application through\nnumerical experiments on both synthetic and real datasets.\n", "date": "2024-08-06", "categories": []}
{"id": "2411.06990", "title": "Causal-discovery-based root-cause analysis and its application in\n  time-series prediction error diagnosis", "abstract": "  Recent rapid advancements of machine learning have greatly enhanced the\naccuracy of prediction models, but most models remain \"black boxes\", making\nprediction error diagnosis challenging, especially with outliers. This lack of\ntransparency hinders trust and reliability in industrial applications.\nHeuristic attribution methods, while helpful, often fail to capture true causal\nrelationships, leading to inaccurate error attributions. Various root-cause\nanalysis methods have been developed using Shapley values, yet they typically\nrequire predefined causal graphs, limiting their applicability for prediction\nerrors in machine learning models. To address these limitations, we introduce\nthe Causal-Discovery-based Root-Cause Analysis (CD-RCA) method that estimates\ncausal relationships between the prediction error and the explanatory\nvariables, without needing a pre-defined causal graph. By simulating synthetic\nerror data, CD-RCA can identify variable contributions to outliers in\nprediction errors by Shapley values. Extensive experiments show CD-RCA\noutperforms current heuristic attribution methods.\n", "date": "2024-11-11", "categories": []}
{"id": "2502.04699", "title": "A Meta-learner for Heterogeneous Effects in Difference-in-Differences", "abstract": "  We address the problem of estimating heterogeneous treatment effects in panel\ndata, adopting the popular Difference-in-Differences (DiD) framework under the\nconditional parallel trends assumption. We propose a novel doubly robust\nmeta-learner for the Conditional Average Treatment Effect on the Treated\n(CATT), reducing the estimation to a convex risk minimization problem involving\na set of auxiliary models. Our framework allows for the flexible estimation of\nthe CATT, when conditioning on any subset of variables of interest using\ngeneric machine learning. Leveraging Neyman orthogonality, our proposed\napproach is robust to estimation errors in the auxiliary models. As a\ngeneralization to our main result, we develop a meta-learning approach for the\nestimation of general conditional functionals under covariate shift. We also\nprovide an extension to the instrumented DiD setting with non-compliance.\nEmpirical results demonstrate the superiority of our approach over existing\nbaselines.\n", "date": "2025-02-07", "categories": []}
{"id": "2502.20099", "title": "Sanity Checking Causal Representation Learning on a Simple Real-World\n  System", "abstract": "  We evaluate methods for causal representation learning (CRL) on a simple,\nreal-world system where these methods are expected to work. The system consists\nof a controlled optical experiment specifically built for this purpose, which\nsatisfies the core assumptions of CRL and where the underlying causal factors\n(the inputs to the experiment) are known, providing a ground truth. We select\nmethods representative of different approaches to CRL and find that they all\nfail to recover the underlying causal factors. To understand the failure modes\nof the evaluated algorithms, we perform an ablation on the data by substituting\nthe real data-generating process with a simpler synthetic equivalent. The\nresults reveal a reproducibility problem, as most methods already fail on this\nsynthetic ablation despite its simple data-generating process. Additionally, we\nobserve that common assumptions on the mixing function are crucial for the\nperformance of some of the methods but do not hold in the real data. Our\nefforts highlight the contrast between the theoretical promise of the state of\nthe art and the challenges in its application. We hope the benchmark serves as\na simple, real-world sanity check to further develop and validate methodology,\nbridging the gap towards CRL methods that work in practice. We make all code\nand datasets publicly available at github.com/simonbing/CRLSanityCheck\n", "date": "2025-02-27", "categories": []}
{"id": "2504.19527", "title": "Identification and Estimation of Long-Term Treatment Effects with\n  Monotone Missing", "abstract": "  Estimating long-term treatment effects has a wide range of applications in\nvarious domains. A key feature in this context is that collecting long-term\noutcomes typically involves a multi-stage process and is subject to monotone\nmissing, where individuals missing at an earlier stage remain missing at\nsubsequent stages. Despite its prevalence, monotone missing has been rarely\nexplored in previous studies on estimating long-term treatment effects. In this\npaper, we address this gap by introducing the sequential missingness assumption\nfor identification. We propose three novel estimation methods, including\ninverse probability weighting, sequential regression imputation, and sequential\nmarginal structural model (SeqMSM). Considering that the SeqMSM method may\nsuffer from high variance due to severe data sparsity caused by monotone\nmissing, we further propose a novel balancing-enhanced approach, BalanceNet, to\nimprove the stability and accuracy of the estimation methods. Extensive\nexperiments on two widely used benchmark datasets demonstrate the effectiveness\nof our proposed methods.\n", "date": "2025-04-28", "categories": []}
{"id": "2402.04869", "title": "Learning by Doing: An Online Causal Reinforcement Learning Framework\n  with Causal-Aware Policy", "abstract": "  As a key component to intuitive cognition and reasoning solutions in human\nintelligence, causal knowledge provides great potential for reinforcement\nlearning (RL) agents' interpretability towards decision-making by helping\nreduce the searching space. However, there is still a considerable gap in\ndiscovering and incorporating causality into RL, which hinders the rapid\ndevelopment of causal RL. In this paper, we consider explicitly modeling the\ngeneration process of states with the causal graphical model, based on which we\naugment the policy. We formulate the causal structure updating into the RL\ninteraction process with active intervention learning of the environment. To\noptimize the derived objective, we propose a framework with theoretical\nperformance guarantees that alternates between two steps: using interventions\nfor causal structure learning during exploration and using the learned causal\nstructure for policy guidance during exploitation. Due to the lack of public\nbenchmarks that allow direct intervention in the state space, we design the\nroot cause localization task in our simulated fault alarm environment and then\nempirically show the effectiveness and robustness of the proposed method\nagainst state-of-the-art baselines. Theoretical analysis shows that our\nperformance improvement attributes to the virtuous cycle of causal-guided\npolicy learning and causal structure learning, which aligns with our\nexperimental results. Codes are available at\nhttps://github.com/DMIRLAB-Group/FaultAlarm_RL.\n", "date": "2024-02-07", "categories": []}
{"id": "2402.14781", "title": "Effective Bayesian Causal Inference via Structural Marginalisation and\n  Autoregressive Orders", "abstract": "  The traditional two-stage approach to causal inference first identifies a\nsingle causal model (or equivalence class of models), which is then used to\nanswer causal queries. However, this neglects any epistemic model uncertainty.\nIn contrast, Bayesian causal inference does incorporate epistemic uncertainty\ninto query estimates via Bayesian marginalisation (posterior averaging) over\nall causal models. While principled, this marginalisation over entire causal\nmodels, i.e., both causal structures (graphs) and mechanisms, poses a\ntremendous computational challenge. In this work, we address this challenge by\ndecomposing structure marginalisation into the marginalisation over (i) causal\norders and (ii) directed acyclic graphs (DAGs) given an order. We can\nmarginalise the latter in closed form by limiting the number of parents per\nvariable and utilising Gaussian processes to model mechanisms. To marginalise\nover orders, we use a sampling-based approximation, for which we devise a novel\nauto-regressive distribution over causal orders (ARCO). Our method outperforms\nstate-of-the-art in structure learning on simulated non-linear additive noise\nbenchmarks, and yields competitive results on real-world data. Furthermore, we\ncan accurately infer interventional distributions and average causal effects.\n", "date": "2024-02-22", "categories": []}
{"id": "2409.01444", "title": "A causal viewpoint on prediction model performance under changes in\n  case-mix: discrimination and calibration respond differently for prognosis\n  and diagnosis predictions", "abstract": "  Prediction models need reliable predictive performance as they inform\nclinical decisions, aiding in diagnosis, prognosis, and treatment planning. The\npredictive performance of these models is typically assessed through\ndiscrimination and calibration. Changes in the distribution of the data impact\nmodel performance and there may be important changes between a model's current\napplication and when and where its performance was last evaluated. In\nhealth-care, a typical change is a shift in case-mix. For example, for\ncardiovascular risk management, a general practitioner sees a different mix of\npatients than a specialist in a tertiary hospital.\n  This work introduces a novel framework that differentiates the effects of\ncase-mix shifts on discrimination and calibration based on the causal direction\nof the prediction task. When prediction is in the causal direction (often the\ncase for prognosis predictions), calibration remains stable under case-mix\nshifts, while discrimination does not. Conversely, when predicting in the\nanti-causal direction (often with diagnosis predictions), discrimination\nremains stable, but calibration does not.\n  A simulation study and empirical validation using cardiovascular disease\nprediction models demonstrate the implications of this framework. The causal\ncase-mix framework provides insights for developing, evaluating and deploying\nprediction models across different clinical settings, emphasizing the\nimportance of understanding the causal structure of the prediction task.\n", "date": "2024-09-02", "categories": []}
{"id": "2504.17166", "title": "Causal rule ensemble approach for multi-arm data", "abstract": "  Heterogeneous treatment effect (HTE) estimation is critical in medical\nresearch. It provides insights into how treatment effects vary among\nindividuals, which can provide statistical evidence for precision medicine.\nWhile most existing methods focus on binary treatment situations, real-world\napplications often involve multiple interventions. However, current HTE\nestimation methods are primarily designed for binary comparisons and often rely\non black-box models, which limit their applicability and interpretability in\nmulti-arm settings. To address these challenges, we propose an interpretable\nmachine learning framework for HTE estimation in multi-arm trials. Our method\nemploys a rule-based ensemble approach consisting of rule generation, rule\nensemble, and HTE estimation, ensuring both predictive accuracy and\ninterpretability. Through extensive simulation studies and real data\napplications, the performance of our method was evaluated against\nstate-of-the-art multi-arm HTE estimation approaches. The results indicate that\nour approach achieved lower bias and higher estimation accuracy compared with\nthose of existing methods. Furthermore, the interpretability of our framework\nallows clearer insights into how covariates influence treatment effects,\nfacilitating clinical decision making. By bridging the gap between accuracy and\ninterpretability, our study contributes a valuable tool for multi-arm HTE\nestimation, supporting precision medicine.\n", "date": "2025-04-24", "categories": []}
{"id": "2406.02584", "title": "A Scoping Review of Earth Observation and Machine Learning for Causal\n  Inference: Implications for the Geography of Poverty", "abstract": "  Earth observation (EO) data such as satellite imagery can have far-reaching\nimpacts on our understanding of the geography of poverty, especially when\ncoupled with machine learning (ML) and computer vision. Early research used\ncomputer vision to predict living conditions in areas with limited data, but\nrecent studies increasingly focus on causal analysis. Despite this shift, the\nuse of EO-ML methods for causal inference lacks thorough documentation, and\nbest practices are still developing. Through a comprehensive scoping review, we\ncatalog the current literature on EO-ML methods in causal analysis. We\nsynthesize five principal approaches to incorporating EO data in causal\nworkflows: (1) outcome imputation for downstream causal analysis, (2) EO image\ndeconfounding, (3) EO-based treatment effect heterogeneity, (4) EO-based\ntransportability analysis, and (5) image-informed causal discovery. Building on\nthese findings, we provide a detailed protocol guiding researchers in\nintegrating EO data into causal analysis -- covering data requirements,\ncomputer vision model selection, and evaluation metrics. While our focus\ncenters on health and living conditions outcomes, our protocol is adaptable to\nother sustainable development domains utilizing EO data.\n", "date": "2024-05-30", "categories": []}
{"id": "2504.15386", "title": "Assessing Surrogate Heterogeneity in Real World Data Using Meta-Learners", "abstract": "  Surrogate markers are most commonly studied within the context of randomized\nclinical trials. However, the need for alternative outcomes extends beyond\nthese settings and may be more pronounced in real-world public health and\nsocial science research, where randomized trials are often impractical.\nResearch on identifying surrogates in real-world non-randomized data is scarce,\nas available statistical approaches for evaluating surrogate markers tend to\nrely on the assumption that treatment is randomized. While the few methods that\nallow for non-randomized treatment/exposure appropriately handle confounding\nindividual characteristics, they do not offer a way to examine surrogate\nheterogeneity with respect to patient characteristics. In this paper, we\npropose a framework to assess surrogate heterogeneity in real-world, i.e.,\nnon-randomized, data and implement this framework using various meta-learners.\nOur approach allows us to quantify heterogeneity in surrogate strength with\nrespect to patient characteristics while accommodating confounders through the\nuse of flexible, off-the-shelf machine learning methods. In addition, we use\nour framework to identify individuals for whom the surrogate is a valid\nreplacement of the primary outcome. We examine the performance of our methods\nvia a simulation study and application to examine heterogeneity in the\nsurrogacy of hemoglobin A1c as a surrogate for fasting plasma glucose.\n", "date": "2025-04-21", "categories": []}
{"id": "2504.15854", "title": "Consistent Causal Inference of Group Effects in Non-Targeted Trials with\n  Finitely Many Effect Levels", "abstract": "  A treatment may be appropriate for some group (the ``sick\" group) on whom it\nhas a positive effect, but it can also have a detrimental effect on subjects\nfrom another group (the ``healthy\" group). In a non-targeted trial both sick\nand healthy subjects may be treated, producing heterogeneous effects within the\ntreated group. Inferring the correct treatment effect on the sick population is\nthen difficult, because the effects on the different groups get tangled. We\npropose an efficient nonparametric approach to estimating the group effects,\ncalled {\\bf PCM} (pre-cluster and merge). We prove its asymptotic consistency\nin a general setting and show, on synthetic data, more than a 10x improvement\nin accuracy over existing state-of-the-art. Our approach applies more generally\nto consistent estimation of functions with a finite range.\n", "date": "2025-04-22", "categories": []}
{"id": "2410.06407", "title": "A Skewness-Based Criterion for Addressing Heteroscedastic Noise in\n  Causal Discovery", "abstract": "  Real-world data often violates the equal-variance assumption\n(homoscedasticity), making it essential to account for heteroscedastic noise in\ncausal discovery. In this work, we explore heteroscedastic symmetric noise\nmodels (HSNMs), where the effect $Y$ is modeled as $Y = f(X) + \\sigma(X)N$,\nwith $X$ as the cause and $N$ as independent noise following a symmetric\ndistribution. We introduce a novel criterion for identifying HSNMs based on the\nskewness of the score (i.e., the gradient of the log density) of the data\ndistribution. This criterion establishes a computationally tractable\nmeasurement that is zero in the causal direction but nonzero in the anticausal\ndirection, enabling the causal direction discovery. We extend this\nskewness-based criterion to the multivariate setting and propose SkewScore, an\nalgorithm that handles heteroscedastic noise without requiring the extraction\nof exogenous noise. We also conduct a case study on the robustness of SkewScore\nin a bivariate model with a latent confounder, providing theoretical insights\ninto its performance. Empirical studies further validate the effectiveness of\nthe proposed method.\n", "date": "2024-10-08", "categories": []}
{"id": "2504.13263", "title": "Causal-Copilot: An Autonomous Causal Analysis Agent", "abstract": "  Causal analysis plays a foundational role in scientific discovery and\nreliable decision-making, yet it remains largely inaccessible to domain experts\ndue to its conceptual and algorithmic complexity. This disconnect between\ncausal methodology and practical usability presents a dual challenge: domain\nexperts are unable to leverage recent advances in causal learning, while causal\nresearchers lack broad, real-world deployment to test and refine their methods.\nTo address this, we introduce Causal-Copilot, an autonomous agent that\noperationalizes expert-level causal analysis within a large language model\nframework. Causal-Copilot automates the full pipeline of causal analysis for\nboth tabular and time-series data -- including causal discovery, causal\ninference, algorithm selection, hyperparameter optimization, result\ninterpretation, and generation of actionable insights. It supports interactive\nrefinement through natural language, lowering the barrier for non-specialists\nwhile preserving methodological rigor. By integrating over 20 state-of-the-art\ncausal analysis techniques, our system fosters a virtuous cycle -- expanding\naccess to advanced causal methods for domain experts while generating rich,\nreal-world applications that inform and advance causal theory. Empirical\nevaluations demonstrate that Causal-Copilot achieves superior performance\ncompared to existing baselines, offering a reliable, scalable, and extensible\nsolution that bridges the gap between theoretical sophistication and real-world\napplicability in causal analysis. A live interactive demo of Causal-Copilot is\navailable at https://causalcopilot.com/.\n", "date": "2025-04-17", "categories": []}
{"id": "2504.14530", "title": "Causality for Natural Language Processing", "abstract": "  Causal reasoning is a cornerstone of human intelligence and a critical\ncapability for artificial systems aiming to achieve advanced understanding and\ndecision-making. This thesis delves into various dimensions of causal reasoning\nand understanding in large language models (LLMs). It encompasses a series of\nstudies that explore the causal inference skills of LLMs, the mechanisms behind\ntheir performance, and the implications of causal and anticausal learning for\nnatural language processing (NLP) tasks. Additionally, it investigates the\napplication of causal reasoning in text-based computational social science,\nspecifically focusing on political decision-making and the evaluation of\nscientific impact through citations. Through novel datasets, benchmark tasks,\nand methodological frameworks, this work identifies key challenges and\nopportunities to improve the causal capabilities of LLMs, providing a\ncomprehensive foundation for future research in this evolving field.\n", "date": "2025-04-20", "categories": []}
{"id": "2504.14937", "title": "Causal DAG Summarization (Full Version)", "abstract": "  Causal inference aids researchers in discovering cause-and-effect\nrelationships, leading to scientific insights. Accurate causal estimation\nrequires identifying confounding variables to avoid false discoveries. Pearl's\ncausal model uses causal DAGs to identify confounding variables, but incorrect\nDAGs can lead to unreliable causal conclusions. However, for high dimensional\ndata, the causal DAGs are often complex beyond human verifiability. Graph\nsummarization is a logical next step, but current methods for general-purpose\ngraph summarization are inadequate for causal DAG summarization. This paper\naddresses these challenges by proposing a causal graph summarization objective\nthat balances graph simplification for better understanding while retaining\nessential causal information for reliable inference. We develop an efficient\ngreedy algorithm and show that summary causal DAGs can be directly used for\ninference and are more robust to misspecification of assumptions, enhancing\nrobustness for causal inference. Experimenting with six real-life datasets, we\ncompared our algorithm to three existing solutions, showing its effectiveness\nin handling high-dimensional data and its ability to generate summary DAGs that\nensure both reliable causal inference and robustness against misspecifications.\n", "date": "2025-04-21", "categories": []}
{"id": "2504.13733", "title": "Dynamic Regularized CBDT: Variance-Calibrated Causal Boosting for\n  Interpretable Heterogeneous Treatment Effects", "abstract": "  Heterogeneous treatment effect estimation in high-stakes applications demands\nmodels that simultaneously optimize precision, interpretability, and\ncalibration. Many existing tree-based causal inference techniques, however,\nexhibit high estimation errors when applied to observational data because they\nstruggle to capture complex interactions among factors and rely on static\nregularization schemes. In this work, we propose Dynamic Regularized Causal\nBoosted Decision Trees (CBDT), a novel framework that integrates variance\nregularization and average treatment effect calibration into the loss function\nof gradient boosted decision trees. Our approach dynamically updates the\nregularization parameters using gradient statistics to better balance the\nbias-variance tradeoff. Extensive experiments on standard benchmark datasets\nand real-world clinical data demonstrate that the proposed method significantly\nimproves estimation accuracy while maintaining reliable coverage of true\ntreatment effects. In an intensive care unit patient triage study, the method\nsuccessfully identified clinically actionable rules and achieved high accuracy\nin treatment effect estimation. The results validate that dynamic\nregularization can effectively tighten error bounds and enhance both predictive\nperformance and model interpretability.\n", "date": "2025-04-18", "categories": []}
{"id": "2405.01744", "title": "ALCM: Autonomous LLM-Augmented Causal Discovery Framework", "abstract": "  To perform effective causal inference in high-dimensional datasets,\ninitiating the process with causal discovery is imperative, wherein a causal\ngraph is generated based on observational data. However, obtaining a complete\nand accurate causal graph poses a formidable challenge, recognized as an NP-\nhard problem. Recently, the advent of Large Language Models (LLMs) has ushered\nin a new era, indicating their emergent capabilities and widespread\napplicability in facilitating causal reasoning across diverse domains, such as\nmedicine, finance, and science. The expansive knowledge base of LLMs holds the\npotential to elevate the field of causal reasoning by offering\ninterpretability, making inferences, generalizability, and uncovering novel\ncausal structures. In this paper, we introduce a new framework, named\nAutonomous LLM-Augmented Causal Discovery Framework (ALCM), to synergize\ndata-driven causal discovery algorithms and LLMs, automating the generation of\na more resilient, accurate, and explicable causal graph. The ALCM consists of\nthree integral components: causal structure learning, causal wrapper, and\nLLM-driven causal refiner. These components autonomously collaborate within a\ndynamic environment to address causal discovery questions and deliver plausible\ncausal graphs. We evaluate the ALCM framework by implementing two\ndemonstrations on seven well-known datasets. Experimental results demonstrate\nthat ALCM outperforms existing LLM methods and conventional data-driven causal\nreasoning mechanisms. This study not only shows the effectiveness of the ALCM\nbut also underscores new research directions in leveraging the causal reasoning\ncapabilities of LLMs.\n", "date": "2024-05-02", "categories": []}
{"id": "2410.13054", "title": "Systems with Switching Causal Relations: A Meta-Causal Perspective", "abstract": "  Most work on causality in machine learning assumes that causal relationships\nare driven by a constant underlying process. However, the flexibility of\nagents' actions or tipping points in the environmental process can change the\nqualitative dynamics of the system. As a result, new causal relationships may\nemerge, while existing ones change or disappear, resulting in an altered causal\ngraph. To analyze these qualitative changes on the causal graph, we propose the\nconcept of meta-causal states, which groups classical causal models into\nclusters based on equivalent qualitative behavior and consolidates specific\nmechanism parameterizations. We demonstrate how meta-causal states can be\ninferred from observed agent behavior, and discuss potential methods for\ndisentangling these states from unlabeled data. Finally, we direct our analysis\ntowards the application of a dynamical system, showing that meta-causal states\ncan also emerge from inherent system dynamics, and thus constitute more than a\ncontext-dependent framework in which mechanisms emerge only as a result of\nexternal factors.\n", "date": "2024-10-16", "categories": []}
{"id": "2504.12594", "title": "Meta-Dependence in Conditional Independence Testing", "abstract": "  Constraint-based causal discovery algorithms utilize many statistical tests\nfor conditional independence to uncover networks of causal dependencies. These\napproaches to causal discovery rely on an assumed correspondence between the\ngraphical properties of a causal structure and the conditional independence\nproperties of observed variables, known as the causal Markov condition and\nfaithfulness. Finite data yields an empirical distribution that is \"close\" to\nthe actual distribution. Across these many possible empirical distributions,\nthe correspondence to the graphical properties can break down for different\nconditional independencies, and multiple violations can occur at the same time.\nWe study this \"meta-dependence\" between conditional independence properties\nusing the following geometric intuition: each conditional independence property\nconstrains the space of possible joint distributions to a manifold. The\n\"meta-dependence\" between conditional independences is informed by the position\nof these manifolds relative to the true probability distribution. We provide a\nsimple-to-compute measure of this meta-dependence using information projections\nand consolidate our findings empirically using both synthetic and real-world\ndata.\n", "date": "2025-04-17", "categories": []}
{"id": "2406.16899", "title": "Prompting or Fine-tuning? Exploring Large Language Models for Causal\n  Graph Validation", "abstract": "  This study explores the capability of Large Language Models (LLMs) to\nevaluate causality in causal graphs generated by conventional statistical\ncausal discovery methods-a task traditionally reliant on manual assessment by\nhuman subject matter experts. To bridge this gap in causality assessment, LLMs\nare employed to evaluate the causal relationships by determining whether a\ncausal connection between variable pairs can be inferred from textual context.\nOur study compares two approaches: (1) prompting-based method for zero-shot and\nfew-shot causal inference and, (2) fine-tuning language models for the causal\nrelation prediction task. While prompt-based LLMs have demonstrated versatility\nacross various NLP tasks, our experiments on biomedical and general-domain\ndatasets show that fine-tuned models consistently outperform them, achieving up\nto a 20.5-point improvement in F1 score-even when using smaller-parameter\nlanguage models. These findings provide valuable insights into the strengths\nand limitations of both approaches for causal graph evaluation.\n", "date": "2024-05-29", "categories": []}
{"id": "2504.10936", "title": "Can LLMs Leverage Observational Data? Towards Data-Driven Causal\n  Discovery with LLMs", "abstract": "  Causal discovery traditionally relies on statistical methods applied to\nobservational data, often requiring large datasets and assumptions about\nunderlying causal structures. Recent advancements in Large Language Models\n(LLMs) have introduced new possibilities for causal discovery by providing\ndomain expert knowledge. However, it remains unclear whether LLMs can\neffectively process observational data for causal discovery. In this work, we\nexplore the potential of LLMs for data-driven causal discovery by integrating\nobservational data for LLM-based reasoning. Specifically, we examine whether\nLLMs can effectively utilize observational data through two prompting\nstrategies: pairwise prompting and breadth first search (BFS)-based prompting.\nIn both approaches, we incorporate the observational data directly into the\nprompt to assess LLMs' ability to infer causal relationships from such data.\nExperiments on benchmark datasets show that incorporating observational data\nenhances causal discovery, boosting F1 scores by up to 0.11 point using both\npairwise and BFS LLM-based prompting, while outperforming traditional\nstatistical causal discovery baseline by up to 0.52 points. Our findings\nhighlight the potential and limitations of LLMs for data-driven causal\ndiscovery, demonstrating their ability to move beyond textual metadata and\neffectively interpret and utilize observational data for more informed causal\nreasoning. Our studies lays the groundwork for future advancements toward fully\nLLM-driven causal discovery.\n", "date": "2025-04-15", "categories": []}
{"id": "2412.19711", "title": "Causal machine learning for heterogeneous treatment effects in the\n  presence of missing outcome data", "abstract": "  When estimating heterogeneous treatment effects, missing outcome data can\ncomplicate treatment effect estimation, causing certain subgroups of the\npopulation to be poorly represented. In this work, we discuss this commonly\noverlooked problem and consider the impact that missing at random (MAR) outcome\ndata has on causal machine learning estimators for the conditional average\ntreatment effect (CATE). We propose two de-biased machine learning estimators\nfor the CATE, the mDR-learner and mEP-learner, which address the issue of\nunder-representation by integrating inverse probability of censoring weights\ninto the DR-learner and EP-learner respectively. We show that under reasonable\nconditions, these estimators are oracle efficient, and illustrate their\nfavorable performance through simulated data settings, comparing them to\nexisting CATE estimators, including comparison to estimators which use common\nmissing data techniques. We present an example of their application using the\nGBSG2 trial, exploring treatment effect heterogeneity when comparing hormonal\ntherapies to non-hormonal therapies among breast cancer patients post surgery,\nand offer guidance on the decisions a practitioner must make when implementing\nthese estimators.\n", "date": "2024-12-27", "categories": []}
{"id": "2504.08836", "title": "Double Machine Learning for Causal Inference under Shared-State\n  Interference", "abstract": "  Researchers and practitioners often wish to measure treatment effects in\nsettings where units interact via markets and recommendation systems. In these\nsettings, units are affected by certain shared states, like prices, algorithmic\nrecommendations or social signals. We formalize this structure, calling it\nshared-state interference, and argue that our formulation captures many\nrelevant applied settings. Our key modeling assumption is that individuals'\npotential outcomes are independent conditional on the shared state. We then\nprove an extension of a double machine learning (DML) theorem providing\nconditions for achieving efficient inference under shared-state interference.\nWe also instantiate our general theorem in several models of interest where it\nis possible to efficiently estimate the average direct effect (ADE) or global\naverage treatment effect (GATE).\n", "date": "2025-04-10", "categories": []}
{"id": "2504.09635", "title": "A Two-Stage Interpretable Matching Framework for Causal Inference", "abstract": "  Matching in causal inference from observational data aims to construct\ntreatment and control groups with similar distributions of covariates, thereby\nreducing confounding and ensuring an unbiased estimation of treatment effects.\nThis matched sample closely mimics a randomized controlled trial (RCT), thus\nimproving the quality of causal estimates. We introduce a novel Two-stage\nInterpretable Matching (TIM) framework for transparent and interpretable\ncovariate matching. In the first stage, we perform exact matching across all\navailable covariates. For treatment and control units without an exact match in\nthe first stage, we proceed to the second stage. Here, we iteratively refine\nthe matching process by removing the least significant confounder in each\niteration and attempting exact matching on the remaining covariates. We learn a\ndistance metric for the dropped covariates to quantify closeness to the\ntreatment unit(s) within the corresponding strata. We used these high- quality\nmatches to estimate the conditional average treatment effects (CATEs). To\nvalidate TIM, we conducted experiments on synthetic datasets with varying\nassociation structures and correlations. We assessed its performance by\nmeasuring bias in CATE estimation and evaluating multivariate overlap between\ntreatment and control groups before and after matching. Additionally, we apply\nTIM to a real-world healthcare dataset from the Centers for Disease Control and\nPrevention (CDC) to estimate the causal effect of high cholesterol on diabetes.\nOur results demonstrate that TIM improves CATE estimates, increases\nmultivariate overlap, and scales effectively to high-dimensional data, making\nit a robust tool for causal inference in observational data.\n", "date": "2025-04-13", "categories": []}
{"id": "2504.10397", "title": "Can LLMs Assist Expert Elicitation for Probabilistic Causal Modeling?", "abstract": "  Objective: This study investigates the potential of Large Language Models\n(LLMs) as an alternative to human expert elicitation for extracting structured\ncausal knowledge and facilitating causal modeling in biometric and healthcare\napplications.\n  Material and Methods: LLM-generated causal structures, specifically Bayesian\nnetworks (BNs), were benchmarked against traditional statistical methods (e.g.,\nBayesian Information Criterion) using healthcare datasets. Validation\ntechniques included structural equation modeling (SEM) to verifying\nrelationships, and measures such as entropy, predictive accuracy, and\nrobustness to compare network structures.\n  Results and Discussion: LLM-generated BNs demonstrated lower entropy than\nexpert-elicited and statistically generated BNs, suggesting higher confidence\nand precision in predictions. However, limitations such as contextual\nconstraints, hallucinated dependencies, and potential biases inherited from\ntraining data require further investigation.\n  Conclusion: LLMs represent a novel frontier in expert elicitation for\nprobabilistic causal modeling, promising to improve transparency and reduce\nuncertainty in the decision-making using such models.\n", "date": "2025-04-14", "categories": []}
{"id": "2503.23820", "title": "When Counterfactual Reasoning Fails: Chaos and Real-World Complexity", "abstract": "  Counterfactual reasoning, a cornerstone of human cognition and\ndecision-making, is often seen as the 'holy grail' of causal learning, with\napplications ranging from interpreting machine learning models to promoting\nalgorithmic fairness. While counterfactual reasoning has been extensively\nstudied in contexts where the underlying causal model is well-defined,\nreal-world causal modeling is often hindered by model and parameter\nuncertainty, observational noise, and chaotic behavior. The reliability of\ncounterfactual analysis in such settings remains largely unexplored. In this\nwork, we investigate the limitations of counterfactual reasoning within the\nframework of Structural Causal Models. Specifically, we empirically investigate\n\\emph{counterfactual sequence estimation} and highlight cases where it becomes\nincreasingly unreliable. We find that realistic assumptions, such as low\ndegrees of model uncertainty or chaotic dynamics, can result in\ncounterintuitive outcomes, including dramatic deviations between predicted and\ntrue counterfactual trajectories. This work urges caution when applying\ncounterfactual reasoning in settings characterized by chaos and uncertainty.\nFurthermore, it raises the question of whether certain systems may pose\nfundamental limitations on the ability to answer counterfactual questions about\ntheir behavior.\n", "date": "2025-03-31", "categories": []}
{"id": "2504.07257", "title": "Better Decisions through the Right Causal World Model", "abstract": "  Reinforcement learning (RL) agents have shown remarkable performances in\nvarious environments, where they can discover effective policies directly from\nsensory inputs. However, these agents often exploit spurious correlations in\nthe training data, resulting in brittle behaviours that fail to generalize to\nnew or slightly modified environments. To address this, we introduce the Causal\nObject-centric Model Extraction Tool (COMET), a novel algorithm designed to\nlearn the exact interpretable causal world models (CWMs). COMET first extracts\nobject-centric state descriptions from observations and identifies the\nenvironment's internal states related to the depicted objects' properties.\nUsing symbolic regression, it models object-centric transitions and derives\ncausal relationships governing object dynamics. COMET further incorporates\nlarge language models (LLMs) for semantic inference, annotating causal\nvariables to enhance interpretability.\n  By leveraging these capabilities, COMET constructs CWMs that align with the\ntrue causal structure of the environment, enabling agents to focus on\ntask-relevant features. The extracted CWMs mitigate the danger of shortcuts,\npermitting the development of RL systems capable of better planning and\ndecision-making across dynamic scenarios. Our results, validated in Atari\nenvironments such as Pong and Freeway, demonstrate the accuracy and robustness\nof COMET, highlighting its potential to bridge the gap between object-centric\nreasoning and causal inference in reinforcement learning.\n", "date": "2025-04-09", "categories": []}
{"id": "2504.07921", "title": "Note on the identification of total effect in Cluster-DAGs with cycles", "abstract": "  In this note, we discuss the identifiability of a total effect in\ncluster-DAGs, allowing for cycles within the cluster-DAG (while still assuming\nthe associated underlying DAG to be acyclic). This is presented into two key\nresults: first, restricting the cluster-DAG to clusters containing at most four\nnodes; second, adapting the notion of d-separation. We provide a graphical\ncriterion to address the identifiability problem.\n", "date": "2025-04-10", "categories": []}
{"id": "2209.14664", "title": "Causal inference in drug discovery and development", "abstract": "  To discover new drugs is to seek and to prove causality. As an emerging\napproach leveraging human knowledge and creativity, data, and machine\nintelligence, causal inference holds the promise of reducing cognitive bias and\nimproving decision making in drug discovery. While it has been applied across\nthe value chain, the concepts and practice of causal inference remain obscure\nto many practitioners. This article offers a non-technical introduction to\ncausal inference, reviews its recent applications, and discusses opportunities\nand challenges of adopting the causal language in drug discovery and\ndevelopment.\n", "date": "2022-09-29", "categories": []}
{"id": "2310.15117", "title": "Causal Order: The Key to Leveraging Imperfect Experts in Causal\n  Inference", "abstract": "  Large Language Models (LLMs) have been used as experts to infer causal\ngraphs, often by repeatedly applying a pairwise prompt that asks about the\ncausal relationship of each variable pair. However, such experts, including\nhuman domain experts, cannot distinguish between direct and indirect effects\ngiven a pairwise prompt. Therefore, instead of the graph, we propose that\ncausal order be used as a more stable output interface for utilizing expert\nknowledge. Even when querying a perfect expert with a pairwise prompt, we show\nthat the inferred graph can have significant errors whereas the causal order is\nalways correct. In practice, however, LLMs are imperfect experts and we find\nthat pairwise prompts lead to multiple cycles. Hence, we propose the triplet\nmethod, a novel querying strategy that introduces an auxiliary variable for\nevery variable pair and instructs the LLM to avoid cycles within this triplet.\nIt then uses a voting-based ensemble method that results in higher accuracy and\nfewer cycles while ensuring cost efficiency. Across multiple real-world graphs,\nsuch a triplet-based method yields a more accurate order than the pairwise\nprompt, using both LLMs and human annotators. The triplet method enhances\nrobustness by repeatedly querying an expert with different auxiliary variables,\nenabling smaller models like Phi-3 and Llama-3 8B Instruct to surpass GPT-4\nwith pairwise prompting. For practical usage, we show how the expert-provided\ncausal order from the triplet method can be used to reduce error in downstream\ngraph discovery and effect inference tasks.\n", "date": "2023-10-23", "categories": []}
{"id": "2504.02694", "title": "Semiparametric Counterfactual Regression", "abstract": "  We study counterfactual regression, which aims to map input features to\noutcomes under hypothetical scenarios that differ from those observed in the\ndata. This is particularly useful for decision-making when adapting to sudden\nshifts in treatment patterns is essential. We propose a doubly robust-style\nestimator for counterfactual regression within a generalizable framework that\naccommodates a broad class of risk functions and flexible constraints, drawing\non tools from semiparametric theory and stochastic optimization. Our approach\nuses incremental interventions to enhance adaptability while maintaining\nconsistency with standard methods. We formulate the target estimand as the\noptimal solution to a stochastic optimization problem and develop an efficient\nestimation strategy, where we can leverage rapid development of modern\noptimization algorithms. We go on to analyze the rates of convergence and\ncharacterize the asymptotic distributions. Our analysis shows that the proposed\nestimators can achieve $\\sqrt{n}$-consistency and asymptotic normality for a\nbroad class of problems. Numerical illustrations highlight their effectiveness\nin adapting to unseen counterfactual scenarios while maintaining parametric\nconvergence rates.\n", "date": "2025-04-03", "categories": []}
{"id": "2504.03753", "title": "MMCE: A Framework for Deep Monotonic Modeling of Multiple Causal Effects", "abstract": "  When we plan to use money as an incentive to change the behavior of a person\n(such as making riders to deliver more orders or making consumers to buy more\nitems), the common approach of this problem is to adopt a two-stage framework\nin order to maximize ROI under cost constraints. In the first stage, the\nindividual price response curve is obtained. In the second stage, business\ngoals and resource constraints are formally expressed and modeled as an\noptimization problem. The first stage is very critical. It can answer a very\nimportant question. This question is how much incremental results can\nincentives bring, which is the basis of the second stage. Usually, the causal\nmodeling is used to obtain the curve. In the case of only observational data,\ncausal modeling and evaluation are very challenging. In some business\nscenarios, multiple causal effects need to be obtained at the same time. This\npaper proposes a new observational data modeling and evaluation framework,\nwhich can simultaneously model multiple causal effects and greatly improve the\nmodeling accuracy under some abnormal distributions. In the absence of RCT\ndata, evaluation seems impossible. This paper summarizes three priors to\nillustrate the necessity and feasibility of qualitative evaluation of cognitive\ntesting. At the same time, this paper innovatively proposes the conditions\nunder which observational data can be considered as an evaluation dataset. Our\napproach is very groundbreaking. It is the first to propose a modeling\nframework that simultaneously obtains multiple causal effects. The offline\nanalysis and online experimental results show the effectiveness of the results\nand significantly improve the effectiveness of the allocation strategies\ngenerated in real world marketing activities.\n", "date": "2025-04-02", "categories": []}
{"id": "2504.04013", "title": "Spatially-Heterogeneous Causal Bayesian Networks for Seismic\n  Multi-Hazard Estimation: A Variational Approach with Gaussian Processes and\n  Normalizing Flows", "abstract": "  Post-earthquake hazard and impact estimation are critical for effective\ndisaster response, yet current approaches face significant limitations.\nTraditional models employ fixed parameters regardless of geographical context,\nmisrepresenting how seismic effects vary across diverse landscapes, while\nremote sensing technologies struggle to distinguish between co-located hazards.\nWe address these challenges with a spatially-aware causal Bayesian network that\ndecouples co-located hazards by modeling their causal relationships with\nlocation-specific parameters. Our framework integrates sensing observations,\nlatent variables, and spatial heterogeneity through a novel combination of\nGaussian Processes with normalizing flows, enabling us to capture how same\nearthquake produces different effects across varied geological and\ntopographical features. Evaluations across three earthquakes demonstrate\nSpatial-VCBN achieves Area Under the Curve (AUC) improvements of up to 35.2%\nover existing methods. These results highlight the critical importance of\nmodeling spatial heterogeneity in causal mechanisms for accurate disaster\nassessment, with direct implications for improving emergency response resource\nallocation.\n", "date": "2025-04-05", "categories": []}
{"id": "2504.04015", "title": "Multi-resolution Score-Based Variational Graphical Diffusion for Causal\n  Disaster System Modeling and Inference", "abstract": "  Complex systems with intricate causal dependencies challenge accurate\nprediction. Effective modeling requires precise physical process\nrepresentation, integration of interdependent factors, and incorporation of\nmulti-resolution observational data. These systems manifest in both static\nscenarios with instantaneous causal chains and temporal scenarios with evolving\ndynamics, complicating modeling efforts. Current methods struggle to\nsimultaneously handle varying resolutions, capture physical relationships,\nmodel causal dependencies, and incorporate temporal dynamics, especially with\ninconsistently sampled data from diverse sources. We introduce Temporal-SVGDM:\nScore-based Variational Graphical Diffusion Model for Multi-resolution\nobservations. Our framework constructs individual SDEs for each variable at its\nnative resolution, then couples these SDEs through a causal score mechanism\nwhere parent nodes inform child nodes' evolution. This enables unified modeling\nof both immediate causal effects in static scenarios and evolving dependencies\nin temporal scenarios. In temporal models, state representations are processed\nthrough a sequence prediction model to predict future states based on\nhistorical patterns and causal relationships. Experiments on real-world\ndatasets demonstrate improved prediction accuracy and causal understanding\ncompared to existing methods, with robust performance under varying levels of\nbackground knowledge. Our model exhibits graceful degradation across different\ndisaster types, successfully handling both static earthquake scenarios and\ntemporal hurricane and wildfire scenarios, while maintaining superior\nperformance even with limited data.\n", "date": "2025-04-05", "categories": []}
{"id": "2407.05625", "title": "New User Event Prediction Through the Lens of Causal Inference", "abstract": "  Modeling and analysis for event series generated by users of heterogeneous\nbehavioral patterns are closely involved in our daily lives, including credit\ncard fraud detection, online platform user recommendation, and social network\nanalysis. The most commonly adopted approach to this task is to assign users to\nbehavior-based categories and analyze each of them separately. However, this\nrequires extensive data to fully understand the user behavior, presenting\nchallenges in modeling newcomers without significant historical knowledge. In\nthis work, we propose a novel discrete event prediction framework for new users\nwith limited history, without needing to know the user's category. We treat the\nuser event history as the \"treatment\" for future events and the user category\nas the key confounder. Thus, the prediction problem can be framed as\ncounterfactual outcome estimation, where each event is re-weighted by its\ninverse propensity score. We demonstrate the improved performance of the\nproposed framework with a numerical simulation study and two real-world\napplications, including Netflix rating prediction and seller contact prediction\nfor customer support at Amazon.\n", "date": "2024-07-08", "categories": []}
{"id": "2504.02456", "title": "The Amenability Framework: Rethinking Causal Ordering Without Estimating\n  Causal Effects", "abstract": "  Who should we prioritize for intervention when we cannot estimate\nintervention effects? In many applied domains (e.g., advertising, customer\nretention, and behavioral nudging) prioritization is guided by predictive\nmodels that estimate outcome probabilities rather than causal effects. This\npaper investigates when these predictions (scores) can effectively rank\nindividuals by their intervention effects, particularly when direct effect\nestimation is infeasible or unreliable. We propose a conceptual framework based\non amenability: an individual's latent proclivity to be influenced by an\nintervention. We then formalize conditions under which predictive scores serve\nas effective proxies for amenability. These conditions justify using non-causal\nscores for intervention prioritization, even when the scores do not directly\nestimate effects. We further show that, under plausible assumptions, predictive\nmodels can outperform causal effect estimators in ranking individuals by\nintervention effects. Empirical evidence from an advertising context supports\nour theoretical findings, demonstrating that predictive modeling can offer a\nmore robust approach to targeting than effect estimation. Our framework\nsuggests a shift in focus, from estimating effects to inferring who is\namenable, as a practical and theoretically grounded strategy for prioritizing\ninterventions in resource-constrained environments.\n", "date": "2025-04-03", "categories": []}
{"id": "2504.01398", "title": "Cause or Trigger? From Philosophy to Causal Modeling", "abstract": "  Not much has been written about the role of triggers in the literature on\ncausal reasoning, causal modeling, or philosophy. In this paper, we focus on\ndescribing triggers and causes in the metaphysical sense and on\ncharacterizations that differentiate them from each other. We carry out a\nphilosophical analysis of these differences. From this, we formulate a\ndefinition that clearly differentiates triggers from causes and can be used for\ncausal reasoning in natural sciences. We propose a mathematical model and the\nCause-Trigger algorithm, which, based on given data to observable processes, is\nable to determine whether a process is a cause or a trigger of an effect. The\npossibility to distinguish triggers from causes directly from data makes the\nalgorithm a useful tool in natural sciences using observational data, but also\nfor real-world scenarios. For example, knowing the processes that trigger\ncauses of a tropical storm could give politicians time to develop actions such\nas evacuation the population. Similarly, knowing the triggers of processes that\ncause global warming could help politicians focus on effective actions. We\ndemonstrate our algorithm on the climatological data of two recent cyclones,\nFreddy and Zazu. The Cause-Trigger algorithm detects processes that trigger\nhigh wind speed in both storms during their cyclogenesis. The findings obtained\nagree with expert knowledge.\n", "date": "2025-04-02", "categories": []}
{"id": "2504.01424", "title": "On the Role of Priors in Bayesian Causal Learning", "abstract": "  In this work, we investigate causal learning of independent causal mechanisms\nfrom a Bayesian perspective. Confirming previous claims from the literature, we\nshow in a didactically accessible manner that unlabeled data (i.e., cause\nrealizations) do not improve the estimation of the parameters defining the\nmechanism. Furthermore, we observe the importance of choosing an appropriate\nprior for the cause and mechanism parameters, respectively. Specifically, we\nshow that a factorized prior results in a factorized posterior, which resonates\nwith Janzing and Sch\\\"olkopf's definition of independent causal mechanisms via\nthe Kolmogorov complexity of the involved distributions and with the concept of\nparameter independence of Heckerman et al.\n", "date": "2025-04-02", "categories": []}
{"id": "2504.01702", "title": "A Causal Inference Framework for Data Rich Environments", "abstract": "  We propose a formal model for counterfactual estimation with unobserved\nconfounding in \"data-rich\" settings, i.e., where there are a large number of\nunits and a large number of measurements per unit. Our model provides a bridge\nbetween the structural causal model view of causal inference common in the\ngraphical models literature with that of the latent factor model view common in\nthe potential outcomes literature. We show how classic models for potential\noutcomes and treatment assignments fit within our framework. We provide an\nidentification argument for the average treatment effect, the average treatment\neffect on the treated, and the average treatment effect on the untreated. For\nany estimator that has a fast enough estimation error rate for a certain\nnuisance parameter, we establish it is consistent for these various causal\nparameters. We then show principal component regression is one such estimator\nthat leads to consistent estimation, and we analyze the minimal smoothness\nrequired of the potential outcomes function for consistency.\n", "date": "2025-04-02", "categories": []}
{"id": "2311.02467", "title": "Individualized Policy Evaluation and Learning under Clustered Network\n  Interference", "abstract": "  Although there is now a large literature on policy evaluation and learning,\nmuch of the prior work assumes that the treatment assignment of one unit does\nnot affect the outcome of another unit. Unfortunately, ignoring interference\ncan lead to biased policy evaluation and ineffective learned policies. For\nexample, treating influential individuals who have many friends can generate\npositive spillover effects, thereby improving the overall performance of an\nindividualized treatment rule (ITR). We consider the problem of evaluating and\nlearning an optimal ITR under clustered network interference (also known as\npartial interference), where clusters of units are sampled from a population\nand units may influence one another within each cluster. Unlike previous\nmethods that impose strong restrictions on spillover effects, such as anonymous\ninterference, the proposed methodology only assumes a semiparametric structural\nmodel, where each unit's outcome is an additive function of individual\ntreatments within the cluster. Under this model, we propose an estimator that\ncan be used to evaluate the empirical performance of an ITR. We show that this\nestimator is substantially more efficient than the standard inverse probability\nweighting estimator, which does not impose any assumption about spillover\neffects. We derive the finite-sample regret bound for a learned ITR, showing\nthat the use of our efficient evaluation estimator leads to the improved\nperformance of learned policies. We consider both experimental and\nobservational studies, and for the latter, we develop a doubly robust estimator\nthat is semiparametrically efficient and yields an optimal regret bound.\nFinally, we conduct simulation and empirical studies to illustrate the\nadvantages of the proposed methodology.\n", "date": "2023-11-04", "categories": []}
{"id": "2503.09194", "title": "Addressing pitfalls in implicit unobserved confounding synthesis using\n  explicit block hierarchical ancestral sampling", "abstract": "  Unbiased data synthesis is crucial for evaluating causal discovery algorithms\nin the presence of unobserved confounding, given the scarcity of real-world\ndatasets. A common approach, implicit parameterization, encodes unobserved\nconfounding by modifying the off-diagonal entries of the idiosyncratic\ncovariance matrix while preserving positive definiteness. Within this approach,\nwe identify that state-of-the-art protocols have two distinct issues that\nhinder unbiased sampling from the complete space of causal models: first, we\ngive a detailed analysis of use of diagonally dominant constructions restricts\nthe spectrum of partial correlation matrices; and second, the restriction of\npossible graphical structures when sampling bidirected edges, unnecessarily\nruling out valid causal models. To address these limitations, we propose an\nimproved explicit modeling approach for unobserved confounding, leveraging\nblock-hierarchical ancestral generation of ground truth causal graphs.\nAlgorithms for converting the ground truth DAG into ancestral graph is provided\nso that the output of causal discovery algorithms could be compared with. We\ndraw connections between implicit and explicit parameterization, prove that our\napproach fully covers the space of causal models, including those generated by\nthe implicit parameterization, thus enabling more robust evaluation of methods\nfor causal discovery and inference.\n", "date": "2025-03-12", "categories": []}
{"id": "2402.01929", "title": "Sample, estimate, aggregate: A recipe for causal discovery foundation\n  models", "abstract": "  Causal discovery, the task of inferring causal structure from data, has the\npotential to uncover mechanistic insights from biological experiments,\nespecially those involving perturbations. However, causal discovery algorithms\nover larger sets of variables tend to be brittle against misspecification or\nwhen data are limited. For example, single-cell transcriptomics measures\nthousands of genes, but the nature of their relationships is not known, and\nthere may be as few as tens of cells per intervention setting. To mitigate\nthese challenges, we propose a foundation model-inspired approach: a supervised\nmodel trained on large-scale, synthetic data to predict causal graphs from\nsummary statistics -- like the outputs of classical causal discovery algorithms\nrun over subsets of variables and other statistical hints like inverse\ncovariance. Our approach is enabled by the observation that typical errors in\nthe outputs of a discovery algorithm remain comparable across datasets.\nTheoretically, we show that the model architecture is well-specified, in the\nsense that it can recover a causal graph consistent with graphs over subsets.\nEmpirically, we train the model to be robust to misspecification and\ndistribution shift using diverse datasets. Experiments on biological and\nsynthetic data confirm that this model generalizes well beyond its training\nset, runs on graphs with hundreds of variables in seconds, and can be easily\nadapted to different underlying data assumptions.\n", "date": "2024-02-02", "categories": []}
{"id": "2503.23605", "title": "Partial Transportability for Domain Generalization", "abstract": "  A fundamental task in AI is providing performance guarantees for predictions\nmade in unseen domains. In practice, there can be substantial uncertainty about\nthe distribution of new data, and corresponding variability in the performance\nof existing predictors. Building on the theory of partial identification and\ntransportability, this paper introduces new results for bounding the value of a\nfunctional of the target distribution, such as the generalization error of a\nclassifier, given data from source domains and assumptions about the data\ngenerating mechanisms, encoded in causal diagrams. Our contribution is to\nprovide the first general estimation technique for transportability problems,\nadapting existing parameterization schemes such Neural Causal Models to encode\nthe structural constraints necessary for cross-population inference. We\ndemonstrate the expressiveness and consistency of this procedure and further\npropose a gradient-based optimization scheme for making scalable inferences in\npractice. Our results are corroborated with experiments.\n", "date": "2025-03-30", "categories": []}
{"id": "2312.15574", "title": "Clustered Switchback Designs for Experimentation Under Spatio-temporal\n  Interference", "abstract": "  We consider experimentation in the presence of non-stationarity, inter-unit\n(spatial) interference, and carry-over effects (temporal interference), where\nwe wish to estimate the global average treatment effect (GATE), the difference\nbetween average outcomes having exposed all units at all times to treatment or\nto control. We suppose spatial interference is described by a graph, where a\nunit's outcome depends on its neighborhood's treatments, and that temporal\ninterference is described by an MDP, where the transition kernel under either\ntreatment (action) satisfies a rapid mixing condition. We propose a clustered\nswitchback design, where units are grouped into clusters and time steps are\ngrouped into blocks, and each whole cluster-block combination is assigned a\nsingle random treatment. Under this design, we show that for graphs that admit\ngood clustering, a truncated Horvitz-Thompson estimator achieves a $\\tilde\nO(1/NT)$ mean squared error (MSE), matching the lower bound up to logarithmic\nterms for sparse graphs. Our results simultaneously generalize the results from\n\\citet{hu2022switchback,ugander2013graph} and \\citet{leung2022rate}. Simulation\nstudies validate the favorable performance of our approach.\n", "date": "2023-12-25", "categories": []}
{"id": "2402.08514", "title": "Counterfactual Influence in Markov Decision Processes", "abstract": "  Our work addresses a fundamental problem in the context of counterfactual\ninference for Markov Decision Processes (MDPs). Given an MDP path $\\tau$, this\nkind of inference allows us to derive counterfactual paths $\\tau'$ describing\nwhat-if versions of $\\tau$ obtained under different action sequences than those\nobserved in $\\tau$. However, as the counterfactual states and actions deviate\nfrom the observed ones over time, the observation $\\tau$ may no longer\ninfluence the counterfactual world, meaning that the analysis is no longer\ntailored to the individual observation, resulting in interventional outcomes\nrather than counterfactual ones. Even though this issue specifically affects\nthe popular Gumbel-max structural causal model used for MDP counterfactuals, it\nhas remained overlooked until now. In this work, we introduce a formal\ncharacterisation of influence based on comparing counterfactual and\ninterventional distributions. We devise an algorithm to construct\ncounterfactual models that automatically satisfy influence constraints.\nLeveraging such models, we derive counterfactual policies that are not just\noptimal for a given reward structure but also remain tailored to the observed\npath. Even though there is an unavoidable trade-off between policy optimality\nand strength of influence constraints, our experiments demonstrate that it is\npossible to derive (near-)optimal policies while remaining under the influence\nof the observation.\n", "date": "2024-02-13", "categories": []}
{"id": "2412.04981", "title": "Causal discovery with endogenous context variables", "abstract": "  Causal systems often exhibit variations of the underlying causal mechanisms\nbetween the variables of the system. Often, these changes are driven by\ndifferent environments or internal states in which the system operates, and we\nrefer to context variables as those variables that indicate this change in\ncausal mechanisms. An example are the causal relations in soil\nmoisture-temperature interactions and their dependence on soil moisture\nregimes: Dry soil triggers a dependence of soil moisture on latent heat, while\nenvironments with wet soil do not feature such a feedback, making it a\ncontext-specific property. Crucially, a regime or context variable such as soil\nmoisture need not be exogenous and can be influenced by the dynamical system\nvariables - precipitation can make a dry soil wet - leading to joint systems\nwith endogenous context variables. In this work we investigate the assumptions\nfor constraint-based causal discovery of context-specific information in\nsystems with endogenous context variables. We show that naive approaches such\nas learning different regime graphs on masked data, or pooling all data, can\nlead to uninformative results. We propose an adaptive constraint-based\ndiscovery algorithm and give a detailed discussion on the connection to\nstructural causal models, including sufficiency assumptions, which allow to\nprove the soundness of our algorithm and to interpret the results causally.\nNumerical experiments demonstrate the performance of the proposed method over\nalternative baselines, but they also unveil current limitations of our method.\n", "date": "2024-12-06", "categories": []}
{"id": "2503.21629", "title": "ClusterSC: Advancing Synthetic Control with Donor Selection", "abstract": "  In causal inference with observational studies, synthetic control (SC) has\nemerged as a prominent tool. SC has traditionally been applied to\naggregate-level datasets, but more recent work has extended its use to\nindividual-level data. As they contain a greater number of observed units, this\nshift introduces the curse of dimensionality to SC. To address this, we propose\nCluster Synthetic Control (ClusterSC), based on the idea that groups of\nindividuals may exist where behavior aligns internally but diverges between\ngroups. ClusterSC incorporates a clustering step to select only the relevant\ndonors for the target. We provide theoretical guarantees on the improvements\ninduced by ClusterSC, supported by empirical demonstrations on synthetic and\nreal-world datasets. The results indicate that ClusterSC consistently\noutperforms classical SC approaches.\n", "date": "2025-03-27", "categories": []}
{"id": "2305.17479", "title": "Inferring Individual Direct Causal Effects Under Heterogeneous Peer\n  Influence", "abstract": "  Causal inference in networks should account for interference, which occurs\nwhen a unit's outcome is influenced by treatments or outcomes of peers.\nHeterogeneous peer influence (HPI) occurs when a unit's outcome is influenced\ndifferently by different peers based on their attributes and relationships, or\nwhen each unit has a different susceptibility to peer influence. Existing\nsolutions to estimating direct causal effects under interference consider\neither homogeneous influence from peers or specific heterogeneous influence\nmechanisms (e.g., based on local neighborhood structure). This paper presents a\nmethodology for estimating individual direct causal effects in the presence of\nHPI where the mechanism of influence is not known a priori. We propose a\nstructural causal model for networks that can capture different possible\nassumptions about network structure, interference conditions, and causal\ndependence and enables reasoning about identifiability in the presence of HPI.\nWe find potential heterogeneous contexts using the causal model and propose a\nnovel graph neural network-based estimator to estimate individual direct causal\neffects. We show that state-of-the-art methods for individual direct effect\nestimation produce biased results in the presence of HPI, and that our proposed\nestimator is robust.\n", "date": "2023-05-27", "categories": []}
{"id": "2503.20546", "title": "Regression-Based Estimation of Causal Effects in the Presence of\n  Selection Bias and Confounding", "abstract": "  We consider the problem of estimating the expected causal effect $E[Y|do(X)]$\nfor a target variable $Y$ when treatment $X$ is set by intervention, focusing\non continuous random variables. In settings without selection bias or\nconfounding, $E[Y|do(X)] = E[Y|X]$, which can be estimated using standard\nregression methods. However, regression fails when systematic missingness\ninduced by selection bias, or confounding distorts the data. Boeken et al.\n[2023] show that when training data is subject to selection, proxy variables\nunaffected by this process can, under certain constraints, be used to correct\nfor selection bias to estimate $E[Y|X]$, and hence $E[Y|do(X)]$, reliably. When\ndata is additionally affected by confounding, however, this equality is no\nlonger valid.\n  Building on these results, we consider a more general setting and propose a\nframework that incorporates both selection bias and confounding. Specifically,\nwe derive theoretical conditions ensuring identifiability and recoverability of\ncausal effects under access to external data and proxy variables. We further\nintroduce a two-step regression estimator (TSR), capable of exploiting proxy\nvariables to adjust for selection bias while accounting for confounding. We\nshow that TSR coincides with prior work if confounding is absent, but achieves\na lower variance. Extensive simulation studies validate TSR's correctness for\nscenarios which may include both selection bias and confounding with proxy\nvariables.\n", "date": "2025-03-26", "categories": []}
{"id": "2409.17027", "title": "Counterfactual Token Generation in Large Language Models", "abstract": "  \"Sure, I am happy to generate a story for you: Captain Lyra stood at the helm\nof her trusty ship, the Maelstrom's Fury, gazing out at the endless sea. [...]\nLyra's eyes welled up with tears as she realized the bitter truth - she had\nsacrificed everything for fleeting riches, and lost the love of her crew, her\nfamily, and herself.\" Although this story, generated by a large language model,\nis captivating, one may wonder -- how would the story have unfolded if the\nmodel had chosen \"Captain Maeve\" as the protagonist instead? We cannot know.\nState-of-the-art large language models are stateless -- they maintain no\ninternal memory or state. Given a prompt, they generate a sequence of tokens as\nan output using an autoregressive process. As a consequence, they cannot reason\nabout counterfactual alternatives to tokens they have generated in the past. In\nthis work, our goal is to enhance them with this functionality. To this end, we\ndevelop a causal model of token generation that builds upon the Gumbel-Max\nstructural causal model. Our model allows any large language model to perform\ncounterfactual token generation at almost no cost in comparison with vanilla\ntoken generation, it is embarrassingly simple to implement, and it does not\nrequire any fine-tuning nor prompt engineering. We implement our model on Llama\n3 8B-Instruct and Ministral-8B-Instruct and conduct a qualitative and a\nquantitative analysis of counterfactually generated text. We conclude with a\ndemonstrative application of counterfactual token generation for bias\ndetection, unveiling interesting insights about the model of the world\nconstructed by large language models.\n", "date": "2024-09-25", "categories": []}
{"id": "2410.16870", "title": "Federated Causal Inference: Multi-Study ATE Estimation beyond\n  Meta-Analysis", "abstract": "  We study Federated Causal Inference, an approach to estimate treatment\neffects from decentralized data across centers. We compare three classes of\nAverage Treatment Effect (ATE) estimators derived from the Plug-in G-Formula,\nranging from simple meta-analysis to one-shot and multi-shot federated\nlearning, the latter leveraging the full data to learn the outcome model\n(albeit requiring more communication). Focusing on Randomized Controlled Trials\n(RCTs), we derive the asymptotic variance of these estimators for linear\nmodels. Our results provide practical guidance on selecting the appropriate\nestimator for various scenarios, including heterogeneity in sample sizes,\ncovariate distributions, treatment assignment schemes, and center effects. We\nvalidate these findings with a simulation study.\n", "date": "2024-10-22", "categories": []}
{"id": "2503.19218", "title": "Analytic DAG Constraints for Differentiable DAG Learning", "abstract": "  Recovering the underlying Directed Acyclic Graph (DAG) structures from\nobservational data presents a formidable challenge, partly due to the\ncombinatorial nature of the DAG-constrained optimization problem. Recently,\nresearchers have identified gradient vanishing as one of the primary obstacles\nin differentiable DAG learning and have proposed several DAG constraints to\nmitigate this issue. By developing the necessary theory to establish a\nconnection between analytic functions and DAG constraints, we demonstrate that\nanalytic functions from the set $\\{f(x) = c_0 + \\sum_{i=1}^{\\infty}c_ix^i |\n\\forall i > 0, c_i > 0; r = \\lim_{i\\rightarrow \\infty}c_{i}/c_{i+1} > 0\\}$ can\nbe employed to formulate effective DAG constraints. Furthermore, we establish\nthat this set of functions is closed under several functional operators,\nincluding differentiation, summation, and multiplication. Consequently, these\noperators can be leveraged to create novel DAG constraints based on existing\nones. Using these properties, we design a series of DAG constraints and develop\nan efficient algorithm to evaluate them. Experiments in various settings\ndemonstrate that our DAG constraints outperform previous state-of-the-art\ncomparators. Our implementation is available at\nhttps://github.com/zzhang1987/AnalyticDAGLearning.\n", "date": "2025-03-24", "categories": []}
{"id": "2503.19554", "title": "Causal Bayesian Optimization with Unknown Graphs", "abstract": "  Causal Bayesian Optimization (CBO) is a methodology designed to optimize an\noutcome variable by leveraging known causal relationships through targeted\ninterventions. Traditional CBO methods require a fully and accurately specified\ncausal graph, which is a limitation in many real-world scenarios where such\ngraphs are unknown. To address this, we propose a new method for the CBO\nframework that operates without prior knowledge of the causal graph. Consistent\nwith causal bandit theory, we demonstrate through theoretical analysis and that\nfocusing on the direct causal parents of the target variable is sufficient for\noptimization, and provide empirical validation in the context of CBO.\nFurthermore we introduce a new method that learns a Bayesian posterior over the\ndirect parents of the target variable. This allows us to optimize the outcome\nvariable while simultaneously learning the causal structure. Our contributions\ninclude a derivation of the closed-form posterior distribution for the linear\ncase. In the nonlinear case where the posterior is not tractable, we present a\nGaussian Process (GP) approximation that still enables CBO by inferring the\nparents of the outcome variable. The proposed method performs competitively\nwith existing benchmarks and scales well to larger graphs, making it a\npractical tool for real-world applications where causal information is\nincomplete.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.19873", "title": "Identification of Average Treatment Effects in Nonparametric Panel\n  Models", "abstract": "  This paper studies identification of average treatment effects in a panel\ndata setting. It introduces a novel nonparametric factor model and proves\nidentification of average treatment effects. The identification proof is based\non the introduction of a consistent estimator. Underlying the proof is a result\nthat there is a consistent estimator for the expected outcome in the absence of\nthe treatment for each unit and time period; this result can be applied more\nbroadly, for example in problems of decompositions of group-level differences\nin outcomes, such as the much-studied gender wage gap.\n", "date": "2025-03-25", "categories": []}
{"id": "2306.04777", "title": "Invariant Causal Set Covering Machines", "abstract": "  Rule-based models, such as decision trees, appeal to practitioners due to\ntheir interpretable nature. However, the learning algorithms that produce such\nmodels are often vulnerable to spurious associations and thus, they are not\nguaranteed to extract causally-relevant insights. In this work, we build on\nideas from the invariant causal prediction literature to propose Invariant\nCausal Set Covering Machines, an extension of the classical Set Covering\nMachine algorithm for conjunctions/disjunctions of binary-valued rules that\nprovably avoids spurious associations. We demonstrate both theoretically and\nempirically that our method can identify the causal parents of a variable of\ninterest in polynomial time.\n", "date": "2023-06-07", "categories": []}
{"id": "2307.10299", "title": "Causality-oriented robustness: exploiting general noise interventions", "abstract": "  Since distribution shifts are common in real-world applications, there is a\npressing need to develop prediction models that are robust against such shifts.\nExisting frameworks, such as empirical risk minimization or distributionally\nrobust optimization, either lack generalizability for unseen distributions or\nrely on postulated distance measures. Alternatively, causality offers a\ndata-driven and structural perspective to robust predictions. However, the\nassumptions necessary for causal inference can be overly stringent, and the\nrobustness offered by such causal models often lacks flexibility. In this\npaper, we focus on causality-oriented robustness and propose Distributional\nRobustness via Invariant Gradients (DRIG), a method that exploits general noise\ninterventions in training data for robust predictions against unseen\ninterventions, and naturally interpolates between in-distribution prediction\nand causality. In a linear setting, we prove that DRIG yields predictions that\nare robust among a data-dependent class of distribution shifts. Furthermore, we\nshow that our framework includes anchor regression as a special case, and that\nit yields prediction models that protect against more diverse perturbations. We\nestablish finite-sample results and extend our approach to semi-supervised\ndomain adaptation to further improve prediction performance. Finally, we\nempirically validate our methods on synthetic simulations and on single-cell\nand intensive health care datasets.\n", "date": "2023-07-18", "categories": []}
{"id": "2407.18755", "title": "Score matching through the roof: linear, nonlinear, and latent variables\n  causal discovery", "abstract": "  Causal discovery from observational data holds great promise, but existing\nmethods rely on strong assumptions about the underlying causal structure, often\nrequiring full observability of all relevant variables. We tackle these\nchallenges by leveraging the score function $\\nabla \\log p(X)$ of observed\nvariables for causal discovery and propose the following contributions. First,\nwe fine-tune the existing identifiability results with the score on additive\nnoise models, showing that their assumption of nonlinearity of the causal\nmechanisms is not necessary. Second, we establish conditions for inferring\ncausal relations from the score even in the presence of hidden variables; this\nresult is two-faced: we demonstrate the score's potential to infer the\nequivalence class of causal graphs with hidden variables (while previous\nresults are restricted to the fully observable setting), and we provide\nsufficient conditions for identifying direct causes in latent variable models.\nBuilding on these insights, we propose a flexible algorithm suited for causal\ndiscovery on linear, nonlinear, and latent variable models, which we\nempirically validate.\n", "date": "2024-07-26", "categories": []}
{"id": "2503.17452", "title": "CausalRivers -- Scaling up benchmarking of causal discovery for\n  real-world time-series", "abstract": "  Causal discovery, or identifying causal relationships from observational\ndata, is a notoriously challenging task, with numerous methods proposed to\ntackle it. Despite this, in-the-wild evaluation of these methods is still\nlacking, as works frequently rely on synthetic data evaluation and sparse\nreal-world examples under critical theoretical assumptions. Real-world causal\nstructures, however, are often complex, making it hard to decide on a proper\ncausal discovery strategy. To bridge this gap, we introduce CausalRivers, the\nlargest in-the-wild causal discovery benchmarking kit for time-series data to\ndate. CausalRivers features an extensive dataset on river discharge that covers\nthe eastern German territory (666 measurement stations) and the state of\nBavaria (494 measurement stations). It spans the years 2019 to 2023 with a\n15-minute temporal resolution. Further, we provide additional data from a flood\naround the Elbe River, as an event with a pronounced distributional shift.\nLeveraging multiple sources of information and time-series meta-data, we\nconstructed two distinct causal ground truth graphs (Bavaria and eastern\nGermany). These graphs can be sampled to generate thousands of subgraphs to\nbenchmark causal discovery across diverse and challenging settings. To\ndemonstrate the utility of CausalRivers, we evaluate several causal discovery\napproaches through a set of experiments to identify areas for improvement.\nCausalRivers has the potential to facilitate robust evaluations and comparisons\nof causal discovery methods. Besides this primary purpose, we also expect that\nthis dataset will be relevant for connected areas of research, such as\ntime-series forecasting and anomaly detection. Based on this, we hope to push\nbenchmark-driven method development that fosters advanced techniques for causal\ndiscovery, as is the case for many other areas of machine learning.\n", "date": "2025-03-21", "categories": []}
{"id": "2503.17569", "title": "Fairness-Driven LLM-based Causal Discovery with Active Learning and\n  Dynamic Scoring", "abstract": "  Causal discovery (CD) plays a pivotal role in numerous scientific fields by\nclarifying the causal relationships that underlie phenomena observed in diverse\ndisciplines. Despite significant advancements in CD algorithms that enhance\nbias and fairness analyses in machine learning, their application faces\nchallenges due to the high computational demands and complexities of\nlarge-scale data. This paper introduces a framework that leverages Large\nLanguage Models (LLMs) for CD, utilizing a metadata-based approach akin to the\nreasoning processes of human experts. By shifting from pairwise queries to a\nmore scalable breadth-first search (BFS) strategy, the number of required\nqueries is reduced from quadratic to linear in terms of variable count, thereby\naddressing scalability concerns inherent in previous approaches. This method\nutilizes an Active Learning (AL) and a Dynamic Scoring Mechanism that\nprioritizes queries based on their potential information gain, combining mutual\ninformation, partial correlation, and LLM confidence scores to refine the\ncausal graph more efficiently and accurately. This BFS query strategy reduces\nthe required number of queries significantly, thereby addressing scalability\nconcerns inherent in previous approaches. This study provides a more scalable\nand efficient solution for leveraging LLMs in fairness-driven CD, highlighting\nthe effects of the different parameters on performance. We perform fairness\nanalyses on the inferred causal graphs, identifying direct and indirect effects\nof sensitive attributes on outcomes. A comparison of these analyses against\nthose from graphs produced by baseline methods highlights the importance of\naccurate causal graph construction in understanding bias and ensuring fairness\nin machine learning systems.\n", "date": "2025-03-21", "categories": []}
{"id": "2502.04210", "title": "Algorithmic causal structure emerging through compression", "abstract": "  We explore the relationship between causality, symmetry, and compression. We\nbuild on and generalize the known connection between learning and compression\nto a setting where causal models are not identifiable. We propose a framework\nwhere causality emerges as a consequence of compressing data across multiple\nenvironments. We define algorithmic causality as an alternative definition of\ncausality when traditional assumptions for causal identifiability do not hold.\nWe demonstrate how algorithmic causal and symmetric structures can emerge from\nminimizing upper bounds on Kolmogorov complexity, without knowledge of\nintervention targets. We hypothesize that these insights may also provide a\nnovel perspective on the emergence of causality in machine learning models,\nsuch as large language models, where causal relationships may not be explicitly\nidentifiable.\n", "date": "2025-02-06", "categories": []}
{"id": "2503.16544", "title": "Causal Discovery and Counterfactual Reasoning to Optimize Persuasive\n  Dialogue Policies", "abstract": "  Tailoring persuasive conversations to users leads to more effective\npersuasion. However, existing dialogue systems often struggle to adapt to\ndynamically evolving user states. This paper presents a novel method that\nleverages causal discovery and counterfactual reasoning for optimizing system\npersuasion capability and outcomes. We employ the Greedy Relaxation of the\nSparsest Permutation (GRaSP) algorithm to identify causal relationships between\nuser and system utterance strategies, treating user strategies as states and\nsystem strategies as actions. GRaSP identifies user strategies as causal\nfactors influencing system responses, which inform Bidirectional Conditional\nGenerative Adversarial Networks (BiCoGAN) in generating counterfactual\nutterances for the system. Subsequently, we use the Dueling Double Deep\nQ-Network (D3QN) model to utilize counterfactual data to determine the best\npolicy for selecting system utterances. Our experiments with the\nPersuasionForGood dataset show measurable improvements in persuasion outcomes\nusing our approach over baseline methods. The observed increase in cumulative\nrewards and Q-values highlights the effectiveness of causal discovery in\nenhancing counterfactual reasoning and optimizing reinforcement learning\npolicies for online dialogue systems.\n", "date": "2025-03-19", "categories": []}
{"id": "2409.19608", "title": "Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via\n  Diffusion Model", "abstract": "  Spatio-temporal (ST) prediction has garnered a De facto attention in earth\nsciences, such as meteorological prediction, human mobility perception.\nHowever, the scarcity of data coupled with the high expenses involved in sensor\ndeployment results in notable data imbalances. Furthermore, models that are\nexcessively customized and devoid of causal connections further undermine the\ngeneralizability and interpretability. To this end, we establish a causal\nframework for ST predictions, termed CaPaint, which targets to identify causal\nregions in data and endow model with causal reasoning ability in a two-stage\nprocess. Going beyond this process, we utilize the back-door adjustment to\nspecifically address the sub-regions identified as non-causal in the upstream\nphase. Specifically, we employ a novel image inpainting technique. By using a\nfine-tuned unconditional Diffusion Probabilistic Model (DDPM) as the generative\nprior, we in-fill the masks defined as environmental parts, offering the\npossibility of reliable extrapolation for potential data distributions. CaPaint\novercomes the high complexity dilemma of optimal ST causal discovery models by\nreducing the data generation complexity from exponential to quasi-linear\nlevels. Extensive experiments conducted on five real-world ST benchmarks\ndemonstrate that integrating the CaPaint concept allows models to achieve\nimprovements ranging from 4.3% to 77.3%. Moreover, compared to traditional\nmainstream ST augmenters, CaPaint underscores the potential of diffusion models\nin ST enhancement, offering a novel paradigm for this field. Our project is\navailable at https://anonymous.4open.science/r/12345-DFCC.\n", "date": "2024-09-29", "categories": []}
{"id": "2410.19464", "title": "LOCAL: Learning with Orientation Matrix to Infer Causal Structure from\n  Time Series Data", "abstract": "  Discovering the underlying Directed Acyclic Graph (DAG) from time series\nobservational data is highly challenging due to the dynamic nature and complex\nnonlinear interactions between variables. Existing methods typically search for\nthe optimal DAG by optimizing an objective function but face scalability\nchallenges, as their computational demands grow exponentially with the\ndimensional expansion of variables. To this end, we propose LOCAL, a highly\nefficient, easy-to-implement, and constraint-free method for recovering dynamic\ncausal structures. LOCAL is the first attempt to formulate a quasi-maximum\nlikelihood-based score function for learning the dynamic DAG equivalent to the\nground truth. Building on this, we introduce two adaptive modules that enhance\nthe algebraic characterization of acyclicity: Asymptotic Causal Mask Learning\n(ACML) and Dynamic Graph Parameter Learning (DGPL). ACML constructs causal\nmasks using learnable priority vectors and the Gumbel-Sigmoid function,\nensuring DAG formation while optimizing computational efficiency. DGPL\ntransforms causal learning into decomposed matrix products, capturing dynamic\ncausal structure in high-dimensional data and improving interpretability.\nExtensive experiments on synthetic and real-world datasets demonstrate that\nLOCAL significantly outperforms existing methods and highlight LOCAL's\npotential as a robust and efficient method for dynamic causal discovery.\n", "date": "2024-10-25", "categories": []}
{"id": "2503.15890", "title": "Time After Time: Deep-Q Effect Estimation for Interventions on When and\n  What to do", "abstract": "  Problems in fields such as healthcare, robotics, and finance requires\nreasoning about the value both of what decision or action to take and when to\ntake it. The prevailing hope is that artificial intelligence will support such\ndecisions by estimating the causal effect of policies such as how to treat\npatients or how to allocate resources over time. However, existing methods for\nestimating the effect of a policy struggle with \\emph{irregular time}. They\neither discretize time, or disregard the effect of timing policies. We present\na new deep-Q algorithm that estimates the effect of both when and what to do\ncalled Earliest Disagreement Q-Evaluation (EDQ). EDQ makes use of recursion for\nthe Q-function that is compatible with flexible sequence models, such as\ntransformers. EDQ provides accurate estimates under standard assumptions. We\nvalidate the approach through experiments on survival time and tumor growth\ntasks.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.16206", "title": "Interpretable Neural Causal Models with TRAM-DAGs", "abstract": "  The ultimate goal of most scientific studies is to understand the underlying\ncausal mechanism between the involved variables. Structural causal models\n(SCMs) are widely used to represent such causal mechanisms. Given an SCM,\ncausal queries on all three levels of Pearl's causal hierarchy can be answered:\n$L_1$ observational, $L_2$ interventional, and $L_3$ counterfactual. An\nessential aspect of modeling the SCM is to model the dependency of each\nvariable on its causal parents. Traditionally this is done by parametric\nstatistical models, such as linear or logistic regression models. This allows\nto handle all kinds of data types and fit interpretable models but bears the\nrisk of introducing a bias. More recently neural causal models came up using\nneural networks (NNs) to model the causal relationships, allowing the\nestimation of nearly any underlying functional form without bias. However,\ncurrent neural causal models are generally restricted to continuous variables\nand do not yield an interpretable form of the causal relationships.\nTransformation models range from simple statistical regressions to complex\nnetworks and can handle continuous, ordinal, and binary data. Here, we propose\nto use TRAMs to model the functional relationships in SCMs allowing us to\nbridge the gap between interpretability and flexibility in causal modeling. We\ncall this method TRAM-DAG and assume currently that the underlying directed\nacyclic graph is known. For the fully observed case, we benchmark TRAM-DAGs\nagainst state-of-the-art statistical and NN-based causal models. We show that\nTRAM-DAGs are interpretable but also achieve equal or superior performance in\nqueries ranging from $L_1$ to $L_3$ in the causal hierarchy. For the continuous\ncase, TRAM-DAGs allow for counterfactual queries for three common causal\nstructures, including unobserved confounding.\n", "date": "2025-03-20", "categories": []}
{"id": "2207.04053", "title": "On the Need and Applicability of Causality for Fairness: A Unified\n  Framework for AI Auditing and Legal Analysis", "abstract": "  As Artificial Intelligence (AI) increasingly influences decisions in critical\nsocietal sectors, understanding and establishing causality becomes essential\nfor evaluating the fairness of automated systems. This article explores the\nsignificance of causal reasoning in addressing algorithmic discrimination,\nemphasizing both legal and societal perspectives. By reviewing landmark cases\nand regulatory frameworks, particularly within the European Union, we\nillustrate the challenges inherent in proving causal claims when confronted\nwith opaque AI decision-making processes. The discussion outlines practical\nobstacles and methodological limitations in applying causal inference to\nreal-world fairness scenarios, proposing actionable solutions to enhance\ntransparency, accountability, and fairness in algorithm-driven decisions.\n", "date": "2022-07-08", "categories": []}
{"id": "2411.07414", "title": "Comparing Targeting Strategies for Maximizing Social Welfare with\n  Limited Resources", "abstract": "  Machine learning is increasingly used to select which individuals receive\nlimited-resource interventions in domains such as human services, education,\ndevelopment, and more. However, it is often not apparent what the right\nquantity is for models to predict. Policymakers rarely have access to data from\na randomized controlled trial (RCT) that would enable accurate estimates of\nwhich individuals would benefit more from the intervention, while observational\ndata creates a substantial risk of bias in treatment effect estimates.\nPractitioners instead commonly use a technique termed ``risk-based targeting\"\nwhere the model is just used to predict each individual's status quo outcome\n(an easier, non-causal task). Those with higher predicted risk are offered\ntreatment. There is currently almost no empirical evidence to inform which\nchoices lead to the most effective machine learning-informed targeting\nstrategies in social domains. In this work, we use data from 5 real-world RCTs\nin a variety of domains to empirically assess such choices. We find that when\ntreatment effects can be estimated with high accuracy (which we simulate by\nallowing the model to partially observe outcomes in advance), treatment effect\nbased targeting substantially outperforms risk-based targeting, even when\ntreatment effect estimates are biased. Moreover, these results hold even when\nthe policymaker has strong normative preferences for assisting higher-risk\nindividuals. However, the features and data actually available in most RCTs we\nexamine do not suffice for accurate estimates of heterogeneous treatment\neffects. Our results suggest treatment effect targeting has significant\npotential benefits, but realizing these benefits requires improvements to data\ncollection and model training beyond what is currently common in practice.\n", "date": "2024-11-11", "categories": []}
{"id": "2503.14795", "title": "The Hardness of Validating Observational Studies with Experimental Data", "abstract": "  Observational data is often readily available in large quantities, but can\nlead to biased causal effect estimates due to the presence of unobserved\nconfounding. Recent works attempt to remove this bias by supplementing\nobservational data with experimental data, which, when available, is typically\non a smaller scale due to the time and cost involved in running a randomised\ncontrolled trial. In this work, we prove a theorem that places fundamental\nlimits on this ``best of both worlds'' approach. Using the framework of\nimpossible inference, we show that although it is possible to use experimental\ndata to \\emph{falsify} causal effect estimates from observational data, in\ngeneral it is not possible to \\emph{validate} such estimates. Our theorem\nproves that while experimental data can be used to detect bias in observational\nstudies, without additional assumptions on the smoothness of the correction\nfunction, it can not be used to remove it. We provide a practical example of\nsuch an assumption, developing a novel Gaussian Process based approach to\nconstruct intervals which contain the true treatment effect with high\nprobability, both inside and outside of the support of the experimental data.\nWe demonstrate our methodology on both simulated and semi-synthetic datasets\nand make the \\href{https://github.com/Jakefawkes/Obs_and_exp_data}{code\navailable}.\n", "date": "2025-03-19", "categories": []}
{"id": "2503.15436", "title": "An extensive simulation study evaluating the interaction of resampling\n  techniques across multiple causal discovery contexts", "abstract": "  Despite the accelerating presence of exploratory causal analysis in modern\nscience and medicine, the available non-experimental methods for validating\ncausal models are not well characterized. One of the most popular methods is to\nevaluate the stability of model features after resampling the data, similar to\nresampling methods for estimating confidence intervals in statistics. Many\naspects of this approach have received little to no attention, however, such as\nwhether the choice of resampling method should depend on the sample size,\nalgorithms being used, or algorithm tuning parameters. We present theoretical\nresults proving that certain resampling methods closely emulate the assignment\nof specific values to algorithm tuning parameters. We also report the results\nof extensive simulation experiments, which verify the theoretical result and\nprovide substantial data to aid researchers in further characterizing\nresampling in the context of causal discovery analysis. Together, the\ntheoretical work and simulation results provide specific guidance on how\nresampling methods and tuning parameters should be selected in practice.\n", "date": "2025-03-19", "categories": []}
{"id": "2301.12351", "title": "Emerging Synergies in Causality and Deep Generative Models: A Survey", "abstract": "  In the field of artificial intelligence (AI), the quest to understand and\nmodel data-generating processes (DGPs) is of paramount importance. Deep\ngenerative models (DGMs) have proven adept in capturing complex data\ndistributions but often fall short in generalization and interpretability. On\nthe other hand, causality offers a structured lens to comprehend the mechanisms\ndriving data generation and highlights the causal-effect dynamics inherent in\nthese processes. While causality excels in interpretability and the ability to\nextrapolate, it grapples with intricacies of high-dimensional spaces.\nRecognizing the synergistic potential, we delve into the confluence of\ncausality and DGMs. We elucidate the integration of causal principles within\nDGMs, investigate causal identification using DGMs, and navigate an emerging\nresearch frontier of causality in large-scale generative models, particularly\ngenerative large language models (LLMs). We offer insights into methodologies,\nhighlight open challenges, and suggest future directions, positioning our\ncomprehensive review as an essential guide in this swiftly emerging and\nevolving area.\n", "date": "2023-01-29", "categories": []}
{"id": "2308.04585", "title": "Kernel Single Proxy Control for Deterministic Confounding", "abstract": "  We consider the problem of causal effect estimation with an unobserved\nconfounder, where we observe a single proxy variable that is associated with\nthe confounder. Although it has been shown that the recovery of an average\ncausal effect is impossible in general from a single proxy variable, we show\nthat causal recovery is possible if the outcome is generated deterministically.\nThis generalizes existing work on causal methods with a single proxy variable\nto the continuous treatment setting. We propose two kernel-based methods for\nthis setting: the first based on the two-stage regression approach, and the\nsecond based on a maximum moment restriction approach. We prove that both\napproaches can consistently estimate the causal effect, and we empirically\ndemonstrate that we can successfully recover the causal effect on challenging\nsynthetic benchmarks.\n", "date": "2023-08-08", "categories": []}
{"id": "2406.01384", "title": "Extending Structural Causal Models for Autonomous Vehicles to Simplify\n  Temporal System Construction & Enable Dynamic Interactions Between Agents", "abstract": "  In this work we aim to bridge the divide between autonomous vehicles and\ncausal reasoning. Autonomous vehicles have come to increasingly interact with\nhuman drivers, and in many cases may pose risks to the physical or mental\nwell-being of those they interact with. Meanwhile causal models, despite their\ninherent transparency and ability to offer contrastive explanations, have found\nlimited usage within such systems. As such, we first identify the challenges\nthat have limited the integration of structural causal models within autonomous\nvehicles. We then introduce a number of theoretical extensions to the\nstructural causal model formalism in order to tackle these challenges. This\naugments these models to possess greater levels of modularisation and\nencapsulation, as well presenting temporal causal model representation with\nconstant space complexity. We also prove through the extensions we have\nintroduced that dynamically mutable sets (e.g. varying numbers of autonomous\nvehicles across time) can be used within a structural causal model while\nmaintaining a relaxed form of causal stationarity. Finally we discuss the\napplication of the extensions in the context of the autonomous vehicle and\nservice robotics domain along with potential directions for future work.\n", "date": "2024-06-03", "categories": []}
{"id": "2410.12730", "title": "Counterfactual Generative Modeling with Variational Causal Inference", "abstract": "  Estimating an individual's counterfactual outcomes under interventions is a\nchallenging task for traditional causal inference and supervised learning\napproaches when the outcome is high-dimensional (e.g. gene expressions, facial\nimages) and covariates are relatively limited. In this case, to predict one's\noutcomes under counterfactual treatments, it is crucial to leverage individual\ninformation contained in the observed outcome in addition to the covariates.\nPrior works using variational inference in counterfactual generative modeling\nhave been focusing on neural adaptations and model variants within the\nconditional variational autoencoder formulation, which we argue is\nfundamentally ill-suited to the notion of counterfactual in causal inference.\nIn this work, we present a novel variational Bayesian causal inference\nframework and its theoretical backings to properly handle counterfactual\ngenerative modeling tasks, through which we are able to conduct counterfactual\nsupervision end-to-end during training without any counterfactual samples, and\nencourage disentangled exogenous noise abduction that aids the correct\nidentification of causal effect in counterfactual generations. In experiments,\nwe demonstrate the advantage of our framework compared to state-of-the-art\nmodels in counterfactual generative modeling on multiple benchmarks.\n", "date": "2024-10-16", "categories": []}
{"id": "2503.13833", "title": "Causal Discovery from Data Assisted by Large Language Models", "abstract": "  Knowledge driven discovery of novel materials necessitates the development of\nthe causal models for the property emergence. While in classical physical\nparadigm the causal relationships are deduced based on the physical principles\nor via experiment, rapid accumulation of observational data necessitates\nlearning causal relationships between dissimilar aspects of materials structure\nand functionalities based on observations. For this, it is essential to\nintegrate experimental data with prior domain knowledge. Here we demonstrate\nthis approach by combining high-resolution scanning transmission electron\nmicroscopy (STEM) data with insights derived from large language models (LLMs).\nBy fine-tuning ChatGPT on domain-specific literature, such as arXiv papers on\nferroelectrics, and combining obtained information with data-driven causal\ndiscovery, we construct adjacency matrices for Directed Acyclic Graphs (DAGs)\nthat map the causal relationships between structural, chemical, and\npolarization degrees of freedom in Sm-doped BiFeO3 (SmBFO). This approach\nenables us to hypothesize how synthesis conditions influence material\nproperties, particularly the coercive field (E0), and guides experimental\nvalidation. The ultimate objective of this work is to develop a unified\nframework that integrates LLM-driven literature analysis with data-driven\ndiscovery, facilitating the precise engineering of ferroelectric materials by\nestablishing clear connections between synthesis conditions and their resulting\nmaterial properties.\n", "date": "2025-03-18", "categories": []}
{"id": "2503.14442", "title": "Inducing Causal Structure for Interpretable Neural Networks Applied to\n  Glucose Prediction for T1DM Patients", "abstract": "  Causal abstraction techniques such as Interchange Intervention Training (IIT)\nhave been proposed to infuse neural network with expert knowledge encoded in\ncausal models, but their application to real-world problems remains limited.\nThis article explores the application of IIT in predicting blood glucose levels\nin Type 1 Diabetes Mellitus (T1DM) patients. The study utilizes an acyclic\nversion of the simglucose simulator approved by the FDA to train a Multi-Layer\nPerceptron (MLP) model, employing IIT to impose causal relationships. Results\nshow that the model trained with IIT effectively abstracted the causal\nstructure and outperformed the standard one in terms of predictive performance\nacross different prediction horizons (PHs) post-meal. Furthermore, the\nbreakdown of the counterfactual loss can be leveraged to explain which part of\nthe causal mechanism are more or less effectively captured by the model. These\npreliminary results suggest the potential of IIT in enhancing predictive models\nin healthcare by effectively complying with expert knowledge.\n", "date": "2025-03-18", "categories": []}
{"id": "2406.11601", "title": "Standardizing Structural Causal Models", "abstract": "  Synthetic datasets generated by structural causal models (SCMs) are commonly\nused for benchmarking causal structure learning algorithms. However, the\nvariances and pairwise correlations in SCM data tend to increase along the\ncausal ordering. Several popular algorithms exploit these artifacts, possibly\nleading to conclusions that do not generalize to real-world settings. Existing\nmetrics like $\\operatorname{Var}$-sortability and\n$\\operatorname{R^2}$-sortability quantify these patterns, but they do not\nprovide tools to remedy them. To address this, we propose\ninternally-standardized structural causal models (iSCMs), a modification of\nSCMs that introduces a standardization operation at each variable during the\ngenerative process. By construction, iSCMs are not\n$\\operatorname{Var}$-sortable. We also find empirical evidence that they are\nmostly not $\\operatorname{R^2}$-sortable for commonly-used graph families.\nMoreover, contrary to the post-hoc standardization of data generated by\nstandard SCMs, we prove that linear iSCMs are less identifiable from prior\nknowledge on the weights and do not collapse to deterministic relationships in\nlarge systems, which may make iSCMs a useful model in causal inference beyond\nthe benchmarking problem studied here. Our code is publicly available at:\nhttps://github.com/werkaaa/iscm.\n", "date": "2024-06-17", "categories": []}
{"id": "2406.17714", "title": "Compositional Models for Estimating Causal Effects", "abstract": "  Many real-world systems can be usefully represented as sets of interacting\ncomponents. Examples include computational systems, such as query processors\nand compilers, natural systems, such as cells and ecosystems, and social\nsystems, such as families and organizations. However, current approaches to\nestimating potential outcomes and causal effects typically treat such systems\nas single units, represent them with a fixed set of variables, and assume a\nhomogeneous data-generating process. In this work, we study a compositional\napproach for estimating individual-level potential outcomes and causal effects\nin structured systems, where each unit is represented by an instance-specific\ncomposition of multiple heterogeneous components. The compositional approach\ndecomposes unit-level causal queries into more fine-grained queries, explicitly\nmodeling how unit-level interventions affect component-level outcomes to\ngenerate a unit's outcome. We demonstrate this approach using modular neural\nnetwork architectures and show that it provides benefits for causal effect\nestimation from observational data, such as accurate causal effect estimation\nfor structured units, increased sample efficiency, improved overlap between\ntreatment and control groups, and compositional generalization to units with\nunseen combinations of components. Remarkably, our results show that\ncompositional modeling can improve the accuracy of causal estimation even when\ncomponent-level outcomes are unobserved. We also create and use a set of\nreal-world evaluation environments for the empirical evaluation of\ncompositional approaches for causal effect estimation and demonstrate the role\nof composition structure, varying amounts of component-level data access, and\ncomponent heterogeneity in the performance of compositional models as compared\nto the non-compositional approaches.\n", "date": "2024-06-25", "categories": []}
{"id": "2411.02134", "title": "Optimizing Multi-Scale Representations to Detect Effect Heterogeneity\n  Using Earth Observation and Computer Vision: Applications to Two Anti-Poverty\n  RCTs", "abstract": "  Earth Observation (EO) data are increasingly used in policy analysis by\nenabling granular estimation of conditional average treatment effects (CATE).\nHowever, a challenge in EO-based causal inference is determining the scale of\nthe input satellite imagery -- balancing the trade-off between capturing\nfine-grained individual heterogeneity in smaller images and broader contextual\ninformation in larger ones. This paper introduces Multi-Scale Representation\nConcatenation, a set of composable procedures that transform arbitrary\nsingle-scale EO-based CATE estimation algorithms into multi-scale ones. We\nbenchmark the performance of Multi-Scale Representation Concatenation on a CATE\nestimation pipeline that combines Vision Transformer (ViT) models (which encode\nimages) with Causal Forests (CFs) to obtain CATE estimates from those\nencodings. We first perform simulation studies where the causal mechanism is\nknown, showing that our multi-scale approach captures information relevant to\neffect heterogeneity that single-scale ViT models fail to capture as measured\nby $R^2$. We then apply the multi-scale method to two randomized controlled\ntrials (RCTs) conducted in Peru and Uganda using Landsat satellite imagery. As\nwe do not have access to ground truth CATEs in the RCT analysis, the Rank\nAverage Treatment Effect Ratio (RATE Ratio) measure is employed to assess\nperformance. Results indicate that Multi-Scale Representation Concatenation\nimproves the performance of deep learning models in EO-based CATE estimation\nwithout the complexity of designing new multi-scale architectures for a\nspecific use case. The application of Multi-Scale Representation Concatenation\ncould have meaningful policy benefits -- e.g., potentially increasing the\nimpact of poverty alleviation programs without additional resource expenditure.\n", "date": "2024-11-04", "categories": []}
{"id": "2411.06518", "title": "Causal Representation Learning from Multimodal Biomedical Observations", "abstract": "  Prevalent in biomedical applications (e.g., human phenotype research),\nmultimodal datasets can provide valuable insights into the underlying\nphysiological mechanisms. However, current machine learning (ML) models\ndesigned to analyze these datasets often lack interpretability and\nidentifiability guarantees, which are essential for biomedical research. Recent\nadvances in causal representation learning have shown promise in identifying\ninterpretable latent causal variables with formal theoretical guarantees.\nUnfortunately, most current work on multimodal distributions either relies on\nrestrictive parametric assumptions or yields only coarse identification\nresults, limiting their applicability to biomedical research that favors a\ndetailed understanding of the mechanisms.\n  In this work, we aim to develop flexible identification conditions for\nmultimodal data and principled methods to facilitate the understanding of\nbiomedical datasets. Theoretically, we consider a nonparametric latent\ndistribution (c.f., parametric assumptions in previous work) that allows for\ncausal relationships across potentially different modalities. We establish\nidentifiability guarantees for each latent component, extending the subspace\nidentification results from previous work. Our key theoretical contribution is\nthe structural sparsity of causal connections between modalities, which, as we\nwill discuss, is natural for a large collection of biomedical systems.\nEmpirically, we present a practical framework to instantiate our theoretical\ninsights. We demonstrate the effectiveness of our approach through extensive\nexperiments on both numerical and synthetic datasets. Results on a real-world\nhuman phenotype dataset are consistent with established biomedical research,\nvalidating our theoretical and methodological framework.\n", "date": "2024-11-10", "categories": []}
{"id": "2503.11870", "title": "Counterfactual Realizability", "abstract": "  It is commonly believed that, in a real-world environment, samples can only\nbe drawn from observational and interventional distributions, corresponding to\nLayers 1 and 2 of the Pearl Causal Hierarchy. Layer 3, representing\ncounterfactual distributions, is believed to be inaccessible by definition.\nHowever, Bareinboim, Forney, and Pearl (2015) introduced a procedure that\nallows an agent to sample directly from a counterfactual distribution, leaving\nopen the question of what other counterfactual quantities can be estimated\ndirectly via physical experimentation. We resolve this by introducing a formal\ndefinition of realizability, the ability to draw samples from a distribution,\nand then developing a complete algorithm to determine whether an arbitrary\ncounterfactual distribution is realizable given fundamental physical\nconstraints, such as the inability to go back in time and subject the same unit\nto a different experimental condition. We illustrate the implications of this\nnew framework for counterfactual data collection using motivating examples from\ncausal fairness and causal reinforcement learning. While the baseline approach\nin these motivating settings typically follows an interventional or\nobservational strategy, we show that a counterfactual strategy provably\ndominates both.\n", "date": "2025-03-14", "categories": []}
{"id": "2503.12784", "title": "Causal Feature Learning in the Social Sciences", "abstract": "  Variable selection poses a significant challenge in causal modeling,\nparticularly within the social sciences, where constructs often rely on\ninter-related factors such as age, socioeconomic status, gender, and race.\nIndeed, it has been argued that such attributes must be modeled as macro-level\nabstractions of lower-level manipulable features, in order to preserve the\nmodularity assumption essential to causal inference. This paper accordingly\nextends the theoretical framework of Causal Feature Learning (CFL).\nEmpirically, we apply the CFL algorithm to diverse social science datasets,\nevaluating how CFL-derived macrostates compare with traditional microstates in\ndownstream modeling tasks.\n", "date": "2025-03-17", "categories": []}
{"id": "2412.17080", "title": "Aligning Graphical and Functional Causal Abstractions", "abstract": "  Causal abstractions allow us to relate causal models on different levels of\ngranularity. To ensure that the models agree on cause and effect, frameworks\nfor causal abstractions define notions of consistency. Two distinct methods for\ncausal abstraction are common in the literature: (i) graphical abstractions,\nsuch as Cluster DAGs, which relate models on a structural level, and (ii)\nfunctional abstractions, like $\\alpha$-abstractions, which relate models by\nmaps between variables and their ranges. In this paper we will align the\nnotions of graphical and functional consistency and show an equivalence between\nthe class of Cluster DAGs, consistent $\\alpha$-abstractions with the range of\nabstracted variables mapped bijectively, and constructive $\\tau$-abstractions.\nFurthermore, we extend this alignment and the expressivity of graphical\nabstractions by introducing Partial Cluster DAGs. Our results provide a\nrigorous bridge between the functional and graphical frameworks and allow for\nadoption and transfer of results between them.\n", "date": "2024-12-22", "categories": []}
{"id": "2503.10834", "title": "On the Identifiability of Causal Abstractions", "abstract": "  Causal representation learning (CRL) enhances machine learning models'\nrobustness and generalizability by learning structural causal models associated\nwith data-generating processes. We focus on a family of CRL methods that uses\ncontrastive data pairs in the observable space, generated before and after a\nrandom, unknown intervention, to identify the latent causal model. (Brehmer et\nal., 2022) showed that this is indeed possible, given that all latent variables\ncan be intervened on individually. However, this is a highly restrictive\nassumption in many systems. In this work, we instead assume interventions on\narbitrary subsets of latent variables, which is more realistic. We introduce a\ntheoretical framework that calculates the degree to which we can identify a\ncausal model, given a set of possible interventions, up to an abstraction that\ndescribes the system at a higher level of granularity.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.11046", "title": "Measuring Similarity in Causal Graphs: A Framework for Semantic and\n  Structural Analysis", "abstract": "  Causal graphs are commonly used to understand and model complex systems.\nResearchers often construct these graphs from different perspectives, leading\nto significant variations for the same problem. Comparing causal graphs is,\ntherefore, essential for evaluating assumptions, integrating insights, and\nresolving disagreements. The rise of AI tools has further amplified this need,\nas they are increasingly used to generate hypothesized causal graphs by\nsynthesizing information from various sources such as prior research and\ncommunity inputs, providing the potential for automating and scaling causal\nmodeling for complex systems. Similar to humans, these tools also produce\ninconsistent results across platforms, versions, and iterations. Despite its\nimportance, research on causal graph comparison remains scarce. Existing\nmethods often focus solely on structural similarities, assuming identical\nvariable names, and fail to capture nuanced semantic relationships, which is\nessential for causal graph comparison. We address these gaps by investigating\nmethods for comparing causal graphs from both semantic and structural\nperspectives. First, we reviewed over 40 existing metrics and, based on\npredefined criteria, selected nine for evaluation from two threads of machine\nlearning: four semantic similarity metrics and five learning graph kernels. We\ndiscuss the usability of these metrics in simple examples to illustrate their\nstrengths and limitations. We then generated a synthetic dataset of 2,000\ncausal graphs using generative AI based on a reference diagram. Our findings\nreveal that each metric captures a different aspect of similarity, highlighting\nthe need to use multiple metrics.\n", "date": "2025-03-14", "categories": []}
{"id": "2503.07811", "title": "A primer on optimal transport for causal inference with observational\n  data", "abstract": "  The theory of optimal transportation has developed into a powerful and\nelegant framework for comparing probability distributions, with wide-ranging\napplications in all areas of science. The fundamental idea of analyzing\nprobabilities by comparing their underlying state space naturally aligns with\nthe core idea of causal inference, where understanding and quantifying\ncounterfactual states is paramount. Despite this intuitive connection, explicit\nresearch at the intersection of optimal transport and causal inference is only\nbeginning to develop. Yet, many foundational models in causal inference have\nimplicitly relied on optimal transport principles for decades, without\nrecognizing the underlying connection. Therefore, the goal of this review is to\noffer an introduction to the surprisingly deep existing connections between\noptimal transport and the identification of causal effects with observational\ndata -- where optimal transport is not just a set of potential tools, but\nactually builds the foundation of model assumptions. As a result, this review\nis intended to unify the language and notation between different areas of\nstatistics, mathematics, and econometrics, by pointing out these existing\nconnections, and to explore novel problems and directions for future work in\nboth areas derived from this realization.\n", "date": "2025-03-10", "categories": []}
{"id": "2503.09719", "title": "Towards Causal Model-Based Policy Optimization", "abstract": "  Real-world decision-making problems are often marked by complex, uncertain\ndynamics that can shift or break under changing conditions. Traditional\nModel-Based Reinforcement Learning (MBRL) approaches learn predictive models of\nenvironment dynamics from queried trajectories and then use these models to\nsimulate rollouts for policy optimization. However, such methods do not account\nfor the underlying causal mechanisms that govern the environment, and thus\ninadvertently capture spurious correlations, making them sensitive to\ndistributional shifts and limiting their ability to generalize. The same\nnaturally holds for model-free approaches. In this work, we introduce Causal\nModel-Based Policy Optimization (C-MBPO), a novel framework that integrates\ncausal learning into the MBRL pipeline to achieve more robust, explainable, and\ngeneralizable policy learning algorithms.\n  Our approach centers on first inferring a Causal Markov Decision Process\n(C-MDP) by learning a local Structural Causal Model (SCM) of both the state and\nreward transition dynamics from trajectories gathered online. C-MDPs differ\nfrom classic MDPs in that we can decompose causal dependencies in the\nenvironment dynamics via specifying an associated Causal Bayesian Network.\nC-MDPs allow for targeted interventions and counterfactual reasoning, enabling\nthe agent to distinguish between mere statistical correlations and causal\nrelationships. The learned SCM is then used to simulate counterfactual\non-policy transitions and rewards under hypothetical actions (or\n``interventions\"), thereby guiding policy optimization more effectively. The\nresulting policy learned by C-MBPO can be shown to be robust to a class of\ndistributional shifts that affect spurious, non-causal relationships in the\ndynamics. We demonstrate this through some simple experiments involving near\nand far OOD dynamics drifts.\n", "date": "2025-03-12", "categories": []}
{"id": "2503.09859", "title": "An Asymmetric Independence Model for Causal Discovery on Path Spaces", "abstract": "  We develop the theory linking 'E-separation' in directed mixed graphs (DMGs)\nwith conditional independence relations among coordinate processes in\nstochastic differential equations (SDEs), where causal relationships are\ndetermined by \"which variables enter the governing equation of which other\nvariables\". We prove a global Markov property for cyclic SDEs, which naturally\nextends to partially observed cyclic SDEs, because our asymmetric independence\nmodel is closed under marginalization. We then characterize the class of graphs\nthat encode the same set of independence relations, yielding a result analogous\nto the seminal 'same skeleton and v-structures' result for directed acyclic\ngraphs (DAGs). In the fully observed case, we show that each such equivalence\nclass of graphs has a greatest element as a parsimonious representation and\ndevelop algorithms to identify this greatest element from data. We conjecture\nthat a greatest element also exists under partial observations, which we verify\ncomputationally for graphs with up to four nodes.\n", "date": "2025-03-12", "categories": []}
{"id": "2403.01865", "title": "Out-of-distribution robustness for multivariate analysis via causal\n  regularisation", "abstract": "  We propose a regularisation strategy of classical machine learning algorithms\nrooted in causality that ensures robustness against distribution shifts.\nBuilding upon the anchor regression framework, we demonstrate how incorporating\na straightforward regularisation term into the loss function of classical\nmultivariate analysis algorithms, such as (orthonormalized) partial least\nsquares, reduced-rank regression, and multiple linear regression, enables\nout-of-distribution generalisation. Our framework allows users to efficiently\nverify the compatibility of a loss function with the regularisation strategy.\nEstimators for selected algorithms are provided, showcasing consistency and\nefficacy in synthetic and real-world climate science problems. The empirical\nvalidation highlights the versatility of anchor regularisation, emphasizing its\ncompatibility with multivariate analysis approaches and its role in enhancing\nreplicability while guarding against distribution shifts. The extended anchor\nframework advances causal inference methodologies, addressing the need for\nreliable out-of-distribution generalisation.\n", "date": "2024-03-04", "categories": []}
{"id": "2405.14001", "title": "Nondeterministic Causal Models", "abstract": "  I generalize acyclic deterministic structural causal models to the\nnondeterministic case and argue that this offers an improved semantics for\ncounterfactuals. The standard, deterministic, semantics developed by Halpern\n(and based on the initial proposal of Galles & Pearl) assumes that for each\nassignment of values to parent variables there is a unique assignment to their\nchild variable, and it assumes that the actual world (an assignment of values\nto all variables of a model) specifies a unique counterfactual world for each\nintervention. Both assumptions are unrealistic, and therefore I drop both of\nthem in my proposal. I do so by allowing multi-valued functions in the\nstructural equations. In addition, I adjust the semantics so that the solutions\nto the equations that obtained in the actual world are preserved in any\ncounterfactual world. I provide a sound and complete axiomatization of the\nresulting logic and compare it to the standard one by Halpern and to more\nrecent proposals that are closer to mine. Finally, I extend these models to the\nprobabilistic case and show that they open up the way to identifying\ncounterfactuals even in Causal Bayesian Networks.\n", "date": "2024-05-22", "categories": []}
{"id": "2407.10448", "title": "Spectral Representation for Causal Estimation with Hidden Confounders", "abstract": "  We address the problem of causal effect estimation where hidden confounders\nare present, with a focus on two settings: instrumental variable regression\nwith additional observed confounders, and proxy causal learning. Our approach\nuses a singular value decomposition of a conditional expectation operator,\nfollowed by a saddle-point optimization problem, which, in the context of IV\nregression, can be thought of as a neural net generalization of the seminal\napproach due to Darolles et al. [2011]. Saddle-point formulations have gathered\nconsiderable attention recently, as they can avoid double sampling bias and are\namenable to modern function approximation methods. We provide experimental\nvalidation in various settings, and show that our approach outperforms existing\nmethods on common benchmarks.\n", "date": "2024-07-15", "categories": []}
{"id": "2503.07849", "title": "Actual Causation and Nondeterministic Causal Models", "abstract": "  In (Beckers, 2025) I introduced nondeterministic causal models as a\ngeneralization of Pearl's standard deterministic causal models. I here take\nadvantage of the increased expressivity offered by these models to offer a\nnovel definition of actual causation (that also applies to deterministic\nmodels). Instead of motivating the definition by way of (often subjective)\nintuitions about examples, I proceed by developing it based entirely on the\nunique function that it can fulfil in communicating and learning a causal\nmodel. First I generalize the more basic notion of counterfactual dependence,\nsecond I show how this notion has a vital role to play in the logic of causal\ndiscovery, third I introduce the notion of a structural simplification of a\ncausal model, and lastly I bring both notions together in my definition of\nactual causation. Although novel, the resulting definition arrives at verdicts\nthat are almost identical to those of my previous definition (Beckers, 2021,\n2022).\n", "date": "2025-03-10", "categories": []}
{"id": "2412.20436", "title": "Treatment Effect Estimation for Graph-Structured Targets", "abstract": "  Treatment effect estimation, which helps understand the causality between\ntreatment and outcome variable, is a central task in decision-making across\nvarious domains. While most studies focus on treatment effect estimation on\nindividual targets, in specific contexts, there is a necessity to comprehend\nthe treatment effect on a group of targets, especially those that have\nrelationships represented as a graph structure between them. In such cases, the\nfocus of treatment assignment is prone to depend on a particular node of the\ngraph, such as the one with the highest degree, thus resulting in an\nobservational bias from a small part of the entire graph. Whereas a bias tends\nto be caused by the small part, straightforward extensions of previous studies\ncannot provide efficient bias mitigation owing to the use of the entire graph\ninformation. In this study, we propose Graph-target Treatment Effect Estimation\n(GraphTEE), a framework designed to estimate treatment effects specifically on\ngraph-structured targets. GraphTEE aims to mitigate observational bias by\nfocusing on confounding variable sets and consider a new regularization\nframework. Additionally, we provide a theoretical analysis on how GraphTEE\nperforms better in terms of bias mitigation. Experiments on synthetic and\nsemi-synthetic datasets demonstrate the effectiveness of our proposed method.\n", "date": "2024-12-29", "categories": []}
{"id": "2502.17030", "title": "Your Assumed DAG is Wrong and Here's How To Deal With It", "abstract": "  Assuming a directed acyclic graph (DAG) that represents prior knowledge of\ncausal relationships between variables is a common starting point for\ncause-effect estimation. Existing literature typically invokes hypothetical\ndomain expert knowledge or causal discovery algorithms to justify this\nassumption. In practice, neither may propose a single DAG with high confidence.\nDomain experts are hesitant to rule out dependencies with certainty or have\nongoing disputes about relationships; causal discovery often relies on\nuntestable assumptions itself or only provides an equivalence class of DAGs and\nis commonly sensitive to hyperparameter and threshold choices. We propose an\nefficient, gradient-based optimization method that provides bounds for causal\nqueries over a collection of causal graphs -- compatible with imperfect prior\nknowledge -- that may still be too large for exhaustive enumeration. Our bounds\nachieve good coverage and sharpness for causal queries such as average\ntreatment effects in linear and non-linear synthetic settings as well as on\nreal-world data. Our approach aims at providing an easy-to-use and widely\napplicable rebuttal to the valid critique of `What if your assumed DAG is\nwrong?'.\n", "date": "2025-02-24", "categories": []}
{"id": "2503.05985", "title": "Black Box Causal Inference: Effect Estimation via Meta Prediction", "abstract": "  Causal inference and the estimation of causal effects plays a central role in\ndecision-making across many areas, including healthcare and economics.\nEstimating causal effects typically requires an estimator that is tailored to\neach problem of interest. But developing estimators can take significant effort\nfor even a single causal inference setting. For example, algorithms for\nregression-based estimators, propensity score methods, and doubly robust\nmethods were designed across several decades to handle causal estimation with\nobserved confounders. Similarly, several estimators have been developed to\nexploit instrumental variables (IVs), including two-stage least-squares (TSLS),\ncontrol functions, and the method-of-moments. In this work, we instead frame\ncausal inference as a dataset-level prediction problem, offloading algorithm\ndesign to the learning process. The approach we introduce, called black box\ncausal inference (BBCI), builds estimators in a black-box manner by learning to\npredict causal effects from sampled dataset-effect pairs. We demonstrate\naccurate estimation of average treatment effects (ATEs) and conditional average\ntreatment effects (CATEs) with BBCI across several causal inference problems\nwith known identification, including problems with less developed estimators.\n", "date": "2025-03-07", "categories": []}
{"id": "2503.06156", "title": "Double Debiased Machine Learning for Mediation Analysis with Continuous\n  Treatments", "abstract": "  Uncovering causal mediation effects is of significant value to practitioners\nseeking to isolate the direct treatment effect from the potential mediated\neffect. We propose a double machine learning (DML) algorithm for mediation\nanalysis that supports continuous treatments. To estimate the target mediated\nresponse curve, our method uses a kernel-based doubly robust moment function\nfor which we prove asymptotic Neyman orthogonality. This allows us to obtain\nasymptotic normality with nonparametric convergence rate while allowing for\nnonparametric or parametric estimation of the nuisance parameters. We then\nderive an optimal bandwidth strategy along with a procedure for estimating\nasymptotic confidence intervals. Finally, to illustrate the benefits of our\nmethod, we provide a numerical evaluation of our approach on a simulation along\nwith an application to real-world medical data to analyze the effect of\nglycemic control on cognitive functions.\n", "date": "2025-03-08", "categories": []}
{"id": "2503.06395", "title": "Causal Discovery and Inference towards Urban Elements and Associated\n  Factors", "abstract": "  To uncover the city's fundamental functioning mechanisms, it is important to\nacquire a deep understanding of complicated relationships among citizens,\nlocation, and mobility behaviors. Previous research studies have applied direct\ncorrelation analysis to investigate such relationships. Nevertheless, due to\nthe ubiquitous confounding effects, empirical correlation analysis may not\naccurately reflect underlying causal relationships among basic urban elements.\nIn this paper, we propose a novel urban causal computing framework to\ncomprehensively explore causalities and confounding effects among a variety of\nfactors across different types of urban elements. In particular, we design a\nreinforcement learning algorithm to discover the potential causal graph, which\ndepicts the causal relations between urban factors. The causal graph further\nserves as the guidance for estimating causal effects between pair-wise urban\nfactors by propensity score matching. After removing the confounding effects\nfrom correlations, we leverage significance levels of causal effects in\ndownstream urban mobility prediction tasks. Experimental studies on open-source\nurban datasets show that the discovered causal graph demonstrates a\nhierarchical structure, where citizens affect locations, and they both cause\nchanges in urban mobility behaviors. Experimental results in urban mobility\nprediction tasks further show that the proposed method can effectively reduce\nconfounding effects and enhance performance of urban computing tasks.\n", "date": "2025-03-09", "categories": []}
{"id": "2503.07302", "title": "When Selection Meets Intervention: Additional Complexities in Causal\n  Discovery", "abstract": "  We address the common yet often-overlooked selection bias in interventional\nstudies, where subjects are selectively enrolled into experiments. For\ninstance, participants in a drug trial are usually patients of the relevant\ndisease; A/B tests on mobile applications target existing users only, and gene\nperturbation studies typically focus on specific cell types, such as cancer\ncells. Ignoring this bias leads to incorrect causal discovery results. Even\nwhen recognized, the existing paradigm for interventional causal discovery\nstill fails to address it. This is because subtle differences in when and where\ninterventions happen can lead to significantly different statistical patterns.\nWe capture this dynamic by introducing a graphical model that explicitly\naccounts for both the observed world (where interventions are applied) and the\ncounterfactual world (where selection occurs while interventions have not been\napplied). We characterize the Markov property of the model, and propose a\nprovably sound algorithm to identify causal relations as well as selection\nmechanisms up to the equivalence class, from data with soft interventions and\nunknown targets. Through synthetic and real-world experiments, we demonstrate\nthat our algorithm effectively identifies true causal relations despite the\npresence of selection bias.\n", "date": "2025-03-10", "categories": []}
{"id": "2503.07475", "title": "Sample Complexity of Nonparametric Closeness Testing for Continuous\n  Distributions and Its Application to Causal Discovery with Hidden Confounding", "abstract": "  We study the problem of closeness testing for continuous distributions and\nits implications for causal discovery. Specifically, we analyze the sample\ncomplexity of distinguishing whether two multidimensional continuous\ndistributions are identical or differ by at least $\\epsilon$ in terms of\nKullback-Leibler (KL) divergence under non-parametric assumptions. To this end,\nwe propose an estimator of KL divergence which is based on the von Mises\nexpansion. Our closeness test attains optimal parametric rates under smoothness\nassumptions. Equipped with this test, which serves as a building block of our\ncausal discovery algorithm to identify the causal structure between two\nmultidimensional random variables, we establish sample complexity guarantees\nfor our causal discovery method. To the best of our knowledge, this work is the\nfirst work that provides sample complexity guarantees for distinguishing cause\nand effect in multidimensional non-linear models with non-Gaussian continuous\nvariables in the presence of unobserved confounding.\n", "date": "2025-03-10", "categories": []}
{"id": "2411.01371", "title": "Network Causal Effect Estimation In Graphical Models Of Contagion And\n  Latent Confounding", "abstract": "  A key question in many network studies is whether the observed correlations\nbetween units are primarily due to contagion or latent confounding. Here, we\nstudy this question using a segregated graph (Shpitser, 2015) representation of\nthese mechanisms, and examine how uncertainty about the true underlying\nmechanism impacts downstream computation of network causal effects,\nparticularly under full interference -- settings where we only have a single\nrealization of a network and each unit may depend on any other unit in the\nnetwork. Under certain assumptions about asymptotic growth of the network, we\nderive likelihood ratio tests that can be used to identify whether different\nsets of variables -- confounders, treatments, and outcomes -- across units\nexhibit dependence due to contagion or latent confounding. We then propose\nnetwork causal effect estimation strategies that provide unbiased and\nconsistent estimates if the dependence mechanisms are either known or correctly\ninferred using our proposed tests. Together, the proposed methods allow network\neffect estimation in a wider range of full interference scenarios that have not\nbeen considered in prior work. We evaluate the effectiveness of our methods\nwith synthetic data and the validity of our assumptions using real-world\nnetworks.\n", "date": "2024-11-02", "categories": []}
{"id": "2411.07180", "title": "Gumbel Counterfactual Generation From Language Models", "abstract": "  Understanding and manipulating the causal generation mechanisms in language\nmodels is essential for controlling their behavior. Previous work has primarily\nrelied on techniques such as representation surgery -- e.g., model ablations or\nmanipulation of linear subspaces tied to specific concepts -- to\n\\emph{intervene} on these models. To understand the impact of interventions\nprecisely, it is useful to examine \\emph{counterfactuals} -- e.g., how a given\nsentence would have appeared had it been generated by the model following a\nspecific intervention. We highlight that counterfactual reasoning is\nconceptually distinct from interventions, as articulated in Pearl's causal\nhierarchy. Based on this observation, we propose a framework for generating\ntrue string counterfactuals by reformulating language models as a structural\nequation model using the Gumbel-max trick, which we called Gumbel\ncounterfactual generation. This reformulation allows us to model the joint\ndistribution over original strings and their counterfactuals resulting from the\nsame instantiation of the sampling noise. We develop an algorithm based on\nhindsight Gumbel sampling that allows us to infer the latent noise variables\nand generate counterfactuals of observed strings. Our experiments demonstrate\nthat the approach produces meaningful counterfactuals while at the same time\nshowing that commonly used intervention techniques have considerable undesired\nside effects.\n", "date": "2024-11-11", "categories": []}
{"id": "2503.04347", "title": "Large Language Models for Zero-shot Inference of Causal Structures in\n  Biology", "abstract": "  Genes, proteins and other biological entities influence one another via\ncausal molecular networks. Causal relationships in such networks are mediated\nby complex and diverse mechanisms, through latent variables, and are often\nspecific to cellular context. It remains challenging to characterise such\nnetworks in practice. Here, we present a novel framework to evaluate large\nlanguage models (LLMs) for zero-shot inference of causal relationships in\nbiology. In particular, we systematically evaluate causal claims obtained from\nan LLM using real-world interventional data. This is done over one hundred\nvariables and thousands of causal hypotheses. Furthermore, we consider several\nprompting and retrieval-augmentation strategies, including large, and\npotentially conflicting, collections of scientific articles. Our results show\nthat with tailored augmentation and prompting, even relatively small LLMs can\ncapture meaningful aspects of causal structure in biological systems. This\nsupports the notion that LLMs could act as orchestration tools in biological\ndiscovery, by helping to distil current knowledge in ways amenable to\ndownstream analysis. Our approach to assessing LLMs with respect to\nexperimental data is relevant for a broad range of problems at the intersection\nof causal learning, LLMs and scientific discovery.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.04358", "title": "Learning Causal Response Representations through Direct Effect Analysis", "abstract": "  We propose a novel approach for learning causal response representations. Our\nmethod aims to extract directions in which a multidimensional outcome is most\ndirectly caused by a treatment variable. By bridging conditional independence\ntesting with causal representation learning, we formulate an optimisation\nproblem that maximises the evidence against conditional independence between\nthe treatment and outcome, given a conditioning set. This formulation employs\nflexible regression models tailored to specific applications, creating a\nversatile framework. The problem is addressed through a generalised eigenvalue\ndecomposition. We show that, under mild assumptions, the distribution of the\nlargest eigenvalue can be bounded by a known $F$-distribution, enabling\ntestable conditional independence. We also provide theoretical guarantees for\nthe optimality of the learned representation in terms of signal-to-noise ratio\nand Fisher information maximisation. Finally, we demonstrate the empirical\neffectiveness of our approach in simulation and real-world experiments. Our\nresults underscore the utility of this framework in uncovering direct causal\neffects within complex, multivariate settings.\n", "date": "2025-03-06", "categories": []}
{"id": "2409.02772", "title": "Unifying Causal Representation Learning with the Invariance Principle", "abstract": "  Causal representation learning (CRL) aims at recovering latent causal\nvariables from high-dimensional observations to solve causal downstream tasks,\nsuch as predicting the effect of new interventions or more robust\nclassification. A plethora of methods have been developed, each tackling\ncarefully crafted problem settings that lead to different types of\nidentifiability. These different settings are widely assumed to be important\nbecause they are often linked to different rungs of Pearl's causal hierarchy,\neven though this correspondence is not always exact. This work shows that\ninstead of strictly conforming to this hierarchical mapping, many causal\nrepresentation learning approaches methodologically align their representations\nwith inherent data symmetries. Identification of causal variables is guided by\ninvariance principles that are not necessarily causal. This result allows us to\nunify many existing approaches in a single method that can mix and match\ndifferent assumptions, including non-causal ones, based on the invariance\nrelevant to the problem at hand. It also significantly benefits applicability,\nwhich we demonstrate by improving treatment effect estimation on real-world\nhigh-dimensional ecological data. Overall, this paper clarifies the role of\ncausal assumptions in the discovery of causal variables and shifts the focus to\npreserving data symmetries.\n", "date": "2024-09-04", "categories": []}
{"id": "2412.16577", "title": "A Meta-Learning Approach to Bayesian Causal Discovery", "abstract": "  Discovering a unique causal structure is difficult due to both inherent\nidentifiability issues, and the consequences of finite data. As such,\nuncertainty over causal structures, such as those obtained from a Bayesian\nposterior, are often necessary for downstream tasks. Finding an accurate\napproximation to this posterior is challenging, due to the large number of\npossible causal graphs, as well as the difficulty in the subproblem of finding\nposteriors over the functional relationships of the causal edges. Recent works\nhave used meta-learning to view the problem of estimating the maximum\na-posteriori causal graph as supervised learning. Yet, these methods are\nlimited when estimating the full posterior as they fail to encode key\nproperties of the posterior, such as correlation between edges and permutation\nequivariance with respect to nodes. Further, these methods also cannot reliably\nsample from the posterior over causal structures. To address these limitations,\nwe propose a Bayesian meta learning model that allows for sampling causal\nstructures from the posterior and encodes these key properties. We compare our\nmeta-Bayesian causal discovery against existing Bayesian causal discovery\nmethods, demonstrating the advantages of directly learning a posterior over\ncausal structure.\n", "date": "2024-12-21", "categories": []}
{"id": "2412.18180", "title": "PCM Selector: Penalized Covariate-Mediator Selection Operator for\n  Evaluating Linear Causal Effects", "abstract": "  For a data-generating process for random variables that can be described with\na linear structural equation model, we consider a situation in which (i) a set\nof covariates satisfying the back-door criterion cannot be observed or (ii)\nsuch a set can be observed, but standard statistical estimation methods cannot\nbe applied to estimate causal effects because of\nmulticollinearity/high-dimensional data problems. We propose a novel two-stage\npenalized regression approach, the penalized covariate-mediator selection\noperator (PCM Selector), to estimate the causal effects in such scenarios.\nUnlike existing penalized regression analyses, when a set of intermediate\nvariables is available, PCM Selector provides a consistent or less biased\nestimator of the causal effect. In addition, PCM Selector provides a variable\nselection procedure for intermediate variables to obtain better estimation\naccuracy of the causal effects than does the back-door criterion.\n", "date": "2024-12-24", "categories": []}
{"id": "2503.03486", "title": "Differentially Private Learners for Heterogeneous Treatment Effects", "abstract": "  Patient data is widely used to estimate heterogeneous treatment effects and\nthus understand the effectiveness and safety of drugs. Yet, patient data\nincludes highly sensitive information that must be kept private. In this work,\nwe aim to estimate the conditional average treatment effect (CATE) from\nobservational data under differential privacy. Specifically, we present\nDP-CATE, a novel framework for CATE estimation that is Neyman-orthogonal and\nfurther ensures differential privacy of the estimates. Our framework is highly\ngeneral: it applies to any two-stage CATE meta-learner with a Neyman-orthogonal\nloss function, and any machine learning model can be used for nuisance\nestimation. We further provide an extension of our DP-CATE, where we employ\nRKHS regression to release the complete CATE function while ensuring\ndifferential privacy. We demonstrate our DP-CATE across various experiments\nusing synthetic and real-world datasets. To the best of our knowledge, we are\nthe first to provide a framework for CATE estimation that is Neyman-orthogonal\nand differentially private.\n", "date": "2025-03-05", "categories": []}
{"id": "2503.03634", "title": "Feature Matching Intervention: Leveraging Observational Data for Causal\n  Representation Learning", "abstract": "  A major challenge in causal discovery from observational data is the absence\nof perfect interventions, making it difficult to distinguish causal features\nfrom spurious ones. We propose an innovative approach, Feature Matching\nIntervention (FMI), which uses a matching procedure to mimic perfect\ninterventions. We define causal latent graphs, extending structural causal\nmodels to latent feature space, providing a framework that connects FMI with\ncausal graph learning. Our feature matching procedure emulates perfect\ninterventions within these causal latent graphs. Theoretical results\ndemonstrate that FMI exhibits strong out-of-distribution (OOD)\ngeneralizability. Experiments further highlight FMI's superior performance in\neffectively identifying causal features solely from observational data.\n", "date": "2025-03-05", "categories": []}
{"id": "2503.03724", "title": "Deep Causal Behavioral Policy Learning: Applications to Healthcare", "abstract": "  We present a deep learning-based approach to studying dynamic clinical\nbehavioral regimes in diverse non-randomized healthcare settings. Our proposed\nmethodology - deep causal behavioral policy learning (DC-BPL) - uses deep\nlearning algorithms to learn the distribution of high-dimensional clinical\naction paths, and identifies the causal link between these action paths and\npatient outcomes. Specifically, our approach: (1) identifies the causal effects\nof provider assignment on clinical outcomes; (2) learns the distribution of\nclinical actions a given provider would take given evolving patient\ninformation; (3) and combines these steps to identify the optimal provider for\na given patient type and emulate that provider's care decisions. Underlying\nthis strategy, we train a large clinical behavioral model (LCBM) on electronic\nhealth records data using a transformer architecture, and demonstrate its\nability to estimate clinical behavioral policies. We propose a novel\ninterpretation of a behavioral policy learned using the LCBM: that it is an\nefficient encoding of complex, often implicit, knowledge used to treat a\npatient. This allows us to learn a space of policies that are critical to a\nwide range of healthcare applications, in which the vast majority of clinical\nknowledge is acquired tacitly through years of practice and only a tiny\nfraction of information relevant to patient care is written down (e.g. in\ntextbooks, studies or standardized guidelines).\n", "date": "2025-03-05", "categories": []}
{"id": "2212.03683", "title": "Neighborhood Adaptive Estimators for Causal Inference under Network\n  Interference", "abstract": "  Estimating causal effects has become an integral part of most applied fields.\nIn this work we consider the violation of the classical no-interference\nassumption with units connected by a network. For tractability, we consider a\nknown network that describes how interference may spread. Unlike previous work\nthe radius (and intensity) of the interference experienced by a unit is unknown\nand can depend on different (local) sub-networks and the assigned treatments.\nWe study estimators for the average direct treatment effect on the treated in\nsuch a setting under additive treatment effects. We establish rates of\nconvergence and distributional results. The proposed estimators considers all\npossible radii for each (local) treatment assignment pattern. In contrast to\nprevious work, we approximate the relevant network interference patterns that\nlead to good estimates of the interference. To handle feature engineering, a\nkey innovation is to propose the use of synthetic treatments to decouple the\ndependence. We provide simulations, an empirical illustration and insights for\nthe general study of interference.\n", "date": "2022-12-07", "categories": []}
{"id": "2405.19985", "title": "Targeted Sequential Indirect Experiment Design", "abstract": "  Scientific hypotheses typically concern specific aspects of complex,\nimperfectly understood or entirely unknown mechanisms, such as the effect of\ngene expression levels on phenotypes or how microbial communities influence\nenvironmental health. Such queries are inherently causal (rather than purely\nassociational), but in many settings, experiments can not be conducted directly\non the target variables of interest, but are indirect. Therefore, they perturb\nthe target variable, but do not remove potential confounding factors. If,\nadditionally, the resulting experimental measurements are multi-dimensional and\nthe studied mechanisms nonlinear, the query of interest is generally not\nidentified. We develop an adaptive strategy to design indirect experiments that\noptimally inform a targeted query about the ground truth mechanism in terms of\nsequentially narrowing the gap between an upper and lower bound on the query.\nWhile the general formulation consists of a bi-level optimization procedure, we\nderive an efficiently estimable analytical kernel-based estimator of the bounds\nfor the causal effect, a query of key interest, and demonstrate the efficacy of\nour approach in confounded, multivariate, nonlinear synthetic settings.\n", "date": "2024-05-30", "categories": []}
{"id": "2406.06348", "title": "Causal Discovery over High-Dimensional Structured Hypothesis Spaces with\n  Causal Graph Partitioning", "abstract": "  The aim in many sciences is to understand the mechanisms that underlie the\nobserved distribution of variables, starting from a set of initial hypotheses.\nCausal discovery allows us to infer mechanisms as sets of cause and effect\nrelationships in a generalized way -- without necessarily tailoring to a\nspecific domain. Causal discovery algorithms search over a structured\nhypothesis space, defined by the set of directed acyclic graphs, to find the\ngraph that best explains the data. For high-dimensional problems, however, this\nsearch becomes intractable and scalable algorithms for causal discovery are\nneeded to bridge the gap. In this paper, we define a novel causal graph\npartition that allows for divide-and-conquer causal discovery with theoretical\nguarantees. We leverage the idea of a superstructure -- a set of learned or\nexisting candidate hypotheses -- to partition the search space. We prove under\ncertain assumptions that learning with a causal graph partition always yields\nthe Markov Equivalence Class of the true causal graph. We show our algorithm\nachieves comparable accuracy and a faster time to solution for\nbiologically-tuned synthetic networks and networks up to ${10^4}$ variables.\nThis makes our method applicable to gene regulatory network inference and other\ndomains with high-dimensional structured hypothesis spaces.\n", "date": "2024-06-10", "categories": []}
{"id": "2503.01139", "title": "Can Large Language Models Help Experimental Design for Causal Discovery?", "abstract": "  Designing proper experiments and selecting optimal intervention targets is a\nlongstanding problem in scientific or causal discovery. Identifying the\nunderlying causal structure from observational data alone is inherently\ndifficult. Obtaining interventional data, on the other hand, is crucial to\ncausal discovery, yet it is usually expensive and time-consuming to gather\nsufficient interventional data to facilitate causal discovery. Previous\napproaches commonly utilize uncertainty or gradient signals to determine the\nintervention targets. However, numerical-based approaches may yield suboptimal\nresults due to the inaccurate estimation of the guiding signals at the\nbeginning when with limited interventional data. In this work, we investigate a\ndifferent approach, whether we can leverage Large Language Models (LLMs) to\nassist with the intervention targeting in causal discovery by making use of the\nrich world knowledge about the experimental design in LLMs. Specifically, we\npresent Large Language Model Guided Intervention Targeting (LeGIT) -- a robust\nframework that effectively incorporates LLMs to augment existing numerical\napproaches for the intervention targeting in causal discovery. Across 4\nrealistic benchmark scales, LeGIT demonstrates significant improvements and\nrobustness over existing methods and even surpasses humans, which demonstrates\nthe usefulness of LLMs in assisting with experimental design for scientific\ndiscovery.\n", "date": "2025-03-03", "categories": []}
{"id": "2503.02093", "title": "Correlation to Causation: A Causal Deep Learning Framework for Arctic\n  Sea Ice Prediction", "abstract": "  Traditional machine learning and deep learning techniques rely on\ncorrelation-based learning, often failing to distinguish spurious associations\nfrom true causal relationships, which limits robustness, interpretability, and\ngeneralizability. To address these challenges, we propose a causality-driven\ndeep learning framework that integrates Multivariate Granger Causality (MVGC)\nand PCMCI+ causal discovery algorithms with a hybrid deep learning\narchitecture. Using 43 years (1979-2021) of daily and monthly Arctic Sea Ice\nExtent (SIE) and ocean-atmospheric datasets, our approach identifies causally\nsignificant factors, prioritizes features with direct influence, reduces\nfeature overhead, and improves computational efficiency. Experiments\ndemonstrate that integrating causal features enhances the deep learning model's\npredictive accuracy and interpretability across multiple lead times. Beyond SIE\nprediction, the proposed framework offers a scalable solution for dynamic,\nhigh-dimensional systems, advancing both theoretical understanding and\npractical applications in predictive modeling.\n", "date": "2025-03-03", "categories": []}
{"id": "2402.18477", "title": "Signature Kernel Conditional Independence Tests in Causal Discovery for\n  Stochastic Processes", "abstract": "  Inferring the causal structure underlying stochastic dynamical systems from\nobservational data holds great promise in domains ranging from science and\nhealth to finance. Such processes can often be accurately modeled via\nstochastic differential equations (SDEs), which naturally imply causal\nrelationships via \"which variables enter the differential of which other\nvariables\". In this paper, we develop conditional independence (CI) constraints\non coordinate processes over selected intervals that are Markov with respect to\nthe acyclic dependence graph (allowing self-loops) induced by a general SDE\nmodel. We then provide a sound and complete causal discovery algorithm, capable\nof handling both fully and partially observed data, and uniquely recovering the\nunderlying or induced ancestral graph by exploiting time directionality\nassuming a CI oracle. Finally, to make our algorithm practically usable, we\nalso propose a flexible, consistent signature kernel-based CI test to infer\nthese constraints from data. We extensively benchmark the CI test in isolation\nand as part of our causal discovery algorithms, outperforming existing\napproaches in SDE models and beyond.\n", "date": "2024-02-28", "categories": []}
{"id": "2502.18960", "title": "Nonparametric Heterogeneous Long-term Causal Effect Estimation via Data\n  Combination", "abstract": "  Long-term causal inference has drawn increasing attention in many scientific\ndomains. Existing methods mainly focus on estimating average long-term causal\neffects by combining long-term observational data and short-term experimental\ndata. However, it is still understudied how to robustly and effectively\nestimate heterogeneous long-term causal effects, significantly limiting\npractical applications. In this paper, we propose several two-stage style\nnonparametric estimators for heterogeneous long-term causal effect estimation,\nincluding propensity-based, regression-based, and multiple robust estimators.\nWe conduct a comprehensive theoretical analysis of their asymptotic properties\nunder mild assumptions, with the ultimate goal of building a better\nunderstanding of the conditions under which some estimators can be expected to\nperform better. Extensive experiments across several semi-synthetic and\nreal-world datasets validate the theoretical results and demonstrate the\neffectiveness of the proposed estimators.\n", "date": "2025-02-26", "categories": []}
{"id": "2503.00725", "title": "Causal Inference on Outcomes Learned from Text", "abstract": "  We propose a machine-learning tool that yields causal inference on text in\nrandomized trials. Based on a simple econometric framework in which text may\ncapture outcomes of interest, our procedure addresses three questions: First,\nis the text affected by the treatment? Second, which outcomes is the effect on?\nAnd third, how complete is our description of causal effects? To answer all\nthree questions, our approach uses large language models (LLMs) that suggest\nsystematic differences across two groups of text documents and then provides\nvalid inference based on costly validation. Specifically, we highlight the need\nfor sample splitting to allow for statistical validation of LLM outputs, as\nwell as the need for human labeling to validate substantive claims about how\ndocuments differ across groups. We illustrate the tool in a proof-of-concept\napplication using abstracts of academic manuscripts.\n", "date": "2025-03-02", "categories": []}
{"id": "2503.01475", "title": "ProRCA: A Causal Python Package for Actionable Root Cause Analysis in\n  Real-world Business Scenarios", "abstract": "  Root Cause Analysis (RCA) is becoming ever more critical as modern systems\ngrow in complexity, volume of data, and interdependencies. While traditional\nRCA methods frequently rely on correlation-based or rule-based techniques,\nthese approaches can prove inadequate in highly dynamic, multi-layered\nenvironments. In this paper, we present a pathway-tracing package built on the\nDoWhy causal inference library. Our method integrates conditional anomaly\nscoring, noise-based attribution, and depth-first path exploration to reveal\nmulti-hop causal chains. By systematically tracing entire causal pathways from\nan observed anomaly back to the initial triggers, our approach provides a\ncomprehensive, end-to-end RCA solution. Experimental evaluations with synthetic\nanomaly injections demonstrate the package's ability to accurately isolate\ntriggers and rank root causes by their overall significance.\n", "date": "2025-03-03", "categories": []}
{"id": "2208.06269", "title": "Probabilistic Variational Causal Approach in Observational Studies", "abstract": "  In this paper, we introduce a new causal methodology that accounts for the\nrarity and frequency of events in observational studies based on their\nrelevance to the underlying problem. Specifically, we propose a direct causal\neffect metric called the Probabilistic vAriational Causal Effect (PACE) and its\nvariations adhering to certain postulates applicable to both non-binary and\nbinary treatments. The PACE metric is derived by integrating the concept of\ntotal variation, representing the purely causal component, with interventions\non the treatment value, combined with the probabilities of hypothetical\ntransitioning between treatment levels. PACE features a parameter $d$, where\nlower values of $d$ correspond to scenarios emphasizing rare treatment values,\nwhile higher values of $d$ focus on situations where the causal impact of more\nfrequent treatment levels is more relevant. Thus, instead of a single causal\neffect value, we provide a causal effect function of the degree $d$.\nAdditionally, we introduce positive and negative PACE to measure the respective\npositive and negative causal changes in the outcome as exposure values shift.\nWe also consider normalized versions of PACE, referred to MEAN PACE.\nFurthermore, we provide an identifiability criterion for PACE to handle\ncounterfactual challenges in observational studies, and we define several\ngeneralizations of our methodology. Lastly, we compare our framework with other\nwell-known causal frameworks through the analysis of various examples.\n", "date": "2022-08-12", "categories": []}
{"id": "2404.11922", "title": "Redefining the Shortest Path Problem Formulation of the Linear\n  Non-Gaussian Acyclic Model: Pairwise Likelihood Ratios, Prior Knowledge, and\n  Path Enumeration", "abstract": "  Effective causal discovery is essential for learning the causal graph from\nobservational data. The linear non-Gaussian acyclic model (LiNGAM) operates\nunder the assumption of a linear data generating process with non-Gaussian\nnoise in determining the causal graph. Its assumption of unmeasured confounders\nbeing absent, however, poses practical limitations. In response, empirical\nresearch has shown that the reformulation of LiNGAM as a shortest path problem\n(LiNGAM-SPP) addresses this limitation. Within LiNGAM-SPP, mutual information\nis chosen to serve as the measure of independence. A challenge is introduced -\nparameter tuning is now needed due to its reliance on kNN mutual information\nestimators. The paper proposes a threefold enhancement to the LiNGAM-SPP\nframework.\n  First, the need for parameter tuning is eliminated by using the pairwise\nlikelihood ratio in lieu of kNN-based mutual information. This substitution is\nvalidated on a general data generating process and benchmark real-world data\nsets, outperforming existing methods especially when given a larger set of\nfeatures. The incorporation of prior knowledge is then enabled by a\nnode-skipping strategy implemented on the graph representation of all causal\norderings to eliminate violations based on the provided input of relative\norderings. Flexibility relative to existing approaches is achieved. Last among\nthe three enhancements is the utilization of the distribution of paths in the\ngraph representation of all causal orderings. From this, crucial properties of\nthe true causal graph such as the presence of unmeasured confounders and\nsparsity may be inferred. To some extent, the expected performance of the\ncausal discovery algorithm may be predicted. The refinements above advance the\npracticality and performance of LiNGAM-SPP, showcasing the potential of\ngraph-search-based methodologies in advancing causal discovery.\n", "date": "2024-04-18", "categories": []}
{"id": "2502.19551", "title": "Extremely Greedy Equivalence Search", "abstract": "  The goal of causal discovery is to learn a directed acyclic graph from data.\nOne of the most well-known methods for this problem is Greedy Equivalence\nSearch (GES). GES searches for the graph by incrementally and greedily adding\nor removing edges to maximize a model selection criterion. It has strong\ntheoretical guarantees on infinite data but can fail in practice on finite\ndata. In this paper, we first identify some of the causes of GES's failure,\nfinding that it can get blocked in local optima, especially in denser graphs.\nWe then propose eXtremely Greedy Equivalent Search (XGES), which involves a new\nheuristic to improve the search strategy of GES while retaining its theoretical\nguarantees. In particular, XGES favors deleting edges early in the search over\ninserting edges, which reduces the possibility of the search ending in local\noptima. A further contribution of this work is an efficient algorithmic\nformulation of XGES (and GES). We benchmark XGES on simulated datasets with\nknown ground truth. We find that XGES consistently outperforms GES in\nrecovering the correct graphs, and it is 10 times faster. XGES implementations\nin Python and C++ are available at https://github.com/ANazaret/XGES.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.20153", "title": "Transfer Learning in Latent Contextual Bandits with Covariate Shift\n  Through Causal Transportability", "abstract": "  Transferring knowledge from one environment to another is an essential\nability of intelligent systems. Nevertheless, when two environments are\ndifferent, naively transferring all knowledge may deteriorate the performance,\na phenomenon known as negative transfer. In this paper, we address this issue\nwithin the framework of multi-armed bandits from the perspective of causal\ninference. Specifically, we consider transfer learning in latent contextual\nbandits, where the actual context is hidden, but a potentially high-dimensional\nproxy is observable. We further consider a covariate shift in the context\nacross environments. We show that naively transferring all knowledge for\nclassical bandit algorithms in this setting led to negative transfer. We then\nleverage transportability theory from causal inference to develop algorithms\nthat explicitly transfer effective knowledge for estimating the causal effects\nof interest in the target environment. Besides, we utilize variational\nautoencoders to approximate causal effects under the presence of a\nhigh-dimensional proxy. We test our algorithms on synthetic and semi-synthetic\ndatasets, empirically demonstrating consistently improved learning efficiency\nacross different proxies compared to baseline algorithms, showing the\neffectiveness of our causal framework in transferring knowledge.\n", "date": "2025-02-27", "categories": []}
{"id": "2405.03342", "title": "Doubly Robust Causal Effect Estimation under Networked Interference via\n  Targeted Learning", "abstract": "  Causal effect estimation under networked interference is an important but\nchallenging problem. Available parametric methods are limited in their model\nspace, while previous semiparametric methods, e.g., leveraging neural networks\nto fit only one single nuisance function, may still encounter misspecification\nproblems under networked interference without appropriate assumptions on the\ndata generation process. To mitigate bias stemming from misspecification, we\npropose a novel doubly robust causal effect estimator under networked\ninterference, by adapting the targeted learning technique to the training of\nneural networks. Specifically, we generalize the targeted learning technique\ninto the networked interference setting and establish the condition under which\nan estimator achieves double robustness. Based on the condition, we devise an\nend-to-end causal effect estimator by transforming the identified theoretical\ncondition into a targeted loss. Moreover, we provide a theoretical analysis of\nour designed estimator, revealing a faster convergence rate compared to a\nsingle nuisance model. Extensive experimental results on two real-world\nnetworks with semisynthetic data demonstrate the effectiveness of our proposed\nestimators.\n", "date": "2024-05-06", "categories": []}
{"id": "2301.07690", "title": "CaRE: Finding Root Causes of Configuration Issues in Highly-Configurable\n  Robots", "abstract": "  Robotic systems have subsystems with a combinatorially large configuration\nspace and hundreds or thousands of possible software and hardware configuration\noptions interacting non-trivially. The configurable parameters are set to\ntarget specific objectives, but they can cause functional faults when\nincorrectly configured. Finding the root cause of such faults is challenging\ndue to the exponentially large configuration space and the dependencies between\nthe robot's configuration settings and performance. This paper proposes CaRE --\na method for diagnosing the root cause of functional faults through the lens of\ncausality. CaRE abstracts the causal relationships between various\nconfiguration options and the robot's performance objectives by learning a\ncausal structure and estimating the causal effects of options on robot\nperformance indicators. We demonstrate CaRE's efficacy by finding the root\ncause of the observed functional faults and validating the diagnosed root cause\nby conducting experiments in both physical robots (Husky and Turtlebot 3) and\nin simulation (Gazebo). Furthermore, we demonstrate that the causal models\nlearned from robots in simulation (e.g., Husky in Gazebo) are transferable to\nphysical robots across different platforms (e.g., Husky and Turtlebot 3).\n", "date": "2023-01-18", "categories": []}
{"id": "2502.17538", "title": "Policy Learning with a Natural Language Action Space: A Causal Approach", "abstract": "  This paper introduces a novel causal framework for multi-stage\ndecision-making in natural language action spaces where outcomes are only\nobserved after a sequence of actions. While recent approaches like Proximal\nPolicy Optimization (PPO) can handle such delayed-reward settings in\nhigh-dimensional action spaces, they typically require multiple models (policy,\nvalue, and reward) and substantial training data. Our approach employs\nQ-learning to estimate Dynamic Treatment Regimes (DTR) through a single model,\nenabling data-efficient policy learning via gradient ascent on language\nembeddings. A key technical contribution of our approach is a decoding strategy\nthat translates optimized embeddings back into coherent natural language. We\nevaluate our approach on mental health intervention, hate speech countering,\nand sentiment transfer tasks, demonstrating significant improvements over\ncompetitive baselines across multiple metrics. Notably, our method achieves\nsuperior transfer strength while maintaining content preservation and fluency,\nas validated through human evaluation. Our work provides a practical foundation\nfor learning optimal policies in complex language tasks where training data is\nlimited.\n", "date": "2025-02-24", "categories": []}
{"id": "2502.18240", "title": "Causal AI-based Root Cause Identification: Research to Practice at Scale", "abstract": "  Modern applications are built as large, distributed systems spanning numerous\nmodules, teams, and data centers. Despite robust engineering and recovery\nstrategies, failures and performance issues remain inevitable, risking\nsignificant disruptions and affecting end users. Rapid and accurate root cause\nidentification is therefore vital to ensure system reliability and maintain key\nservice metrics.\n  We have developed a novel causality-based Root Cause Identification (RCI)\nalgorithm that emphasizes causation over correlation. This algorithm has been\nintegrated into IBM Instana-bridging research to practice at scale-and is now\nin production use by enterprise customers. By leveraging \"causal AI,\" Instana\nstands apart from typical Application Performance Management (APM) tools,\npinpointing issues in near real-time. This paper highlights Instana's advanced\nfailure diagnosis capabilities, discussing both the theoretical underpinnings\nand practical implementations of the RCI algorithm. Real-world examples\nillustrate how our causality-based approach enhances reliability and\nperformance in today's complex system landscapes.\n", "date": "2025-02-25", "categories": []}
{"id": "2502.15697", "title": "Robust Uplift Modeling with Large-Scale Contexts for Real-time Marketing", "abstract": "  Improving user engagement and platform revenue is crucial for online\nmarketing platforms. Uplift modeling is proposed to solve this problem, which\napplies different treatments (e.g., discounts, bonus) to satisfy corresponding\nusers. Despite progress in this field, limitations persist. Firstly, most of\nthem focus on scenarios where only user features exist. However, in real-world\nscenarios, there are rich contexts available in the online platform (e.g.,\nshort videos, news), and the uplift model needs to infer an incentive for each\nuser on the specific item, which is called real-time marketing. Thus, only\nconsidering the user features will lead to biased prediction of the responses,\nwhich may cause the cumulative error for uplift prediction. Moreover, due to\nthe large-scale contexts, directly concatenating the context features with the\nuser features will cause a severe distribution shift in the treatment and\ncontrol groups. Secondly, capturing the interaction relationship between the\nuser features and context features can better predict the user response. To\nsolve the above limitations, we propose a novel model-agnostic Robust Uplift\nModeling with Large-Scale Contexts (UMLC) framework for Real-time Marketing.\nOur UMLC includes two customized modules. 1) A response-guided context grouping\nmodule for extracting context features information and condensing value space\nthrough clusters. 2) A feature interaction module for obtaining better uplift\nprediction. Specifically, this module contains two parts: a user-context\ninteraction component for better modeling the response; a treatment-feature\ninteraction component for discovering the treatment assignment sensitive\nfeature of each instance to better predict the uplift. Moreover, we conduct\nextensive experiments on a synthetic dataset and a real-world product dataset\nto verify the effectiveness and compatibility of our UMLC.\n", "date": "2025-01-04", "categories": []}
{"id": "2502.16056", "title": "Since Faithfulness Fails: The Performance Limits of Neural Causal\n  Discovery", "abstract": "  Neural causal discovery methods have recently improved in terms of\nscalability and computational efficiency. However, our systematic evaluation\nhighlights significant room for improvement in their accuracy when uncovering\ncausal structures. We identify a fundamental limitation: neural networks cannot\nreliably distinguish between existing and non-existing causal relationships in\nthe finite sample regime. Our experiments reveal that neural networks, as used\nin contemporary causal discovery approaches, lack the precision needed to\nrecover ground-truth graphs, even for small graphs and relatively large sample\nsizes. Furthermore, we identify the faithfulness property as a critical\nbottleneck: (i) it is likely to be violated across any reasonable dataset size\nrange, and (ii) its violation directly undermines the performance of neural\ndiscovery methods. These findings lead us to conclude that progress within the\ncurrent paradigm is fundamentally constrained, necessitating a paradigm shift\nin this domain.\n", "date": "2025-02-22", "categories": []}
{"id": "2502.16172", "title": "Practical programming research of Linear DML model based on the simplest\n  Python code: From the standpoint of novice researchers", "abstract": "  This paper presents linear DML models for causal inference using the simplest\nPython code on a Jupyter notebook based on an Anaconda platform and compares\nthe performance of different DML models. The results show that current Library\nAPI technology is not yet sufficient to enable novice Python users to build\nqualified and high-quality DML models with the simplest coding approach. Novice\nusers attempting to perform DML causal inference using Python still have to\nimprove their mathematical and computer knowledge to adapt to more flexible DML\nprogramming. Additionally, the issue of mismatched outcome variable dimensions\nis also widespread when building linear DML models in Jupyter notebook.\n", "date": "2025-02-22", "categories": []}
{"id": "2502.07891", "title": "The Observational Partial Order of Causal Structures with Latent\n  Variables", "abstract": "  For two causal structures with the same set of visible variables, one is said\nto observationally dominate the other if the set of distributions over the\nvisible variables realizable by the first contains the set of distributions\nover the visible variables realizable by the second. Knowing such dominance\nrelations is useful for adjudicating between these structures given\nobservational data. We here consider the problem of determining the partial\norder of equivalence classes of causal structures with latent variables\nrelative to observational dominance. We provide a complete characterization of\nthe dominance order in the case of three visible variables, and a partial\ncharacterization in the case of four visible variables. Our techniques also\nhelp to identify which observational equivalence classes have a set of\nrealizable distributions that is characterized by nontrivial inequality\nconstraints, analogous to Bell inequalities and instrumental inequalities. We\nfind evidence that as one increases the number of visible variables, the\nequivalence classes satisfying nontrivial inequality constraints become\nubiquitous. (Because such classes are the ones for which there can be a\ndifference in the distributions that are quantumly and classically realizable,\nthis implies that the potential for quantum-classical gaps is also ubiquitous.)\nFurthermore, we find evidence that constraint-based causal discovery algorithms\nthat rely solely on conditional independence constraints have a significantly\nweaker distinguishing power among observational equivalence classes than\nalgorithms that go beyond these (i.e., algorithms that also leverage nested\nMarkov constraints and inequality constraints).\n", "date": "2025-02-11", "categories": []}
{"id": "2410.07021", "title": "Do Contemporary Causal Inference Models Capture Real-World\n  Heterogeneity? Findings from a Large-Scale Benchmark", "abstract": "  We present unexpected findings from a large-scale benchmark study evaluating\nConditional Average Treatment Effect (CATE) estimation algorithms, i.e., CATE\nmodels. By running 16 modern CATE models on 12 datasets and 43,200 sampled\nvariants generated through diverse observational sampling strategies, we find\nthat: (a) 62\\% of CATE estimates have a higher Mean Squared Error (MSE) than a\ntrivial zero-effect predictor, rendering them ineffective; (b) in datasets with\nat least one useful CATE estimate, 80\\% still have higher MSE than a\nconstant-effect model; and (c) Orthogonality-based models outperform other\nmodels only 30\\% of the time, despite widespread optimism about their\nperformance. These findings highlight significant challenges in current CATE\nmodels and underscore the need for broader evaluation and methodological\nimprovements.\n  Our findings stem from a novel application of \\textit{observational\nsampling}, originally developed to evaluate Average Treatment Effect (ATE)\nestimates from observational methods with experiment data. To adapt\nobservational sampling for CATE evaluation, we introduce a statistical\nparameter, $Q$, equal to MSE minus a constant and preserves the ranking of\nmodels by their MSE. We then derive a family of sample statistics, collectively\ncalled $\\hat{Q}$, that can be computed from real-world data. When used in\nobservational sampling, $\\hat{Q}$ is an unbiased estimator of $Q$ and\nasymptotically selects the model with the smallest MSE. To ensure the benchmark\nreflects real-world heterogeneity, we handpick datasets where outcomes come\nfrom field rather than simulation. By integrating observational sampling, new\nstatistics, and real-world datasets, the benchmark provides new insights into\nCATE model performance and reveals gaps in capturing real-world heterogeneity,\nemphasizing the need for more robust benchmarks.\n", "date": "2024-10-09", "categories": []}
{"id": "2412.14497", "title": "Disentangled Graph Autoencoder for Treatment Effect Estimation", "abstract": "  Treatment effect estimation from observational data has attracted significant\nattention across various research fields. However, many widely used methods\nrely on the unconfoundedness assumption, which is often unrealistic due to the\ninability to observe all confounders, thereby overlooking the influence of\nlatent confounders. To address this limitation, recent approaches have utilized\nauxiliary network information to infer latent confounders, relaxing this\nassumption. However, these methods often treat observed variables and networks\nas proxies only for latent confounders, which can result in inaccuracies when\ncertain variables influence treatment without affecting outcomes, or vice\nversa. This conflation of distinct latent factors undermines the precision of\ntreatment effect estimation. To overcome this challenge, we propose a novel\ndisentangled variational graph autoencoder for treatment effect estimation on\nnetworked observational data. Our graph encoder disentangles latent factors\ninto instrumental, confounding, adjustment, and noisy factors, while enforcing\nfactor independence using the Hilbert-Schmidt Independence Criterion. Extensive\nexperiments on multiple networked datasets demonstrate that our method\noutperforms state-of-the-art approaches.\n", "date": "2024-12-19", "categories": []}
{"id": "2502.14755", "title": "Multi-Objective Causal Bayesian Optimization", "abstract": "  In decision-making problems, the outcome of an intervention often depends on\nthe causal relationships between system components and is highly costly to\nevaluate. In such settings, causal Bayesian optimization (CBO) can exploit the\ncausal relationships between the system variables and sequentially perform\ninterventions to approach the optimum with minimal data. Extending CBO to the\nmulti-outcome setting, we propose Multi-Objective Causal Bayesian Optimization\n(MO-CBO), a paradigm for identifying Pareto-optimal interventions within a\nknown multi-target causal graph. We first derive a graphical characterization\nfor potentially optimal sets of variables to intervene upon. Showing that any\nMO-CBO problem can be decomposed into several traditional multi-objective\noptimization tasks, we then introduce an algorithm that sequentially balances\nexploration across these tasks using relative hypervolume improvement. The\nproposed method will be validated on both synthetic and real-world causal\ngraphs, demonstrating its superiority over traditional (non-causal)\nmulti-objective Bayesian optimization in settings where causal information is\navailable.\n", "date": "2025-02-20", "categories": []}
{"id": "2311.01412", "title": "Causal Temporal Regime Structure Learning", "abstract": "  Understanding causal relationships in multivariate time series is essential\nfor predicting and controlling dynamic systems in fields like economics,\nneuroscience, and climate science. However, existing causal discovery methods\noften assume stationarity, limiting their effectiveness when time series\nconsist of sequential regimes, consecutive temporal segments with unknown\nboundaries and changing causal structures. In this work, we firstly introduce a\nframework to describe and model such time series. Then, we present CASTOR, a\nnovel method that concurrently learns the Directed Acyclic Graph (DAG) for each\nregime while determining the number of regimes and their sequential\narrangement. CASTOR optimizes the data log-likelihood using an\nexpectation-maximization algorithm, alternating between assigning regime\nindices (expectation step) and inferring causal relationships in each regime\n(maximization step). We establish the identifiability of the regimes and DAGs\nwithin our framework. Extensive experiments show that CASTOR consistently\noutperforms existing causal discovery models in detecting different regimes and\nlearning their DAGs across various settings, including linear and nonlinear\ncausal relationships, on both synthetic and real world datasets.\n", "date": "2023-11-02", "categories": []}
{"id": "2502.13701", "title": "Causes and Strategies in Multiagent Systems", "abstract": "  Causality plays an important role in daily processes, human reasoning, and\nartificial intelligence. There has however not been much research on causality\nin multi-agent strategic settings. In this work, we introduce a systematic way\nto build a multi-agent system model, represented as a concurrent game\nstructure, for a given structural causal model. In the obtained so-called\ncausal concurrent game structure, transitions correspond to interventions on\nagent variables of the given causal model. The Halpern and Pearl framework of\ncausality is used to determine the effects of a certain value for an agent\nvariable on other variables. The causal concurrent game structure allows us to\nanalyse and reason about causal effects of agents' strategic decisions. We\nformally investigate the relation between causal concurrent game structures and\nthe original structural causal models.\n", "date": "2025-02-19", "categories": []}
{"id": "2406.09207", "title": "Investigating potential causes of Sepsis with Bayesian network structure\n  learning", "abstract": "  Sepsis is a life-threatening and serious global health issue. This study\ncombines knowledge with available hospital data to investigate the potential\ncauses of Sepsis that can be affected by policy decisions. We investigate the\nunderlying causal structure of this problem by combining clinical expertise\nwith score-based, constraint-based, and hybrid structure learning algorithms. A\nnovel approach to model averaging and knowledge-based constraints was\nimplemented to arrive at a consensus structure for causal inference. The\nstructure learning process highlighted the importance of exploring data-driven\napproaches alongside clinical expertise. This includes discovering unexpected,\nalthough reasonable, relationships from a clinical perspective. Hypothetical\ninterventions on Chronic Obstructive Pulmonary Disease, Alcohol dependence, and\nDiabetes suggest that the presence of any of these risk factors in patients\nincreases the likelihood of Sepsis. This finding, alongside measuring the\neffect of these risk factors on Sepsis, has potential policy implications.\nRecognising the importance of prediction in improving health outcomes related\nto Sepsis, the model is also assessed in its ability to predict Sepsis by\nevaluating accuracy, sensitivity, and specificity. These three indicators all\nhad results around 70%, and the AUC was 80%, which means the causal structure\nof the model is reasonably accurate given that the models were trained on data\navailable for commissioning purposes only.\n", "date": "2024-06-13", "categories": []}
{"id": "2410.03514", "title": "Stabilized Neural Prediction of Potential Outcomes in Continuous Time", "abstract": "  Patient trajectories from electronic health records are widely used to\nestimate conditional average potential outcomes (CAPOs) of treatments over\ntime, which then allows to personalize care. Yet, existing neural methods for\nthis purpose have a key limitation: while some adjust for time-varying\nconfounding, these methods assume that the time series are recorded in discrete\ntime. In other words, they are constrained to settings where measurements and\ntreatments are conducted at fixed time steps, even though this is unrealistic\nin medical practice. In this work, we aim to estimate CAPOs in continuous time.\nThe latter is of direct practical relevance because it allows for modeling\npatient trajectories where measurements and treatments take place at arbitrary,\nirregular timestamps. We thus propose a new method called stabilized continuous\ntime inverse propensity network (SCIP-Net). For this, we further derive\nstabilized inverse propensity weights for robust estimation of the CAPOs. To\nthe best of our knowledge, our SCIP-Net is the first neural method that\nperforms proper adjustments for time-varying confounding in continuous time.\n", "date": "2024-10-04", "categories": []}
{"id": "2410.15939", "title": "CausalGraph2LLM: Evaluating LLMs for Causal Queries", "abstract": "  Causality is essential in scientific research, enabling researchers to\ninterpret true relationships between variables. These causal relationships are\noften represented by causal graphs, which are directed acyclic graphs. With the\nrecent advancements in Large Language Models (LLMs), there is an increasing\ninterest in exploring their capabilities in causal reasoning and their\npotential use to hypothesize causal graphs. These tasks necessitate the LLMs to\nencode the causal graph effectively for subsequent downstream tasks. In this\npaper, we introduce CausalGraph2LLM, a comprehensive benchmark comprising over\n700k queries across diverse causal graph settings to evaluate the causal\nreasoning capabilities of LLMs. We categorize the causal queries into two\ntypes: graph-level and node-level queries. We benchmark both open-sourced and\npropriety models for our study. Our findings reveal that while LLMs show\npromise in this domain, they are highly sensitive to the encoding used. Even\ncapable models like GPT-4 and Gemini-1.5 exhibit sensitivity to encoding, with\ndeviations of about $60\\%$. We further demonstrate this sensitivity for\ndownstream causal intervention tasks. Moreover, we observe that LLMs can often\ndisplay biases when presented with contextual information about a causal graph,\npotentially stemming from their parametric memory.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.16676", "title": "CausalEval: Towards Better Causal Reasoning in Language Models", "abstract": "  Causal reasoning (CR) is a crucial aspect of intelligence, essential for\nproblem-solving, decision-making, and understanding the world. While language\nmodels (LMs) can generate rationales for their outputs, their ability to\nreliably perform causal reasoning remains uncertain, often falling short in\ntasks requiring a deep understanding of causality. In this paper, we introduce\nCausalEval, a comprehensive review of research aimed at enhancing LMs for\ncausal reasoning, coupled with an empirical evaluation of current models and\nmethods. We categorize existing methods based on the role of LMs: either as\nreasoning engines or as helpers providing knowledge or data to traditional CR\nmethods, followed by a detailed discussion of methodologies in each category.\nWe then assess the performance of current LMs and various enhancement methods\non a range of causal reasoning tasks, providing key findings and in-depth\nanalysis. Finally, we present insights from current studies and highlight\npromising directions for future research. We aim for this work to serve as a\ncomprehensive resource, fostering further advancements in causal reasoning with\nLMs.\n", "date": "2024-10-22", "categories": []}
{"id": "2502.12393", "title": "Time Series Treatment Effects Analysis with Always-Missing Controls", "abstract": "  Estimating treatment effects in time series data presents a significant\nchallenge, especially when the control group is always unobservable. For\nexample, in analyzing the effects of Christmas on retail sales, we lack direct\nobservation of what would have occurred in late December without the Christmas\nimpact. To address this, we try to recover the control group in the event\nperiod while accounting for confounders and temporal dependencies. Experimental\nresults on the M5 Walmart retail sales data demonstrate robust estimation of\nthe potential outcome of the control group as well as accurate predicted\nholiday effect. Furthermore, we provided theoretical guarantees for the\nestimated treatment effect, proving its consistency and asymptotic normality.\nThe proposed methodology is applicable not only to this always-missing control\nscenario but also in other conventional time series causal inference settings.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.12707", "title": "CausalMan: A physics-based simulator for large-scale causality", "abstract": "  A comprehensive understanding of causality is critical for navigating and\noperating within today's complex real-world systems. The absence of realistic\ncausal models with known data generating processes complicates fair\nbenchmarking. In this paper, we present the CausalMan simulator, modeled after\na real-world production line. The simulator features a diverse range of linear\nand non-linear mechanisms and challenging-to-predict behaviors, such as\ndiscrete mode changes. We demonstrate the inadequacy of many state-of-the-art\napproaches and analyze the significant differences in their performance and\ntractability, both in terms of runtime and memory complexity. As a\ncontribution, we will release the CausalMan large-scale simulator. We present\ntwo derived datasets, and perform an extensive evaluation of both.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.12874", "title": "Testing for Causal Fairness", "abstract": "  Causality is widely used in fairness analysis to prevent discrimination on\nsensitive attributes, such as genders in career recruitment and races in crime\nprediction. However, the current data-based Potential Outcomes Framework (POF)\noften leads to untrustworthy fairness analysis results when handling\nhigh-dimensional data. To address this, we introduce a distribution-based POF\nthat transform fairness analysis into Distributional Closeness Testing (DCT) by\nintervening on sensitive attributes. We define counterfactual closeness\nfairness as the null hypothesis of DCT, where a sensitive attribute is\nconsidered fair if its factual and counterfactual potential outcome\ndistributions are sufficiently close. We introduce the Norm-Adaptive Maximum\nMean Discrepancy Treatment Effect (N-TE) as a statistic for measuring\ndistributional closeness and apply DCT using the empirical estimator of NTE,\nreferred to Counterfactual Fairness-CLOseness Testing ($\\textrm{CF-CLOT}$). To\nensure the trustworthiness of testing results, we establish the testing\nconsistency of N-TE through rigorous theoretical analysis. $\\textrm{CF-CLOT}$\ndemonstrates sensitivity in fairness analysis through the flexibility of the\ncloseness parameter $\\epsilon$. Unfair sensitive attributes have been\nsuccessfully tested by $\\textrm{CF-CLOT}$ in extensive experiments across\nvarious real-world scenarios, which validate the consistency of the testing.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.13132", "title": "Learning to Defer for Causal Discovery with Imperfect Experts", "abstract": "  Integrating expert knowledge, e.g. from large language models, into causal\ndiscovery algorithms can be challenging when the knowledge is not guaranteed to\nbe correct. Expert recommendations may contradict data-driven results, and\ntheir reliability can vary significantly depending on the domain or specific\nquery. Existing methods based on soft constraints or inconsistencies in\npredicted causal relationships fail to account for these variations in\nexpertise. To remedy this, we propose L2D-CD, a method for gauging the\ncorrectness of expert recommendations and optimally combining them with\ndata-driven causal discovery results. By adapting learning-to-defer (L2D)\nalgorithms for pairwise causal discovery (CD), we learn a deferral function\nthat selects whether to rely on classical causal discovery methods using\nnumerical data or expert recommendations based on textual meta-data. We\nevaluate L2D-CD on the canonical T\\\"ubingen pairs dataset and demonstrate its\nsuperior performance compared to both the causal discovery method and the\nexpert used in isolation. Moreover, our approach identifies domains where the\nexpert's performance is strong or weak. Finally, we outline a strategy for\ngeneralizing this approach to causal discovery on graphs with more than two\nvariables, paving the way for further research in this area.\n", "date": "2025-02-18", "categories": []}
{"id": "2310.14691", "title": "Identifiability of total effects from abstractions of time series causal\n  graphs", "abstract": "  We study the problem of identifiability of the total effect of an\nintervention from observational time series in the situation, common in\npractice, where one only has access to abstractions of the true causal graph.\nWe consider here two abstractions: the extended summary causal graph, which\nconflates all lagged causal relations but distinguishes between lagged and\ninstantaneous relations, and the summary causal graph which does not give any\nindication about the lag between causal relations. We show that the total\neffect is always identifiable in extended summary causal graphs and provide\nsufficient conditions for identifiability in summary causal graphs. We\nfurthermore provide adjustment sets allowing to estimate the total effect\nwhenever it is identifiable.\n", "date": "2023-10-23", "categories": []}
{"id": "2402.11068", "title": "Large Language Models for Causal Discovery: Current Landscape and Future\n  Directions", "abstract": "  Causal discovery (CD) and Large Language Models (LLMs) have emerged as\ntransformative fields in artificial intelligence that have evolved largely\nindependently. While CD specializes in uncovering cause-effect relationships\nfrom data, and LLMs excel at natural language processing and generation, their\nintegration presents unique opportunities for advancing causal understanding.\nThis survey examines how LLMs are transforming CD across three key dimensions:\ndirect causal extraction from text, integration of domain knowledge into\nstatistical methods, and refinement of causal structures. We systematically\nanalyze approaches that leverage LLMs for CD tasks, highlighting their\ninnovative use of metadata and natural language for causal inference. Our\nanalysis reveals both LLMs' potential to enhance traditional CD methods and\ntheir current limitations as imperfect expert systems. We identify key research\ngaps, outline evaluation frameworks and benchmarks for LLM-based causal\ndiscovery, and advocate future research efforts for leveraging LLMs in\ncausality research. As the first comprehensive examination of the synergy\nbetween LLMs and CD, this work lays the groundwork for future advances in the\nfield.\n", "date": "2024-02-16", "categories": []}
{"id": "2411.01292", "title": "Causal reasoning in difference graphs", "abstract": "  Understanding causal mechanisms across different populations is essential for\ndesigning effective public health interventions. Recently, difference graphs\nhave been introduced as a tool to visually represent causal variations between\ntwo distinct populations. While there has been progress in inferring these\ngraphs from data through causal discovery methods, there remains a gap in\nsystematically leveraging their potential to enhance causal reasoning. This\npaper addresses that gap by establishing conditions for identifying causal\nchanges and effects using difference graphs. It specifically focuses on\nidentifying total causal changes and total effects in a nonparametric setting,\nas well as direct causal changes and direct effects in a linear setting. In\ndoing so, it provides a novel approach to causal reasoning that holds potential\nfor various public health applications.\n", "date": "2024-11-02", "categories": []}
{"id": "2502.10883", "title": "Learning Identifiable Structures Helps Avoid Bias in DNN-based\n  Supervised Causal Learning", "abstract": "  Causal discovery is a structured prediction task that aims to predict causal\nrelations among variables based on their data samples. Supervised Causal\nLearning (SCL) is an emerging paradigm in this field. Existing Deep Neural\nNetwork (DNN)-based methods commonly adopt the \"Node-Edge approach\", in which\nthe model first computes an embedding vector for each variable-node, then uses\nthese variable-wise representations to concurrently and independently predict\nfor each directed causal-edge. In this paper, we first show that this\narchitecture has some systematic bias that cannot be mitigated regardless of\nmodel size and data size. We then propose SiCL, a DNN-based SCL method that\npredicts a skeleton matrix together with a v-tensor (a third-order tensor\nrepresenting the v-structures). According to the Markov Equivalence Class (MEC)\ntheory, both the skeleton and the v-structures are identifiable causal\nstructures under the canonical MEC setting, so predictions about skeleton and\nv-structures do not suffer from the identifiability limit in causal discovery,\nthus SiCL can avoid the systematic bias in Node-Edge architecture, and enable\nconsistent estimators for causal discovery. Moreover, SiCL is also equipped\nwith a specially designed pairwise encoder module with a unidirectional\nattention layer to model both internal and external relationships of pairs of\nnodes. Experimental results on both synthetic and real-world benchmarks show\nthat SiCL significantly outperforms other DNN-based SCL approaches.\n", "date": "2025-02-15", "categories": []}
{"id": "2502.11680", "title": "Spectral structure learning for clinical time series", "abstract": "  We develop and evaluate a structure learning algorithm for clinical time\nseries. Clinical time series are multivariate time series observed in multiple\npatients and irregularly sampled, challenging existing structure learning\nalgorithms. We assume that our times series are realizations of StructGP, a\nk-dimensional multi-output or multi-task stationary Gaussian process (GP), with\nindependent patients sharing the same covariance function. StructGP encodes\nordered conditional relations between time series, represented in a directed\nacyclic graph. We implement an adapted NOTEARS algorithm, which based on a\ndifferentiable definition of acyclicity, recovers the graph by solving a series\nof continuous optimization problems. Simulation results show that up to mean\ndegree 3 and 20 tasks, we reach a median recall of 0.93% [IQR, 0.86, 0.97]\nwhile keeping a median precision of 0.71% [0.57-0.84], for recovering directed\nedges. We further show that the regularization path is key to identifying the\ngraph. With StructGP, we proposed a model of time series dependencies, that\nflexibly adapt to different time series regularity, while enabling us to learn\nthese dependencies from observations.\n", "date": "2025-02-17", "categories": []}
{"id": "2410.13914", "title": "Exogenous Matching: Learning Good Proposals for Tractable Counterfactual\n  Estimation", "abstract": "  We propose an importance sampling method for tractable and efficient\nestimation of counterfactual expressions in general settings, named Exogenous\nMatching. By minimizing a common upper bound of counterfactual estimators, we\ntransform the variance minimization problem into a conditional distribution\nlearning problem, enabling its integration with existing conditional\ndistribution modeling approaches. We validate the theoretical results through\nexperiments under various types and settings of Structural Causal Models (SCMs)\nand demonstrate the outperformance on counterfactual estimation tasks compared\nto other existing importance sampling methods. We also explore the impact of\ninjecting structural prior knowledge (counterfactual Markov boundaries) on the\nresults. Finally, we apply this method to identifiable proxy SCMs and\ndemonstrate the unbiasedness of the estimates, empirically illustrating the\napplicability of the method to practical scenarios.\n", "date": "2024-10-17", "categories": []}
{"id": "2502.08776", "title": "Treatment response as a latent variable", "abstract": "  Scientists often need to analyze the samples in a study that responded to\ntreatment in order to refine their hypotheses and find potential causal drivers\nof response. Natural variation in outcomes makes teasing apart responders from\nnon-responders a statistical inference problem. To handle latent responses, we\nintroduce the causal two-groups (C2G) model, a causal extension of the\nclassical two-groups model. The C2G model posits that treated samples may or\nmay not experience an effect, according to some prior probability. We propose\ntwo empirical Bayes procedures for the causal two-groups model, one under\nsemi-parametric conditions and another under fully nonparametric conditions.\nThe semi-parametric model assumes additive treatment effects and is\nidentifiable from observed data. The nonparametric model is unidentifiable, but\nwe show it can still be used to test for response in each treated sample. We\nshow empirically and theoretically that both methods for selecting responders\ncontrol the false discovery rate at the target level with near-optimal power.\nWe also propose two novel estimands of interest and provide a strategy for\nderiving estimand intervals in the unidentifiable nonparametric model. On a\ncancer immunotherapy dataset, the nonparametric C2G model recovers\nclinically-validated predictive biomarkers of both positive and negative\noutcomes. Code is available at https://github.com/tansey-lab/causal2groups.\n", "date": "2025-02-12", "categories": []}
{"id": "2502.08858", "title": "Estimating Probabilities of Causation with Machine Learning Models", "abstract": "  Probabilities of causation play a crucial role in modern decision-making.\nThis paper addresses the challenge of predicting probabilities of causation for\nsubpopulations with insufficient data using machine learning models. Tian and\nPearl first defined and derived tight bounds for three fundamental\nprobabilities of causation: the probability of necessity and sufficiency (PNS),\nthe probability of sufficiency (PS), and the probability of necessity (PN).\nHowever, estimating these probabilities requires both experimental and\nobservational distributions specific to each subpopulation, which are often\nunavailable or impractical to obtain with limited population-level data. We\nassume that the probabilities of causation for each subpopulation are\ndetermined by its characteristics. To estimate these probabilities for\nsubpopulations with insufficient data, we propose using machine learning models\nthat draw insights from subpopulations with sufficient data. Our evaluation of\nmultiple machine learning models indicates that, given sufficient\npopulation-level data and an appropriate choice of machine learning model and\nactivation function, PNS can be effectively predicted. Through simulation\nstudies, we show that our multilayer perceptron (MLP) model with the Mish\nactivation function achieves a mean absolute error (MAE) of approximately 0.02\nin predicting PNS for 32,768 subpopulations using data from around 2,000\nsubpopulations.\n", "date": "2025-02-13", "categories": []}
{"id": "2502.09226", "title": "Generating Causally Compliant Counterfactual Explanations using ASP", "abstract": "  This research is focused on generating achievable counterfactual\nexplanations. Given a negative outcome computed by a machine learning model or\na decision system, the novel CoGS approach generates (i) a counterfactual\nsolution that represents a positive outcome and (ii) a path that will take us\nfrom the negative outcome to the positive one, where each node in the path\nrepresents a change in an attribute (feature) value. CoGS computes paths that\nrespect the causal constraints among features. Thus, the counterfactuals\ncomputed by CoGS are realistic. CoGS utilizes rule-based machine learning\nalgorithms to model causal dependencies between features. The paper discusses\nthe current status of the research and the preliminary results obtained.\n", "date": "2025-02-13", "categories": []}
{"id": "2209.05935", "title": "Variational Causal Inference", "abstract": "  Estimating an individual's potential outcomes under counterfactual treatments\nis a challenging task for traditional causal inference and supervised learning\napproaches when the outcome is high-dimensional (e.g. gene expressions, impulse\nresponses, human faces) and covariates are relatively limited. In this case, to\nconstruct one's outcome under a counterfactual treatment, it is crucial to\nleverage individual information contained in its observed factual outcome on\ntop of the covariates. We propose a deep variational Bayesian framework that\nrigorously integrates two main sources of information for outcome construction\nunder a counterfactual treatment: one source is the individual features\nembedded in the high-dimensional factual outcome; the other source is the\nresponse distribution of similar subjects (subjects with the same covariates)\nthat factually received this treatment of interest.\n", "date": "2022-09-13", "categories": []}
{"id": "2502.07857", "title": "SNAP: Sequential Non-Ancestor Pruning for Targeted Causal Effect\n  Estimation With an Unknown Graph", "abstract": "  Causal discovery can be computationally demanding for large numbers of\nvariables. If we only wish to estimate the causal effects on a small subset of\ntarget variables, we might not need to learn the causal graph for all\nvariables, but only a small subgraph that includes the targets and their\nadjustment sets. In this paper, we focus on identifying causal effects between\ntarget variables in a computationally and statistically efficient way. This\ntask combines causal discovery and effect estimation, aligning the discovery\nobjective with the effects to be estimated. We show that definite non-ancestors\nof the targets are unnecessary to learn causal relations between the targets\nand to identify efficient adjustments sets. We sequentially identify and prune\nthese definite non-ancestors with our Sequential Non-Ancestor Pruning (SNAP)\nframework, which can be used either as a preprocessing step to standard causal\ndiscovery methods, or as a standalone sound and complete causal discovery\nalgorithm. Our results on synthetic and real data show that both approaches\nsubstantially reduce the number of independence tests and the computation time\nwithout compromising the quality of causal effect estimations.\n", "date": "2025-02-11", "categories": []}
{"id": "2312.11582", "title": "Shapley-PC: Constraint-based Causal Structure Learning with a Shapley\n  Inspired Framework", "abstract": "  Causal Structure Learning (CSL), also referred to as causal discovery,\namounts to extracting causal relations among variables in data. CSL enables the\nestimation of causal effects from observational data alone, avoiding the need\nto perform real life experiments. Constraint-based CSL leverages conditional\nindependence tests to perform causal discovery. We propose Shapley-PC, a novel\nmethod to improve constraint-based CSL algorithms by using Shapley values over\nthe possible conditioning sets, to decide which variables are responsible for\nthe observed conditional (in)dependences. We prove soundness, completeness and\nasymptotic consistency of Shapley-PC and run a simulation study showing that\nour proposed algorithm is superior to existing versions of PC.\n", "date": "2023-12-18", "categories": []}
{"id": "2406.14302", "title": "Identifiable Exchangeable Mechanisms for Causal Structure and\n  Representation Learning", "abstract": "  Identifying latent representations or causal structures is important for good\ngeneralization and downstream task performance. However, both fields have been\ndeveloped rather independently. We observe that several methods in both\nrepresentation and causal structure learning rely on the same data-generating\nprocess (DGP), namely, exchangeable but not i.i.d. (independent and identically\ndistributed) data. We provide a unified framework, termed Identifiable\nExchangeable Mechanisms (IEM), for representation and structure learning under\nthe lens of exchangeability. IEM provides new insights that let us relax the\nnecessary conditions for causal structure identification in exchangeable\nnon--i.i.d. data. We also demonstrate the existence of a duality condition in\nidentifiable representation learning, leading to new identifiability results.\nWe hope this work will pave the way for further research in causal\nrepresentation learning.\n", "date": "2024-06-20", "categories": []}
{"id": "2407.16975", "title": "On the Parameter Identifiability of Partially Observed Linear Causal\n  Models", "abstract": "  Linear causal models are important tools for modeling causal dependencies and\nyet in practice, only a subset of the variables can be observed. In this paper,\nwe examine the parameter identifiability of these models by investigating\nwhether the edge coefficients can be recovered given the causal structure and\npartially observed data. Our setting is more general than that of prior\nresearch - we allow all variables, including both observed and latent ones, to\nbe flexibly related, and we consider the coefficients of all edges, whereas\nmost existing works focus only on the edges between observed variables.\nTheoretically, we identify three types of indeterminacy for the parameters in\npartially observed linear causal models. We then provide graphical conditions\nthat are sufficient for all parameters to be identifiable and show that some of\nthem are provably necessary. Methodologically, we propose a novel\nlikelihood-based parameter estimation method that addresses the variance\nindeterminacy of latent variables in a specific way and can asymptotically\nrecover the underlying parameters up to trivial indeterminacy. Empirical\nstudies on both synthetic and real-world datasets validate our identifiability\ntheory and the effectiveness of the proposed method in the finite-sample\nregime. Code: https://github.com/dongxinshuai/scm-identify.\n", "date": "2024-07-24", "categories": []}
{"id": "2409.09822", "title": "Causal Inference with Large Language Model: A Survey", "abstract": "  Causal inference has been a pivotal challenge across diverse domains such as\nmedicine and economics, demanding a complicated integration of human knowledge,\nmathematical reasoning, and data mining capabilities. Recent advancements in\nnatural language processing (NLP), particularly with the advent of large\nlanguage models (LLMs), have introduced promising opportunities for traditional\ncausal inference tasks. This paper reviews recent progress in applying LLMs to\ncausal inference, encompassing various tasks spanning different levels of\ncausation. We summarize the main causal problems and approaches, and present a\ncomparison of their evaluation results in different causal scenarios.\nFurthermore, we discuss key findings and outline directions for future\nresearch, underscoring the potential implications of integrating LLMs in\nadvancing causal inference methodologies.\n", "date": "2024-09-15", "categories": []}
{"id": "2412.03727", "title": "Online Experimental Design With Estimation-Regret Trade-off Under\n  Network Interference", "abstract": "  Network interference has attracted significant attention in the field of\ncausal inference, encapsulating various sociological behaviors where the\ntreatment assigned to one individual within a network may affect the outcomes\nof others, such as their neighbors. A key challenge in this setting is that\nstandard causal inference methods often assume independent treatment effects\namong individuals, which may not hold in networked environments. To estimate\ninterference-aware causal effects, a traditional approach is to inherit the\nindependent settings, where practitioners randomly assign experimental\nparticipants into different groups and compare their outcomes. While effective\nin offline settings, this strategy becomes problematic in sequential\nexperiments, where suboptimal decision persists, leading to substantial regret.\nTo address this issue, we introduce a unified interference-aware framework for\nonline experimental design. Compared to existing studies, we extend the\ndefinition of arm space by utilizing the statistical concept of exposure\nmapping, which allows for a more flexible and context-aware representation of\ntreatment effects in networked settings. Crucially, we establish a\nPareto-optimal trade-off between estimation accuracy and regret under the\nnetwork concerning both time period and arm space, which remains superior to\nbaseline models even without network interference. Furthermore, we propose an\nalgorithmic implementation and discuss its generalization across different\nlearning settings and network topology.\n", "date": "2024-12-04", "categories": []}
{"id": "2501.11622", "title": "Causal Learning for Heterogeneous Subgroups Based on Nonlinear Causal\n  Kernel Clustering", "abstract": "  Due to the challenge posed by multi-source and heterogeneous data collected\nfrom diverse environments, causal relationships among features can exhibit\nvariations influenced by different time spans, regions, or strategies. This\ndiversity makes a single causal model inadequate for accurately representing\ncomplex causal relationships in all observational data, a crucial consideration\nin causal learning. To address this challenge, the nonlinear Causal Kernel\nClustering method is introduced for heterogeneous subgroup causal learning,\nhighlighting variations in causal relationships across diverse subgroups. The\nmain component for clustering heterogeneous subgroups lies in the construction\nof the $u$-centered sample mapping function with the property of unbiased\nestimation, which assesses the differences in potential nonlinear causal\nrelationships in various samples and supported by causal identifiability\ntheory. Experimental results indicate that the method performs well in\nidentifying heterogeneous subgroups and enhancing causal learning, leading to a\nreduction in prediction error.\n", "date": "2025-01-20", "categories": []}
{"id": "2501.13763", "title": "Integrating Causality with Neurochaos Learning: Proposed Approach and\n  Research Agenda", "abstract": "  Deep learning implemented via neural networks, has revolutionized machine\nlearning by providing methods for complex tasks such as object\ndetection/classification and prediction. However, architectures based on deep\nneural networks have started to yield diminishing returns, primarily due to\ntheir statistical nature and inability to capture causal structure in the\ntraining data. Another issue with deep learning is its high energy consumption,\nwhich is not that desirable from a sustainability perspective.\n  Therefore, alternative approaches are being considered to address these\nissues, both of which are inspired by the functioning of the human brain. One\napproach is causal learning, which takes into account causality among the items\nin the dataset on which the neural network is trained. It is expected that this\nwill help minimize the spurious correlations that are prevalent in the learned\nrepresentations of deep neural networks. The other approach is Neurochaos\nLearning, a recent development, which draws its inspiration from the nonlinear\nchaotic firing intrinsic to neurons in biological neural networks\n(brain/central nervous system). Both approaches have shown improved results\nover just deep learning alone.\n  To that end, in this position paper, we investigate how causal and neurochaos\nlearning approaches can be integrated together to produce better results,\nespecially in domains that contain linked data. We propose an approach for this\nintegration to enhance classification, prediction and reinforcement learning.\nWe also propose a set of research questions that need to be investigated in\norder to make this integration a reality.\n", "date": "2025-01-23", "categories": []}
{"id": "2310.01766", "title": "Learning Causal Alignment for Reliable Disease Diagnosis", "abstract": "  Aligning the decision-making process of machine learning algorithms with that\nof experienced radiologists is crucial for reliable diagnosis. While existing\nmethods have attempted to align their diagnosis behaviors to those of\nradiologists reflected in the training data, this alignment is primarily\nassociational rather than causal, resulting in pseudo-correlations that may not\ntransfer well. In this paper, we propose a causality-based alignment framework\ntowards aligning the model's decision process with that of experts.\nSpecifically, we first employ counterfactual generation to identify the causal\nchain of model decisions. To align this causal chain with that of experts, we\npropose a causal alignment loss that enforces the model to focus on causal\nfactors underlying each decision step in the whole causal chain. To optimize\nthis loss that involves the counterfactual generator as an implicit function of\nthe model's parameters, we employ the implicit function theorem equipped with\nthe conjugate gradient method for efficient estimation. We demonstrate the\neffectiveness of our method on two medical diagnosis applications, showcasing\nfaithful alignment to radiologists.\n", "date": "2023-10-03", "categories": []}
{"id": "2312.02573", "title": "UTBoost: Gradient Boosted Decision Trees for Uplift Modeling", "abstract": "  Uplift modeling comprises a collection of machine learning techniques\ndesigned for managers to predict the incremental impact of specific actions on\ncustomer outcomes. However, accurately estimating this incremental impact poses\nsignificant challenges due to the necessity of determining the difference\nbetween two mutually exclusive outcomes for each individual. In our study, we\nintroduce two novel modifications to the established Gradient Boosting Decision\nTrees (GBDT) technique. These modifications sequentially learn the causal\neffect, addressing the counterfactual dilemma. Each modification innovates upon\nthe existing technique in terms of the ensemble learning method and the\nlearning objective, respectively. Experiments with large-scale datasets\nvalidate the effectiveness of our methods, consistently achieving substantial\nimprovements over baseline models.\n", "date": "2023-12-05", "categories": []}
{"id": "2502.04673", "title": "Optimistic Algorithms for Adaptive Estimation of the Average Treatment\n  Effect", "abstract": "  Estimation and inference for the Average Treatment Effect (ATE) is a\ncornerstone of causal inference and often serves as the foundation for\ndeveloping procedures for more complicated settings. Although traditionally\nanalyzed in a batch setting, recent advances in martingale theory have paved\nthe way for adaptive methods that can enhance the power of downstream\ninference. Despite these advances, progress in understanding and developing\nadaptive algorithms remains in its early stages. Existing work either focus on\nasymptotic analyses that overlook exploration-exploitation tradeoffs relevant\nin finite-sample regimes or rely on simpler but suboptimal estimators. In this\nwork, we address these limitations by studying adaptive sampling procedures\nthat take advantage of the asymptotically optimal Augmented Inverse Probability\nWeighting (AIPW) estimator. Our analysis uncovers challenges obscured by\nasymptotic approaches and introduces a novel algorithmic design principle\nreminiscent of optimism in multiarmed bandits. This principled approach enables\nour algorithm to achieve significant theoretical and empirical gains compared\nto prior methods. Our findings mark a step forward in advancing adaptive causal\ninference methods in theory and practice.\n", "date": "2025-02-07", "categories": []}
{"id": "2502.05037", "title": "Leveraging a Simulator for Learning Causal Representations from\n  Post-Treatment Covariates for CATE", "abstract": "  Treatment effect estimation involves assessing the impact of different\ntreatments on individual outcomes. Current methods estimate Conditional Average\nTreatment Effect (CATE) using observational datasets where covariates are\ncollected before treatment assignment and outcomes are observed afterward,\nunder assumptions like positivity and unconfoundedness. In this paper, we\naddress a scenario where both covariates and outcomes are gathered after\ntreatment. We show that post-treatment covariates render CATE unidentifiable,\nand recovering CATE requires learning treatment-independent causal\nrepresentations. Prior work shows that such representations can be learned\nthrough contrastive learning if counterfactual supervision is available in\nobservational data. However, since counterfactuals are rare, other works have\nexplored using simulators that offer synthetic counterfactual supervision. Our\ngoal in this paper is to systematically analyze the role of simulators in\nestimating CATE. We analyze the CATE error of several baselines and highlight\ntheir limitations. We then establish a generalization bound that characterizes\nthe CATE error from jointly training on real and simulated distributions, as a\nfunction of the real-simulator mismatch. Finally, we introduce SimPONet, a\nnovel method whose loss function is inspired from our generalization bound. We\nfurther show how SimPONet adjusts the simulator's influence on the learning\nobjective based on the simulator's relevance to the CATE task. We experiment\nwith various DGPs, by systematically varying the real-simulator distribution\ngap to evaluate SimPONet's efficacy against state-of-the-art CATE baselines.\n", "date": "2025-02-07", "categories": []}
{"id": "2502.05150", "title": "CodeSCM: Causal Analysis for Multi-Modal Code Generation", "abstract": "  In this paper, we propose CodeSCM, a Structural Causal Model (SCM) for\nanalyzing multi-modal code generation using large language models (LLMs). By\napplying interventions to CodeSCM, we measure the causal effects of different\nprompt modalities, such as natural language, code, and input-output examples,\non the model. CodeSCM introduces latent mediator variables to separate the code\nand natural language semantics of a multi-modal code generation prompt. Using\nthe principles of Causal Mediation Analysis on these mediators we quantify\ndirect effects representing the model's spurious leanings. We find that, in\naddition to natural language instructions, input-output examples significantly\ninfluence code generation.\n", "date": "2025-02-07", "categories": []}
{"id": "2405.07373", "title": "From Probability to Counterfactuals: the Increasing Complexity of\n  Satisfiability in Pearl's Causal Hierarchy", "abstract": "  The framework of Pearl's Causal Hierarchy (PCH) formalizes three types of\nreasoning: probabilistic (i.e. purely observational), interventional, and\ncounterfactual, that reflect the progressive sophistication of human thought\nregarding causation. We investigate the computational complexity aspects of\nreasoning in this framework focusing mainly on satisfiability problems\nexpressed in probabilistic and causal languages across the PCH. That is, given\na system of formulas in the standard probabilistic and causal languages, does\nthere exist a model satisfying the formulas?\n  Our main contribution is to prove the exact computational complexities\nshowing that languages allowing addition and marginalization (via the summation\noperator) yield NP^PP, PSPACE-, and NEXP-complete satisfiability problems,\ndepending on the level of the PCH. These are the first results to demonstrate a\nstrictly increasing complexity across the PCH: from probabilistic to causal and\ncounterfactual reasoning. On the other hand, in the case of full languages,\ni.e. allowing addition, marginalization, and multiplication, we show that the\nsatisfiability for the counterfactual level remains the same as for the\nprobabilistic and causal levels, solving an open problem in the field.\n", "date": "2024-05-12", "categories": []}
{"id": "2407.05287", "title": "Model-agnostic meta-learners for estimating heterogeneous treatment\n  effects over time", "abstract": "  Estimating heterogeneous treatment effects (HTEs) over time is crucial in\nmany disciplines such as personalized medicine. For example, electronic health\nrecords are commonly collected over several time periods and then used to\npersonalize treatment decisions. Existing works for this task have mostly\nfocused on model-based learners (i.e., learners that adapt specific\nmachine-learning models). In contrast, model-agnostic learners -- so-called\nmeta-learners -- are largely unexplored. In our paper, we propose several\nmeta-learners that are model-agnostic and thus can be used in combination with\narbitrary machine learning models (e.g., transformers) to estimate HTEs over\ntime. Here, our focus is on learners that can be obtained via weighted\npseudo-outcome regressions, which allows for efficient estimation by targeting\nthe treatment effect directly. We then provide a comprehensive theoretical\nanalysis that characterizes the different learners and that allows us to offer\ninsights into when specific learners are preferable. Finally, we confirm our\ntheoretical insights through numerical experiments. In sum, while meta-learners\nare already state-of-the-art for the static setting, we are the first to\npropose a comprehensive set of meta-learners for estimating HTEs in the\ntime-varying setting.\n", "date": "2024-07-07", "categories": []}
{"id": "2502.03802", "title": "MXMap: A Multivariate Cross Mapping Framework for Causal Discovery in\n  Dynamical Systems", "abstract": "  Convergent Cross Mapping (CCM) is a powerful method for detecting causality\nin coupled nonlinear dynamical systems, providing a model-free approach to\ncapture dynamic causal interactions. Partial Cross Mapping (PCM) was introduced\nas an extension of CCM to address indirect causality in three-variable systems\nby comparing cross-mapping quality between direct cause-effect mapping and\nindirect mapping through an intermediate conditioning variable. However, PCM\nremains limited to univariate delay embeddings in its cross-mapping processes.\nIn this work, we extend PCM to the multivariate setting, introducing multiPCM,\nwhich leverages multivariate embeddings to more effectively distinguish\nindirect causal relationships. We further propose a multivariate cross-mapping\nframework (MXMap) for causal discovery in dynamical systems. This two-phase\nframework combines (1) pairwise CCM tests to establish an initial causal graph\nand (2) multiPCM to refine the graph by pruning indirect causal connections.\nThrough experiments on simulated data and the ERA5 Reanalysis weather dataset,\nwe demonstrate the effectiveness of MXMap. Additionally, MXMap is compared\nagainst several baseline methods, showing advantages in accuracy and causal\ngraph refinement.\n", "date": "2025-02-06", "categories": []}
{"id": "2402.12062", "title": "Causal Equal Protection as Algorithmic Fairness", "abstract": "  By combining the philosophical literature on statistical evidence and the\ninterdisciplinary literature on algorithmic fairness, we revisit recent\nobjections against classification parity in light of causal analyses of\nalgorithmic fairness and the distinction between predictive and diagnostic\nevidence. We focus on trial proceedings as a black-box classification algorithm\nin which defendants are sorted into two groups by convicting or acquitting\nthem. We defend a novel principle, causal equal protection, that combines\nclassification parity with the causal approach. In the do-calculus, causal\nequal protection requires that individuals should not be subject to uneven\nrisks of classification error because of their protected or socially salient\ncharacteristics. The explicit use of protected characteristics, however, may be\nrequired if it equalizes these risks.\n", "date": "2024-02-19", "categories": []}
{"id": "2501.04871", "title": "RieszBoost: Gradient Boosting for Riesz Regression", "abstract": "  Answering causal questions often involves estimating linear functionals of\nconditional expectations, such as the average treatment effect or the effect of\na longitudinal modified treatment policy. By the Riesz representation theorem,\nthese functionals can be expressed as the expected product of the conditional\nexpectation of the outcome and the Riesz representer, a key component in doubly\nrobust estimation methods. Traditionally, the Riesz representer is estimated\nindirectly by deriving its explicit analytical form, estimating its components,\nand substituting these estimates into the known form (e.g., the inverse\npropensity score). However, deriving or estimating the analytical form can be\nchallenging, and substitution methods are often sensitive to practical\npositivity violations, leading to higher variance and wider confidence\nintervals. In this paper, we propose a novel gradient boosting algorithm to\ndirectly estimate the Riesz representer without requiring its explicit\nanalytical form. This method is particularly suited for tabular data, offering\na flexible, nonparametric, and computationally efficient alternative to\nexisting methods for Riesz regression. Through simulation studies, we\ndemonstrate that our algorithm performs on par with or better than indirect\nestimation techniques across a range of functionals, providing a user-friendly\nand robust solution for estimating causal quantities.\n", "date": "2025-01-08", "categories": []}
{"id": "2501.08958", "title": "Kolmogorov-Arnold Networks for Time Series Granger Causality Inference", "abstract": "  We propose the Granger causality inference Kolmogorov-Arnold Networks\n(KANGCI), a novel architecture that extends the recently proposed\nKolmogorov-Arnold Networks (KAN) to the domain of causal inference. By\nextracting base weights from KAN layers and incorporating the sparsity-inducing\npenalty and ridge regularization, KANGCI effectively infers the Granger\ncausality from time series. Additionally, we propose an algorithm based on\ntime-reversed Granger causality that automatically selects causal relationships\nwith better inference performance from the original or time-reversed time\nseries or integrates the results to mitigate spurious connectivities.\nComprehensive experiments conducted on Lorenz-96, Gene regulatory networks,\nfMRI BOLD signals, VAR, and real-world EEG datasets demonstrate that the\nproposed model achieves competitive performance to state-of-the-art methods in\ninferring Granger causality from nonlinear, high-dimensional, and\nlimited-sample time series.\n", "date": "2025-01-15", "categories": []}
{"id": "2502.02701", "title": "Practically Effective Adjustment Variable Selection in Causal Inference", "abstract": "  In the estimation of causal effects, one common method for removing the\ninfluence of confounders is to adjust the variables that satisfy the back-door\ncriterion. However, it is not always possible to uniquely determine sets of\nsuch variables. Moreover, real-world data is almost always limited, which means\nit may be insufficient for statistical estimation. Therefore, we propose\ncriteria for selecting variables from a list of candidate adjustment variables\nalong with an algorithm to prevent accuracy degradation in causal effect\nestimation. We initially focus on directed acyclic graphs (DAGs) and then\noutlines specific steps for applying this method to completed partially\ndirected acyclic graphs (CPDAGs). We also present and prove a theorem on causal\neffect computation possibility in CPDAGs. Finally, we demonstrate the practical\nutility of our method using both existing and artificial data.\n", "date": "2025-02-04", "categories": []}
{"id": "2502.02109", "title": "Causally-informed Deep Learning towards Explainable and Generalizable\n  Outcomes Prediction in Critical Care", "abstract": "  Recent advances in deep learning (DL) have prompted the development of\nhigh-performing early warning score (EWS) systems, predicting clinical\ndeteriorations such as acute kidney injury, acute myocardial infarction, or\ncirculatory failure. DL models have proven to be powerful tools for various\ntasks but come with the cost of lacking interpretability and limited\ngeneralizability, hindering their clinical applications. To develop a practical\nEWS system applicable to various outcomes, we propose causally-informed\nexplainable early prediction model, which leverages causal discovery to\nidentify the underlying causal relationships of prediction and thus owns two\nunique advantages: demonstrating the explicit interpretation of the prediction\nwhile exhibiting decent performance when applied to unfamiliar environments.\nBenefiting from these features, our approach achieves superior accuracy for 6\ndifferent critical deteriorations and achieves better generalizability across\ndifferent patient groups, compared to various baseline algorithms. Besides, we\nprovide explicit causal pathways to serve as references for assistant clinical\ndiagnosis and potential interventions. The proposed approach enhances the\npractical application of deep learning in various medical scenarios.\n", "date": "2025-02-04", "categories": []}
{"id": "2502.02327", "title": "Policy-Guided Causal State Representation for Offline Reinforcement\n  Learning Recommendation", "abstract": "  In offline reinforcement learning-based recommender systems (RLRS), learning\neffective state representations is crucial for capturing user preferences that\ndirectly impact long-term rewards. However, raw state representations often\ncontain high-dimensional, noisy information and components that are not\ncausally relevant to the reward. Additionally, missing transitions in offline\ndata make it challenging to accurately identify features that are most relevant\nto user satisfaction. To address these challenges, we propose Policy-Guided\nCausal Representation (PGCR), a novel two-stage framework for causal feature\nselection and state representation learning in offline RLRS. In the first\nstage, we learn a causal feature selection policy that generates modified\nstates by isolating and retaining only the causally relevant components (CRCs)\nwhile altering irrelevant components. This policy is guided by a reward\nfunction based on the Wasserstein distance, which measures the causal effect of\nstate components on the reward and encourages the preservation of CRCs that\ndirectly influence user interests. In the second stage, we train an encoder to\nlearn compact state representations by minimizing the mean squared error (MSE)\nloss between the latent representations of the original and modified states,\nensuring that the representations focus on CRCs. We provide a theoretical\nanalysis proving the identifiability of causal effects from interventions,\nvalidating the ability of PGCR to isolate critical state components for\ndecision-making. Extensive experiments demonstrate that PGCR significantly\nimproves recommendation performance, confirming its effectiveness for offline\nRL-based recommender systems.\n", "date": "2025-02-04", "categories": []}
{"id": "2405.05025", "title": "Learning Structural Causal Models through Deep Generative Models:\n  Methods, Guarantees, and Challenges", "abstract": "  This paper provides a comprehensive review of deep structural causal models\n(DSCMs), particularly focusing on their ability to answer counterfactual\nqueries using observational data within known causal structures. It delves into\nthe characteristics of DSCMs by analyzing the hypotheses, guarantees, and\napplications inherent to the underlying deep learning components and structural\ncausal models, fostering a finer understanding of their capabilities and\nlimitations in addressing different counterfactual queries. Furthermore, it\nhighlights the challenges and open questions in the field of deep structural\ncausal modeling. It sets the stages for researchers to identify future work\ndirections and for practitioners to get an overview in order to find out the\nmost appropriate methods for their needs.\n", "date": "2024-05-08", "categories": []}
{"id": "2405.13888", "title": "Marrying Causal Representation Learning with Dynamical Systems for\n  Science", "abstract": "  Causal representation learning promises to extend causal models to hidden\ncausal variables from raw entangled measurements. However, most progress has\nfocused on proving identifiability results in different settings, and we are\nnot aware of any successful real-world application. At the same time, the field\nof dynamical systems benefited from deep learning and scaled to countless\napplications but does not allow parameter identification. In this paper, we\ndraw a clear connection between the two and their key assumptions, allowing us\nto apply identifiable methods developed in causal representation learning to\ndynamical systems. At the same time, we can leverage scalable differentiable\nsolvers developed for differential equations to build models that are both\nidentifiable and practical. Overall, we learn explicitly controllable models\nthat isolate the trajectory-specific parameters for further downstream tasks\nsuch as out-of-distribution classification or treatment effect estimation. We\nexperiment with a wind simulator with partially known factors of variation. We\nalso apply the resulting model to real-world climate data and successfully\nanswer downstream causal questions in line with existing literature on climate\nchange.\n", "date": "2024-05-22", "categories": []}
{"id": "2411.00945", "title": "Higher-Order Causal Message Passing for Experimentation with Complex\n  Interference", "abstract": "  Accurate estimation of treatment effects is essential for decision-making\nacross various scientific fields. This task, however, becomes challenging in\nareas like social sciences and online marketplaces, where treating one\nexperimental unit can influence outcomes for others through direct or indirect\ninteractions. Such interference can lead to biased treatment effect estimates,\nparticularly when the structure of these interactions is unknown. We address\nthis challenge by introducing a new class of estimators based on causal\nmessage-passing, specifically designed for settings with pervasive, unknown\ninterference. Our estimator draws on information from the sample mean and\nvariance of unit outcomes and treatments over time, enabling efficient use of\nobserved data to estimate the evolution of the system state. Concretely, we\nconstruct non-linear features from the moments of unit outcomes and treatments\nand then learn a function that maps these features to future mean and variance\nof unit outcomes. This allows for the estimation of the treatment effect over\ntime. Extensive simulations across multiple domains, using synthetic and real\nnetwork data, demonstrate the efficacy of our approach in estimating total\ntreatment effect dynamics, even in cases where interference exhibits\nnon-monotonic behavior in the probability of treatment.\n", "date": "2024-11-01", "categories": []}
{"id": "2411.01045", "title": "Fighting Spurious Correlations in Text Classification via a Causal\n  Learning Perspective", "abstract": "  In text classification tasks, models often rely on spurious correlations for\npredictions, incorrectly associating irrelevant features with the target\nlabels. This issue limits the robustness and generalization of models,\nespecially when faced with out-of-distribution data where such spurious\ncorrelations no longer hold. To address this challenge, we propose the Causally\nCalibrated Robust Classifier (CCR), which aims to reduce models' reliance on\nspurious correlations and improve model robustness. Our approach integrates a\ncausal feature selection method based on counterfactual reasoning, along with\nan unbiased inverse propensity weighting (IPW) loss function. By focusing on\nselecting causal features, we ensure that the model relies less on spurious\nfeatures during prediction. We theoretically justify our approach and\nempirically show that CCR achieves state-of-the-art performance among methods\nwithout group labels, and in some cases, it can compete with the models that\nutilize group labels.\n", "date": "2024-11-01", "categories": []}
{"id": "2501.19335", "title": "What is causal about causal models and representations?", "abstract": "  Causal Bayesian networks are 'causal' models since they make predictions\nabout interventional distributions. To connect such causal model predictions to\nreal-world outcomes, we must determine which actions in the world correspond to\nwhich interventions in the model. For example, to interpret an action as an\nintervention on a treatment variable, the action will presumably have to a)\nchange the distribution of treatment in a way that corresponds to the\nintervention, and b) not change other aspects, such as how the outcome depends\non the treatment; while the marginal distributions of some variables may change\nas an effect. We introduce a formal framework to make such requirements for\ndifferent interpretations of actions as interventions precise. We prove that\nthe seemingly natural interpretation of actions as interventions is circular:\nUnder this interpretation, every causal Bayesian network that correctly models\nthe observational distribution is trivially also interventionally valid, and no\naction yields empirical data that could possibly falsify such a model. We prove\nan impossibility result: No interpretation exists that is non-circular and\nsimultaneously satisfies a set of natural desiderata. Instead, we examine\nnon-circular interpretations that may violate some desiderata and show how this\nmay in turn enable the falsification of causal models. By rigorously examining\nhow a causal Bayesian network could be a 'causal' model of the world instead of\nmerely a mathematical object, our formal framework contributes to the\nconceptual foundations of causal representation learning, causal discovery, and\ncausal abstraction, while also highlighting some limitations of existing\napproaches.\n", "date": "2025-01-31", "categories": []}
{"id": "2502.00501", "title": "Optimizing Feature Selection in Causal Inference: A Three-Stage\n  Computational Framework for Unbiased Estimation", "abstract": "  Feature selection is an important but challenging task in causal inference\nfor obtaining unbiased estimates of causal quantities. Properly selected\nfeatures in causal inference not only significantly reduce the time required to\nimplement a matching algorithm but, more importantly, can also reduce the bias\nand variance when estimating causal quantities. When feature selection\ntechniques are applied in causal inference, the crucial criterion is to select\nvariables that, when used for matching, can achieve an unbiased and robust\nestimation of causal quantities. Recent research suggests that balancing only\non treatment-associated variables introduces bias while balancing on spurious\nvariables increases variance. To address this issue, we propose an enhanced\nthree-stage framework that shows a significant improvement in selecting the\ndesired subset of variables compared to the existing state-of-the-art feature\nselection framework for causal inference, resulting in lower bias and variance\nin estimating the causal quantity. We evaluated our proposed framework using a\nstate-of-the-art synthetic data across various settings and observed superior\nperformance within a feasible computation time, ensuring scalability for\nlarge-scale datasets. Finally, to demonstrate the applicability of our proposed\nmethodology using large-scale real-world data, we evaluated an important US\nhealthcare policy related to the opioid epidemic crisis: whether opioid use\ndisorder has a causal relationship with suicidal behavior.\n", "date": "2025-02-01", "categories": []}
{"id": "2501.18859", "title": "A Deep Spatio-Temporal Architecture for Dynamic Effective Connectivity\n  Network Analysis Based on Dynamic Causal Discovery", "abstract": "  Dynamic effective connectivity networks (dECNs) reveal the changing directed\nbrain activity and the dynamic causal influences among brain regions, which\nfacilitate the identification of individual differences and enhance the\nunderstanding of human brain. Although the existing causal discovery methods\nhave shown promising results in effective connectivity network analysis, they\noften overlook the dynamics of causality, in addition to the incorporation of\nspatio-temporal information in brain activity data. To address these issues, we\npropose a deep spatio-temporal fusion architecture, which employs a dynamic\ncausal deep encoder to incorporate spatio-temporal information into dynamic\ncausality modeling, and a dynamic causal deep decoder to verify the discovered\ncausality. The effectiveness of the proposed method is first illustrated with\nsimulated data. Then, experimental results from Philadelphia Neurodevelopmental\nCohort (PNC) demonstrate the superiority of the proposed method in inferring\ndECNs, which reveal the dynamic evolution of directed flow between brain\nregions. The analysis shows the difference of dECNs between young adults and\nchildren. Specifically, the directed brain functional networks transit from\nfluctuating undifferentiated systems to more stable specialized networks as one\ngrows. This observation provides further evidence on the modularization and\nadaptation of brain networks during development, leading to higher cognitive\nabilities observed in young adults.\n", "date": "2025-01-31", "categories": []}
{"id": "2408.13556", "title": "What if? Causal Machine Learning in Supply Chain Risk Management", "abstract": "  The penultimate goal for developing machine learning models in supply chain\nmanagement is to make optimal interventions. However, most machine learning\nmodels identify correlations in data rather than inferring causation, making it\ndifficult to systematically plan for better outcomes. In this article, we\npropose and evaluate the use of causal machine learning for developing supply\nchain risk intervention models, and demonstrate its use with a case study in\nsupply chain risk management in the maritime engineering sector. Our findings\nhighlight that causal machine learning enhances decision-making processes by\nidentifying changes that can be achieved under different supply chain\ninterventions, allowing \"what-if\" scenario planning. We therefore propose\ndifferent machine learning developmental pathways for for predicting risk, and\nplanning for interventions to minimise risk and outline key steps for supply\nchain researchers to explore causal machine learning.\n", "date": "2024-08-24", "categories": []}
{"id": "2501.18337", "title": "Unfaithful Probability Distributions in Binary Triple of Causality\n  Directed Acyclic Graph", "abstract": "  Faithfulness is the foundation of probability distribution and graph in\ncausal discovery and causal inference. In this paper, several unfaithful\nprobability distribution examples are constructed in three--vertices binary\ncausality directed acyclic graph (DAG) structure, which are not faithful to\ncausal DAGs described in J.M.,Robins,et al. Uniform consistency in causal\ninference. Biometrika (2003),90(3): 491--515. And the general unfaithful\nprobability distribution with multiple independence and conditional\nindependence in binary triple causal DAG is given.\n", "date": "2025-01-30", "categories": []}
{"id": "2403.09580", "title": "Algorithmic syntactic causal identification", "abstract": "  Causal identification in causal Bayes nets (CBNs) is an important tool in\ncausal inference allowing the derivation of interventional distributions from\nobservational distributions where this is possible in principle. However, most\nexisting formulations of causal identification using techniques such as\nd-separation and do-calculus are expressed within the mathematical language of\nclassical probability theory on CBNs. However, there are many causal settings\nwhere probability theory and hence current causal identification techniques are\ninapplicable such as relational databases, dataflow programs such as hardware\ndescription languages, distributed systems and most modern machine learning\nalgorithms. We show that this restriction can be lifted by replacing the use of\nclassical probability theory with the alternative axiomatic foundation of\nsymmetric monoidal categories. In this alternative axiomatization, we show how\nan unambiguous and clean distinction can be drawn between the general syntax of\ncausal models and any specific semantic implementation of that causal model.\nThis allows a purely syntactic algorithmic description of general causal\nidentification by a translation of recent formulations of the general ID\nalgorithm through fixing. Our description is given entirely in terms of the\nnon-parametric ADMG structure specifying a causal model and the algebraic\nsignature of the corresponding monoidal category, to which a sequence of\nmanipulations is then applied so as to arrive at a modified monoidal category\nin which the desired, purely syntactic interventional causal model, is\nobtained. We use this idea to derive purely syntactic analogues of classical\nback-door and front-door causal adjustment, and illustrate an application to a\nmore complex causal model.\n", "date": "2024-03-14", "categories": []}
{"id": "2501.17354", "title": "Fundamental Computational Limits in Pursuing Invariant Causal Prediction\n  and Invariance-Guided Regularization", "abstract": "  Pursuing invariant prediction from heterogeneous environments opens the door\nto learning causality in a purely data-driven way and has several applications\nin causal discovery and robust transfer learning. However, existing methods\nsuch as ICP [Peters et al., 2016] and EILLS [Fan et al., 2024] that can attain\nsample-efficient estimation are based on exponential time algorithms. In this\npaper, we show that such a problem is intrinsically hard in computation: the\ndecision problem, testing whether a non-trivial prediction-invariant solution\nexists across two environments, is NP-hard even for the linear causal\nrelationship. In the world where P$\\neq$NP, our results imply that the\nestimation error rate can be arbitrarily slow using any computationally\nefficient algorithm. This suggests that pursuing causality is fundamentally\nharder than detecting associations when no prior assumption is pre-offered.\n  Given there is almost no hope of computational improvement under the worst\ncase, this paper proposes a method capable of attaining both computationally\nand statistically efficient estimation under additional conditions.\nFurthermore, our estimator is a distributionally robust estimator with an\nellipse-shaped uncertain set where more uncertainty is placed on spurious\ndirections than invariant directions, resulting in a smooth interpolation\nbetween the most predictive solution and the causal solution by varying the\ninvariance hyper-parameter. Non-asymptotic results and empirical applications\nsupport the claim.\n", "date": "2025-01-29", "categories": []}
{"id": "2501.16399", "title": "Detecting clinician implicit biases in diagnoses using proximal causal\n  inference", "abstract": "  Clinical decisions to treat and diagnose patients are affected by implicit\nbiases formed by racism, ableism, sexism, and other stereotypes. These biases\nreflect broader systemic discrimination in healthcare and risk marginalizing\nalready disadvantaged groups. Existing methods for measuring implicit biases\nrequire controlled randomized testing and only capture individual attitudes\nrather than outcomes. However, the \"big-data\" revolution has led to the\navailability of large observational medical datasets, like EHRs and biobanks,\nthat provide the opportunity to investigate discrepancies in patient health\noutcomes. In this work, we propose a causal inference approach to detect the\neffect of clinician implicit biases on patient outcomes in large-scale medical\ndata. Specifically, our method uses proximal mediation to disentangle\npathway-specific effects of a patient's sociodemographic attribute on a\nclinician's diagnosis decision. We test our method on real-world data from the\nUK Biobank. Our work can serve as a tool that initiates conversation and brings\nawareness to unequal health outcomes caused by implicit biases.\n", "date": "2025-01-27", "categories": []}
{"id": "2208.10671", "title": "Cardinality-Regularized Hawkes-Granger Model", "abstract": "  We propose a new sparse Granger-causal learning framework for temporal event\ndata. We focus on a specific class of point processes called the Hawkes\nprocess. We begin by pointing out that most of the existing sparse causal\nlearning algorithms for the Hawkes process suffer from a singularity in maximum\nlikelihood estimation. As a result, their sparse solutions can appear only as\nnumerical artifacts. In this paper, we propose a mathematically well-defined\nsparse causal learning framework based on a cardinality-regularized Hawkes\nprocess, which remedies the pathological issues of existing approaches. We\nleverage the proposed algorithm for the task of instance-wise causal event\nanalysis, where sparsity plays a critical role. We validate the proposed\nframework with two real use-cases, one from the power grid and the other from\nthe cloud data center management domain.\n", "date": "2022-08-23", "categories": []}
{"id": "2411.07978", "title": "A Note on Doubly Robust Estimator in Regression Discontinuity Designs", "abstract": "  This note introduces a doubly robust (DR) estimator for regression\ndiscontinuity (RD) designs. RD designs provide a quasi-experimental framework\nfor estimating treatment effects, where treatment assignment depends on whether\na running variable surpasses a predefined cutoff. A common approach in RD\nestimation is the use of nonparametric regression methods, such as local linear\nregression. However, the validity of these methods still relies on the\nconsistency of the nonparametric estimators. In this study, we propose the\nDR-RD estimator, which combines two distinct estimators for the conditional\nexpected outcomes. The primary advantage of the DR-RD estimator lies in its\nability to ensure the consistency of the treatment effect estimation as long as\nat least one of the two estimators is consistent. Consequently, our DR-RD\nestimator enhances robustness of treatment effect estimators in RD designs.\n", "date": "2024-11-12", "categories": []}
{"id": "2501.08888", "title": "A Partial Initialization Strategy to Mitigate the Overfitting Problem in\n  CATE Estimation with Hidden Confounding", "abstract": "  Estimating the conditional average treatment effect (CATE) from observational\ndata plays a crucial role in areas such as e-commerce, healthcare, and\neconomics. Existing studies mainly rely on the strong ignorability assumption\nthat there are no hidden confounders, whose existence cannot be tested from\nobservational data and can invalidate any causal conclusion. In contrast, data\ncollected from randomized controlled trials (RCT) do not suffer from\nconfounding but are usually limited by a small sample size. To avoid\noverfitting caused by the small-scale RCT data, we propose a novel two-stage\npretraining-finetuning (TSPF) framework with a partial parameter initialization\nstrategy to estimate the CATE in the presence of hidden confounding. In the\nfirst stage, a foundational representation of covariates is trained to estimate\ncounterfactual outcomes through large-scale observational data. In the second\nstage, we propose to train an augmented representation of the covariates, which\nis concatenated with the foundational representation obtained in the first\nstage to adjust for the hidden confounding. Rather than training a separate\nnetwork from scratch, part of the prediction heads are initialized from the\nfirst stage. The superiority of our approach is validated on two datasets with\nextensive experiments.\n", "date": "2025-01-15", "categories": []}
{"id": "2501.14837", "title": "A Semiparametric Bayesian Method for Instrumental Variable Analysis with\n  Partly Interval-Censored Time-to-Event Outcome", "abstract": "  This paper develops a semiparametric Bayesian instrumental variable analysis\nmethod for estimating the causal effect of an endogenous variable when dealing\nwith unobserved confounders and measurement errors with partly\ninterval-censored time-to-event data, where event times are observed exactly\nfor some subjects but left-censored, right-censored, or interval-censored for\nothers. Our method is based on a two-stage Dirichlet process mixture\ninstrumental variable (DPMIV) model which simultaneously models the first-stage\nrandom error term for the exposure variable and the second-stage random error\nterm for the time-to-event outcome using a bivariate Gaussian mixture of the\nDirichlet process (DPM) model. The DPM model can be broadly understood as a\nmixture model with an unspecified number of Gaussian components, which relaxes\nthe normal error assumptions and allows the number of mixture components to be\ndetermined by the data. We develop an MCMC algorithm for the DPMIV model\ntailored for partly interval-censored data and conduct extensive simulations to\nassess the performance of our DPMIV method in comparison with some competing\nmethods. Our simulations revealed that our proposed method is robust under\ndifferent error distributions and can have superior performance over its\nparametric counterpart under various scenarios. We further demonstrate the\neffectiveness of our approach on an UK Biobank data to investigate the causal\neffect of systolic blood pressure on time-to-development of cardiovascular\ndisease from the onset of diabetes mellitus.\n", "date": "2025-01-23", "categories": []}
{"id": "2501.14933", "title": "Conformal Inference of Individual Treatment Effects Using Conditional\n  Density Estimates", "abstract": "  In an era where diverse and complex data are increasingly accessible, the\nprecise prediction of individual treatment effects (ITE) becomes crucial across\nfields such as healthcare, economics, and public policy. Current\nstate-of-the-art approaches, while providing valid prediction intervals through\nConformal Quantile Regression (CQR) and related techniques, often yield overly\nconservative prediction intervals. In this work, we introduce a conformal\ninference approach to ITE using the conditional density of the outcome given\nthe covariates. We leverage the reference distribution technique to efficiently\nestimate the conditional densities as the score functions under a two-stage\nconformal ITE framework. We show that our prediction intervals are not only\nmarginally valid but are narrower than existing methods. Experimental results\nfurther validate the usefulness of our method.\n", "date": "2025-01-24", "categories": []}
{"id": "2501.14997", "title": "Causal Discovery via Bayesian Optimization", "abstract": "  Existing score-based methods for directed acyclic graph (DAG) learning from\nobservational data struggle to recover the causal graph accurately and\nsample-efficiently. To overcome this, in this study, we propose DrBO (DAG\nrecovery via Bayesian Optimization)-a novel DAG learning framework leveraging\nBayesian optimization (BO) to find high-scoring DAGs. We show that, by\nsophisticatedly choosing the promising DAGs to explore, we can find\nhigher-scoring ones much more efficiently. To address the scalability issues of\nconventional BO in DAG learning, we replace Gaussian Processes commonly\nemployed in BO with dropout neural networks, trained in a continual manner,\nwhich allows for (i) flexibly modeling the DAG scores without overfitting, (ii)\nincorporation of uncertainty into the estimated scores, and (iii) scaling with\nthe number of evaluations. As a result, DrBO is computationally efficient and\ncan find the accurate DAG in fewer trials and less time than existing\nstate-of-the-art methods. This is demonstrated through an extensive set of\nempirical evaluations on many challenging settings with both synthetic and real\ndata. Our implementation is available at https://github.com/baosws/DrBO.\n", "date": "2025-01-25", "categories": []}
{"id": "2501.15973", "title": "Integrating Probabilistic Trees and Causal Networks for Clinical and\n  Epidemiological Data", "abstract": "  Healthcare decision-making requires not only accurate predictions but also\ninsights into how factors influence patient outcomes. While traditional Machine\nLearning (ML) models excel at predicting outcomes, such as identifying high\nrisk patients, they are limited in addressing what-if questions about\ninterventions. This study introduces the Probabilistic Causal Fusion (PCF)\nframework, which integrates Causal Bayesian Networks (CBNs) and Probability\nTrees (PTrees) to extend beyond predictions. PCF leverages causal relationships\nfrom CBNs to structure PTrees, enabling both the quantification of factor\nimpacts and simulation of hypothetical interventions. PCF was validated on\nthree real-world healthcare datasets i.e. MIMIC-IV, Framingham Heart Study, and\nDiabetes, chosen for their clinically diverse variables. It demonstrated\npredictive performance comparable to traditional ML models while providing\nadditional causal reasoning capabilities. To enhance interpretability, PCF\nincorporates sensitivity analysis and SHapley Additive exPlanations (SHAP).\nSensitivity analysis quantifies the influence of causal parameters on outcomes\nsuch as Length of Stay (LOS), Coronary Heart Disease (CHD), and Diabetes, while\nSHAP highlights the importance of individual features in predictive modeling.\nBy combining causal reasoning with predictive modeling, PCF bridges the gap\nbetween clinical intuition and data-driven insights. Its ability to uncover\nrelationships between modifiable factors and simulate hypothetical scenarios\nprovides clinicians with a clearer understanding of causal pathways. This\napproach supports more informed, evidence-based decision-making, offering a\nrobust framework for addressing complex questions in diverse healthcare\nsettings.\n", "date": "2025-01-27", "categories": []}
{"id": "2501.14006", "title": "Asymmetrical Latent Representation for Individual Treatment Effect\n  Modeling", "abstract": "  Conditional Average Treatment Effect (CATE) estimation, at the heart of\ncounterfactual reasoning, is a crucial challenge for causal modeling both\ntheoretically and applicatively, in domains such as healthcare, sociology, or\nadvertising. Borrowing domain adaptation principles, a popular design maps the\nsample representation to a latent space that balances control and treated\npopulations while enabling the prediction of the potential outcomes. This paper\npresents a new CATE estimation approach based on the asymmetrical search for\ntwo latent spaces called Asymmetrical Latent Representation for Individual\nTreatment Effect (ALRITE), where the two latent spaces are respectively\nintended to optimize the counterfactual prediction accuracy on the control and\nthe treated samples. Under moderate assumptions, ALRITE admits an upper bound\non the precision of the estimation of heterogeneous effects (PEHE), and the\napproach is empirically successfully validated compared to the state-of-the-art\n", "date": "2025-01-23", "categories": []}
{"id": "2410.00150", "title": "What If We Had Used a Different App? Reliable Counterfactual KPI\n  Analysis in Wireless Systems", "abstract": "  In modern wireless network architectures, such as Open Radio Access Network\n(O-RAN), the operation of the radio access network (RAN) is managed by\napplications, or apps for short, deployed at intelligent controllers. These\napps are selected from a given catalog based on current contextual information.\nFor instance, a scheduling app may be selected on the basis of current traffic\nand network conditions. Once an app is chosen and run, it is no longer possible\nto directly test the key performance indicators (KPIs) that would have been\nobtained with another app. In other words, we can never simultaneously observe\nboth the actual KPI, obtained by the selected app, and the counterfactual KPI,\nwhich would have been attained with another app, for the same network\ncondition, making individual-level counterfactual KPIs analysis particularly\nchallenging. This what-if analysis, however, would be valuable to monitor and\noptimize the network operation, e.g., to identify suboptimal app selection\nstrategies. This paper addresses the problem of estimating the values of KPIs\nthat would have been obtained if a different app had been implemented by the\nRAN. To this end, we propose a conformal-prediction-based counterfactual\nanalysis method for wireless systems that provides reliable error bars for the\nestimated KPIs, despite the inherent covariate shift between logged and test\ndata. Experimental results for medium access control-layer apps and for\nphysical-layer apps demonstrate the merits of the proposed method.\n", "date": "2024-09-30", "categories": []}
{"id": "2501.10423", "title": "Do we actually understand the impact of renewables on electricity\n  prices? A causal inference approach", "abstract": "  The energy transition is profoundly reshaping electricity market dynamics. It\nmakes it essential to understand how renewable energy generation actually\nimpacts electricity prices, among all other market drivers. These insights are\ncritical to design policies and market interventions that ensure affordable,\nreliable, and sustainable energy systems. However, identifying causal effects\nfrom observational data is a major challenge, requiring innovative causal\ninference approaches that go beyond conventional regression analysis only. We\nbuild upon the state of the art by developing and applying a local partially\nlinear double machine learning approach. Its application yields the first\nrobust causal evidence on the distinct and non-linear effects of wind and solar\npower generation on UK wholesale electricity prices, revealing key insights\nthat have eluded previous analyses. We find that, over 2018-2024, wind power\ngeneration has a U-shaped effect on prices: at low penetration levels, a 1 GWh\nincrease in energy generation reduces prices by up to 7 GBP/MWh, but this\neffect gets close to none at mid-penetration levels (20-30%) before\nintensifying again. Solar power places substantial downward pressure on prices\nat very low penetration levels (up to 9 GBP/MWh per 1 GWh increase in energy\ngeneration), though its impact weakens quite rapidly. We also uncover a\ncritical trend where the price-reducing effects of both wind and solar power\nhave become more pronounced over time (from 2018 to 2024), highlighting their\ngrowing influence on electricity markets amid rising penetration. Our study\nprovides both novel analysis approaches and actionable insights to guide\npolicymakers in appraising the way renewables impact electricity markets.\n", "date": "2025-01-10", "categories": []}
{"id": "2501.10693", "title": "Distributionally Robust Policy Evaluation and Learning for Continuous\n  Treatment with Observational Data", "abstract": "  Using offline observational data for policy evaluation and learning allows\ndecision-makers to evaluate and learn a policy that connects characteristics\nand interventions. Most existing literature has focused on either discrete\ntreatment spaces or assumed no difference in the distributions between the\npolicy-learning and policy-deployed environments. These restrict applications\nin many real-world scenarios where distribution shifts are present with\ncontinuous treatment. To overcome these challenges, this paper focuses on\ndeveloping a distributionally robust policy under a continuous treatment\nsetting. The proposed distributionally robust estimators are established using\nthe Inverse Probability Weighting (IPW) method extended from the discrete one\nfor policy evaluation and learning under continuous treatments. Specifically,\nwe introduce a kernel function into the proposed IPW estimator to mitigate the\nexclusion of observations that can occur in the standard IPW method to\ncontinuous treatments. We then provide finite-sample analysis that guarantees\nthe convergence of the proposed distributionally robust policy evaluation and\nlearning estimators. The comprehensive experiments further verify the\neffectiveness of our approach when distribution shifts are present.\n", "date": "2025-01-18", "categories": []}
{"id": "2501.10235", "title": "SpaceTime: Causal Discovery from Non-Stationary Time Series", "abstract": "  Understanding causality is challenging and often complicated by changing\ncausal relationships over time and across environments. Climate patterns, for\nexample, shift over time with recurring seasonal trends, while also depending\non geographical characteristics such as ecosystem variability. Existing methods\nfor discovering causal graphs from time series either assume stationarity, do\nnot permit both temporal and spatial distribution changes, or are unaware of\nlocations with the same causal relationships. In this work, we therefore unify\nthe three tasks of causal graph discovery in the non-stationary multi-context\nsetting, of reconstructing temporal regimes, and of partitioning datasets and\ntime intervals into those where invariant causal relationships hold. To\nconstruct a consistent score that forms the basis of our method, we employ the\nMinimum Description Length principle. Our resulting algorithm SPACETIME\nsimultaneously accounts for heterogeneity across space and non-stationarity\nover time. Given multiple time series, it discovers regime changepoints and a\ntemporal causal graph using non-parametric functional modeling and kernelized\ndiscrepancy testing. We also show that our method provides insights into\nreal-world phenomena such as river-runoff measured at different catchments and\nbiosphere-atmosphere interactions across ecosystems.\n", "date": "2025-01-17", "categories": []}
{"id": "2501.09163", "title": "Towards Understanding Extrapolation: a Causal Lens", "abstract": "  Canonical work handling distribution shifts typically necessitates an entire\ntarget distribution that lands inside the training distribution. However,\npractical scenarios often involve only a handful of target samples, potentially\nlying outside the training support, which requires the capability of\nextrapolation. In this work, we aim to provide a theoretical understanding of\nwhen extrapolation is possible and offer principled methods to achieve it\nwithout requiring an on-support target distribution. To this end, we formulate\nthe extrapolation problem with a latent-variable model that embodies the\nminimal change principle in causal mechanisms. Under this formulation, we cast\nthe extrapolation problem into a latent-variable identification problem. We\nprovide realistic conditions on shift properties and the estimation objectives\nthat lead to identification even when only one off-support target sample is\navailable, tackling the most challenging scenarios. Our theory reveals the\nintricate interplay between the underlying manifold's smoothness and the shift\nproperties. We showcase how our theoretical results inform the design of\npractical adaptation algorithms. Through experiments on both synthetic and\nreal-world data, we validate our theoretical findings and their practical\nimplications.\n", "date": "2025-01-15", "categories": []}
{"id": "2501.09304", "title": "Finding the Trigger: Causal Abductive Reasoning on Video Events", "abstract": "  This paper introduces a new problem, Causal Abductive Reasoning on Video\nEvents (CARVE), which involves identifying causal relationships between events\nin a video and generating hypotheses about causal chains that account for the\noccurrence of a target event. To facilitate research in this direction, we\ncreate two new benchmark datasets with both synthetic and realistic videos,\naccompanied by trigger-target labels generated through a novel counterfactual\nsynthesis approach. To explore the challenge of solving CARVE, we present a\nCausal Event Relation Network (CERN) that examines the relationships between\nvideo events in temporal and semantic spaces to efficiently determine the\nroot-cause trigger events. Through extensive experiments, we demonstrate the\ncritical roles of event relational representation learning and interaction\nmodeling in solving video causal reasoning challenges. The introduction of the\nCARVE task, along with the accompanying datasets and the CERN framework, will\nadvance future research on video causal reasoning and significantly facilitate\nvarious applications, including video surveillance, root-cause analysis and\nmovie content management.\n", "date": "2025-01-16", "categories": []}
{"id": "2406.00519", "title": "Learning Discrete Concepts in Latent Hierarchical Models", "abstract": "  Learning concepts from natural high-dimensional data (e.g., images) holds\npotential in building human-aligned and interpretable machine learning models.\nDespite its encouraging prospect, formalization and theoretical insights into\nthis crucial task are still lacking. In this work, we formalize concepts as\ndiscrete latent causal variables that are related via a hierarchical causal\nmodel that encodes different abstraction levels of concepts embedded in\nhigh-dimensional data (e.g., a dog breed and its eye shapes in natural images).\nWe formulate conditions to facilitate the identification of the proposed causal\nmodel, which reveals when learning such concepts from unsupervised data is\npossible. Our conditions permit complex causal hierarchical structures beyond\nlatent trees and multi-level directed acyclic graphs in prior work and can\nhandle high-dimensional, continuous observed variables, which is well-suited\nfor unstructured data modalities such as images. We substantiate our\ntheoretical claims with synthetic data experiments. Further, we discuss our\ntheory's implications for understanding the underlying mechanisms of latent\ndiffusion models and provide corresponding empirical evidence for our\ntheoretical insights.\n", "date": "2024-06-01", "categories": []}
{"id": "2501.08426", "title": "Causal vs. Anticausal merging of predictors", "abstract": "  We study the differences arising from merging predictors in the causal and\nanticausal directions using the same data. In particular we study the\nasymmetries that arise in a simple model where we merge the predictors using\none binary variable as target and two continuous variables as predictors. We\nuse Causal Maximum Entropy (CMAXENT) as inductive bias to merge the predictors,\nhowever, we expect similar differences to hold also when we use other merging\nmethods that take into account asymmetries between cause and effect. We show\nthat if we observe all bivariate distributions, the CMAXENT solution reduces to\na logistic regression in the causal direction and Linear Discriminant Analysis\n(LDA) in the anticausal direction. Furthermore, we study how the decision\nboundaries of these two solutions differ whenever we observe only some of the\nbivariate distributions implications for Out-Of-Variable (OOV) generalisation.\n", "date": "2025-01-14", "categories": []}
{"id": "2403.20287", "title": "Benchmarking Counterfactual Image Generation", "abstract": "  Generative AI has revolutionised visual content editing, empowering users to\neffortlessly modify images and videos. However, not all edits are equal. To\nperform realistic edits in domains such as natural image or medical imaging,\nmodifications must respect causal relationships inherent to the data generation\nprocess. Such image editing falls into the counterfactual image generation\nregime. Evaluating counterfactual image generation is substantially complex:\nnot only it lacks observable ground truths, but also requires adherence to\ncausal constraints. Although several counterfactual image generation methods\nand evaluation metrics exist, a comprehensive comparison within a unified\nsetting is lacking. We present a comparison framework to thoroughly benchmark\ncounterfactual image generation methods. We integrate all models that have been\nused for the task at hand and expand them to novel datasets and causal graphs,\ndemonstrating the superiority of Hierarchical VAEs across most datasets and\nmetrics. Our framework is implemented in a user-friendly Python package that\ncan be extended to incorporate additional SCMs, causal methods, generative\nmodels, and datasets for the community to build on. Code:\nhttps://github.com/gulnazaki/counterfactual-benchmark.\n", "date": "2024-03-29", "categories": []}
{"id": "2405.14496", "title": "Hybrid Top-Down Global Causal Discovery with Local Search for Linear and\n  Nonlinear Additive Noise Models", "abstract": "  Learning the unique directed acyclic graph corresponding to an unknown causal\nmodel is a challenging task. Methods based on functional causal models can\nidentify a unique graph, but either suffer from the curse of dimensionality or\nimpose strong parametric assumptions. To address these challenges, we propose a\nnovel hybrid approach for global causal discovery in observational data that\nleverages local causal substructures. We first present a topological sorting\nalgorithm that leverages ancestral relationships in linear structural causal\nmodels to establish a compact top-down hierarchical ordering, encoding more\ncausal information than linear orderings produced by existing methods. We\ndemonstrate that this approach generalizes to nonlinear settings with arbitrary\nnoise. We then introduce a nonparametric constraint-based algorithm that prunes\nspurious edges by searching for local conditioning sets, achieving greater\naccuracy than current methods. We provide theoretical guarantees for\ncorrectness and worst-case polynomial time complexities, with empirical\nvalidation on synthetic data.\n", "date": "2024-05-23", "categories": []}
{"id": "2501.06534", "title": "Dynamic Causal Structure Discovery and Causal Effect Estimation", "abstract": "  To represent the causal relationships between variables, a directed acyclic\ngraph (DAG) is widely utilized in many areas, such as social sciences,\nepidemics, and genetics. Many causal structure learning approaches are\ndeveloped to learn the hidden causal structure utilizing deep-learning\napproaches. However, these approaches have a hidden assumption that the causal\nrelationship remains unchanged over time, which may not hold in real life. In\nthis paper, we develop a new framework to model the dynamic causal graph where\nthe causal relations are allowed to be time-varying. We incorporate the basis\napproximation method into the score-based causal discovery approach to capture\nthe dynamic pattern of the causal graphs. Utilizing the autoregressive model\nstructure, we could capture both contemporaneous and time-lagged causal\nrelationships while allowing them to vary with time. We propose an algorithm\nthat could provide both past-time estimates and future-time predictions on the\ncausal graphs, and conduct simulations to demonstrate the usefulness of the\nproposed method. We also apply the proposed method for the covid-data analysis,\nand provide causal estimates on how policy restriction's effect changes.\n", "date": "2025-01-11", "categories": []}
{"id": "2501.07206", "title": "A data-driven approach to discover and quantify systemic lupus\n  erythematosus etiological heterogeneity from electronic health records", "abstract": "  Systemic lupus erythematosus (SLE) is a complex heterogeneous disease with\nmany manifestational facets. We propose a data-driven approach to discover\nprobabilistic independent sources from multimodal imperfect EHR data. These\nsources represent exogenous variables in the data generation process causal\ngraph that estimate latent root causes of the presence of SLE in the health\nrecord. We objectively evaluated the sources against the original variables\nfrom which they were discovered by training supervised models to discriminate\nSLE from negative health records using a reduced set of labelled instances. We\nfound 19 predictive sources with high clinical validity and whose EHR\nsignatures define independent factors of SLE heterogeneity. Using the sources\nas input patient data representation enables models to provide with rich\nexplanations that better capture the clinical reasons why a particular record\nis (not) an SLE case. Providers may be willing to trade patient-level\ninterpretability for discrimination especially in challenging cases.\n", "date": "2025-01-13", "categories": []}
{"id": "2501.05498", "title": "Generative Flow Networks: Theory and Applications to Structure Learning", "abstract": "  Without any assumptions about data generation, multiple causal models may\nexplain our observations equally well. To avoid selecting a single arbitrary\nmodel that could result in unsafe decisions if it does not match reality, it is\ntherefore essential to maintain a notion of epistemic uncertainty about our\npossible candidates. This thesis studies the problem of structure learning from\na Bayesian perspective, approximating the posterior distribution over the\nstructure of a causal model, represented as a directed acyclic graph (DAG),\ngiven data. It introduces Generative Flow Networks (GFlowNets), a novel class\nof probabilistic models designed for modeling distributions over discrete and\ncompositional objects such as graphs. They treat generation as a sequential\ndecision making problem, constructing samples of a target distribution defined\nup to a normalization constant piece by piece. In the first part of this\nthesis, we present the mathematical foundations of GFlowNets, their connections\nto existing domains of machine learning and statistics such as variational\ninference and reinforcement learning, and their extensions beyond discrete\nproblems. In the second part of this thesis, we show how GFlowNets can\napproximate the posterior distribution over DAG structures of causal Bayesian\nNetworks, along with the parameters of its causal mechanisms, given\nobservational and experimental data.\n", "date": "2025-01-09", "categories": []}
{"id": "2501.06077", "title": "Explainable Federated Bayesian Causal Inference and Its Application in\n  Advanced Manufacturing", "abstract": "  Causal inference has recently gained notable attention across various fields\nlike biology, healthcare, and environmental science, especially within\nexplainable artificial intelligence (xAI) systems, for uncovering the causal\nrelationships among multiple variables and outcomes. Yet, it has not been fully\nrecognized and deployed in the manufacturing systems. In this paper, we\nintroduce an explainable, scalable, and flexible federated Bayesian learning\nframework, \\texttt{xFBCI}, designed to explore causality through treatment\neffect estimation in distributed manufacturing systems. By leveraging federated\nBayesian learning, we efficiently estimate posterior of local parameters to\nderive the propensity score for each client without accessing local private\ndata. These scores are then used to estimate the treatment effect using\npropensity score matching (PSM). Through simulations on various datasets and a\nreal-world Electrohydrodynamic (EHD) printing data, we demonstrate that our\napproach outperforms standard Bayesian causal inference methods and several\nstate-of-the-art federated learning benchmarks.\n", "date": "2025-01-10", "categories": []}
{"id": "2310.05549", "title": "A New Transformation Approach for Uplift Modeling with Binary Outcome", "abstract": "  Uplift modeling has been used effectively in fields such as marketing and\ncustomer retention, to target those customers who are more likely to respond\ndue to the campaign or treatment. Essentially, it is a machine learning\ntechnique that predicts the gain from performing some action with respect to\nnot taking it. A popular class of uplift models is the transformation approach\nthat redefines the target variable with the original treatment indicator. These\ntransformation approaches only need to train and predict the difference in\noutcomes directly. The main drawback of these approaches is that in general it\ndoes not use the information in the treatment indicator beyond the construction\nof the transformed outcome and usually is not efficient. In this paper, we\ndesign a novel transformed outcome for the case of the binary target variable\nand unlock the full value of the samples with zero outcome. From a practical\nperspective, our new approach is flexible and easy to use. Experimental results\non synthetic and real-world datasets obviously show that our new approach\noutperforms the traditional one. At present, our new approach has already been\napplied to precision marketing in a China nation-wide financial holdings group.\n", "date": "2023-10-09", "categories": []}
{"id": "2501.04061", "title": "Causal Machine Learning Methods for Estimating Personalised Treatment\n  Effects -- Insights on validity from two large trials", "abstract": "  Causal machine learning (ML) methods hold great promise for advancing\nprecision medicine by estimating personalized treatment effects. However, their\nreliability remains largely unvalidated in empirical settings. In this study,\nwe assessed the internal and external validity of 17 mainstream causal\nheterogeneity ML methods -- including metalearners, tree-based methods, and\ndeep learning methods -- using data from two large randomized controlled\ntrials: the International Stroke Trial (N=19,435) and the Chinese Acute Stroke\nTrial (N=21,106). Our findings reveal that none of the ML methods reliably\nvalidated their performance, neither internal nor external, showing significant\ndiscrepancies between training and test data on the proposed evaluation\nmetrics. The individualized treatment effects estimated from training data\nfailed to generalize to the test data, even in the absence of distribution\nshifts. These results raise concerns about the current applicability of causal\nML models in precision medicine, and highlight the need for more robust\nvalidation techniques to ensure generalizability.\n", "date": "2025-01-07", "categories": []}
{"id": "2501.04724", "title": "Guiding Treatment Strategies: The Role of Adjuvant Anti-Her2 Neu Therapy\n  and Skin/Nipple Involvement in Local Recurrence-Free Survival in Breast\n  Cancer Patients", "abstract": "  This study explores how causal inference models, specifically the Linear\nNon-Gaussian Acyclic Model (LiNGAM), can extract causal relationships between\ndemographic factors, treatments, conditions, and outcomes from observational\npatient data, enabling insights beyond correlation. Unlike traditional\nrandomized controlled trials (RCTs), which establish causal relationships\nwithin narrowly defined populations, our method leverages broader observational\ndata, improving generalizability. Using over 40 features in the Duke MRI Breast\nCancer dataset, we found that Adjuvant Anti-Her2 Neu Therapy increased local\nrecurrence-free survival by 169 days, while Skin/Nipple involvement reduced it\nby 351 days. These findings highlight the therapy's importance for\nHer2-positive patients and the need for targeted interventions for high-risk\ncases, informing personalized treatment strategies.\n", "date": "2025-01-04", "categories": []}
{"id": "2501.04898", "title": "Optimality and Adaptivity of Deep Neural Features for Instrumental\n  Variable Regression", "abstract": "  We provide a convergence analysis of deep feature instrumental variable\n(DFIV) regression (Xu et al., 2021), a nonparametric approach to IV regression\nusing data-adaptive features learned by deep neural networks in two stages. We\nprove that the DFIV algorithm achieves the minimax optimal learning rate when\nthe target structural function lies in a Besov space. This is shown under\nstandard nonparametric IV assumptions, and an additional smoothness assumption\non the regularity of the conditional distribution of the covariate given the\ninstrument, which controls the difficulty of Stage 1. We further demonstrate\nthat DFIV, as a data-adaptive algorithm, is superior to fixed-feature (kernel\nor sieve) IV methods in two ways. First, when the target function possesses low\nspatial homogeneity (i.e., it has both smooth and spiky/discontinuous regions),\nDFIV still achieves the optimal rate, while fixed-feature methods are shown to\nbe strictly suboptimal. Second, comparing with kernel-based two-stage\nregression estimators, DFIV is provably more data efficient in the Stage 1\nsamples.\n", "date": "2025-01-09", "categories": []}
{"id": "2501.05197", "title": "An Algorithmic Approach for Causal Health Equity: A Look at Race\n  Differentials in Intensive Care Unit (ICU) Outcomes", "abstract": "  The new era of large-scale data collection and analysis presents an\nopportunity for diagnosing and understanding the causes of health inequities.\nIn this study, we describe a framework for systematically analyzing health\ndisparities using causal inference. The framework is illustrated by\ninvestigating racial and ethnic disparities in intensive care unit (ICU)\noutcome between majority and minority groups in Australia (Indigenous vs.\nNon-Indigenous) and the United States (African-American vs. White). We\ndemonstrate that commonly used statistical measures for quantifying inequity\nare insufficient, and focus on attributing the observed disparity to the causal\nmechanisms that generate it. We find that minority patients are younger at\nadmission, have worse chronic health, are more likely to be admitted for urgent\nand non-elective reasons, and have higher illness severity. At the same time,\nhowever, we find a protective direct effect of belonging to a minority group,\nwith minority patients showing improved survival compared to their majority\ncounterparts, with all other variables kept equal. We demonstrate that this\nprotective effect is related to the increased probability of being admitted to\nICU, with minority patients having an increased risk of ICU admission. We also\nfind that minority patients, while showing improved survival, are more likely\nto be readmitted to ICU. Thus, due to worse access to primary health care,\nminority patients are more likely to end up in ICU for preventable conditions,\ncausing a reduction in the mortality rates and creating an effect that appears\nto be protective. Since the baseline risk of ICU admission may serve as proxy\nfor lack of access to primary care, we developed the Indigenous Intensive Care\nEquity (IICE) Radar, a monitoring system for tracking the over-utilization of\nICU resources by the Indigenous population of Australia across geographical\nareas.\n", "date": "2025-01-09", "categories": []}
{"id": "2301.12553", "title": "Asymptotic Inference for Multi-Stage Stationary Treatment Policy with\n  Variable Selection", "abstract": "  Dynamic treatment regimes or policies are a sequence of decision functions\nover multiple stages that are tailored to individual features. One important\nclass of treatment policies in practice, namely multi-stage stationary\ntreatment policies, prescribes treatment assignment probabilities using the\nsame decision function across stages, where the decision is based on the same\nset of features consisting of time-evolving variables (e.g., routinely\ncollected disease biomarkers). Although there has been extensive literature on\nconstructing valid inference for the value function associated with dynamic\ntreatment policies, little work has focused on the policies themselves,\nespecially in the presence of high-dimensional feature variables. We aim to\nfill the gap in this work. Specifically, we first estimate the multi-stage\nstationary treatment policy using an augmented inverse probability weighted\nestimator for the value function to increase asymptotic efficiency, and further\napply a penalty to select important feature variables. We then construct\none-step improvements of the policy parameter estimators for valid inference.\nTheoretically, we show that the improved estimators are asymptotically normal,\neven if nuisance parameters are estimated at a slow convergence rate and the\ndimension of the feature variables increases with the sample size. Our\nnumerical studies demonstrate that the proposed method estimates a sparse\npolicy with a near-optimal value function and conducts valid inference for the\npolicy parameters.\n", "date": "2023-01-29", "categories": []}
{"id": "2501.02961", "title": "A Point Process Model for Optimizing Repeated Personalized Action\n  Delivery to Users", "abstract": "  This paper provides a formalism for an important class of causal inference\nproblems inspired by user-advertiser interaction in online advertiser. Then\nthis formalism is specialized to an extension of temporal marked point\nprocesses and the neural point processes are suggested as practical solutions\nto some interesting special cases.\n", "date": "2025-01-06", "categories": []}
{"id": "2310.18634", "title": "SSL Framework for Causal Inconsistency between Structures and\n  Representations", "abstract": "  The cross-pollination between causal discovery and deep learning has led to\nincreasingly extensive interactions. It results in a large number of deep\nlearning data types (such as images, text, etc.) extending into the field of\ncausal discovery, and a multitude of deep learning tasks have begun to utilize\ncausal discovery to explore the internal causal structure and causal\nrepresentation of data. In this paper, we first identified that a complex data\ntype, ``Indefinite Data\", has conflicts between causal relationships expressed\nby the causal structure and causal representation generated by deep learning\nmodels, a phenomenon referred to as causal inconsistency. We thoroughly\nanalyzed related work to explain why only Indefinite Data exhibits causal\ninconsistency while other data types do not. Furthermore, to alleviate causal\ninconsistency, we proposed a self-supervised learning (SSL) framework based on\nintervention, hoping to provide more causal information from different\nintervention views to promote consistency between structure and representation.\nExtensive experiments have shown that the SSL framework enhances causal\nconsistency and can further improve causal structure and representation\nlearning performance. Additionally, we extended the SSL framework to three\ndifferent downstream tasks and LLM instructions. The quantitative results of\nthese applications all reflect the performance improvement brought about by\ncausal consistency.\n", "date": "2023-10-28", "categories": []}
{"id": "2312.08174", "title": "Double Machine Learning for Static Panel Models with Fixed Effects", "abstract": "  Recent advances in causal inference have seen the development of methods\nwhich make use of the predictive power of machine learning algorithms. In this\npaper, we develop novel double machine learning (DML) procedures for panel data\nin which these algorithms are used to approximate high-dimensional and\nnonlinear nuisance functions of the covariates. Our new procedures are\nextensions of the well-known correlated random effects, within-group and\nfirst-difference estimators from linear to nonlinear panel models,\nspecifically, Robinson (1988)'s partially linear regression model with fixed\neffects and unspecified nonlinear confounding. Our simulation study assesses\nthe performance of these procedures using different machine learning\nalgorithms. We use our procedures to re-estimate the impact of minimum wage on\nvoting behaviour in the UK. From our results, we recommend the use of\nfirst-differencing because it imposes the fewest constraints on the\ndistribution of the fixed effects, and an ensemble learning strategy to ensure\noptimum estimator accuracy.\n", "date": "2023-12-13", "categories": []}
{"id": "2501.00119", "title": "Post Launch Evaluation of Policies in a High-Dimensional Setting", "abstract": "  A/B tests, also known as randomized controlled experiments (RCTs), are the\ngold standard for evaluating the impact of new policies, products, or\ndecisions. However, these tests can be costly in terms of time and resources,\npotentially exposing users, customers, or other test subjects (units) to\ninferior options. This paper explores practical considerations in applying\nmethodologies inspired by \"synthetic control\" as an alternative to traditional\nA/B testing in settings with very large numbers of units, involving up to\nhundreds of millions of units, which is common in modern applications such as\ne-commerce and ride-sharing platforms. This method is particularly valuable in\nsettings where the treatment affects only a subset of units, leaving many units\nunaffected. In these scenarios, synthetic control methods leverage data from\nunaffected units to estimate counterfactual outcomes for treated units. After\nthe treatment is implemented, these estimates can be compared to actual\noutcomes to measure the treatment effect. A key challenge in creating accurate\ncounterfactual outcomes is interpolation bias, a well-documented phenomenon\nthat occurs when control units differ significantly from treated units. To\naddress this, we propose a two-phase approach: first using nearest neighbor\nmatching based on unit covariates to select similar control units, then\napplying supervised learning methods suitable for high-dimensional data to\nestimate counterfactual outcomes. Testing using six large-scale experiments\ndemonstrates that this approach successfully improves estimate accuracy.\nHowever, our analysis reveals that machine learning bias -- which arises from\nmethods that trade off bias for variance reduction -- can impact results and\naffect conclusions about treatment effects. We document this bias in\nlarge-scale experimental settings and propose effective de-biasing techniques\nto address this challenge.\n", "date": "2024-12-30", "categories": []}
{"id": "2501.00673", "title": "Controlled Causal Hallucinations Can Estimate Phantom Nodes in\n  Multiexpert Mixtures of Fuzzy Cognitive Maps", "abstract": "  An adaptive multiexpert mixture of feedback causal models can approximate\nmissing or phantom nodes in large-scale causal models. The result gives a\nscalable form of \\emph{big knowledge}. The mixed model approximates a sampled\ndynamical system by approximating its main limit-cycle equilibria. Each expert\nfirst draws a fuzzy cognitive map (FCM) with at least one missing causal node\nor variable. FCMs are directed signed partial-causality cyclic graphs. They mix\nnaturally through convex combination to produce a new causal feedback FCM.\nSupervised learning helps each expert FCM estimate its phantom node by\ncomparing the FCM's partial equilibrium with the complete multi-node\nequilibrium. Such phantom-node estimation allows partial control over these\ncausal hallucinations and helps approximate the future trajectory of the\ndynamical system. But the approximation can be computationally heavy. Mixing\nthe tuned expert FCMs gives a practical way to find several phantom nodes and\nthereby better approximate the feedback system's true equilibrium behavior.\n", "date": "2024-12-31", "categories": []}
{"id": "2501.00854", "title": "A Graphical Approach to State Variable Selection in Off-policy Learning", "abstract": "  Sequential decision problems are widely studied across many areas of science.\nA key challenge when learning policies from historical data - a practice\ncommonly referred to as off-policy learning - is how to ``identify'' the impact\nof a policy of interest when the observed data are not randomized. Off-policy\nlearning has mainly been studied in two settings: dynamic treatment regimes\n(DTRs), where the focus is on controlling confounding in medical problems with\nshort decision horizons, and offline reinforcement learning (RL), where the\nfocus is on dimension reduction in closed systems such as games. The gap\nbetween these two well studied settings has limited the wider application of\noff-policy learning to many real-world problems. Using the theory for causal\ninference based on acyclic directed mixed graph (ADMGs), we provide a set of\ngraphical identification criteria in general decision processes that encompass\nboth DTRs and MDPs. We discuss how our results relate to the often implicit\ncausal assumptions made in the DTR and RL literatures and further clarify\nseveral common misconceptions. Finally, we present a realistic simulation study\nfor the dynamic pricing problem encountered in container logistics, and\ndemonstrate how violations of our graphical criteria can lead to suboptimal\npolicies.\n", "date": "2025-01-01", "categories": []}
{"id": "2304.09010", "title": "Causal Flow-based Variational Auto-Encoder for Disentangled Causal\n  Representation Learning", "abstract": "  Disentangled representation learning aims to learn low-dimensional\nrepresentations where each dimension corresponds to an underlying generative\nfactor. While the Variational Auto-Encoder (VAE) is widely used for this\npurpose, most existing methods assume independence among factors, a\nsimplification that does not hold in many real-world scenarios where factors\nare often interdependent and exhibit causal relationships. To overcome this\nlimitation, we propose the Disentangled Causal Variational Auto-Encoder\n(DCVAE), a novel supervised VAE framework that integrates causal flows into the\nrepresentation learning process, enabling the learning of more meaningful and\ninterpretable disentangled representations. We evaluate DCVAE on both synthetic\nand real-world datasets, demonstrating its superior ability in causal\ndisentanglement and intervention experiments. Furthermore, DCVAE outperforms\nstate-of-the-art methods in various downstream tasks, highlighting its\npotential for learning true causal structures among factors.\n", "date": "2023-04-18", "categories": []}
{"id": "2412.20289", "title": "Causal Discovery on Dependent Binary Data", "abstract": "  The assumption of independence between observations (units) in a dataset is\nprevalent across various methodologies for learning causal graphical models.\nHowever, this assumption often finds itself in conflict with real-world data,\nposing challenges to accurate structure learning. We propose a\ndecorrelation-based approach for causal graph learning on dependent binary\ndata, where the local conditional distribution is defined by a latent utility\nmodel with dependent errors across units. We develop a pairwise maximum\nlikelihood method to estimate the covariance matrix for the dependence among\nthe units. Then, leveraging the estimated covariance matrix, we develop an\nEM-like iterative algorithm to generate and decorrelate samples of the latent\nutility variables, which serve as decorrelated data. Any standard causal\ndiscovery method can be applied on the decorrelated data to learn the\nunderlying causal graph. We demonstrate that the proposed decorrelation\napproach significantly improves the accuracy in causal graph learning, through\nnumerical experiments on both synthetic and real-world datasets.\n", "date": "2024-12-28", "categories": []}
{"id": "2409.12797", "title": "Efficient Identification of Direct Causal Parents via Invariance and\n  Minimum Error Testing", "abstract": "  Invariant causal prediction (ICP) is a popular technique for finding causal\nparents (direct causes) of a target via exploiting distribution shifts and\ninvariance testing (Peters et al., 2016). However, since ICP needs to run an\nexponential number of tests and fails to identify parents when distribution\nshifts only affect a few variables, applying ICP to practical large scale\nproblems is challenging. We propose MMSE-ICP and fastICP, two approaches which\nemploy an error inequality to address the identifiability problem of ICP. The\ninequality states that the minimum prediction error of the predictor using\ncausal parents is the smallest among all predictors which do not use\ndescendants. fastICP is an efficient approximation tailored for large problems\nas it exploits the inequality and a heuristic to run fewer tests. MMSE-ICP and\nfastICP not only outperform competitive baselines in many simulations but also\nachieve state-of-the-art result on a large scale real data benchmark.\n", "date": "2024-09-19", "categories": []}
{"id": "2409.17647", "title": "MECD: Unlocking Multi-Event Causal Discovery in Video Reasoning", "abstract": "  Video causal reasoning aims to achieve a high-level understanding of video\ncontent from a causal perspective. However, current video reasoning tasks are\nlimited in scope, primarily executed in a question-answering paradigm and\nfocusing on short videos containing only a single event and simple causal\nrelationships, lacking comprehensive and structured causality analysis for\nvideos with multiple events. To fill this gap, we introduce a new task and\ndataset, Multi-Event Causal Discovery (MECD). It aims to uncover the causal\nrelationships between events distributed chronologically across long videos.\nGiven visual segments and textual descriptions of events, MECD requires\nidentifying the causal associations between these events to derive a\ncomprehensive, structured event-level video causal diagram explaining why and\nhow the final result event occurred. To address MECD, we devise a novel\nframework inspired by the Granger Causality method, using an efficient\nmask-based event prediction model to perform an Event Granger Test, which\nestimates causality by comparing the predicted result event when premise events\nare masked versus unmasked. Furthermore, we integrate causal inference\ntechniques such as front-door adjustment and counterfactual inference to\naddress challenges in MECD like causality confounding and illusory causality.\nExperiments validate the effectiveness of our framework in providing causal\nrelationships in multi-event videos, outperforming GPT-4o and VideoLLaVA by\n5.7% and 4.1%, respectively.\n", "date": "2024-09-26", "categories": []}
{"id": "2410.07751", "title": "Learning Low-Level Causal Relations using a Simulated Robotic Arm", "abstract": "  Causal learning allows humans to predict the effect of their actions on the\nknown environment and use this knowledge to plan the execution of more complex\nactions. Such knowledge also captures the behaviour of the environment and can\nbe used for its analysis and the reasoning behind the behaviour. This type of\nknowledge is also crucial in the design of intelligent robotic systems with\ncommon sense. In this paper, we study causal relations by learning the forward\nand inverse models based on data generated by a simulated robotic arm involved\nin two sensorimotor tasks. As a next step, we investigate feature attribution\nmethods for the analysis of the forward model, which reveals the low-level\ncausal effects corresponding to individual features of the state vector related\nto both the arm joints and the environment features. This type of analysis\nprovides solid ground for dimensionality reduction of the state\nrepresentations, as well as for the aggregation of knowledge towards the\nexplainability of causal effects at higher levels.\n", "date": "2024-10-10", "categories": []}
{"id": "2410.10044", "title": "DAG-aware Transformer for Causal Effect Estimation", "abstract": "  Causal inference is a critical task across fields such as healthcare,\neconomics, and the social sciences. While recent advances in machine learning,\nespecially those based on the deep-learning architectures, have shown potential\nin estimating causal effects, existing approaches often fall short in handling\ncomplex causal structures and lack adaptability across various causal\nscenarios. In this paper, we present a novel transformer-based method for\ncausal inference that overcomes these challenges. The core innovation of our\nmodel lies in its integration of causal Directed Acyclic Graphs (DAGs) directly\ninto the attention mechanism, enabling it to accurately model the underlying\ncausal structure. This allows for flexible estimation of both average treatment\neffects (ATE) and conditional average treatment effects (CATE). Extensive\nexperiments on both synthetic and real-world datasets demonstrate that our\napproach surpasses existing methods in estimating causal effects across a wide\nrange of scenarios. The flexibility and robustness of our model make it a\nvaluable tool for researchers and practitioners tackling complex causal\ninference problems.\n", "date": "2024-10-13", "categories": []}
{"id": "2412.17508", "title": "An efficient search-and-score algorithm for ancestral graphs using\n  multivariate information scores", "abstract": "  We propose a greedy search-and-score algorithm for ancestral graphs, which\ninclude directed as well as bidirected edges, originating from unobserved\nlatent variables. The normalized likelihood score of ancestral graphs is\nestimated in terms of multivariate information over relevant ``ac-connected\nsubsets'' of vertices, C, that are connected through collider paths confined to\nthe ancestor set of C. For computational efficiency, the proposed two-step\nalgorithm relies on local information scores limited to the close surrounding\nvertices of each node (step 1) and edge (step 2). This computational strategy,\nalthough restricted to information contributions from ac-connected subsets\ncontaining up to two-collider paths, is shown to outperform state-of-the-art\ncausal discovery methods on challenging benchmark datasets.\n", "date": "2024-12-23", "categories": []}
{"id": "2412.19578", "title": "Graph-attention-based Casual Discovery with Trust Region-navigated\n  Clipping Policy Optimization", "abstract": "  In many domains of empirical sciences, discovering the causal structure\nwithin variables remains an indispensable task. Recently, to tackle with\nunoriented edges or latent assumptions violation suffered by conventional\nmethods, researchers formulated a reinforcement learning (RL) procedure for\ncausal discovery, and equipped REINFORCE algorithm to search for the\nbest-rewarded directed acyclic graph. The two keys to the overall performance\nof the procedure are the robustness of RL methods and the efficient encoding of\nvariables. However, on the one hand, REINFORCE is prone to local convergence\nand unstable performance during training. Neither trust region policy\noptimization, being computationally-expensive, nor proximal policy optimization\n(PPO), suffering from aggregate constraint deviation, is decent alternative for\ncombinatory optimization problems with considerable individual subactions. We\npropose a trust region-navigated clipping policy optimization method for causal\ndiscovery that guarantees both better search efficiency and steadiness in\npolicy optimization, in comparison with REINFORCE, PPO and our prioritized\nsampling-guided REINFORCE implementation. On the other hand, to boost the\nefficient encoding of variables, we propose a refined graph attention encoder\ncalled SDGAT that can grasp more feature information without priori\nneighbourhood information. With these improvements, the proposed method\noutperforms former RL method in both synthetic and benchmark datasets in terms\nof output results and optimization robustness.\n", "date": "2024-12-27", "categories": []}
{"id": "2311.17303", "title": "Enhancing the Performance of Neural Networks Through Causal Discovery\n  and Integration of Domain Knowledge", "abstract": "  In this paper, we develop a generic methodology to encode hierarchical\ncausality structure among observed variables into a neural network in order to\nimprove its predictive performance. The proposed methodology, called\ncausality-informed neural network (CINN), leverages three coherent steps to\nsystematically map the structural causal knowledge into the layer-to-layer\ndesign of neural network while strictly preserving the orientation of every\ncausal relationship. In the first step, CINN discovers causal relationships\nfrom observational data via directed acyclic graph (DAG) learning, where causal\ndiscovery is recast as a continuous optimization problem to avoid the\ncombinatorial nature. In the second step, the discovered hierarchical causality\nstructure among observed variables is systematically encoded into neural\nnetwork through a dedicated architecture and customized loss function. By\ncategorizing variables in the causal DAG as root, intermediate, and leaf nodes,\nthe hierarchical causal DAG is translated into CINN with a one-to-one\ncorrespondence between nodes in the causal DAG and units in the CINN while\nmaintaining the relative order among these nodes. Regarding the loss function,\nboth intermediate and leaf nodes in the DAG graph are treated as target outputs\nduring CINN training so as to drive co-learning of causal relationships among\ndifferent types of nodes. As multiple loss components emerge in CINN, we\nleverage the projection of conflicting gradients to mitigate gradient\ninterference among the multiple learning tasks. Computational experiments\nacross a broad spectrum of UCI data sets demonstrate substantial advantages of\nCINN in predictive performance over other state-of-the-art methods. In\naddition, an ablation study underscores the value of integrating structural and\nquantitative causal knowledge in enhancing the neural network's predictive\nperformance incrementally.\n", "date": "2023-11-29", "categories": []}
{"id": "2410.23620", "title": "Identifiability Guarantees for Causal Disentanglement from Purely\n  Observational Data", "abstract": "  Causal disentanglement aims to learn about latent causal factors behind data,\nholding the promise to augment existing representation learning methods in\nterms of interpretability and extrapolation. Recent advances establish\nidentifiability results assuming that interventions on (single) latent factors\nare available; however, it remains debatable whether such assumptions are\nreasonable due to the inherent nature of intervening on latent variables.\nAccordingly, we reconsider the fundamentals and ask what can be learned using\njust observational data.\n  We provide a precise characterization of latent factors that can be\nidentified in nonlinear causal models with additive Gaussian noise and linear\nmixing, without any interventions or graphical restrictions. In particular, we\nshow that the causal variables can be identified up to a layer-wise\ntransformation and that further disentanglement is not possible. We transform\nthese theoretical results into a practical algorithm consisting of solving a\nquadratic program over the score estimation of the observed data. We provide\nsimulation results to support our theoretical guarantees and demonstrate that\nour algorithm can derive meaningful causal representations from purely\nobservational data.\n", "date": "2024-10-31", "categories": []}
{"id": "2402.05639", "title": "Nonparametric Instrumental Variable Regression through Stochastic\n  Approximate Gradients", "abstract": "  Instrumental variables (IVs) provide a powerful strategy for identifying\ncausal effects in the presence of unobservable confounders. Within the\nnonparametric setting (NPIV), recent methods have been based on nonlinear\ngeneralizations of Two-Stage Least Squares and on minimax formulations derived\nfrom moment conditions or duality. In a novel direction, we show how to\nformulate a functional stochastic gradient descent algorithm to tackle NPIV\nregression by directly minimizing the populational risk. We provide theoretical\nsupport in the form of bounds on the excess risk, and conduct numerical\nexperiments showcasing our method's superior stability and competitive\nperformance relative to current state-of-the-art alternatives. This algorithm\nenables flexible estimator choices, such as neural networks or kernel based\nmethods, as well as non-quadratic loss functions, which may be suitable for\nstructural equations beyond the setting of continuous outcomes and additive\nnoise. Finally, we demonstrate this flexibility of our framework by presenting\nhow it naturally addresses the important case of binary outcomes, which has\nreceived far less attention by recent developments in the NPIV literature.\n", "date": "2024-02-08", "categories": []}
{"id": "2404.05545", "title": "Evaluating Interventional Reasoning Capabilities of Large Language\n  Models", "abstract": "  Numerous decision-making tasks require estimating causal effects under\ninterventions on different parts of a system. As practitioners consider using\nlarge language models (LLMs) to automate decisions, studying their causal\nreasoning capabilities becomes crucial. A recent line of work evaluates LLMs\nability to retrieve commonsense causal facts, but these evaluations do not\nsufficiently assess how LLMs reason about interventions. Motivated by the role\nthat interventions play in causal inference, in this paper, we conduct\nempirical analyses to evaluate whether LLMs can accurately update their\nknowledge of a data-generating process in response to an intervention. We\ncreate benchmarks that span diverse causal graphs (e.g., confounding,\nmediation) and variable types, and enable a study of intervention-based\nreasoning. These benchmarks allow us to isolate the ability of LLMs to\naccurately predict changes resulting from their ability to memorize facts or\nfind other shortcuts. We evaluate six LLMs on the benchmarks, finding that GPT\nmodels show promising accuracy at predicting the intervention effects.\n", "date": "2024-04-08", "categories": []}
{"id": "2411.11590", "title": "Robust Causal Analysis of Linear Cyclic Systems With Hidden Confounders", "abstract": "  We live in a world full of complex systems which we need to improve our\nunderstanding of. To accomplish this, purely probabilistic investigations are\noften not enough. They are only the first step and must be followed by learning\nthe system's underlying mechanisms. This is what the discipline of causality is\nconcerned with. Many of those complex systems contain feedback loops which\nmeans that our methods have to allow for cyclic causal relations. Furthermore,\nsystems are rarely sufficiently isolated, which means that there are usually\nhidden confounders, i.e., unmeasured variables that each causally affects more\nthan one measured variable. Finally, data is often distorted by contaminating\nprocesses, and we need to apply methods that are robust against such\ndistortions. That's why we consider the robustness of LLC, see \\cite{llc}, one\nof the few causal analysis methods that can deal with cyclic models with hidden\nconfounders. Following a theoretical analysis of LLC's robustness properties,\nwe also provide robust extensions of LLC. To facilitate reproducibility and\nfurther research in this field, we make the source code publicly available.\n", "date": "2024-11-18", "categories": []}
{"id": "2412.16691", "title": "From Correlation to Causation: Understanding Climate Change through\n  Causal Analysis and LLM Interpretations", "abstract": "  This research presents a three-step causal inference framework that\nintegrates correlation analysis, machine learning-based causality discovery,\nand LLM-driven interpretations to identify socioeconomic factors influencing\ncarbon emissions and contributing to climate change. The approach begins with\nidentifying correlations, progresses to causal analysis, and enhances decision\nmaking through LLM-generated inquiries about the context of climate change. The\nproposed framework offers adaptable solutions that support data-driven\npolicy-making and strategic decision-making in climate-related contexts,\nuncovering causal relationships within the climate change domain.\n", "date": "2024-12-21", "categories": []}
{"id": "2412.16939", "title": "Image Quality Assessment: Investigating Causal Perceptual Effects with\n  Abductive Counterfactual Inference", "abstract": "  Existing full-reference image quality assessment (FR-IQA) methods often fail\nto capture the complex causal mechanisms that underlie human perceptual\nresponses to image distortions, limiting their ability to generalize across\ndiverse scenarios. In this paper, we propose an FR-IQA method based on\nabductive counterfactual inference to investigate the causal relationships\nbetween deep network features and perceptual distortions. First, we explore the\ncausal effects of deep features on perception and integrate causal reasoning\nwith feature comparison, constructing a model that effectively handles complex\ndistortion types across different IQA scenarios. Second, the analysis of the\nperceptual causal correlations of our proposed method is independent of the\nbackbone architecture and thus can be applied to a variety of deep networks.\nThrough abductive counterfactual experiments, we validate the proposed causal\nrelationships, confirming the model's superior perceptual relevance and\ninterpretability of quality scores. The experimental results demonstrate the\nrobustness and effectiveness of the method, providing competitive quality\npredictions across multiple benchmarks. The source code is available at\nhttps://anonymous.4open.science/r/DeepCausalQuality-25BC.\n", "date": "2024-12-22", "categories": []}
{"id": "2405.14848", "title": "Local Causal Discovery for Structural Evidence of Direct Discrimination", "abstract": "  Identifying the causal pathways of unfairness is a critical objective for\nimproving policy design and algorithmic decision-making. Prior work in causal\nfairness analysis often requires knowledge of the causal graph, hindering\npractical applications in complex or low-knowledge domains. Moreover, global\ndiscovery methods that learn causal structure from data can display unstable\nperformance on finite samples, preventing robust fairness conclusions. To\nmitigate these challenges, we introduce local discovery for direct\ndiscrimination (LD3): a method that uncovers structural evidence of direct\nunfairness by identifying the causal parents of an outcome variable. LD3\nperforms a linear number of conditional independence tests relative to variable\nset size, and allows for latent confounding under the sufficient condition that\nall parents of the outcome are observed. We show that LD3 returns a valid\nadjustment set (VAS) under a new graphical criterion for the weighted\ncontrolled direct effect, a qualitative indicator of direct discrimination. LD3\nlimits unnecessary adjustment, providing interpretable VAS for assessing\nunfairness. We use LD3 to analyze causal fairness in two complex decision\nsystems: criminal recidivism prediction and liver transplant allocation. LD3\nwas more time-efficient and returned more plausible results on real-world data\nthan baselines, which took 46$\\times$ to 5870$\\times$ longer to execute.\n", "date": "2024-05-23", "categories": []}
{"id": "2407.07934", "title": "Identifying Macro Conditional Independencies and Macro Total Effects in\n  Summary Causal Graphs with Latent Confounding", "abstract": "  Understanding causal relations in dynamic systems is essential in\nepidemiology. While causal inference methods have been extensively studied,\nthey often rely on fully specified causal graphs, which may not always be\navailable in complex dynamic systems. Partially specified causal graphs, and in\nparticular summary causal graphs (SCGs), provide a simplified representation of\ncausal relations between time series when working spacio-temporal data,\nomitting temporal information and focusing on causal structures between\nclusters of of temporal variables. Unlike fully specified causal graphs, SCGs\ncan contain cycles, which complicate their analysis and interpretation. In\naddition, their cluster-based nature introduces new challenges concerning the\ntypes of queries of interest: macro queries, which involve relationships\nbetween clusters represented as vertices in the graph, and micro queries, which\npertain to relationships between variables that are not directly visible\nthrough the vertices of the graph. In this paper, we first clearly distinguish\nbetween macro conditional independencies and micro conditional independencies\nand between macro total effects and micro total effects. Then, we demonstrate\nthe soundness and completeness of the d-separation to identify macro\nconditional independencies in SCGs. Furthermore, we establish that the\ndo-calculus is sound and complete for identifying macro total effects in SCGs.\nFinally, we give a graphical characterization for the non-identifiability of\nmacro total effects in SCGs.\n", "date": "2024-07-10", "categories": []}
{"id": "2412.02579", "title": "Factored space models: Towards causality between levels of abstraction", "abstract": "  Causality plays an important role in understanding intelligent behavior, and\nthere is a wealth of literature on mathematical models for causality, most of\nwhich is focused on causal graphs. Causal graphs are a powerful tool for a wide\nrange of applications, in particular when the relevant variables are known and\nat the same level of abstraction. However, the given variables can also be\nunstructured data, like pixels of an image. Meanwhile, the causal variables,\nsuch as the positions of objects in the image, can be arbitrary deterministic\nfunctions of the given variables. Moreover, the causal variables may form a\nhierarchy of abstractions, in which the macro-level variables are deterministic\nfunctions of the micro-level variables. Causal graphs are limited when it comes\nto modeling this kind of situation. In the presence of deterministic\nrelationships there is generally no causal graph that satisfies both the Markov\ncondition and the faithfulness condition. We introduce factored space models as\nan alternative to causal graphs which naturally represent both probabilistic\nand deterministic relationships at all levels of abstraction. Moreover, we\nintroduce structural independence and establish that it is equivalent to\nstatistical independence in every distribution that factorizes over the\nfactored space. This theorem generalizes the classical soundness and\ncompleteness theorem for d-separation.\n", "date": "2024-12-03", "categories": []}
{"id": "2412.15373", "title": "Granger Causality Detection with Kolmogorov-Arnold Networks", "abstract": "  Discovering causal relationships in time series data is central in many\nscientific areas, ranging from economics to climate science. Granger causality\nis a powerful tool for causality detection. However, its original formulation\nis limited by its linear form and only recently nonlinear machine-learning\ngeneralizations have been introduced. This study contributes to the definition\nof neural Granger causality models by investigating the application of\nKolmogorov-Arnold networks (KANs) in Granger causality detection and comparing\ntheir capabilities against multilayer perceptrons (MLP). In this work, we\ndevelop a framework called Granger Causality KAN (GC-KAN) along with a tailored\ntraining approach designed specifically for Granger causality detection. We\ntest this framework on both Vector Autoregressive (VAR) models and chaotic\nLorenz-96 systems, analysing the ability of KANs to sparsify input features by\nidentifying Granger causal relationships, providing a concise yet accurate\nmodel for Granger causality detection. Our findings show the potential of KANs\nto outperform MLPs in discerning interpretable Granger causal relationships,\nparticularly for the ability of identifying sparse Granger causality patterns\nin high-dimensional settings, and more generally, the potential of AI in\ncausality discovery for the dynamical laws in physical systems.\n", "date": "2024-12-19", "categories": []}
{"id": "2305.09565", "title": "Toward Falsifying Causal Graphs Using a Permutation-Based Test", "abstract": "  Understanding causal relationships among the variables of a system is\nparamount to explain and control its behavior. For many real-world systems,\nhowever, the true causal graph is not readily available and one must resort to\npredictions made by algorithms or domain experts. Therefore, metrics that\nquantitatively assess the goodness of a causal graph provide helpful checks\nbefore using it in downstream tasks. Existing metrics provide an\n$\\textit{absolute}$ number of inconsistencies between the graph and the\nobserved data, and without a baseline, practitioners are left to answer the\nhard question of how many such inconsistencies are acceptable or expected.\nHere, we propose a novel consistency metric by constructing a baseline through\nnode permutations. By comparing the number of inconsistencies with those on the\nbaseline, we derive an interpretable metric that captures whether the graph is\nsignificantly better than random. Evaluating on both simulated and real data\nsets from various domains, including biology and cloud monitoring, we\ndemonstrate that the true graph is not falsified by our metric, whereas the\nwrong graphs given by a hypothetical user are likely to be falsified.\n", "date": "2023-05-16", "categories": []}
{"id": "2408.05428", "title": "Generalized Encouragement-Based Instrumental Variables for\n  Counterfactual Regression", "abstract": "  In causal inference, encouragement designs (EDs) are widely used to analyze\ncausal effects, when randomized controlled trials (RCTs) are impractical or\ncompliance to treatment cannot be perfectly enforced. Unlike RCTs, which\ndirectly allocate treatments, EDs randomly assign encouragement policies that\npositively motivate individuals to engage in a specific treatment. These random\nencouragements act as instrumental variables (IVs), facilitating the\nidentification of causal effects through leveraging exogenous perturbations in\ndiscrete treatment scenarios. However, real-world applications of encouragement\ndesigns often face challenges such as incomplete randomization, limited\nexperimental data, and significantly fewer encouragements compared to\ntreatments, hindering precise causal effect estimation. To address this, this\npaper introduces novel theories and algorithms for identifying the Conditional\nAverage Treatment Effect (CATE) using variations in encouragement. Further, by\nleveraging both observational and encouragement data, we propose a generalized\nIV estimator, named Encouragement-based Counterfactual Regression (EnCounteR),\nto effectively estimate the causal effects. Extensive experiments on both\nsynthetic and real-world datasets demonstrate the superiority of EnCounteR over\nexisting methods.\n", "date": "2024-08-10", "categories": []}
{"id": "2412.14491", "title": "Mediation Analysis for Probabilities of Causation", "abstract": "  Probabilities of causation (PoC) offer valuable insights for informed\ndecision-making. This paper introduces novel variants of PoC-controlled direct,\nnatural direct, and natural indirect probability of necessity and sufficiency\n(PNS). These metrics quantify the necessity and sufficiency of a treatment for\nproducing an outcome, accounting for different causal pathways. We develop\nidentification theorems for these new PoC measures, allowing for their\nestimation from observational data. We demonstrate the practical application of\nour results through an analysis of a real-world psychology dataset.\n", "date": "2024-12-19", "categories": []}
{"id": "2210.07416", "title": "GLACIAL: Granger and Learning-based Causality Analysis for Longitudinal\n  Imaging Studies", "abstract": "  The Granger framework is useful for discovering causal relations in\ntime-varying signals. However, most Granger causality (GC) methods are\ndeveloped for densely sampled timeseries data. A substantially different\nsetting, particularly common in medical imaging, is the longitudinal study\ndesign, where multiple subjects are followed and sparsely observed over time.\nLongitudinal studies commonly track several biomarkers, which are likely\ngoverned by nonlinear dynamics that might have subject-specific idiosyncrasies\nand exhibit both direct and indirect causes. Furthermore, real-world\nlongitudinal data often suffer from widespread missingness. GC methods are not\nwell-suited to handle these issues. In this paper, we propose an approach named\nGLACIAL (Granger and LeArning-based CausalIty Analysis for Longitudinal\nstudies) to fill this methodological gap by marrying GC with a multi-task\nneural forecasting model. GLACIAL treats subjects as independent samples and\nuses the model's average prediction accuracy on hold-out subjects to probe\ncausal links. Input dropout and model interpolation are used to efficiently\nlearn nonlinear dynamic relationships between a large number of variables and\nto handle missing values respectively. Extensive simulations and experiments on\na real longitudinal medical imaging dataset show GLACIAL beating competitive\nbaselines and confirm its utility. Our code is available at\nhttps://github.com/mnhng/GLACIAL.\n", "date": "2022-10-13", "categories": []}
{"id": "2412.13952", "title": "Prompting Strategies for Enabling Large Language Models to Infer\n  Causation from Correlation", "abstract": "  The reasoning abilities of Large Language Models (LLMs) are attracting\nincreasing attention. In this work, we focus on causal reasoning and address\nthe task of establishing causal relationships based on correlation information,\na highly challenging problem on which several LLMs have shown poor performance.\nWe introduce a prompting strategy for this problem that breaks the original\ntask into fixed subquestions, with each subquestion corresponding to one step\nof a formal causal discovery algorithm, the PC algorithm. The proposed\nprompting strategy, PC-SubQ, guides the LLM to follow these algorithmic steps,\nby sequentially prompting it with one subquestion at a time, augmenting the\nnext subquestion's prompt with the answer to the previous one(s). We evaluate\nour approach on an existing causal benchmark, Corr2Cause: our experiments\nindicate a performance improvement across five LLMs when comparing PC-SubQ to\nbaseline prompting strategies. Results are robust to causal query\nperturbations, when modifying the variable names or paraphrasing the\nexpressions.\n", "date": "2024-12-18", "categories": []}
{"id": "2412.12401", "title": "Causally Consistent Normalizing Flow", "abstract": "  Causal inconsistency arises when the underlying causal graphs captured by\ngenerative models like \\textit{Normalizing Flows} (NFs) are inconsistent with\nthose specified in causal models like \\textit{Struct Causal Models} (SCMs).\nThis inconsistency can cause unwanted issues including the unfairness problem.\nPrior works to achieve causal consistency inevitably compromise the\nexpressiveness of their models by disallowing hidden layers. In this work, we\nintroduce a new approach: \\textbf{C}ausally \\textbf{C}onsistent\n\\textbf{N}ormalizing \\textbf{F}low (CCNF). To the best of our knowledge, CCNF\nis the first causally consistent generative model that can approximate any\ndistribution with multiple layers. CCNF relies on two novel constructs: a\nsequential representation of SCMs and partial causal transformations. These\nconstructs allow CCNF to inherently maintain causal consistency without\nsacrificing expressiveness. CCNF can handle all forms of causal inference\ntasks, including interventions and counterfactuals. Through experiments, we\nshow that CCNF outperforms current approaches in causal inference. We also\nempirically validate the practical utility of CCNF by applying it to real-world\ndatasets and show how CCNF addresses challenges like unfairness effectively.\n", "date": "2024-12-16", "categories": []}
{"id": "2409.19377", "title": "Interpretable, multi-dimensional Evaluation Framework for Causal\n  Discovery from observational i.i.d. Data", "abstract": "  Nonlinear causal discovery from observational data imposes strict\nidentifiability assumptions on the formulation of structural equations utilized\nin the data generating process. The evaluation of structure learning methods\nunder assumption violations requires a rigorous and interpretable approach,\nwhich quantifies both the structural similarity of the estimation with the\nground truth and the capacity of the discovered graphs to be used for causal\ninference. Motivated by the lack of unified performance assessment framework,\nwe introduce an interpretable, six-dimensional evaluation metric, i.e.,\ndistance to optimal solution (DOS), which is specifically tailored to the field\nof causal discovery. Furthermore, this is the first research to assess the\nperformance of structure learning algorithms from seven different families on\nincreasing percentage of non-identifiable, nonlinear causal patterns, inspired\nby real-world processes. Our large-scale simulation study, which incorporates\nseven experimental factors, shows that besides causal order-based methods,\namortized causal discovery delivers results with comparatively high proximity\nto the optimal solution.\n", "date": "2024-09-28", "categories": []}
{"id": "2412.11104", "title": "ABC3: Active Bayesian Causal Inference with Cohn Criteria in Randomized\n  Experiments", "abstract": "  In causal inference, randomized experiment is a de facto method to overcome\nvarious theoretical issues in observational study. However, the experimental\ndesign requires expensive costs, so an efficient experimental design is\nnecessary. We propose ABC3, a Bayesian active learning policy for causal\ninference. We show a policy minimizing an estimation error on conditional\naverage treatment effect is equivalent to minimizing an integrated posterior\nvariance, similar to Cohn criteria \\citep{cohn1994active}. We theoretically\nprove ABC3 also minimizes an imbalance between the treatment and control groups\nand the type 1 error probability. Imbalance-minimizing characteristic is\nespecially notable as several works have emphasized the importance of achieving\nbalance. Through extensive experiments on real-world data sets, ABC3 achieves\nthe highest efficiency, while empirically showing the theoretical results hold.\n", "date": "2024-12-15", "categories": []}
{"id": "2402.14027", "title": "Learning causation event conjunction sequences", "abstract": "  This is an examination of some methods that learn causations in event\nsequences. A causation is defined as a conjunction of one or more cause events\noccurring in an arbitrary order, with possible intervening non-causal events,\nthat lead to an effect. The methods include recurrent and non-recurrent\nartificial neural networks (ANNs), as well as a histogram-based algorithm. An\nattention recurrent ANN performed the best of the ANNs, while the histogram\nalgorithm was significantly superior to all the ANNs.\n", "date": "2024-02-17", "categories": []}
{"id": "2404.06969", "title": "A Fixed-Point Approach for Causal Generative Modeling", "abstract": "  We propose a novel formalism for describing Structural Causal Models (SCMs)\nas fixed-point problems on causally ordered variables, eliminating the need for\nDirected Acyclic Graphs (DAGs), and establish the weakest known conditions for\ntheir unique recovery given the topological ordering (TO). Based on this, we\ndesign a two-stage causal generative model that first infers in a zero-shot\nmanner a valid TO from observations, and then learns the generative SCM on the\nordered variables. To infer TOs, we propose to amortize the learning of TOs on\nsynthetically generated datasets by sequentially predicting the leaves of\ngraphs seen during training. To learn SCMs, we design a transformer-based\narchitecture that exploits a new attention mechanism enabling the modeling of\ncausal structures, and show that this parameterization is consistent with our\nformalism. Finally, we conduct an extensive evaluation of each method\nindividually, and show that when combined, our model outperforms various\nbaselines on generated out-of-distribution problems. The code is available on\n\\href{https://github.com/microsoft/causica/tree/main/research_experiments/fip}{Github}.\n", "date": "2024-04-10", "categories": []}
{"id": "2405.19225", "title": "Synthetic Potential Outcomes and Causal Mixture Identifiability", "abstract": "  Heterogeneous data from multiple populations, sub-groups, or sources is often\nrepresented as a ``mixture model'' with a single latent class influencing all\nof the observed covariates. Heterogeneity can be resolved at multiple levels by\ngrouping populations according to different notions of similarity. This paper\nproposes grouping with respect to the causal response of an intervention or\nperturbation on the system. This definition is distinct from previous notions,\nsuch as similar covariate values (e.g. clustering) or similar correlations\nbetween covariates (e.g. Gaussian mixture models). To solve the problem, we\n``synthetically sample'' from a counterfactual distribution using higher-order\nmulti-linear moments of the observable data. To understand how these ``causal\nmixtures'' fit in with more classical notions, we develop a hierarchy of\nmixture identifiability.\n", "date": "2024-05-29", "categories": []}
{"id": "2408.17053", "title": "Estimating Conditional Average Treatment Effects via Sufficient\n  Representation Learning", "abstract": "  Estimating the conditional average treatment effects (CATE) is very important\nin causal inference and has a wide range of applications across many fields. In\nthe estimation process of CATE, the unconfoundedness assumption is typically\nrequired to ensure the identifiability of the regression problems. When\nestimating CATE using high-dimensional data, there have been many variable\nselection methods and neural network approaches based on representation\nlearning, while these methods do not provide a way to verify whether the subset\nof variables after dimensionality reduction or the learned representations\nstill satisfy the unconfoundedness assumption during the estimation process,\nwhich can lead to ineffective estimates of the treatment effects. Additionally,\nthese methods typically use data from only the treatment or control group when\nestimating the regression functions for each group. This paper proposes a novel\nneural network approach named \\textbf{CrossNet} to learn a sufficient\nrepresentation for the features, based on which we then estimate the CATE,\nwhere cross indicates that in estimating the regression functions, we used data\nfrom their own group as well as cross-utilized data from another group.\nNumerical simulations and empirical results demonstrate that our method\noutperforms the competitive approaches.\n", "date": "2024-08-30", "categories": []}
{"id": "2412.09843", "title": "Learning Structural Causal Models from Ordering: Identifiable Flow\n  Models", "abstract": "  In this study, we address causal inference when only observational data and a\nvalid causal ordering from the causal graph are available. We introduce a set\nof flow models that can recover component-wise, invertible transformation of\nexogenous variables. Our flow-based methods offer flexible model design while\nmaintaining causal consistency regardless of the number of discretization\nsteps. We propose design improvements that enable simultaneous learning of all\ncausal mechanisms and reduce abduction and prediction complexity to linear O(n)\nrelative to the number of layers, independent of the number of causal\nvariables. Empirically, we demonstrate that our method outperforms previous\nstate-of-the-art approaches and delivers consistent performance across a wide\nrange of structural causal models in answering observational, interventional,\nand counterfactual questions. Additionally, our method achieves a significant\nreduction in computational time compared to existing diffusion-based\ntechniques, making it practical for large structural causal models.\n", "date": "2024-12-13", "categories": []}
{"id": "2412.10009", "title": "Class flipping for uplift modeling and Heterogeneous Treatment Effect\n  estimation on imbalanced RCT data", "abstract": "  Uplift modeling and Heterogeneous Treatment Effect (HTE) estimation aim at\npredicting the causal effect of an action, such as a medical treatment or a\nmarketing campaign on a specific individual. In this paper, we focus on data\nfrom Randomized Controlled Experiments which guarantee causal interpretation of\nthe outcomes. Class and treatment imbalance are important problems in uplift\nmodeling/HTE, but classical undersampling or oversampling based approaches are\nhard to apply in this case since they distort the predicted effect. Calibration\nmethods have been proposed in the past, however, they do not guarantee correct\npredictions. In this work, we propose an approach alternative to undersampling,\nbased on flipping the class value of selected records. We show that the\nproposed approach does not distort the predicted effect and does not require\ncalibration. The method is especially useful for models based on class variable\ntransformation (modified outcome models). We address those models separately,\ndesigning a transformation scheme which guarantees correct predictions and\naddresses also the problem of treatment imbalance which is especially important\nfor those models. Experiments fully confirm our theoretical results.\nAdditionally, we demonstrate that our method is a viable alternative also for\nstandard classification problems.\n", "date": "2024-12-13", "categories": []}
{"id": "2407.18607", "title": "Using GPT-4 to guide causal machine learning", "abstract": "  Since its introduction to the public, ChatGPT has had an unprecedented\nimpact. While some experts praised AI advancements and highlighted their\npotential risks, others have been critical about the accuracy and usefulness of\nLarge Language Models (LLMs). In this paper, we are interested in the ability\nof LLMs to identify causal relationships. We focus on the well-established\nGPT-4 (Turbo) and evaluate its performance under the most restrictive\nconditions, by isolating its ability to infer causal relationships based solely\non the variable labels without being given any other context by humans,\ndemonstrating the minimum level of effectiveness one can expect when it is\nprovided with label-only information. We show that questionnaire participants\njudge the GPT-4 graphs as the most accurate in the evaluated categories,\nclosely followed by knowledge graphs constructed by domain experts, with causal\nMachine Learning (ML) far behind. We use these results to highlight the\nimportant limitation of causal ML, which often produces causal graphs that\nviolate common sense, affecting trust in them. However, we show that pairing\nGPT-4 with causal ML overcomes this limitation, resulting in graphical\nstructures learnt from real data that align more closely with those identified\nby domain experts, compared to structures learnt by causal ML alone. Overall,\nour findings suggest that despite GPT-4 not being explicitly designed to reason\ncausally, it can still be a valuable tool for causal representation, as it\nimproves the causal discovery process of causal ML algorithms that are designed\nto do just that.\n", "date": "2024-07-26", "categories": []}
{"id": "2309.10211", "title": "Loop Polarity Analysis to Avoid Underspecification in Deep Learning", "abstract": "  Deep learning is a powerful set of techniques for detecting complex patterns\nin data. However, when the causal structure of that process is underspecified,\ndeep learning models can be brittle, lacking robustness to shifts in the\ndistribution of the data-generating process. In this paper, we turn to loop\npolarity analysis as a tool for specifying the causal structure of a\ndata-generating process, in order to encode a more robust understanding of the\nrelationship between system structure and system behavior within the deep\nlearning pipeline. We use simulated epidemic data based on an SIR model to\ndemonstrate how measuring the polarity of the different feedback loops that\ncompose a system can lead to more robust inferences on the part of neural\nnetworks, improving the out-of-distribution performance of a deep learning\nmodel and infusing a system-dynamics-inspired approach into the machine\nlearning development pipeline.\n", "date": "2023-09-18", "categories": []}
{"id": "2412.06810", "title": "I See, Therefore I Do: Estimating Causal Effects for Image Treatments", "abstract": "  Causal effect estimation under observational studies is challenging due to\nthe lack of ground truth data and treatment assignment bias. Though various\nmethods exist in literature for addressing this problem, most of them ignore\nmulti-dimensional treatment information by considering it as scalar, either\ncontinuous or discrete. Recently, certain works have demonstrated the utility\nof this rich yet complex treatment information into the estimation process,\nresulting in better causal effect estimation. However, these works have been\ndemonstrated on either graphs or textual treatments. There is a notable gap in\nexisting literature in addressing higher dimensional data such as images that\nhas a wide variety of applications. In this work, we propose a model named NICE\n(Network for Image treatments Causal effect Estimation), for estimating\nindividual causal effects when treatments are images. NICE demonstrates an\neffective way to use the rich multidimensional information present in image\ntreatments that helps in obtaining improved causal effect estimates. To\nevaluate the performance of NICE, we propose a novel semi-synthetic data\nsimulation framework that generates potential outcomes when images serve as\ntreatments. Empirical results on these datasets, under various setups including\nthe zero-shot case, demonstrate that NICE significantly outperforms existing\nmodels that incorporate treatment information for causal effect estimation.\n", "date": "2024-11-28", "categories": []}
{"id": "2401.13009", "title": "Comparative Study of Causal Discovery Methods for Cyclic Models with\n  Hidden Confounders", "abstract": "  Nowadays, the need for causal discovery is ubiquitous. A better understanding\nof not just the stochastic dependencies between parts of a system, but also the\nactual cause-effect relations, is essential for all parts of science. Thus, the\nneed for reliable methods to detect causal directions is growing constantly. In\nthe last 50 years, many causal discovery algorithms have emerged, but most of\nthem are applicable only under the assumption that the systems have no feedback\nloops and that they are causally sufficient, i.e. that there are no unmeasured\nsubsystems that can affect multiple measured variables. This is unfortunate\nsince those restrictions can often not be presumed in practice. Feedback is an\nintegral feature of many processes, and real-world systems are rarely\ncompletely isolated and fully measured. Fortunately, in recent years, several\ntechniques, that can cope with cyclic, causally insufficient systems, have been\ndeveloped. And with multiple methods available, a practical application of\nthose algorithms now requires knowledge of the respective strengths and\nweaknesses. Here, we focus on the problem of causal discovery for sparse linear\nmodels which are allowed to have cycles and hidden confounders. We have\nprepared a comprehensive and thorough comparative study of four causal\ndiscovery techniques: two versions of the LLC method [10] and two variants of\nthe ASP-based algorithm [11]. The evaluation investigates the performance of\nthose techniques for various experiments with multiple interventional setups\nand different dataset sizes.\n", "date": "2024-01-23", "categories": []}
{"id": "2402.09236", "title": "Learning Interpretable Concepts: Unifying Causal Representation Learning\n  and Foundation Models", "abstract": "  To build intelligent machine learning systems, there are two broad\napproaches. One approach is to build inherently interpretable models, as\nendeavored by the growing field of causal representation learning. The other\napproach is to build highly-performant foundation models and then invest\nefforts into understanding how they work. In this work, we relate these two\napproaches and study how to learn human-interpretable concepts from data.\nWeaving together ideas from both fields, we formally define a notion of\nconcepts and show that they can be provably recovered from diverse data.\nExperiments on synthetic data and large language models show the utility of our\nunified approach.\n", "date": "2024-02-14", "categories": []}
{"id": "2403.14713", "title": "Auditing Fairness under Unobserved Confounding", "abstract": "  Many definitions of fairness or inequity involve unobservable causal\nquantities that cannot be directly estimated without strong assumptions. For\ninstance, it is particularly difficult to estimate notions of fairness that\nrely on hard-to-measure concepts such as risk (e.g., quantifying whether\npatients at the same risk level have equal probability of treatment, regardless\nof group membership). Such measurements of risk can be accurately obtained when\nno unobserved confounders have jointly influenced past decisions and outcomes.\nHowever, in the real world, this assumption rarely holds. In this paper, we\nshow that, surprisingly, one can still compute meaningful bounds on treatment\nrates for high-risk individuals (i.e., conditional on their true,\n\\textit{unobserved} negative outcome), even when entirely eliminating or\nrelaxing the assumption that we observe all relevant risk factors used by\ndecision makers. We use the fact that in many real-world settings (e.g., the\nrelease of a new treatment) we have data from prior to any allocation to derive\nunbiased estimates of risk. This result enables us to audit unfair outcomes of\nexisting decision-making systems in a principled manner. We demonstrate the\neffectiveness of our framework with a real-world study of Paxlovid allocation,\nprovably identifying that observed racial inequity cannot be explained by\nunobserved confounders of the same strength as important observed covariates.\n", "date": "2024-03-18", "categories": []}
{"id": "2406.02534", "title": "Enhancing predictive imaging biomarker discovery through treatment\n  effect analysis", "abstract": "  Identifying predictive covariates, which forecast individual treatment\neffectiveness, is crucial for decision-making across different disciplines such\nas personalized medicine. These covariates, referred to as biomarkers, are\nextracted from pre-treatment data, often within randomized controlled trials,\nand should be distinguished from prognostic biomarkers, which are independent\nof treatment assignment. Our study focuses on discovering predictive imaging\nbiomarkers, specific image features, by leveraging pre-treatment images to\nuncover new causal relationships. Unlike labor-intensive approaches relying on\nhandcrafted features prone to bias, we present a novel task of directly\nlearning predictive features from images. We propose an evaluation protocol to\nassess a model's ability to identify predictive imaging biomarkers and\ndifferentiate them from purely prognostic ones by employing statistical testing\nand a comprehensive analysis of image feature attribution. We explore the\nsuitability of deep learning models originally developed for estimating the\nconditional average treatment effect (CATE) for this task, which have been\nassessed primarily for their precision of CATE estimation while overlooking the\nevaluation of imaging biomarker discovery. Our proof-of-concept analysis\ndemonstrates the feasibility and potential of our approach in discovering and\nvalidating predictive imaging biomarkers from synthetic outcomes and real-world\nimage datasets. Our code is available at\n\\url{https://github.com/MIC-DKFZ/predictive_image_biomarker_analysis}.\n", "date": "2024-06-04", "categories": []}
{"id": "2412.05783", "title": "Two-way Deconfounder for Off-policy Evaluation in Causal Reinforcement\n  Learning", "abstract": "  This paper studies off-policy evaluation (OPE) in the presence of unmeasured\nconfounders. Inspired by the two-way fixed effects regression model widely used\nin the panel data literature, we propose a two-way unmeasured confounding\nassumption to model the system dynamics in causal reinforcement learning and\ndevelop a two-way deconfounder algorithm that devises a neural tensor network\nto simultaneously learn both the unmeasured confounders and the system\ndynamics, based on which a model-based estimator can be constructed for\nconsistent policy value estimation. We illustrate the effectiveness of the\nproposed estimator through theoretical results and numerical experiments.\n", "date": "2024-12-08", "categories": []}
{"id": "2412.04641", "title": "Disentangled Representation Learning for Causal Inference with\n  Instruments", "abstract": "  Latent confounders are a fundamental challenge for inferring causal effects\nfrom observational data. The instrumental variable (IV) approach is a practical\nway to address this challenge. Existing IV based estimators need a known IV or\nother strong assumptions, such as the existence of two or more IVs in the\nsystem, which limits the application of the IV approach. In this paper, we\nconsider a relaxed requirement, which assumes there is an IV proxy in the\nsystem without knowing which variable is the proxy. We propose a Variational\nAutoEncoder (VAE) based disentangled representation learning method to learn an\nIV representation from a dataset with latent confounders and then utilise the\nIV representation to obtain an unbiased estimation of the causal effect from\nthe data. Extensive experiments on synthetic and real-world data have\ndemonstrated that the proposed algorithm outperforms the existing IV based\nestimators and VAE-based estimators.\n", "date": "2024-12-05", "categories": []}
{"id": "2412.04799", "title": "Estimating the treatment effect over time under general interference\n  through deep learner integrated TMLE", "abstract": "  Understanding the effects of quarantine policies in populations with\nunderlying social networks is crucial for public health, yet most causal\ninference methods fail here due to their assumption of independent individuals.\nWe introduce DeepNetTMLE, a deep-learning-enhanced Targeted Maximum Likelihood\nEstimation (TMLE) method designed to estimate time-sensitive treatment effects\nin observational data. DeepNetTMLE mitigates bias from time-varying confounders\nunder general interference by incorporating a temporal module and domain\nadversarial training to build intervention-invariant representations. This\nprocess removes associations between current treatments and historical\nvariables, while the targeting step maintains the bias-variance trade-off,\nenhancing the reliability of counterfactual predictions. Using simulations of a\n``Susceptible-Infected-Recovered'' model with varied quarantine coverages, we\nshow that DeepNetTMLE achieves lower bias and more precise confidence intervals\nin counterfactual estimates, enabling optimal quarantine recommendations within\nbudget constraints, surpassing state-of-the-art methods.\n", "date": "2024-12-06", "categories": []}
{"id": "2401.10101", "title": "Bayesian Networks for Causal Analysis in Socioecological Systems", "abstract": "  Causal and counterfactual reasoning are emerging directions in data science\nthat allow us to reason about hypothetical scenarios. This is particularly\nuseful in fields like environmental and ecological sciences, where\ninterventional data are usually not available. Structural causal models are\nprobabilistic models for causal analysis that simplify this kind of reasoning\ndue to their graphical representation. They can be regarded as extensions of\nthe so-called Bayesian networks, a well known modeling tool commonly used in\nenvironmental and ecological problems. The main contribution of this paper is\nto analyze the relations of necessity and sufficiency between the variables of\na socioecological system using counterfactual reasoning with Bayesian networks.\nIn particular, we consider a case study involving socioeconomic factors and\nland-uses in southern Spain. In addition, this paper aims to be a coherent\noverview of the fundamental concepts for applying counterfactual reasoning, so\nthat environmental researchers with a background in Bayesian networks can\neasily take advantage of the structural causal model formalism.\n", "date": "2024-01-18", "categories": []}
{"id": "2406.05937", "title": "Linear Causal Representation Learning from Unknown Multi-node\n  Interventions", "abstract": "  Despite the multifaceted recent advances in interventional causal\nrepresentation learning (CRL), they primarily focus on the stylized assumption\nof single-node interventions. This assumption is not valid in a wide range of\napplications, and generally, the subset of nodes intervened in an\ninterventional environment is fully unknown. This paper focuses on\ninterventional CRL under unknown multi-node (UMN) interventional environments\nand establishes the first identifiability results for general latent causal\nmodels (parametric or nonparametric) under stochastic interventions (soft or\nhard) and linear transformation from the latent to observed space.\nSpecifically, it is established that given sufficiently diverse interventional\nenvironments, (i) identifiability up to ancestors is possible using only soft\ninterventions, and (ii) perfect identifiability is possible using hard\ninterventions. Remarkably, these guarantees match the best-known results for\nmore restrictive single-node interventions. Furthermore, CRL algorithms are\nalso provided that achieve the identifiability guarantees. A central step in\ndesigning these algorithms is establishing the relationships between UMN\ninterventional CRL and score functions associated with the statistical models\nof different interventional environments. Establishing these relationships also\nserves as constructive proof of the identifiability guarantees.\n", "date": "2024-06-09", "categories": []}
{"id": "2411.01295", "title": "Marginal Causal Flows for Validation and Inference", "abstract": "  Investigating the marginal causal effect of an intervention on an outcome\nfrom complex data remains challenging due to the inflexibility of employed\nmodels and the lack of complexity in causal benchmark datasets, which often\nfail to reproduce intricate real-world data patterns. In this paper we\nintroduce Frugal Flows, a novel likelihood-based machine learning model that\nuses normalising flows to flexibly learn the data-generating process, while\nalso directly inferring the marginal causal quantities from observational data.\nWe propose that these models are exceptionally well suited for generating\nsynthetic data to validate causal methods. They can create synthetic datasets\nthat closely resemble the empirical dataset, while automatically and exactly\nsatisfying a user-defined average treatment effect. To our knowledge, Frugal\nFlows are the first generative model to both learn flexible data\nrepresentations and also exactly parameterise quantities such as the average\ntreatment effect and the degree of unobserved confounding. We demonstrate the\nabove with experiments on both simulated and real-world datasets.\n", "date": "2024-11-02", "categories": []}
{"id": "2412.03913", "title": "Graph Disentangle Causal Model: Enhancing Causal Inference in Networked\n  Observational Data", "abstract": "  Estimating individual treatment effects (ITE) from observational data is a\ncritical task across various domains. However, many existing works on ITE\nestimation overlook the influence of hidden confounders, which remain\nunobserved at the individual unit level. To address this limitation,\nresearchers have utilized graph neural networks to aggregate neighbors'\nfeatures to capture the hidden confounders and mitigate confounding bias by\nminimizing the discrepancy of confounder representations between the treated\nand control groups. Despite the success of these approaches, practical\nscenarios often treat all features as confounders and involve substantial\ndifferences in feature distributions between the treated and control groups.\nConfusing the adjustment and confounder and enforcing strict balance on the\nconfounder representations could potentially undermine the effectiveness of\noutcome prediction. To mitigate this issue, we propose a novel framework called\nthe \\textit{Graph Disentangle Causal model} (GDC) to conduct ITE estimation in\nthe network setting. GDC utilizes a causal disentangle module to separate unit\nfeatures into adjustment and confounder representations. Then we design a graph\naggregation module consisting of three distinct graph aggregators to obtain\nadjustment, confounder, and counterfactual confounder representations. Finally,\na causal constraint module is employed to enforce the disentangled\nrepresentations as true causal factors. The effectiveness of our proposed\nmethod is demonstrated by conducting comprehensive experiments on two networked\ndatasets.\n", "date": "2024-12-05", "categories": []}
{"id": "2412.04285", "title": "Deep Causal Inference for Point-referenced Spatial Data with Continuous\n  Treatments", "abstract": "  Causal reasoning is often challenging with spatial data, particularly when\nhandling high-dimensional inputs. To address this, we propose a neural network\n(NN) based framework integrated with an approximate Gaussian process to manage\nspatial interference and unobserved confounding. Additionally, we adopt a\ngeneralized propensity-score-based approach to address partially observed\noutcomes when estimating causal effects with continuous treatments. We evaluate\nour framework using synthetic, semi-synthetic, and real-world data inferred\nfrom satellite imagery. Our results demonstrate that NN-based models\nsignificantly outperform linear spatial regression models in estimating causal\neffects. Furthermore, in real-world case studies, NN-based models offer more\nreasonable predictions of causal effects, facilitating decision-making in\nrelevant applications.\n", "date": "2024-12-05", "categories": []}
{"id": "2406.06671", "title": "Controlling Counterfactual Harm in Decision Support Systems Based on\n  Prediction Sets", "abstract": "  Decision support systems based on prediction sets help humans solve\nmulticlass classification tasks by narrowing down the set of potential label\nvalues to a subset of them, namely a prediction set, and asking them to always\npredict label values from the prediction sets. While this type of systems have\nbeen proven to be effective at improving the average accuracy of the\npredictions made by humans, by restricting human agency, they may cause\nharm$\\unicode{x2014}$a human who has succeeded at predicting the ground-truth\nlabel of an instance on their own may have failed had they used these systems.\nIn this paper, our goal is to control how frequently a decision support system\nbased on prediction sets may cause harm, by design. To this end, we start by\ncharacterizing the above notion of harm using the theoretical framework of\nstructural causal models. Then, we show that, under a natural, albeit\nunverifiable, monotonicity assumption, we can estimate how frequently a system\nmay cause harm using only predictions made by humans on their own. Further, we\nalso show that, under a weaker monotonicity assumption, which can be verified\nexperimentally, we can bound how frequently a system may cause harm again using\nonly predictions made by humans on their own. Building upon these assumptions,\nwe introduce a computational framework to design decision support systems based\non prediction sets that are guaranteed to cause harm less frequently than a\nuser-specified value using conformal risk control. We validate our framework\nusing real human predictions from two different human subject studies and show\nthat, in decision support systems based on prediction sets, there is a\ntrade-off between accuracy and counterfactual harm.\n", "date": "2024-06-10", "categories": []}
{"id": "2412.02900", "title": "MACAW: A Causal Generative Model for Medical Imaging", "abstract": "  Although deep learning techniques show promising results for many\nneuroimaging tasks in research settings, they have not yet found widespread use\nin clinical scenarios. One of the reasons for this problem is that many machine\nlearning models only identify correlations between the input images and the\noutputs of interest, which can lead to many practical problems, such as\nencoding of uninformative biases and reduced explainability. Thus, recent\nresearch is exploring if integrating a priori causal knowledge into deep\nlearning models is a potential avenue to identify these problems. This work\nintroduces a new causal generative architecture named Masked Causal Flow\n(MACAW) for neuroimaging applications. Within this context, three main\ncontributions are described. First, a novel approach that integrates complex\ncausal structures into normalizing flows is proposed. Second, counterfactual\nprediction is performed to identify the changes in effect variables associated\nwith a cause variable. Finally, an explicit Bayesian inference for\nclassification is derived and implemented, providing an inherent uncertainty\nestimation. The feasibility of the proposed method was first evaluated using\nsynthetic data and then using MRI brain data from more than 23000 participants\nof the UK biobank study. The evaluation results show that the proposed method\ncan (1) accurately encode causal reasoning and generate counterfactuals\nhighlighting the structural changes in the brain known to be associated with\naging, (2) accurately predict a subject's age from a single 2D MRI slice, and\n(3) generate new samples assuming other values for subject-specific indicators\nsuch as age, sex, and body mass index. The code for a toy dataset is available\nat the following link: https://github.com/vibujithan/macaw-2D.git.\n", "date": "2024-12-03", "categories": []}
{"id": "2312.00710", "title": "SpaCE: The Spatial Confounding Environment", "abstract": "  Spatial confounding poses a significant challenge in scientific studies\ninvolving spatial data, where unobserved spatial variables can influence both\ntreatment and outcome, possibly leading to spurious associations. To address\nthis problem, we introduce SpaCE: The Spatial Confounding Environment, the\nfirst toolkit to provide realistic benchmark datasets and tools for\nsystematically evaluating causal inference methods designed to alleviate\nspatial confounding. Each dataset includes training data, true counterfactuals,\na spatial graph with coordinates, and smoothness and confounding scores\ncharacterizing the effect of a missing spatial confounder. It also includes\nrealistic semi-synthetic outcomes and counterfactuals, generated using\nstate-of-the-art machine learning ensembles, following best practices for\ncausal inference benchmarks. The datasets cover real treatment and covariates\nfrom diverse domains, including climate, health and social sciences. SpaCE\nfacilitates an automated end-to-end pipeline, simplifying data loading,\nexperimental setup, and evaluating machine learning and causal inference\nmodels. The SpaCE project provides several dozens of datasets of diverse sizes\nand spatial complexity. It is publicly available as a Python package,\nencouraging community feedback and contributions.\n", "date": "2023-12-01", "categories": []}
{"id": "2406.08666", "title": "Interventional Causal Discovery in a Mixture of DAGs", "abstract": "  Causal interactions among a group of variables are often modeled by a single\ncausal graph. In some domains, however, these interactions are best described\nby multiple co-existing causal graphs, e.g., in dynamical systems or genomics.\nThis paper addresses the hitherto unknown role of interventions in learning\ncausal interactions among variables governed by a mixture of causal systems,\neach modeled by one directed acyclic graph (DAG). Causal discovery from\nmixtures is fundamentally more challenging than single-DAG causal discovery.\nTwo major difficulties stem from (i)~an inherent uncertainty about the\nskeletons of the component DAGs that constitute the mixture and (ii)~possibly\ncyclic relationships across these component DAGs. This paper addresses these\nchallenges and aims to identify edges that exist in at least one component DAG\nof the mixture, referred to as the true edges. First, it establishes matching\nnecessary and sufficient conditions on the size of interventions required to\nidentify the true edges. Next, guided by the necessity results, an adaptive\nalgorithm is designed that learns all true edges using $O(n^2)$ interventions,\nwhere $n$ is the number of nodes. Remarkably, the size of the interventions is\noptimal if the underlying mixture model does not contain cycles across its\ncomponents. More generally, the gap between the intervention size used by the\nalgorithm and the optimal size is quantified. It is shown to be bounded by the\ncyclic complexity number of the mixture model, defined as the size of the\nminimal intervention that can break the cycles in the mixture, which is upper\nbounded by the number of cycles among the ancestors of a node.\n", "date": "2024-06-12", "categories": []}
{"id": "2410.22472", "title": "Learning Identifiable Factorized Causal Representations of Cellular\n  Responses", "abstract": "  The study of cells and their responses to genetic or chemical perturbations\npromises to accelerate the discovery of therapeutic targets. However, designing\nadequate and insightful models for such data is difficult because the response\nof a cell to perturbations essentially depends on its biological context (e.g.,\ngenetic background or cell type). For example, while discovering therapeutic\ntargets, one may want to enrich for drugs that specifically target a certain\ncell type. This challenge emphasizes the need for methods that explicitly take\ninto account potential interactions between drugs and contexts. Towards this\ngoal, we propose a novel Factorized Causal Representation (FCR) learning method\nthat reveals causal structure in single-cell perturbation data from several\ncell lines. Based on the framework of identifiable deep generative models, FCR\nlearns multiple cellular representations that are disentangled, comprised of\ncovariate-specific ($\\mathbf{z}_x$), treatment-specific ($\\mathbf{z}_{t}$), and\ninteraction-specific ($\\mathbf{z}_{tx}$) blocks. Based on recent advances in\nnon-linear ICA theory, we prove the component-wise identifiability of\n$\\mathbf{z}_{tx}$ and block-wise identifiability of $\\mathbf{z}_t$ and\n$\\mathbf{z}_x$. Then, we present our implementation of FCR, and empirically\ndemonstrate that it outperforms state-of-the-art baselines in various tasks\nacross four single-cell datasets.\n", "date": "2024-10-29", "categories": []}
{"id": "2412.02439", "title": "Nature versus nurture in galaxy formation: the effect of environment on\n  star formation with causal machine learning", "abstract": "  Understanding how galaxies form and evolve is at the heart of modern\nastronomy. With the advent of large-scale surveys and simulations, remarkable\nprogress has been made in the last few decades. Despite this, the physical\nprocesses behind the phenomena, and particularly their importance, remain far\nfrom known, as correlations have primarily been established rather than the\nunderlying causality. We address this challenge by applying the causal\ninference framework. Specifically, we tackle the fundamental open question of\nwhether galaxy formation and evolution depends more on nature (i.e., internal\nprocesses) or nurture (i.e., external processes), by estimating the causal\neffect of environment on star-formation rate in the IllustrisTNG simulations.\nTo do so, we develop a comprehensive causal model and employ cutting-edge\ntechniques from epidemiology to overcome the long-standing problem of\ndisentangling nature and nurture. We find that the causal effect is negative\nand substantial, with environment suppressing the SFR by a maximal factor of\n$\\sim100$. While the overall effect at $z=0$ is negative, in the early\nuniverse, environment is discovered to have a positive impact, boosting star\nformation by a factor of $\\sim10$ at $z\\sim1$ and by even greater amounts at\nhigher redshifts. Furthermore, we show that: (i) nature also plays an important\nrole, as ignoring it underestimates the causal effect in intermediate-density\nenvironments by a factor of $\\sim2$, (ii) controlling for the stellar mass at a\nsnapshot in time, as is common in the literature, is not only insufficient to\ndisentangle nature and nurture but actually has an adverse effect, though (iii)\nstellar mass is an adequate proxy of the effects of nature. Finally, this work\nmay prove a useful blueprint for extracting causal insights in other fields\nthat deal with dynamical systems with closed feedback loops, such as the\nEarth's climate.\n", "date": "2024-12-03", "categories": []}
{"id": "2303.03092", "title": "Environment Invariant Linear Least Squares", "abstract": "  This paper considers a multi-environment linear regression model in which\ndata from multiple experimental settings are collected. The joint distribution\nof the response variable and covariates may vary across different environments,\nyet the conditional expectations of $y$ given the unknown set of important\nvariables are invariant. Such a statistical model is related to the problem of\nendogeneity, causal inference, and transfer learning. The motivation behind it\nis illustrated by how the goals of prediction and attribution are inherent in\nestimating the true parameter and the important variable set. We construct a\nnovel environment invariant linear least squares (EILLS) objective function, a\nmulti-environment version of linear least-squares regression that leverages the\nabove conditional expectation invariance structure and heterogeneity among\ndifferent environments to determine the true parameter. Our proposed method is\napplicable without any additional structural knowledge and can identify the\ntrue parameter under a near-minimal identification condition. We establish\nnon-asymptotic $\\ell_2$ error bounds on the estimation error for the EILLS\nestimator in the presence of spurious variables. Moreover, we further show that\nthe $\\ell_0$ penalized EILLS estimator can achieve variable selection\nconsistency in high-dimensional regimes. These non-asymptotic results\ndemonstrate the sample efficiency of the EILLS estimator and its capability to\ncircumvent the curse of endogeneity in an algorithmic manner without any prior\nstructural knowledge. To the best of our knowledge, this paper is the first to\nrealize statistically efficient invariance learning in the general linear\nmodel.\n", "date": "2023-03-06", "categories": []}
{"id": "2412.01674", "title": "Causal Discovery by Interventions via Integer Programming", "abstract": "  Causal discovery is essential across various scientific fields to uncover\ncausal structures within data. Traditional methods relying on observational\ndata have limitations due to confounding variables. This paper presents an\noptimization-based approach using integer programming (IP) to design minimal\nintervention sets that ensure causal structure identifiability. Our method\nprovides exact and modular solutions that can be adjusted to different\nexperimental settings and constraints. We demonstrate its effectiveness through\ncomparative analysis across different settings, demonstrating its applicability\nand robustness.\n", "date": "2024-12-02", "categories": []}
{"id": "2310.15580", "title": "Identifiable Latent Polynomial Causal Models Through the Lens of Change", "abstract": "  Causal representation learning aims to unveil latent high-level causal\nrepresentations from observed low-level data. One of its primary tasks is to\nprovide reliable assurance of identifying these latent causal models, known as\nidentifiability. A recent breakthrough explores identifiability by leveraging\nthe change of causal influences among latent causal variables across multiple\nenvironments \\citep{liu2022identifying}. However, this progress rests on the\nassumption that the causal relationships among latent causal variables adhere\nstrictly to linear Gaussian models. In this paper, we extend the scope of\nlatent causal models to involve nonlinear causal relationships, represented by\npolynomial models, and general noise distributions conforming to the\nexponential family. Additionally, we investigate the necessity of imposing\nchanges on all causal parameters and present partial identifiability results\nwhen part of them remains unchanged. Further, we propose a novel empirical\nestimation method, grounded in our theoretical finding, that enables learning\nconsistent latent causal representations. Our experimental results, obtained\nfrom both synthetic and real-world data, validate our theoretical contributions\nconcerning identifiability and consistency.\n", "date": "2023-10-24", "categories": []}
{"id": "2411.19245", "title": "Contrastive representations of high-dimensional, structured treatments", "abstract": "  Estimating causal effects is vital for decision making. In standard causal\neffect estimation, treatments are usually binary- or continuous-valued.\nHowever, in many important real-world settings, treatments can be structured,\nhigh-dimensional objects, such as text, video, or audio. This provides a\nchallenge to traditional causal effect estimation. While leveraging the shared\nstructure across different treatments can help generalize to unseen treatments\nat test time, we show in this paper that using such structure blindly can lead\nto biased causal effect estimation. We address this challenge by devising a\nnovel contrastive approach to learn a representation of the high-dimensional\ntreatments, and prove that it identifies underlying causal factors and discards\nnon-causally relevant factors. We prove that this treatment representation\nleads to unbiased estimates of the causal effect, and empirically validate and\nbenchmark our results on synthetic and real-world datasets.\n", "date": "2024-11-28", "categories": []}
{"id": "2411.19556", "title": "Differentiable Causal Discovery For Latent Hierarchical Causal Models", "abstract": "  Discovering causal structures with latent variables from observational data\nis a fundamental challenge in causal discovery. Existing methods often rely on\nconstraint-based, iterative discrete searches, limiting their scalability to\nlarge numbers of variables. Moreover, these methods frequently assume linearity\nor invertibility, restricting their applicability to real-world scenarios. We\npresent new theoretical results on the identifiability of nonlinear latent\nhierarchical causal models, relaxing previous assumptions in literature about\nthe deterministic nature of latent variables and exogenous noise. Building on\nthese insights, we develop a novel differentiable causal discovery algorithm\nthat efficiently estimates the structure of such models. To the best of our\nknowledge, this is the first work to propose a differentiable causal discovery\nmethod for nonlinear latent hierarchical models. Our approach outperforms\nexisting methods in both accuracy and scalability. We demonstrate its practical\nutility by learning interpretable hierarchical latent structures from\nhigh-dimensional image data and demonstrate its effectiveness on downstream\ntasks.\n", "date": "2024-11-29", "categories": []}
{"id": "2410.06163", "title": "Markov Equivalence and Consistency in Differentiable Structure Learning", "abstract": "  Existing approaches to differentiable structure learning of directed acyclic\ngraphs (DAGs) rely on strong identifiability assumptions in order to guarantee\nthat global minimizers of the acyclicity-constrained optimization problem\nidentifies the true DAG. Moreover, it has been observed empirically that the\noptimizer may exploit undesirable artifacts in the loss function. We explain\nand remedy these issues by studying the behavior of differentiable\nacyclicity-constrained programs under general likelihoods with multiple global\nminimizers. By carefully regularizing the likelihood, it is possible to\nidentify the sparsest model in the Markov equivalence class, even in the\nabsence of an identifiable parametrization. We first study the Gaussian case in\ndetail, showing how proper regularization of the likelihood defines a score\nthat identifies the sparsest model. Assuming faithfulness, it also recovers the\nMarkov equivalence class. These results are then generalized to general models\nand likelihoods, where the same claims hold. These theoretical results are\nvalidated empirically, showing how this can be done using standard\ngradient-based optimizers, thus paving the way for differentiable structure\nlearning under general models and losses.\n", "date": "2024-10-08", "categories": []}
{"id": "2411.17774", "title": "Leaning Time-Varying Instruments for Identifying Causal Effects in\n  Time-Series Data", "abstract": "  Querying causal effects from time-series data is important across various\nfields, including healthcare, economics, climate science, and epidemiology.\nHowever, this task becomes complex in the existence of time-varying latent\nconfounders, which affect both treatment and outcome variables over time and\ncan introduce bias in causal effect estimation. Traditional instrumental\nvariable (IV) methods are limited in addressing such complexities due to the\nneed for predefined IVs or strong assumptions that do not hold in dynamic\nsettings. To tackle these issues, we develop a novel Time-varying Conditional\nInstrumental Variables (CIV) for Debiasing causal effect estimation, referred\nto as TDCIV. TDCIV leverages Long Short-Term Memory (LSTM) and Variational\nAutoencoder (VAE) models to disentangle and learn the representations of\ntime-varying CIV and its conditioning set from proxy variables without prior\nknowledge. Under the assumptions of the Markov property and availability of\nproxy variables, we theoretically establish the validity of these learned\nrepresentations for addressing the biases from time-varying latent confounders,\nthus enabling accurate causal effect estimation. Our proposed TDCIV is the\nfirst to effectively learn time-varying CIV and its associated conditioning set\nwithout relying on domain-specific knowledge.\n", "date": "2024-11-26", "categories": []}
{"id": "2411.17937", "title": "Spatio-temporal Causal Learning for Streamflow Forecasting", "abstract": "  Streamflow plays an essential role in the sustainable planning and management\nof national water resources. Traditional hydrologic modeling approaches\nsimulate streamflow by establishing connections across multiple physical\nprocesses, such as rainfall and runoff. These data, inherently connected both\nspatially and temporally, possess intrinsic causal relations that can be\nleveraged for robust and accurate forecasting. Recently, spatio-temporal graph\nneural networks (STGNNs) have been adopted, excelling in various domains, such\nas urban traffic management, weather forecasting, and pandemic control, and\nthey also promise advances in streamflow management. However, learning causal\nrelationships directly from vast observational data is theoretically and\ncomputationally challenging. In this study, we employ a river flow graph as\nprior knowledge to facilitate the learning of the causal structure and then use\nthe learned causal graph to predict streamflow at targeted sites. The proposed\nmodel, Causal Streamflow Forecasting (CSF) is tested in a real-world study in\nthe Brazos River basin in Texas. Our results demonstrate that our method\noutperforms regular spatio-temporal graph neural networks and achieves higher\ncomputational efficiency compared to traditional simulation methods. By\neffectively integrating river flow graphs with STGNNs, this research offers a\nnovel approach to streamflow prediction, showcasing the potential of combining\nadvanced neural network techniques with domain-specific knowledge for enhanced\nperformance in hydrologic modeling.\n", "date": "2024-11-26", "categories": []}
{"id": "2411.17989", "title": "Regularized Multi-LLMs Collaboration for Enhanced Score-based Causal\n  Discovery", "abstract": "  As the significance of understanding the cause-and-effect relationships among\nvariables increases in the development of modern systems and algorithms,\nlearning causality from observational data has become a preferred and efficient\napproach over conducting randomized control trials. However, purely\nobservational data could be insufficient to reconstruct the true causal graph.\nConsequently, many researchers tried to utilise some form of prior knowledge to\nimprove causal discovery process. In this context, the impressive capabilities\nof large language models (LLMs) have emerged as a promising alternative to the\ncostly acquisition of prior expert knowledge. In this work, we further explore\nthe potential of using LLMs to enhance causal discovery approaches,\nparticularly focusing on score-based methods, and we propose a general\nframework to utilise the capacity of not only one but multiple LLMs to augment\nthe discovery process.\n", "date": "2024-11-27", "categories": []}
{"id": "2410.00509", "title": "Learning Personalized Treatment Decisions in Precision Medicine:\n  Disentangling Treatment Assignment Bias in Counterfactual Outcome Prediction\n  and Biomarker Identification", "abstract": "  Precision medicine has the potential to tailor treatment decisions to\nindividual patients using machine learning (ML) and artificial intelligence\n(AI), but it faces significant challenges due to complex biases in clinical\nobservational data and the high-dimensional nature of biological data. This\nstudy models various types of treatment assignment biases using mutual\ninformation and investigates their impact on ML models for counterfactual\nprediction and biomarker identification. Unlike traditional counterfactual\nbenchmarks that rely on fixed treatment policies, our work focuses on modeling\ndifferent characteristics of the underlying observational treatment policy in\ndistinct clinical settings. We validate our approach through experiments on toy\ndatasets, semi-synthetic tumor cancer genome atlas (TCGA) data, and real-world\nbiological outcomes from drug and CRISPR screens. By incorporating empirical\nbiological mechanisms, we create a more realistic benchmark that reflects the\ncomplexities of real-world data. Our analysis reveals that different biases\nlead to varying model performances, with some biases, especially those\nunrelated to outcome mechanisms, having minimal effect on prediction accuracy.\nThis highlights the crucial need to account for specific biases in clinical\nobservational data in counterfactual ML model development, ultimately enhancing\nthe personalization of treatment decisions in precision medicine.\n", "date": "2024-10-01", "categories": []}
{"id": "2411.15888", "title": "Evaluating Large Language Models for Causal Modeling", "abstract": "  In this paper, we consider the process of transforming causal domain\nknowledge into a representation that aligns more closely with guidelines from\ncausal data science. To this end, we introduce two novel tasks related to\ndistilling causal domain knowledge into causal variables and detecting\ninteraction entities using LLMs. We have determined that contemporary LLMs are\nhelpful tools for conducting causal modeling tasks in collaboration with human\nexperts, as they can provide a wider perspective. Specifically, LLMs, such as\nGPT-4-turbo and Llama3-70b, perform better in distilling causal domain\nknowledge into causal variables compared to sparse expert models, such as\nMixtral-8x22b. On the contrary, sparse expert models such as Mixtral-8x22b\nstand out as the most effective in identifying interaction entities. Finally,\nwe highlight the dependency between the domain where the entities are generated\nand the performance of the chosen LLM for causal modeling.\n", "date": "2024-11-24", "categories": []}
{"id": "2411.14665", "title": "Double Machine Learning for Adaptive Causal Representation in\n  High-Dimensional Data", "abstract": "  Adaptive causal representation learning from observational data is presented,\nintegrated with an efficient sample splitting technique within the\nsemiparametric estimating equation framework. The support points sample\nsplitting (SPSS), a subsampling method based on energy distance, is employed\nfor efficient double machine learning (DML) in causal inference. The support\npoints are selected and split as optimal representative points of the full raw\ndata in a random sample, in contrast to the traditional random splitting, and\nproviding an optimal sub-representation of the underlying data generating\ndistribution. They offer the best representation of a full big dataset, whereas\nthe unit structural information of the underlying distribution via the\ntraditional random data splitting is most likely not preserved. Three machine\nlearning estimators were adopted for causal inference, support vector machine\n(SVM), deep learning (DL), and a hybrid super learner (SL) with deep learning\n(SDL), using SPSS. A comparative study is conducted between the proposed SVM,\nDL, and SDL representations using SPSS, and the benchmark results from\nChernozhukov et al. (2018), which employed random forest, neural network, and\nregression trees with a random k-fold cross-fitting technique on the\n401(k)-pension plan real data. The simulations show that DL with SPSS and the\nhybrid methods of DL and SL with SPSS outperform SVM with SPSS in terms of\ncomputational efficiency and the estimation quality, respectively.\n", "date": "2024-11-22", "categories": []}
{"id": "2411.14341", "title": "Logarithmic Neyman Regret for Adaptive Estimation of the Average\n  Treatment Effect", "abstract": "  Estimation of the Average Treatment Effect (ATE) is a core problem in causal\ninference with strong connections to Off-Policy Evaluation in Reinforcement\nLearning. This paper considers the problem of adaptively selecting the\ntreatment allocation probability in order to improve estimation of the ATE. The\nmajority of prior work on adaptive ATE estimation focus on asymptotic\nguarantees, and in turn overlooks important practical considerations such as\nthe difficulty of learning the optimal treatment allocation as well as\nhyper-parameter selection. Existing non-asymptotic methods are limited by poor\nempirical performance and exponential scaling of the Neyman regret with respect\nto problem parameters. In order to address these gaps, we propose and analyze\nthe Clipped Second Moment Tracking (ClipSMT) algorithm, a variant of an\nexisting algorithm with strong asymptotic optimality guarantees, and provide\nfinite sample bounds on its Neyman regret. Our analysis shows that ClipSMT\nachieves exponential improvements in Neyman regret on two fronts: improving the\ndependence on $T$ from $O(\\sqrt{T})$ to $O(\\log T)$, as well as reducing the\nexponential dependence on problem parameters to a polynomial dependence.\nFinally, we conclude with simulations which show the marked improvement of\nClipSMT over existing approaches.\n", "date": "2024-11-21", "categories": []}
{"id": "2210.05558", "title": "Causal and Counterfactual Views of Missing Data Models", "abstract": "  It is often said that the fundamental problem of causal inference is a\nmissing data problem -- the comparison of responses to two hypothetical\ntreatment assignments is made difficult because for every experimental unit\nonly one potential response is observed. In this paper, we consider the\nimplications of the converse view: that missing data problems are a form of\ncausal inference. We make explicit how the missing data problem of recovering\nthe complete data law from the observed law can be viewed as identification of\na joint distribution over counterfactual variables corresponding to values had\nwe (possibly contrary to fact) been able to observe them. Drawing analogies\nwith causal inference, we show how identification assumptions in missing data\ncan be encoded in terms of graphical models defined over counterfactual and\nobserved variables. We review recent results in missing data identification\nfrom this viewpoint. In doing so, we note interesting similarities and\ndifferences between missing data and causal identification theories.\n", "date": "2022-10-11", "categories": []}
{"id": "2411.12759", "title": "A Novel Approach to Eliminating Hallucinations in Large Language\n  Model-Assisted Causal Discovery", "abstract": "  The increasing use of large language models (LLMs) in causal discovery as a\nsubstitute for human domain experts highlights the need for optimal model\nselection. This paper presents the first hallucination survey of popular LLMs\nfor causal discovery. We show that hallucinations exist when using LLMs in\ncausal discovery so the choice of LLM is important. We propose using Retrieval\nAugmented Generation (RAG) to reduce hallucinations when quality data is\navailable. Additionally, we introduce a novel method employing multiple LLMs\nwith an arbiter in a debate to audit edges in causal graphs, achieving a\ncomparable reduction in hallucinations to RAG.\n", "date": "2024-11-16", "categories": []}
{"id": "2411.12786", "title": "Off-policy estimation with adaptively collected data: the power of\n  online learning", "abstract": "  We consider estimation of a linear functional of the treatment effect using\nadaptively collected data. This task finds a variety of applications including\nthe off-policy evaluation (\\textsf{OPE}) in contextual bandits, and estimation\nof the average treatment effect (\\textsf{ATE}) in causal inference. While a\ncertain class of augmented inverse propensity weighting (\\textsf{AIPW})\nestimators enjoys desirable asymptotic properties including the semi-parametric\nefficiency, much less is known about their non-asymptotic theory with\nadaptively collected data. To fill in the gap, we first establish generic upper\nbounds on the mean-squared error of the class of AIPW estimators that crucially\ndepends on a sequentially weighted error between the treatment effect and its\nestimates. Motivated by this, we also propose a general reduction scheme that\nallows one to produce a sequence of estimates for the treatment effect via\nonline learning to minimize the sequentially weighted estimation error. To\nillustrate this, we provide three concrete instantiations in (\\romannumeral 1)\nthe tabular case; (\\romannumeral 2) the case of linear function approximation;\nand (\\romannumeral 3) the case of general function approximation for the\noutcome model. We then provide a local minimax lower bound to show the\ninstance-dependent optimality of the \\textsf{AIPW} estimator using no-regret\nonline learning algorithms.\n", "date": "2024-11-19", "categories": []}
{"id": "2411.13264", "title": "Transformers with Sparse Attention for Granger Causality", "abstract": "  Temporal causal analysis means understanding the underlying causes behind\nobserved variables over time. Deep learning based methods such as transformers\nare increasingly used to capture temporal dynamics and causal relationships\nbeyond mere correlations. Recent works suggest self-attention weights of\ntransformers as a useful indicator of causal links. We leverage this to propose\na novel modification to the self-attention module to establish causal links\nbetween the variables of multivariate time-series data with varying lag\ndependencies. Our Sparse Attention Transformer captures causal relationships\nusing a two-fold approach - performing temporal attention first followed by\nattention between the variables across the time steps masking them individually\nto compute Granger Causality indices. The key novelty in our approach is the\nability of the model to assert importance and pick the most significant past\ntime instances for its prediction task against manually feeding a fixed time\nlag value. We demonstrate the effectiveness of our approach via extensive\nexperimentation on several synthetic benchmark datasets. Furthermore, we\ncompare the performance of our model with the traditional Vector Autoregression\nbased Granger Causality method that assumes fixed lag length.\n", "date": "2024-11-20", "categories": []}
{"id": "2406.07005", "title": "DecoR: Deconfounding Time Series with Robust Regression", "abstract": "  Causal inference on time series data is a challenging problem, especially in\nthe presence of unobserved confounders. This work focuses on estimating the\ncausal effect between two time series that are confounded by a third,\nunobserved time series. Assuming spectral sparsity of the confounder, we show\nhow in the frequency domain this problem can be framed as an adversarial\noutlier problem. We introduce Deconfounding by Robust regression (DecoR), a\nnovel approach that estimates the causal effect using robust linear regression\nin the frequency domain. Considering two different robust regression\ntechniques, we first improve existing bounds on the estimation error for such\ntechniques. Crucially, our results do not require distributional assumptions on\nthe covariates. We can therefore use them in time series settings. Applying\nthese results to DecoR, we prove, under suitable assumptions, upper bounds for\nthe estimation error of DecoR that imply consistency. We demonstrate DecoR's\neffectiveness through experiments on both synthetic and real-world data from\nEarth system science. The simulation experiments furthermore suggest that DecoR\nis robust with respect to model misspecification.\n", "date": "2024-06-11", "categories": []}
{"id": "2411.11256", "title": "Progressive Generalization Risk Reduction for Data-Efficient Causal\n  Effect Estimation", "abstract": "  Causal effect estimation (CEE) provides a crucial tool for predicting the\nunobserved counterfactual outcome for an entity. As CEE relaxes the requirement\nfor ``perfect'' counterfactual samples (e.g., patients with identical\nattributes and only differ in treatments received) that are impractical to\nobtain and can instead operate on observational data, it is usually used in\nhigh-stake domains like medical treatment effect prediction. Nevertheless, in\nthose high-stake domains, gathering a decently sized, fully labelled\nobservational dataset remains challenging due to hurdles associated with costs,\nethics, expertise and time needed, etc., of which medical treatment surveys are\na typical example. Consequently, if the training dataset is small in scale, low\ngeneralization risks can hardly be achieved on any CEE algorithms.\n  Unlike existing CEE methods that assume the constant availability of a\ndataset with abundant samples, in this paper, we study a more realistic CEE\nsetting where the labelled data samples are scarce at the beginning, while more\ncan be gradually acquired over the course of training -- assuredly under a\nlimited budget considering their expensive nature. Then, the problem naturally\ncomes down to actively selecting the best possible samples to be labelled,\ne.g., identifying the next subset of patients to conduct the treatment survey.\nHowever, acquiring quality data for reducing the CEE risk under limited\nlabelling budgets remains under-explored until now. To fill the gap, we\ntheoretically analyse the generalization risk from an intriguing perspective of\nprogressively shrinking its upper bound, and develop a principled label\nacquisition pipeline exclusively for CEE tasks. With our analysis, we propose\nthe Model Agnostic Causal Active Learning (MACAL) algorithm for batch-wise\nlabel acquisition, which aims to reduce both the CEE model's uncertainty and\nthe post-acquisition ...\n", "date": "2024-11-18", "categories": []}
{"id": "2411.10008", "title": "Graph-based Complexity for Causal Effect by Empirical Plug-in", "abstract": "  This paper focuses on the computational complexity of computing empirical\nplug-in estimates for causal effect queries. Given a causal graph and\nobservational data, any identifiable causal query can be estimated from an\nexpression over the observed variables, called the estimand. The estimand can\nthen be evaluated by plugging in probabilities computed empirically from data.\nIn contrast to conventional wisdom, which assumes that high dimensional\nprobabilistic functions will lead to exponential evaluation time of the\nestimand. We show that computation can be done efficiently, potentially in time\nlinear in the data size, depending on the estimand's hypergraph.\n  In particular, we show that both the treewidth and hypertree width of the\nestimand's structure bound the evaluation complexity of the plug-in estimands,\nanalogous to their role in the complexity of probabilistic inference in\ngraphical models. Often, the hypertree width provides a more effective bound,\nsince the empirical distributions are sparse.\n", "date": "2024-11-15", "categories": []}
{"id": "2212.08322", "title": "ReCo: Reliable Causal Chain Reasoning via Structural Causal Recurrent\n  Neural Networks", "abstract": "  Causal chain reasoning (CCR) is an essential ability for many decision-making\nAI systems, which requires the model to build reliable causal chains by\nconnecting causal pairs. However, CCR suffers from two main transitive\nproblems: threshold effect and scene drift. In other words, the causal pairs to\nbe spliced may have a conflicting threshold boundary or scenario. To address\nthese issues, we propose a novel Reliable Causal chain reasoning\nframework~(ReCo), which introduces exogenous variables to represent the\nthreshold and scene factors of each causal pair within the causal chain, and\nestimates the threshold and scene contradictions across exogenous variables via\nstructural causal recurrent neural networks~(SRNN). Experiments show that ReCo\noutperforms a series of strong baselines on both Chinese and English CCR\ndatasets. Moreover, by injecting reliable causal chain knowledge distilled by\nReCo, BERT can achieve better performances on four downstream causal-related\ntasks than BERT models enhanced by other kinds of knowledge.\n", "date": "2022-12-16", "categories": []}
{"id": "2411.01881", "title": "Causal Discovery and Classification Using Lempel-Ziv Complexity", "abstract": "  Inferring causal relationships in the decision-making processes of machine\nlearning algorithms is a crucial step toward achieving explainable Artificial\nIntelligence (AI). In this research, we introduce a novel causality measure and\na distance metric derived from Lempel-Ziv (LZ) complexity. We explore how the\nproposed causality measure can be used in decision trees by enabling splits\nbased on features that most strongly \\textit{cause} the outcome. We further\nevaluate the effectiveness of the causality-based decision tree and the\ndistance-based decision tree in comparison to a traditional decision tree using\nGini impurity. While the proposed methods demonstrate comparable classification\nperformance overall, the causality-based decision tree significantly\noutperforms both the distance-based decision tree and the Gini-based decision\ntree on datasets generated from causal models. This result indicates that the\nproposed approach can capture insights beyond those of classical decision\ntrees, especially in causally structured data. Based on the features used in\nthe LZ causal measure based decision tree, we introduce a causal strength for\neach features in the dataset so as to infer the predominant causal variables\nfor the occurrence of the outcome.\n", "date": "2024-11-04", "categories": []}
{"id": "2409.15503", "title": "From Text to Treatment Effects: A Meta-Learning Approach to Handling\n  Text-Based Confounding", "abstract": "  One of the central goals of causal machine learning is the accurate\nestimation of heterogeneous treatment effects from observational data. In\nrecent years, meta-learning has emerged as a flexible, model-agnostic paradigm\nfor estimating conditional average treatment effects (CATE) using any\nsupervised model. This paper examines the performance of meta-learners when the\nconfounding variables are expressed in text. Through synthetic data\nexperiments, we show that learners using pre-trained text representations of\nconfounders, in addition to tabular background variables, achieve improved CATE\nestimates compared to those relying solely on the tabular variables,\nparticularly when sufficient data is available. However, due to the entangled\nnature of the text embeddings, these models do not fully match the performance\nof meta-learners with perfect confounder knowledge. These findings highlight\nboth the potential and the limitations of pre-trained text representations for\ncausal inference and open up interesting avenues for future research.\n", "date": "2024-09-23", "categories": []}
{"id": "2411.08861", "title": "Interaction Testing in Variation Analysis", "abstract": "  Relationships of cause and effect are of prime importance for explaining\nscientific phenomena. Often, rather than just understanding the effects of\ncauses, researchers also wish to understand how a cause $X$ affects an outcome\n$Y$ mechanistically -- i.e., what are the causal pathways that are activated\nbetween $X$ and $Y$. For analyzing such questions, a range of methods has been\ndeveloped over decades under the rubric of causal mediation analysis.\nTraditional mediation analysis focuses on decomposing the average treatment\neffect (ATE) into direct and indirect effects, and therefore focuses on the ATE\nas the central quantity. This corresponds to providing explanations for\nassociations in the interventional regime, such as when the treatment $X$ is\nrandomized. Commonly, however, it is of interest to explain associations in the\nobservational regime, and not just in the interventional regime. In this paper,\nwe introduce \\text{variation analysis}, an extension of mediation analysis that\nfocuses on the total variation (TV) measure between $X$ and $Y$, written as\n$\\mathrm{E}[Y \\mid X=x_1] - \\mathrm{E}[Y \\mid X=x_0]$. The TV measure\nencompasses both causal and confounded effects, as opposed to the ATE which\nonly encompasses causal (direct and mediated) variations. In this way, the TV\nmeasure is suitable for providing explanations in the natural regime and\nanswering questions such as ``why is $X$ associated with $Y$?''. Our focus is\non decomposing the TV measure, in a way that explicitly includes direct,\nindirect, and confounded variations. Furthermore, we also decompose the TV\nmeasure to include interaction terms between these different pathways.\nSubsequently, interaction testing is introduced, involving hypothesis tests to\ndetermine if interaction terms are significantly different from zero. If\ninteractions are not significant, more parsimonious decompositions of the TV\nmeasure can be used.\n", "date": "2024-11-13", "categories": []}
{"id": "2304.03247", "title": "A Bayesian Framework for Causal Analysis of Recurrent Events with Timing\n  Misalignment", "abstract": "  Observational studies of recurrent event rates are common in biomedical\nstatistics. Broadly, the goal is to estimate differences in event rates under\ntwo treatments within a defined target population over a specified followup\nwindow. Estimation with observational data is challenging because, while\nmembership in the target population is defined in terms of eligibility\ncriteria, treatment is rarely observed exactly at the time of eligibility.\nAd-hoc solutions to this timing misalignment can induce bias by incorrectly\nattributing prior event counts and person-time to treatment. Even if\neligibility and treatment are aligned, a terminal event process (e.g. death)\noften stops the recurrent event process of interest. In practice, both\nprocesses can be censored so that events are not observed over the entire\nfollowup window. Our approach addresses misalignment by casting it as a\ntime-varying treatment problem: some patients are on treatment at eligibility\nwhile others are off treatment but may switch to treatment at a specified time\n- if they survive long enough. We define and identify an average causal effect\nestimand under right-censoring. Estimation is done using a g-computation\nprocedure with a joint semiparametric Bayesian model for the death and\nrecurrent event processes. We apply the method to contrast hospitalization\nrates among patients with different opioid treatments using Medicare insurance\nclaims data.\n", "date": "2023-04-06", "categories": []}
{"id": "2411.06338", "title": "CRTRE: Causal Rule Generation with Target Trial Emulation Framework", "abstract": "  Causal inference and model interpretability are gaining increasing attention,\nparticularly in the biomedical domain. Despite recent advance, decorrelating\nfeatures in nonlinear environments with human-interpretable representations\nremains underexplored. In this study, we introduce a novel method called causal\nrule generation with target trial emulation framework (CRTRE), which applies\nrandomize trial design principles to estimate the causal effect of association\nrules. We then incorporate such association rules for the downstream\napplications such as prediction of disease onsets. Extensive experiments on six\nhealthcare datasets, including synthetic data, real-world disease collections,\nand MIMIC-III/IV, demonstrate the model's superior performance. Specifically,\nour method achieved a $\\beta$ error of 0.907, outperforming DWR (1.024) and SVM\n(1.141). On real-world datasets, our model achieved accuracies of 0.789, 0.920,\nand 0.300 for Esophageal Cancer, Heart Disease, and Cauda Equina Syndrome\nprediction task, respectively, consistently surpassing baseline models. On the\nICD code prediction tasks, it achieved AUC Macro scores of 92.8 on MIMIC-III\nand 96.7 on MIMIC-IV, outperforming the state-of-the-art models KEPT and MSMN.\nExpert evaluations further validate the model's effectiveness, causality, and\ninterpretability.\n", "date": "2024-11-10", "categories": []}
{"id": "2411.06391", "title": "CausalStock: Deep End-to-end Causal Discovery for News-driven Stock\n  Movement Prediction", "abstract": "  There are two issues in news-driven multi-stock movement prediction tasks\nthat are not well solved in the existing works. On the one hand, \"relation\ndiscovery\" is a pivotal part when leveraging the price information of other\nstocks to achieve accurate stock movement prediction. Given that stock\nrelations are often unidirectional, such as the \"supplier-consumer\"\nrelationship, causal relations are more appropriate to capture the impact\nbetween stocks. On the other hand, there is substantial noise existing in the\nnews data leading to extracting effective information with difficulty. With\nthese two issues in mind, we propose a novel framework called CausalStock for\nnews-driven multi-stock movement prediction, which discovers the temporal\ncausal relations between stocks. We design a lag-dependent temporal causal\ndiscovery mechanism to model the temporal causal graph distribution. Then a\nFunctional Causal Model is employed to encapsulate the discovered causal\nrelations and predict the stock movements. Additionally, we propose a Denoised\nNews Encoder by taking advantage of the excellent text evaluation ability of\nlarge language models (LLMs) to extract useful information from massive news\ndata. The experiment results show that CausalStock outperforms the strong\nbaselines for both news-driven multi-stock movement prediction and multi-stock\nmovement prediction tasks on six real-world datasets collected from the US,\nChina, Japan, and UK markets. Moreover, getting benefit from the causal\nrelations, CausalStock could offer a clear prediction mechanism with good\nexplainability.\n", "date": "2024-11-10", "categories": []}
{"id": "2411.07006", "title": "Estimating Causal Effects in Partially Directed Parametric Causal Factor\n  Graphs", "abstract": "  Lifting uses a representative of indistinguishable individuals to exploit\nsymmetries in probabilistic relational models, denoted as parametric factor\ngraphs, to speed up inference while maintaining exact answers. In this paper,\nwe show how lifting can be applied to causal inference in partially directed\ngraphs, i.e., graphs that contain both directed and undirected edges to\nrepresent causal relationships between random variables. We present partially\ndirected parametric causal factor graphs (PPCFGs) as a generalisation of\npreviously introduced parametric causal factor graphs, which require a fully\ndirected graph. We further show how causal inference can be performed on a\nlifted level in PPCFGs, thereby extending the applicability of lifted causal\ninference to a broader range of models requiring less prior knowledge about\ncausal relationships.\n", "date": "2024-11-11", "categories": []}
{"id": "2411.05625", "title": "Cross-validating causal discovery via Leave-One-Variable-Out", "abstract": "  We propose a new approach to falsify causal discovery algorithms without\nground truth, which is based on testing the causal model on a pair of variables\nthat has been dropped when learning the causal model. To this end, we use the\n\"Leave-One-Variable-Out (LOVO)\" prediction where $Y$ is inferred from $X$\nwithout any joint observations of $X$ and $Y$, given only training data from\n$X,Z_1,\\dots,Z_k$ and from $Z_1,\\dots,Z_k,Y$. We demonstrate that causal models\non the two subsets, in the form of Acyclic Directed Mixed Graphs (ADMGs), often\nentail conclusions on the dependencies between $X$ and $Y$, enabling this type\nof prediction. The prediction error can then be estimated since the joint\ndistribution $P(X, Y)$ is assumed to be available, and $X$ and $Y$ have only\nbeen omitted for the purpose of falsification. After presenting this graphical\nmethod, which is applicable to general causal discovery algorithms, we\nillustrate how to construct a LOVO predictor tailored towards algorithms\nrelying on specific a priori assumptions, such as linear additive noise models.\nSimulations indicate that the LOVO prediction error is indeed correlated with\nthe accuracy of the causal outputs, affirming the method's effectiveness.\n", "date": "2024-11-08", "categories": []}
{"id": "2411.04243", "title": "ION-C: Integration of Overlapping Networks via Constraints", "abstract": "  In many causal learning problems, variables of interest are often not all\nmeasured over the same observations, but are instead distributed across\nmultiple datasets with overlapping variables. Tillman et al. (2008) presented\nthe first algorithm for enumerating the minimal equivalence class of\nground-truth DAGs consistent with all input graphs by exploiting local\nindependence relations, called ION. In this paper, this problem is formulated\nas a more computationally efficient answer set programming (ASP) problem, which\nwe call ION-C, and solved with the ASP system clingo. The ION-C algorithm was\nrun on random synthetic graphs with varying sizes, densities, and degrees of\noverlap between subgraphs, with overlap having the largest impact on runtime,\nnumber of solution graphs, and agreement within the output set. To validate\nION-C on real-world data, we ran the algorithm on overlapping graphs learned\nfrom data from two successive iterations of the European Social Survey (ESS),\nusing a procedure for conducting joint independence tests to prevent\ninconsistencies in the input.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.03919", "title": "A Causal Framework for Precision Rehabilitation", "abstract": "  Precision rehabilitation offers the promise of an evidence-based approach for\noptimizing individual rehabilitation to improve long-term functional outcomes.\nEmerging techniques, including those driven by artificial intelligence, are\nrapidly expanding our ability to quantify the different domains of function\nduring rehabilitation, other encounters with healthcare, and in the community.\nWhile this seems poised to usher rehabilitation into the era of big data and\nshould be a powerful driver of precision rehabilitation, our field lacks a\ncoherent framework to utilize these data and deliver on this promise. We\npropose a framework that builds upon multiple existing pillars to fill this\ngap. Our framework aims to identify the Optimal Dynamic Treatment Regimens\n(ODTR), or the decision-making strategy that takes in the range of available\nmeasurements and biomarkers to identify interventions likely to maximize\nlong-term function. This is achieved by designing and fitting causal models,\nwhich extend the Computational Neurorehabilitation framework using tools from\ncausal inference. These causal models can learn from heterogeneous data from\ndifferent silos, which must include detailed documentation of interventions,\nsuch as using the Rehabilitation Treatment Specification System. The models\nthen serve as digital twins of patient recovery trajectories, which can be used\nto learn the ODTR. Our causal modeling framework also emphasizes quantitatively\nlinking changes across levels of the functioning to ensure that interventions\ncan be precisely selected based on careful measurement of impairments while\nalso being selected to maximize outcomes that are meaningful to patients and\nstakeholders. We believe this approach can provide a unifying framework to\nleverage growing big rehabilitation data and AI-powered measurements to produce\nprecision rehabilitation treatments that can improve clinical outcomes.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.04054", "title": "Partial Structure Discovery is Sufficient for No-regret Learning in\n  Causal Bandits", "abstract": "  Causal knowledge about the relationships among decision variables and a\nreward variable in a bandit setting can accelerate the learning of an optimal\ndecision. Current works often assume the causal graph is known, which may not\nalways be available a priori. Motivated by this challenge, we focus on the\ncausal bandit problem in scenarios where the underlying causal graph is unknown\nand may include latent confounders. While intervention on the parents of the\nreward node is optimal in the absence of latent confounders, this is not\nnecessarily the case in general. Instead, one must consider a set of possibly\noptimal arms/interventions, each being a special subset of the ancestors of the\nreward node, making causal discovery beyond the parents of the reward node\nessential. For regret minimization, we identify that discovering the full\ncausal structure is unnecessary; however, no existing work provides the\nnecessary and sufficient components of the causal graph. We formally\ncharacterize the set of necessary and sufficient latent confounders one needs\nto detect or learn to ensure that all possibly optimal arms are identified\ncorrectly. We also propose a randomized algorithm for learning the causal graph\nwith a limited number of samples, providing a sample complexity guarantee for\nany desired confidence level. In the causal bandit setup, we propose a\ntwo-stage approach. In the first stage, we learn the induced subgraph on\nancestors of the reward, along with a necessary and sufficient subset of latent\nconfounders, to construct the set of possibly optimal arms. The regret incurred\nduring this phase scales polynomially with respect to the number of nodes in\nthe causal graph. The second phase involves the application of a standard\nbandit algorithm, such as the UCB algorithm. We also establish a regret bound\nfor our two-phase approach, which is sublinear in the number of rounds.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.03004", "title": "Controlling for Unobserved Confounding with Large Language Model\n  Classification of Patient Smoking Status", "abstract": "  Causal understanding is a fundamental goal of evidence-based medicine. When\nrandomization is impossible, causal inference methods allow the estimation of\ntreatment effects from retrospective analysis of observational data. However,\nsuch analyses rely on a number of assumptions, often including that of no\nunobserved confounding. In many practical settings, this assumption is violated\nwhen important variables are not explicitly measured in the clinical record.\nPrior work has proposed to address unobserved confounding with machine learning\nby imputing unobserved variables and then correcting for the classifier's\nmismeasurement. When such a classifier can be trained and the necessary\nassumptions are met, this method can recover an unbiased estimate of a causal\neffect. However, such work has been limited to synthetic data, simple\nclassifiers, and binary variables. This paper extends this methodology by using\na large language model trained on clinical notes to predict patients' smoking\nstatus, which would otherwise be an unobserved confounder. We then apply a\nmeasurement error correction on the categorical predicted smoking status to\nestimate the causal effect of transthoracic echocardiography on mortality in\nthe MIMIC dataset.\n", "date": "2024-11-05", "categories": []}
{"id": "2411.03028", "title": "Graph Agnostic Causal Bayesian Optimisation", "abstract": "  We study the problem of globally optimising a target variable of an unknown\ncausal graph on which a sequence of soft or hard interventions can be\nperformed. The problem of optimising the target variable associated with a\ncausal graph is formalised as Causal Bayesian Optimisation (CBO). We study the\nCBO problem under the cumulative regret objective with unknown causal graphs\nfor two settings, namely structural causal models with hard interventions and\nfunction networks with soft interventions. We propose Graph Agnostic Causal\nBayesian Optimisation (GACBO), an algorithm that actively discovers the causal\nstructure that contributes to achieving optimal rewards. GACBO seeks to balance\nexploiting the actions that give the best rewards against exploring the causal\nstructures and functions. To the best of our knowledge, our work is the first\nto study causal Bayesian optimization with cumulative regret objectives in\nscenarios where the graph is unknown or partially known. We show our proposed\nalgorithm outperforms baselines in simulated experiments and real-world\napplications.\n", "date": "2024-11-05", "categories": []}
{"id": "2411.03275", "title": "Causal Responsibility Attribution for Human-AI Collaboration", "abstract": "  As Artificial Intelligence (AI) systems increasingly influence\ndecision-making across various fields, the need to attribute responsibility for\nundesirable outcomes has become essential, though complicated by the complex\ninterplay between humans and AI. Existing attribution methods based on actual\ncausality and Shapley values tend to disproportionately blame agents who\ncontribute more to an outcome and rely on real-world measures of\nblameworthiness that may misalign with responsible AI standards. This paper\npresents a causal framework using Structural Causal Models (SCMs) to\nsystematically attribute responsibility in human-AI systems, measuring overall\nblameworthiness while employing counterfactual reasoning to account for agents'\nexpected epistemic levels. Two case studies illustrate the framework's\nadaptability in diverse human-AI collaboration scenarios.\n", "date": "2024-11-05", "categories": []}
{"id": "2406.06452", "title": "Estimating Heterogeneous Treatment Effects by Combining Weak Instruments\n  and Observational Data", "abstract": "  Accurately predicting conditional average treatment effects (CATEs) is\ncrucial in personalized medicine and digital platform analytics. Since the\ntreatments of interest often cannot be directly randomized, observational data\nis leveraged to learn CATEs, but this approach can incur significant bias from\nunobserved confounding. One strategy to overcome these limitations is to\nleverage instrumental variables (IVs) as latent quasi-experiments, such as\nrandomized intent-to-treat assignments or randomized product recommendations.\nThis approach, on the other hand, can suffer from low compliance,\n$\\textit{i.e.}$, IV weakness. Some subgroups may even exhibit zero compliance,\nmeaning we cannot instrument for their CATEs at all. In this paper, we develop\na novel approach to combine IV and observational data to enable reliable CATE\nestimation in the presence of unobserved confounding in the observational data\nand low compliance in the IV data, including no compliance for some subgroups.\nWe propose a two-stage framework that first learns $\\textit{biased}$ CATEs from\nthe observational data, and then applies a compliance-weighted correction using\nIV data, effectively leveraging IV strength variability across covariates. We\ncharacterize the convergence rates of our method and validate its effectiveness\nthrough a simulation study. Additionally, we demonstrate its utility with real\ndata by analyzing the heterogeneous effects of 401(k) plan participation on\nwealth.\n", "date": "2024-06-10", "categories": []}
{"id": "2410.24059", "title": "Identifying General Mechanism Shifts in Linear Causal Representations", "abstract": "  We consider the linear causal representation learning setting where we\nobserve a linear mixing of $d$ unknown latent factors, which follow a linear\nstructural causal model. Recent work has shown that it is possible to recover\nthe latent factors as well as the underlying structural causal model over them,\nup to permutation and scaling, provided that we have at least $d$ environments,\neach of which corresponds to perfect interventions on a single latent node\n(factor). After this powerful result, a key open problem faced by the community\nhas been to relax these conditions: allow for coarser than perfect single-node\ninterventions, and allow for fewer than $d$ of them, since the number of latent\nfactors $d$ could be very large. In this work, we consider precisely such a\nsetting, where we allow a smaller than $d$ number of environments, and also\nallow for very coarse interventions that can very coarsely \\textit{change the\nentire causal graph over the latent factors}. On the flip side, we relax what\nwe wish to extract to simply the \\textit{list of nodes that have shifted\nbetween one or more environments}. We provide a surprising identifiability\nresult that it is indeed possible, under some very mild standard assumptions,\nto identify the set of shifted nodes. Our identifiability proof moreover is a\nconstructive one: we explicitly provide necessary and sufficient conditions for\na node to be a shifted node, and show that we can check these conditions given\nobserved data. Our algorithm lends itself very naturally to the sample setting\nwhere instead of just interventional distributions, we are provided datasets of\nsamples from each of these distributions. We corroborate our results on both\nsynthetic experiments as well as an interesting psychometric dataset. The code\ncan be found at https://github.com/TianyuCodings/iLCS.\n", "date": "2024-10-31", "categories": []}
{"id": "2411.01250", "title": "Hierarchical and Density-based Causal Clustering", "abstract": "  Understanding treatment effect heterogeneity is vital for scientific and\npolicy research. However, identifying and evaluating heterogeneous treatment\neffects pose significant challenges due to the typically unknown subgroup\nstructure. Recently, a novel approach, causal k-means clustering, has emerged\nto assess heterogeneity of treatment effect by applying the k-means algorithm\nto unknown counterfactual regression functions. In this paper, we expand upon\nthis framework by integrating hierarchical and density-based clustering\nalgorithms. We propose plug-in estimators that are simple and readily\nimplementable using off-the-shelf algorithms. Unlike k-means clustering, which\nrequires the margin condition, our proposed estimators do not rely on strong\nstructural assumptions on the outcome process. We go on to study their rate of\nconvergence, and show that under the minimal regularity conditions, the\nadditional cost of causal clustering is essentially the estimation error of the\noutcome regression functions. Our findings significantly extend the\ncapabilities of the causal clustering framework, thereby contributing to the\nprogression of methodologies for identifying homogeneous subgroups in treatment\nresponse, consequently facilitating more nuanced and targeted interventions.\nThe proposed methods also open up new avenues for clustering with generic\npseudo-outcomes. We explore finite sample properties via simulation, and\nillustrate the proposed methods in voting and employment projection datasets.\n", "date": "2024-11-02", "categories": []}
{"id": "2411.01498", "title": "Educational Effects in Mathematics: Conditional Average Treatment Effect\n  depending on the Number of Treatments", "abstract": "  This study examines the educational effect of the Academic Support Center at\nKogakuin University. Following the initial assessment, it was suggested that\ngroup bias had led to an underestimation of the Center's true impact. To\naddress this issue, the authors applied the theory of causal inference. By\nusing T-learner, the conditional average treatment effect (CATE) of the\nCenter's face-to-face (F2F) personal assistance program was evaluated.\nExtending T-learner, the authors produced a new CATE function that depends on\nthe number of treatments (F2F sessions) and used the estimated function to\npredict the CATE performance of F2F assistance.\n", "date": "2024-11-03", "categories": []}
{"id": "2411.02383", "title": "Linear Causal Bandits: Unknown Graph and Soft Interventions", "abstract": "  Designing causal bandit algorithms depends on two central categories of\nassumptions: (i) the extent of information about the underlying causal graphs\nand (ii) the extent of information about interventional statistical models.\nThere have been extensive recent advances in dispensing with assumptions on\neither category. These include assuming known graphs but unknown interventional\ndistributions, and the converse setting of assuming unknown graphs but access\nto restrictive hard/$\\operatorname{do}$ interventions, which removes the\nstochasticity and ancestral dependencies. Nevertheless, the problem in its\ngeneral form, i.e., unknown graph and unknown stochastic intervention models,\nremains open. This paper addresses this problem and establishes that in a graph\nwith $N$ nodes, maximum in-degree $d$ and maximum causal path length $L$, after\n$T$ interaction rounds the regret upper bound scales as\n$\\tilde{\\mathcal{O}}((cd)^{L-\\frac{1}{2}}\\sqrt{T} + d + RN)$ where $c>1$ is a\nconstant and $R$ is a measure of intervention power. A universal minimax lower\nbound is also established, which scales as $\\Omega(d^{L-\\frac{3}{2}}\\sqrt{T})$.\nImportantly, the graph size $N$ has a diminishing effect on the regret as $T$\ngrows. These bounds have matching behavior in $T$, exponential dependence on\n$L$, and polynomial dependence on $d$ (with the gap $d\\ $). On the algorithmic\naspect, the paper presents a novel way of designing a computationally efficient\nCB algorithm, addressing a challenge that the existing CB algorithms using soft\ninterventions face.\n", "date": "2024-11-04", "categories": []}
{"id": "2402.07419", "title": "Conditional Generative Models are Sufficient to Sample from Any Causal\n  Effect Estimand", "abstract": "  Causal inference from observational data plays critical role in many\napplications in trustworthy machine learning. While sound and complete\nalgorithms exist to compute causal effects, many of them assume access to\nconditional likelihoods, which is difficult to estimate for high-dimensional\n(particularly image) data. Researchers have alleviated this issue by simulating\ncausal relations with neural models. However, when we have high-dimensional\nvariables in the causal graph along with some unobserved confounders, no\nexisting work can effectively sample from the un/conditional interventional\ndistributions. In this work, we show how to sample from any identifiable\ninterventional distribution given an arbitrary causal graph through a sequence\nof push-forward computations of conditional generative models, such as\ndiffusion models. Our proposed algorithm follows the recursive steps of the\nexisting likelihood-based identification algorithms to train a set of\nfeed-forward models, and connect them in a specific way to sample from the\ndesired distribution. We conduct experiments on a Colored MNIST dataset having\nboth the treatment ($X$) and the target variables ($Y$) as images and sample\nfrom $P(y|do(x))$. Our algorithm also enables us to conduct a causal analysis\nto evaluate spurious correlations among input features of generative models\npre-trained on the CelebA dataset. Finally, we generate high-dimensional\ninterventional samples from the MIMIC-CXR dataset involving text and image\nvariables.\n", "date": "2024-02-12", "categories": []}
{"id": "2402.18392", "title": "Unveiling the Potential of Robustness in Selecting Conditional Average\n  Treatment Effect Estimators", "abstract": "  The growing demand for personalized decision-making has led to a surge of\ninterest in estimating the Conditional Average Treatment Effect (CATE). Various\ntypes of CATE estimators have been developed with advancements in machine\nlearning and causal inference. However, selecting the desirable CATE estimator\nthrough a conventional model validation procedure remains impractical due to\nthe absence of counterfactual outcomes in observational data. Existing\napproaches for CATE estimator selection, such as plug-in and pseudo-outcome\nmetrics, face two challenges. First, they must determine the metric form and\nthe underlying machine learning models for fitting nuisance parameters (e.g.,\noutcome function, propensity function, and plug-in learner). Second, they lack\na specific focus on selecting a robust CATE estimator. To address these\nchallenges, this paper introduces a Distributionally Robust Metric (DRM) for\nCATE estimator selection. The proposed DRM is nuisance-free, eliminating the\nneed to fit models for nuisance parameters, and it effectively prioritizes the\nselection of a distributionally robust CATE estimator. The experimental results\nvalidate the effectiveness of the DRM method in selecting CATE estimators that\nare robust to the distribution shift incurred by covariate shift and hidden\nconfounders.\n", "date": "2024-02-28", "categories": []}
{"id": "2410.06726", "title": "Bounds and Sensitivity Analysis of the Causal Effect Under\n  Outcome-Independent MNAR Confounding", "abstract": "  We report assumption-free bounds for any contrast between the probabilities\nof the potential outcome under exposure and non-exposure when the confounders\nare missing not at random. We assume that the missingness mechanism is\noutcome-independent. We also report a sensitivity analysis method to complement\nour bounds.\n", "date": "2024-10-09", "categories": []}
{"id": "2411.00126", "title": "Training and Evaluating Causal Forecasting Models for Time-Series", "abstract": "  Deep learning time-series models are often used to make forecasts that inform\ndownstream decisions. Since these decisions can differ from those in the\ntraining set, there is an implicit requirement that time-series models will\ngeneralize outside of their training distribution. Despite this core\nrequirement, time-series models are typically trained and evaluated on\nin-distribution predictive tasks. We extend the orthogonal statistical learning\nframework to train causal time-series models that generalize better when\nforecasting the effect of actions outside of their training distribution. To\nevaluate these models, we leverage Regression Discontinuity Designs popular in\neconomics to construct a test set of causal treatment effects.\n", "date": "2024-10-31", "categories": []}
{"id": "2411.00213", "title": "Learning Mixtures of Unknown Causal Interventions", "abstract": "  The ability to conduct interventions plays a pivotal role in learning causal\nrelationships among variables, thus facilitating applications across diverse\nscientific disciplines such as genomics, economics, and machine learning.\nHowever, in many instances within these applications, the process of generating\ninterventional data is subject to noise: rather than data being sampled\ndirectly from the intended interventional distribution, interventions often\nyield data sampled from a blend of both intended and unintended interventional\ndistributions.\n  We consider the fundamental challenge of disentangling mixed interventional\nand observational data within linear Structural Equation Models (SEMs) with\nGaussian additive noise without the knowledge of the true causal graph. We\ndemonstrate that conducting interventions, whether do or soft, yields\ndistributions with sufficient diversity and properties conducive to efficiently\nrecovering each component within the mixture. Furthermore, we establish that\nthe sample complexity required to disentangle mixed data inversely correlates\nwith the extent of change induced by an intervention in the equations governing\nthe affected variable values. As a result, the causal graph can be identified\nup to its interventional Markov Equivalence Class, similar to scenarios where\nno noise influences the generation of interventional data. We further support\nour theoretical findings by conducting simulations wherein we perform causal\ndiscovery from such mixed data.\n", "date": "2024-10-31", "categories": []}
{"id": "2402.01607", "title": "Natural Counterfactuals With Necessary Backtracking", "abstract": "  Counterfactual reasoning is pivotal in human cognition and especially\nimportant for providing explanations and making decisions. While Judea Pearl's\ninfluential approach is theoretically elegant, its generation of a\ncounterfactual scenario often requires too much deviation from the observed\nscenarios to be feasible, as we show using simple examples. To mitigate this\ndifficulty, we propose a framework of \\emph{natural counterfactuals} and a\nmethod for generating counterfactuals that are more feasible with respect to\nthe actual data distribution. Our methodology incorporates a certain amount of\nbacktracking when needed, allowing changes in causally preceding variables to\nminimize deviations from realistic scenarios. Specifically, we introduce a\nnovel optimization framework that permits but also controls the extent of\nbacktracking with a naturalness criterion. Empirical experiments demonstrate\nthe effectiveness of our method. The code is available at\nhttps://github.com/GuangyuanHao/natural_counterfactuals.\n", "date": "2024-02-02", "categories": []}
{"id": "2410.22382", "title": "Debiasing Alternative Data for Credit Underwriting Using Causal\n  Inference", "abstract": "  Alternative data provides valuable insights for lenders to evaluate a\nborrower's creditworthiness, which could help expand credit access to\nunderserved groups and lower costs for borrowers. But some forms of alternative\ndata have historically been excluded from credit underwriting because it could\nact as an illegal proxy for a protected class like race or gender, causing\nredlining. We propose a method for applying causal inference to a supervised\nmachine learning model to debias alternative data so that it might be used for\ncredit underwriting. We demonstrate how our algorithm can be used against a\npublic credit dataset to improve model accuracy across different racial groups,\nwhile providing theoretically robust nondiscrimination guarantees.\n", "date": "2024-10-29", "categories": []}
{"id": "2410.23862", "title": "$\\psi$DAG: Projected Stochastic Approximation Iteration for DAG\n  Structure Learning", "abstract": "  Learning the structure of Directed Acyclic Graphs (DAGs) presents a\nsignificant challenge due to the vast combinatorial search space of possible\ngraphs, which scales exponentially with the number of nodes. Recent\nadvancements have redefined this problem as a continuous optimization task by\nincorporating differentiable acyclicity constraints. These methods commonly\nrely on algebraic characterizations of DAGs, such as matrix exponentials, to\nenable the use of gradient-based optimization techniques. Despite these\ninnovations, existing methods often face optimization difficulties due to the\nhighly non-convex nature of DAG constraints and the per-iteration computational\ncomplexity. In this work, we present a novel framework for learning DAGs,\nemploying a Stochastic Approximation approach integrated with Stochastic\nGradient Descent (SGD)-based optimization techniques. Our framework introduces\nnew projection methods tailored to efficiently enforce DAG constraints,\nensuring that the algorithm converges to a feasible local minimum. With its low\niteration complexity, the proposed method is well-suited for handling\nlarge-scale problems with improved computational efficiency. We demonstrate the\neffectiveness and scalability of our framework through comprehensive\nexperimental evaluations, which confirm its superior performance across various\nsettings.\n", "date": "2024-10-31", "categories": []}
{"id": "2410.21917", "title": "Identifiability Analysis of Linear ODE Systems with Hidden Confounders", "abstract": "  The identifiability analysis of linear Ordinary Differential Equation (ODE)\nsystems is a necessary prerequisite for making reliable causal inferences about\nthese systems. While identifiability has been well studied in scenarios where\nthe system is fully observable, the conditions for identifiability remain\nunexplored when latent variables interact with the system. This paper aims to\naddress this gap by presenting a systematic analysis of identifiability in\nlinear ODE systems incorporating hidden confounders. Specifically, we\ninvestigate two cases of such systems. In the first case, latent confounders\nexhibit no causal relationships, yet their evolution adheres to specific\nfunctional forms, such as polynomial functions of time $t$. Subsequently, we\nextend this analysis to encompass scenarios where hidden confounders exhibit\ncausal dependencies, with the causal structure of latent variables described by\na Directed Acyclic Graph (DAG). The second case represents a more intricate\nvariation of the first case, prompting a more comprehensive identifiability\nanalysis. Accordingly, we conduct detailed identifiability analyses of the\nsecond system under various observation conditions, including both continuous\nand discrete observations from single or multiple trajectories. To validate our\ntheoretical results, we perform a series of simulations, which support and\nsubstantiate our findings.\n", "date": "2024-10-29", "categories": []}
{"id": "2410.22481", "title": "Bayesian Counterfactual Prediction Models for HIV Care Retention with\n  Incomplete Outcome and Covariate Information", "abstract": "  Like many chronic diseases, human immunodeficiency virus (HIV) is managed\nover time at regular clinic visits. At each visit, patient features are\nassessed, treatments are prescribed, and a subsequent visit is scheduled. There\nis a need for data-driven methods for both predicting retention and\nrecommending scheduling decisions that optimize retention. Prediction models\ncan be useful for estimating retention rates across a range of scheduling\noptions. However, training such models with electronic health records (EHR)\ninvolves several complexities. First, formal causal inference methods are\nneeded to adjust for observed confounding when estimating retention rates under\ncounterfactual scheduling decisions. Second, competing events such as death\npreclude retention, while censoring events render retention missing. Third,\ninconsistent monitoring of features such as viral load and CD4 count lead to\ncovariate missingness. This paper presents an all-in-one approach for both\npredicting HIV retention and optimizing scheduling while accounting for these\ncomplexities. We formulate and identify causal retention estimands in terms of\npotential return-time under a hypothetical scheduling decision. Flexible\nBayesian approaches are used to model the observed return-time distribution\nwhile accounting for competing and censoring events and form posterior point\nand uncertainty estimates for these estimands. We address the urgent need for\ndata-driven decision support in HIV care by applying our method to EHR from the\nAcademic Model Providing Access to Healthcare (AMPATH) - a consortium of\nclinics that treat HIV in Western Kenya.\n", "date": "2024-10-29", "categories": []}
{"id": "2410.22754", "title": "An Overview of Causal Inference using Kernel Embeddings", "abstract": "  Kernel embeddings have emerged as a powerful tool for representing\nprobability measures in a variety of statistical inference problems. By mapping\nprobability measures into a reproducing kernel Hilbert space (RKHS), kernel\nembeddings enable flexible representations of complex relationships between\nvariables. They serve as a mechanism for efficiently transferring the\nrepresentation of a distribution downstream to other tasks, such as hypothesis\ntesting or causal effect estimation. In the context of causal inference, the\nmain challenges include identifying causal associations and estimating the\naverage treatment effect from observational data, where confounding variables\nmay obscure direct cause-and-effect relationships. Kernel embeddings provide a\nrobust nonparametric framework for addressing these challenges. They allow for\nthe representations of distributions of observational data and their seamless\ntransformation into representations of interventional distributions to estimate\nrelevant causal quantities. We overview recent research that leverages the\nexpressiveness of kernel embeddings in tandem with causal inference.\n", "date": "2024-10-30", "categories": []}
{"id": "2410.23155", "title": "QWO: Speeding Up Permutation-Based Causal Discovery in LiGAMs", "abstract": "  Causal discovery is essential for understanding relationships among variables\nof interest in many scientific domains. In this paper, we focus on\npermutation-based methods for learning causal graphs in Linear Gaussian Acyclic\nModels (LiGAMs), where the permutation encodes a causal ordering of the\nvariables. Existing methods in this setting are not scalable due to their high\ncomputational complexity. These methods are comprised of two main components:\n(i) constructing a specific DAG, $\\mathcal{G}^\\pi$, for a given permutation\n$\\pi$, which represents the best structure that can be learned from the\navailable data while adhering to $\\pi$, and (ii) searching over the space of\npermutations (i.e., causal orders) to minimize the number of edges in\n$\\mathcal{G}^\\pi$. We introduce QWO, a novel approach that significantly\nenhances the efficiency of computing $\\mathcal{G}^\\pi$ for a given permutation\n$\\pi$. QWO has a speed-up of $O(n^2)$ ($n$ is the number of variables) compared\nto the state-of-the-art BIC-based method, making it highly scalable. We show\nthat our method is theoretically sound and can be integrated into existing\nsearch strategies such as GRASP and hill-climbing-based methods to improve\ntheir performance.\n", "date": "2024-10-30", "categories": []}
{"id": "2310.14763", "title": "Externally Valid Policy Evaluation Combining Trial and Observational\n  Data", "abstract": "  Randomized trials are widely considered as the gold standard for evaluating\nthe effects of decision policies. Trial data is, however, drawn from a\npopulation which may differ from the intended target population and this raises\na problem of external validity (aka. generalizability). In this paper we seek\nto use trial data to draw valid inferences about the outcome of a policy on the\ntarget population. Additional covariate data from the target population is used\nto model the sampling of individuals in the trial study. We develop a method\nthat yields certifiably valid trial-based policy evaluations under any\nspecified range of model miscalibrations. The method is nonparametric and the\nvalidity is assured even with finite samples. The certified policy evaluations\nare illustrated using both simulated and real data.\n", "date": "2023-10-23", "categories": []}
{"id": "2401.06687", "title": "Proximal Causal Inference With Text Data", "abstract": "  Recent text-based causal methods attempt to mitigate confounding bias by\nestimating proxies of confounding variables that are partially or imperfectly\nmeasured from unstructured text data. These approaches, however, assume\nanalysts have supervised labels of the confounders given text for a subset of\ninstances, a constraint that is sometimes infeasible due to data privacy or\nannotation costs. In this work, we address settings in which an important\nconfounding variable is completely unobserved. We propose a new causal\ninference method that uses two instances of pre-treatment text data, infers two\nproxies using two zero-shot models on the separate instances, and applies these\nproxies in the proximal g-formula. We prove, under certain assumptions about\nthe instances of text and accuracy of the zero-shot predictions, that our\nmethod of inferring text-based proxies satisfies identification conditions of\nthe proximal g-formula while other seemingly reasonable proposals do not. To\naddress untestable assumptions associated with our method and the proximal\ng-formula, we further propose an odds ratio falsification heuristic that flags\nwhen to proceed with downstream effect estimation using the inferred proxies.\nWe evaluate our method in synthetic and semi-synthetic settings -- the latter\nwith real-world clinical notes from MIMIC-III and open large language models\nfor zero-shot prediction -- and find that our method produces estimates with\nlow bias. We believe that this text-based design of proxies allows for the use\nof proximal causal inference in a wider range of scenarios, particularly those\nfor which obtaining suitable proxies from structured data is difficult.\n", "date": "2024-01-12", "categories": []}
{"id": "2402.11652", "title": "Doubly Robust Inference in Causal Latent Factor Models", "abstract": "  This article introduces a new estimator of average treatment effects under\nunobserved confounding in modern data-rich environments featuring large numbers\nof units and outcomes. The proposed estimator is doubly robust, combining\noutcome imputation, inverse probability weighting, and a novel cross-fitting\nprocedure for matrix completion. We derive finite-sample and asymptotic\nguarantees, and show that the error of the new estimator converges to a\nmean-zero Gaussian distribution at a parametric rate. Simulation results\ndemonstrate the relevance of the formal properties of the estimators analyzed\nin this article.\n", "date": "2024-02-18", "categories": []}
{"id": "2405.14547", "title": "Causal Effect Identification in a Sub-Population with Latent Variables", "abstract": "  The s-ID problem seeks to compute a causal effect in a specific\nsub-population from the observational data pertaining to the same sub\npopulation (Abouei et al., 2023). This problem has been addressed when all the\nvariables in the system are observable. In this paper, we consider an extension\nof the s-ID problem that allows for the presence of latent variables. To tackle\nthe challenges induced by the presence of latent variables in a sub-population,\nwe first extend the classical relevant graphical definitions, such as\nc-components and Hedges, initially defined for the so-called ID problem (Pearl,\n1995; Tian & Pearl, 2002), to their new counterparts. Subsequently, we propose\na sound algorithm for the s-ID problem with latent variables.\n", "date": "2024-05-23", "categories": []}
{"id": "2406.00535", "title": "Causal Contrastive Learning for Counterfactual Regression Over Time", "abstract": "  Estimating treatment effects over time holds significance in various domains,\nincluding precision medicine, epidemiology, economy, and marketing. This paper\nintroduces a unique approach to counterfactual regression over time,\nemphasizing long-term predictions. Distinguishing itself from existing models\nlike Causal Transformer, our approach highlights the efficacy of employing RNNs\nfor long-term forecasting, complemented by Contrastive Predictive Coding (CPC)\nand Information Maximization (InfoMax). Emphasizing efficiency, we avoid the\nneed for computationally expensive transformers. Leveraging CPC, our method\ncaptures long-term dependencies in the presence of time-varying confounders.\nNotably, recent models have disregarded the importance of invertible\nrepresentation, compromising identification assumptions. To remedy this, we\nemploy the InfoMax principle, maximizing a lower bound of mutual information\nbetween sequence data and its representation. Our method achieves\nstate-of-the-art counterfactual estimation results using both synthetic and\nreal-world data, marking the pioneering incorporation of Contrastive Predictive\nEncoding in causal inference.\n", "date": "2024-06-01", "categories": []}
{"id": "2410.21343", "title": "Combining Incomplete Observational and Randomized Data for Heterogeneous\n  Treatment Effects", "abstract": "  Data from observational studies (OSs) is widely available and readily\nobtainable yet frequently contains confounding biases. On the other hand, data\nderived from randomized controlled trials (RCTs) helps to reduce these biases;\nhowever, it is expensive to gather, resulting in a tiny size of randomized\ndata. For this reason, effectively fusing observational data and randomized\ndata to better estimate heterogeneous treatment effects (HTEs) has gained\nincreasing attention. However, existing methods for integrating observational\ndata with randomized data must require \\textit{complete} observational data,\nmeaning that both treated subjects and untreated subjects must be included in\nOSs. This prerequisite confines the applicability of such methods to very\nspecific situations, given that including all subjects, whether treated or\nuntreated, in observational studies is not consistently achievable. In our\npaper, we propose a resilient approach to \\textbf{C}ombine \\textbf{I}ncomplete\n\\textbf{O}bservational data and randomized data for HTE estimation, which we\nabbreviate as \\textbf{CIO}. The CIO is capable of estimating HTEs efficiently\nregardless of the completeness of the observational data, be it full or\npartial. Concretely, a confounding bias function is first derived using the\npseudo-experimental group from OSs, in conjunction with the pseudo-control\ngroup from RCTs, via an effect estimation procedure. This function is\nsubsequently utilized as a corrective residual to rectify the observed outcomes\nof observational data during the HTE estimation by combining the available\nobservational data and the all randomized data. To validate our approach, we\nhave conducted experiments on a synthetic dataset and two semi-synthetic\ndatasets.\n", "date": "2024-10-28", "categories": []}
{"id": "2410.21531", "title": "Deep Learning Methods for the Noniterative Conditional Expectation\n  G-Formula for Causal Inference from Complex Observational Data", "abstract": "  The g-formula can be used to estimate causal effects of sustained treatment\nstrategies using observational data under the identifying assumptions of\nconsistency, positivity, and exchangeability. The non-iterative conditional\nexpectation (NICE) estimator of the g-formula also requires correct estimation\nof the conditional distribution of the time-varying treatment, confounders, and\noutcome. Parametric models, which have been traditionally used for this\npurpose, are subject to model misspecification, which may result in biased\ncausal estimates. Here, we propose a unified deep learning framework for the\nNICE g-formula estimator that uses multitask recurrent neural networks for\nestimation of the joint conditional distributions. Using simulated data, we\nevaluated our model's bias and compared it with that of the parametric\ng-formula estimator. We found lower bias in the estimates of the causal effect\nof sustained treatment strategies on a survival outcome when using the deep\nlearning estimator compared with the parametric NICE estimator in settings with\nsimple and complex temporal dependencies between covariates. These findings\nsuggest that our Deep Learning g-formula estimator may be less sensitive to\nmodel misspecification than the classical parametric NICE estimator when\nestimating the causal effect of sustained treatment strategies from complex\nobservational data.\n", "date": "2024-10-28", "categories": []}
{"id": "2410.22194", "title": "ADAM: An Embodied Causal Agent in Open-World Environments", "abstract": "  In open-world environments like Minecraft, existing agents face challenges in\ncontinuously learning structured knowledge, particularly causality. These\nchallenges stem from the opacity inherent in black-box models and an excessive\nreliance on prior knowledge during training, which impair their\ninterpretability and generalization capability. To this end, we introduce ADAM,\nAn emboDied causal Agent in Minecraft, that can autonomously navigate the open\nworld, perceive multimodal contexts, learn causal world knowledge, and tackle\ncomplex tasks through lifelong learning. ADAM is empowered by four key\ncomponents: 1) an interaction module, enabling the agent to execute actions\nwhile documenting the interaction processes; 2) a causal model module, tasked\nwith constructing an ever-growing causal graph from scratch, which enhances\ninterpretability and diminishes reliance on prior knowledge; 3) a controller\nmodule, comprising a planner, an actor, and a memory pool, which uses the\nlearned causal graph to accomplish tasks; 4) a perception module, powered by\nmultimodal large language models, which enables ADAM to perceive like a human\nplayer. Extensive experiments show that ADAM constructs an almost perfect\ncausal graph from scratch, enabling efficient task decomposition and execution\nwith strong interpretability. Notably, in our modified Minecraft games where no\nprior knowledge is available, ADAM maintains its performance and shows\nremarkable robustness and generalization capability. ADAM pioneers a novel\nparadigm that integrates causal methods and embodied agents in a synergistic\nmanner. Our project page is at https://opencausalab.github.io/ADAM.\n", "date": "2024-10-29", "categories": []}
{"id": "2312.17122", "title": "LLM4Causal: Democratized Causal Tools for Everyone via Large Language\n  Model", "abstract": "  Large Language Models (LLMs) have shown their success in language\nunderstanding and reasoning on general topics. However, their capability to\nperform inference based on user-specified structured data and knowledge in\ncorpus-rare concepts, such as causal decision-making is still limited. In this\nwork, we explore the possibility of fine-tuning an open-sourced LLM into\nLLM4Causal, which can identify the causal task, execute a corresponding\nfunction, and interpret its numerical results based on users' queries and the\nprovided dataset. Meanwhile, we propose a data generation process for more\ncontrollable GPT prompting and present two instruction-tuning datasets: (1)\nCausal-Retrieval-Bench for causal problem identification and input parameter\nextraction for causal function calling and (2) Causal-Interpret-Bench for\nin-context causal interpretation. By conducting end-to-end evaluations and two\nablation studies, we showed that LLM4Causal can deliver end-to-end solutions\nfor causal problems and provide easy-to-understand answers, which significantly\noutperforms the baselines.\n", "date": "2023-12-28", "categories": []}
{"id": "2401.01426", "title": "Modular Learning of Deep Causal Generative Models for High-dimensional\n  Causal Inference", "abstract": "  Sound and complete algorithms have been proposed to compute identifiable\ncausal queries using the causal structure and data. However, most of these\nalgorithms assume accurate estimation of the data distribution, which is\nimpractical for high-dimensional variables such as images. On the other hand,\nmodern deep generative architectures can be trained to sample from\nhigh-dimensional distributions. However, training these networks are typically\nvery costly. Thus, it is desirable to leverage pre-trained models to answer\ncausal queries using such high-dimensional data. To address this, we propose\nmodular training of deep causal generative models that not only makes learning\nmore efficient, but also allows us to utilize large, pre-trained conditional\ngenerative models. To the best of our knowledge, our algorithm, Modular-DCM is\nthe first algorithm that, given the causal structure, uses adversarial training\nto learn the network weights, and can make use of pre-trained models to\nprovably sample from any identifiable causal query in the presence of latent\nconfounders. With extensive experiments on the Colored-MNIST dataset, we\ndemonstrate that our algorithm outperforms the baselines. We also show our\nalgorithm's convergence on the COVIDx dataset and its utility with a causal\ninvariant prediction problem on CelebA-HQ.\n", "date": "2024-01-02", "categories": []}
{"id": "2404.11055", "title": "Do LLMs Think Fast and Slow? A Causal Study on Sentiment Analysis", "abstract": "  Sentiment analysis (SA) aims to identify the sentiment expressed in a text,\nsuch as a product review. Given a review and the sentiment associated with it,\nthis work formulates SA as a combination of two tasks: (1) a causal discovery\ntask that distinguishes whether a review \"primes\" the sentiment (Causal\nHypothesis C1), or the sentiment \"primes\" the review (Causal Hypothesis C2);\nand (2) the traditional prediction task to model the sentiment using the review\nas input. Using the peak-end rule in psychology, we classify a sample as C1 if\nits overall sentiment score approximates an average of all the sentence-level\nsentiments in the review, and C2 if the overall sentiment score approximates an\naverage of the peak and end sentiments. For the prediction task, we use the\ndiscovered causal mechanisms behind the samples to improve LLM performance by\nproposing causal prompts that give the models an inductive bias of the\nunderlying causal graph, leading to substantial improvements by up to 32.13 F1\npoints on zero-shot five-class SA. Our code is at\nhttps://github.com/cogito233/causal-sa\n", "date": "2024-04-17", "categories": []}
{"id": "2405.16069", "title": "IncomeSCM: From tabular data set to time-series simulator and causal\n  estimation benchmark", "abstract": "  Evaluating observational estimators of causal effects demands information\nthat is rarely available: unconfounded interventions and outcomes from the\npopulation of interest, created either by randomization or adjustment. As a\nresult, it is customary to fall back on simulators when creating benchmark\ntasks. Simulators offer great control but are often too simplistic to make\nchallenging tasks, either because they are hand-designed and lack the nuances\nof real-world data, or because they are fit to observational data without\nstructural constraints. In this work, we propose a general, repeatable strategy\nfor turning observational data into sequential structural causal models and\nchallenging estimation tasks by following two simple principles: 1) fitting\nreal-world data where possible, and 2) creating complexity by composing simple,\nhand-designed mechanisms. We implement these ideas in a highly configurable\nsoftware package and apply it to the well-known Adult income data set to\nconstruct the IncomeSCM simulator. From this, we devise multiple estimation\ntasks and sample data sets to compare established estimators of causal effects.\nThe tasks present a suitable challenge, with effect estimates varying greatly\nin quality between methods, despite similar performance in the modeling of\nfactual outcomes, highlighting the need for dedicated causal estimators and\nmodel selection criteria.\n", "date": "2024-05-25", "categories": []}
{"id": "2407.07018", "title": "End-To-End Causal Effect Estimation from Unstructured Natural Language\n  Data", "abstract": "  Knowing the effect of an intervention is critical for human decision-making,\nbut current approaches for causal effect estimation rely on manual data\ncollection and structuring, regardless of the causal assumptions. This\nincreases both the cost and time-to-completion for studies. We show how large,\ndiverse observational text data can be mined with large language models (LLMs)\nto produce inexpensive causal effect estimates under appropriate causal\nassumptions. We introduce NATURAL, a novel family of causal effect estimators\nbuilt with LLMs that operate over datasets of unstructured text. Our estimators\nuse LLM conditional distributions (over variables of interest, given the text\ndata) to assist in the computation of classical estimators of causal effect. We\novercome a number of technical challenges to realize this idea, such as\nautomating data curation and using LLMs to impute missing information. We\nprepare six (two synthetic and four real) observational datasets, paired with\ncorresponding ground truth in the form of randomized trials, which we used to\nsystematically evaluate each step of our pipeline. NATURAL estimators\ndemonstrate remarkable performance, yielding causal effect estimates that fall\nwithin 3 percentage points of their ground truth counterparts, including on\nreal-world Phase 3/4 clinical trials. Our results suggest that unstructured\ntext data is a rich source of causal effect information, and NATURAL is a first\nstep towards an automated pipeline to tap this resource.\n", "date": "2024-07-09", "categories": []}
{"id": "2410.19870", "title": "Causal Order Discovery based on Monotonic SCMs", "abstract": "  In this paper, we consider the problem of causal order discovery within the\nframework of monotonic Structural Causal Models (SCMs), which have gained\nattention for their potential to enable causal inference and causal discovery\nfrom observational data. While existing approaches either assume prior\nknowledge about the causal order or use complex optimization techniques to\nimpose sparsity in the Jacobian of Triangular Monotonic Increasing maps, our\nwork introduces a novel sequential procedure that directly identifies the\ncausal order by iteratively detecting the root variable. This method eliminates\nthe need for sparsity assumptions and the associated optimization challenges,\nenabling the identification of a unique SCM without the need for multiple\nindependence tests to break the Markov equivalence class. We demonstrate the\neffectiveness of our approach in sequentially finding the root variable,\ncomparing it to methods that maximize Jacobian sparsity.\n", "date": "2024-10-24", "categories": []}
{"id": "2410.19923", "title": "Language Agents Meet Causality -- Bridging LLMs and Causal World Models", "abstract": "  Large Language Models (LLMs) have recently shown great promise in planning\nand reasoning applications. These tasks demand robust systems, which arguably\nrequire a causal understanding of the environment. While LLMs can acquire and\nreflect common sense causal knowledge from their pretraining data, this\ninformation is often incomplete, incorrect, or inapplicable to a specific\nenvironment. In contrast, causal representation learning (CRL) focuses on\nidentifying the underlying causal structure within a given environment. We\npropose a framework that integrates CRLs with LLMs to enable causally-aware\nreasoning and planning. This framework learns a causal world model, with causal\nvariables linked to natural language expressions. This mapping provides LLMs\nwith a flexible interface to process and generate descriptions of actions and\nstates in text form. Effectively, the causal world model acts as a simulator\nthat the LLM can query and interact with. We evaluate the framework on causal\ninference and planning tasks across temporal scales and environmental\ncomplexities. Our experiments demonstrate the effectiveness of the approach,\nwith the causally-aware method outperforming LLM-based reasoners, especially\nfor longer planning horizons.\n", "date": "2024-10-25", "categories": []}
{"id": "2410.20089", "title": "Sample Efficient Bayesian Learning of Causal Graphs from Interventions", "abstract": "  Causal discovery is a fundamental problem with applications spanning various\nareas in science and engineering. It is well understood that solely using\nobservational data, one can only orient the causal graph up to its Markov\nequivalence class, necessitating interventional data to learn the complete\ncausal graph. Most works in the literature design causal discovery policies\nwith perfect interventions, i.e., they have access to infinite interventional\nsamples. This study considers a Bayesian approach for learning causal graphs\nwith limited interventional samples, mirroring real-world scenarios where such\nsamples are usually costly to obtain. By leveraging the recent result of\nWien\\\"obst et al. (2023) on uniform DAG sampling in polynomial time, we can\nefficiently enumerate all the cut configurations and their corresponding\ninterventional distributions of a target set, and further track their\nposteriors. Given any number of interventional samples, our proposed algorithm\nrandomly intervenes on a set of target vertices that cut all the edges in the\ngraph and returns a causal graph according to the posterior of each target set.\nWhen the number of interventional samples is large enough, we show\ntheoretically that our proposed algorithm will return the true causal graph\nwith high probability. We compare our algorithm against various baseline\nmethods on simulated datasets, demonstrating its superior accuracy measured by\nthe structural Hamming distance between the learned DAG and the ground truth.\nAdditionally, we present a case study showing how this algorithm could be\nmodified to answer more general causal questions without learning the whole\ngraph. As an example, we illustrate that our method can be used to estimate the\ncausal effect of a variable that cannot be intervened.\n", "date": "2024-10-26", "categories": []}
{"id": "2410.20405", "title": "Causal Modeling in Multi-Context Systems: Distinguishing Multiple\n  Context-Specific Causal Graphs which Account for Observational Support", "abstract": "  Causal structure learning with data from multiple contexts carries both\nopportunities and challenges. Opportunities arise from considering shared and\ncontext-specific causal graphs enabling to generalize and transfer causal\nknowledge across contexts. However, a challenge that is currently understudied\nin the literature is the impact of differing observational support between\ncontexts on the identifiability of causal graphs. Here we study in detail\nrecently introduced [6] causal graph objects that capture both causal\nmechanisms and data support, allowing for the analysis of a larger class of\ncontext-specific changes, characterizing distribution shifts more precisely. We\nthereby extend results on the identifiability of context-specific causal\nstructures and propose a framework to model context-specific independence (CSI)\nwithin structural causal models (SCMs) in a refined way that allows to explore\nscenarios where these graph objects differ. We demonstrate how this framework\ncan help explaining phenomena like anomalies or extreme events, where causal\nmechanisms change or appear to change under different conditions. Our results\ncontribute to the theoretical foundations for understanding causal relations in\nmulti-context systems, with implications for generalization, transfer learning,\nand anomaly detection. Future work may extend this approach to more complex\ndata types, such as time-series.\n", "date": "2024-10-27", "categories": []}
{"id": "2410.20647", "title": "General Causal Imputation via Synthetic Interventions", "abstract": "  Given two sets of elements (such as cell types and drug compounds),\nresearchers typically only have access to a limited subset of their\ninteractions. The task of causal imputation involves using this subset to\npredict unobserved interactions. Squires et al. (2022) have proposed two\nestimators for this task based on the synthetic interventions (SI) estimator:\nSI-A (for actions) and SI-C (for contexts). We extend their work and introduce\na novel causal imputation estimator, generalized synthetic interventions (GSI).\nWe prove the identifiability of this estimator for data generated from a more\ncomplex latent factor model. On synthetic and real data we show empirically\nthat it recovers or outperforms their estimators.\n", "date": "2024-10-28", "categories": []}
{"id": "2410.21141", "title": "LLM-initialized Differentiable Causal Discovery", "abstract": "  The discovery of causal relationships between random variables is an\nimportant yet challenging problem that has applications across many scientific\ndomains. Differentiable causal discovery (DCD) methods are effective in\nuncovering causal relationships from observational data; however, these\napproaches often suffer from limited interpretability and face challenges in\nincorporating domain-specific prior knowledge. In contrast, Large Language\nModels (LLMs)-based causal discovery approaches have recently been shown\ncapable of providing useful priors for causal discovery but struggle with\nformal causal reasoning. In this paper, we propose LLM-DCD, which uses an LLM\nto initialize the optimization of the maximum likelihood objective function of\nDCD approaches, thereby incorporating strong priors into the discovery method.\nTo achieve this initialization, we design our objective function to depend on\nan explicitly defined adjacency matrix of the causal graph as its only\nvariational parameter. Directly optimizing the explicitly defined adjacency\nmatrix provides a more interpretable approach to causal discovery.\nAdditionally, we demonstrate higher accuracy on key benchmarking datasets of\nour approach compared to state-of-the-art alternatives, and provide empirical\nevidence that the quality of the initialization directly impacts the quality of\nthe final output of our DCD approach. LLM-DCD opens up new opportunities for\ntraditional causal discovery methods like DCD to benefit from future\nimprovements in the causal reasoning capabilities of LLMs.\n", "date": "2024-10-28", "categories": []}
{"id": "2410.19469", "title": "Unified Causality Analysis Based on the Degrees of Freedom", "abstract": "  Temporally evolving systems are typically modeled by dynamic equations. A key\nchallenge in accurate modeling is understanding the causal relationships\nbetween subsystems, as well as identifying the presence and influence of\nunobserved hidden drivers on the observed dynamics. This paper presents a\nunified method capable of identifying fundamental causal relationships between\npairs of systems, whether deterministic or stochastic. Notably, the method also\nuncovers hidden common causes beyond the observed variables. By analyzing the\ndegrees of freedom in the system, our approach provides a more comprehensive\nunderstanding of both causal influence and hidden confounders. This unified\nframework is validated through theoretical models and simulations,\ndemonstrating its robustness and potential for broader application.\n", "date": "2024-10-25", "categories": []}
{"id": "2410.19575", "title": "Considerations for Distribution Shift Robustness of Diagnostic Models in\n  Healthcare", "abstract": "  We consider robustness to distribution shifts in the context of diagnostic\nmodels in healthcare, where the prediction target $Y$, e.g., the presence of a\ndisease, is causally upstream of the observations $X$, e.g., a biomarker.\nDistribution shifts may occur, for instance, when the training data is\ncollected in a domain with patients having particular demographic\ncharacteristics while the model is deployed on patients from a different\ndemographic group. In the domain of applied ML for health, it is common to\npredict $Y$ from $X$ without considering further information about the patient.\nHowever, beyond the direct influence of the disease $Y$ on biomarker $X$, a\npredictive model may learn to exploit confounding dependencies (or shortcuts)\nbetween $X$ and $Y$ that are unstable under certain distribution shifts. In\nthis work, we highlight a data generating mechanism common to healthcare\nsettings and discuss how recent theoretical results from the causality\nliterature can be applied to build robust predictive models. We theoretically\nshow why ignoring covariates as well as common invariant learning approaches\nwill in general not yield robust predictors in the studied setting, while\nincluding certain covariates into the prediction model will. In an extensive\nsimulation study, we showcase the robustness (or lack thereof) of different\npredictors under various data generating processes. Lastly, we analyze the\nperformance of the different approaches using the PTB-XL dataset, a public\ndataset of annotated ECG recordings.\n", "date": "2024-10-25", "categories": []}
{"id": "2410.18147", "title": "MEC-IP: Efficient Discovery of Markov Equivalent Classes via Integer\n  Programming", "abstract": "  This paper presents a novel Integer Programming (IP) approach for discovering\nthe Markov Equivalent Class (MEC) of Bayesian Networks (BNs) through\nobservational data. The MEC-IP algorithm utilizes a unique clique-focusing\nstrategy and Extended Maximal Spanning Graphs (EMSG) to streamline the search\nfor MEC, thus overcoming the computational limitations inherent in other\nexisting algorithms. Our numerical results show that not only a remarkable\nreduction in computational time is achieved by our algorithm but also an\nimprovement in causal discovery accuracy is seen across diverse datasets. These\nfindings underscore this new algorithm's potential as a powerful tool for\nresearchers and practitioners in causal discovery and BNSL, offering a\nsignificant leap forward toward the efficient and accurate analysis of complex\ndata structures.\n", "date": "2024-10-22", "categories": []}
{"id": "2410.18396", "title": "Revisiting Differentiable Structure Learning: Inconsistency of $\\ell_1$\n  Penalty and Beyond", "abstract": "  Recent advances in differentiable structure learning have framed the\ncombinatorial problem of learning directed acyclic graphs as a continuous\noptimization problem. Various aspects, including data standardization, have\nbeen studied to identify factors that influence the empirical performance of\nthese methods. In this work, we investigate critical limitations in\ndifferentiable structure learning methods, focusing on settings where the true\nstructure can be identified up to Markov equivalence classes, particularly in\nthe linear Gaussian case. While Ng et al. (2024) highlighted potential\nnon-convexity issues in this setting, we demonstrate and explain why the use of\n$\\ell_1$-penalized likelihood in such cases is fundamentally inconsistent, even\nif the global optimum of the optimization problem can be found. To resolve this\nlimitation, we develop a hybrid differentiable structure learning method based\non $\\ell_0$-penalized likelihood with hard acyclicity constraint, where the\n$\\ell_0$ penalty can be approximated by different techniques including\nGumbel-Softmax. Specifically, we first estimate the underlying moral graph, and\nuse it to restrict the search space of the optimization problem, which helps\nalleviate the non-convexity issue. Experimental results show that the proposed\nmethod enhances empirical performance both before and after data\nstandardization, providing a more reliable path for future advancements in\ndifferentiable structure learning, especially for learning Markov equivalence\nclasses.\n", "date": "2024-10-24", "categories": []}
{"id": "2402.09891", "title": "Do causal predictors generalize better to new domains?", "abstract": "  We study how well machine learning models trained on causal features\ngeneralize across domains. We consider 16 prediction tasks on tabular datasets\ncovering applications in health, employment, education, social benefits, and\npolitics. Each dataset comes with multiple domains, allowing us to test how\nwell a model trained in one domain performs in another. For each prediction\ntask, we select features that have a causal influence on the target of\nprediction. Our goal is to test the hypothesis that models trained on causal\nfeatures generalize better across domains. Without exception, we find that\npredictors using all available features, regardless of causality, have better\nin-domain and out-of-domain accuracy than predictors using causal features.\nMoreover, even the absolute drop in accuracy from one domain to the other is no\nbetter for causal predictors than for models that use all features. In\naddition, we show that recent causal machine learning methods for domain\ngeneralization do not perform better in our evaluation than standard predictors\ntrained on the set of causal features. Likewise, causal discovery algorithms\neither fail to run or select causal variables that perform no better than our\nselection. Extensive robustness checks confirm that our findings are stable\nunder variable misclassification.\n", "date": "2024-02-15", "categories": []}
{"id": "2410.17691", "title": "Longitudinal Causal Image Synthesis", "abstract": "  Clinical decision-making relies heavily on causal reasoning and longitudinal\nanalysis. For example, for a patient with Alzheimer's disease (AD), how will\nthe brain grey matter atrophy in a year if intervened on the A-beta level in\ncerebrospinal fluid? The answer is fundamental to diagnosis and follow-up\ntreatment. However, this kind of inquiry involves counterfactual medical images\nwhich can not be acquired by instrumental or correlation-based image synthesis\nmodels. Yet, such queries require counterfactual medical images, not obtainable\nthrough standard image synthesis models. Hence, a causal longitudinal image\nsynthesis (CLIS) method, enabling the synthesis of such images, is highly\nvaluable. However, building a CLIS model confronts three primary yet unmet\nchallenges: mismatched dimensionality between high-dimensional images and\nlow-dimensional tabular variables, inconsistent collection intervals of\nfollow-up data, and inadequate causal modeling capability of existing causal\ngraph methods for image data. In this paper, we established a tabular-visual\ncausal graph (TVCG) for CLIS overcoming these challenges through a novel\nintegration of generative imaging, continuous-time modeling, and structural\ncausal models combined with a neural network. We train our CLIS based on the\nADNI dataset and evaluate it on two other AD datasets, which illustrate the\noutstanding yet controllable quality of the synthesized images and the\ncontributions of synthesized MRI to the characterization of AD progression,\nsubstantiating the reliability and utility in clinics.\n", "date": "2024-10-23", "categories": []}
{"id": "2311.18460", "title": "Causal Fairness under Unobserved Confounding: A Neural Sensitivity\n  Framework", "abstract": "  Fairness for machine learning predictions is widely required in practice for\nlegal, ethical, and societal reasons. Existing work typically focuses on\nsettings without unobserved confounding, even though unobserved confounding can\nlead to severe violations of causal fairness and, thus, unfair predictions. In\nthis work, we analyze the sensitivity of causal fairness to unobserved\nconfounding. Our contributions are three-fold. First, we derive bounds for\ncausal fairness metrics under different sources of unobserved confounding. This\nenables practitioners to examine the sensitivity of their machine learning\nmodels to unobserved confounding in fairness-critical applications. Second, we\npropose a novel neural framework for learning fair predictions, which allows us\nto offer worst-case guarantees of the extent to which causal fairness can be\nviolated due to unobserved confounding. Third, we demonstrate the effectiveness\nof our framework in a series of experiments, including a real-world case study\nabout predicting prison sentences. To the best of our knowledge, ours is the\nfirst work to study causal fairness under unobserved confounding. To this end,\nour work is of direct practical value as a refutation strategy to ensure the\nfairness of predictions in high-stakes applications.\n", "date": "2023-11-30", "categories": []}
{"id": "2410.14679", "title": "HyperCausalLP: Causal Link Prediction using Hyper-Relational Knowledge\n  Graph", "abstract": "  Causal networks are often incomplete with missing causal links. This is due\nto various issues, such as missing observation data. Recent approaches to the\nissue of incomplete causal networks have used knowledge graph link prediction\nmethods to find the missing links. In the causal link A causes B causes C, the\ninfluence of A to C is influenced by B which is known as a mediator. Existing\napproaches using knowledge graph link prediction do not consider these mediated\ncausal links. This paper presents HyperCausalLP, an approach designed to find\nmissing causal links within a causal network with the help of mediator links.\nThe problem of missing links is formulated as a hyper-relational knowledge\ngraph completion. The approach uses a knowledge graph link prediction model\ntrained on a hyper-relational knowledge graph with the mediators. The approach\nis evaluated on a causal benchmark dataset, CLEVRER-Humans. Results show that\nthe inclusion of knowledge about mediators in causal link prediction using\nhyper-relational knowledge graph improves the performance on an average by\n5.94% mean reciprocal rank.\n", "date": "2024-09-12", "categories": []}
{"id": "2410.14680", "title": "Influence of Backdoor Paths on Causal Link Prediction", "abstract": "  The current method for predicting causal links in knowledge graphs uses\nweighted causal relations. For a given link between cause-effect entities, the\npresence of a confounder affects the causal link prediction, which can lead to\nspurious and inaccurate results. We aim to block these confounders using\nbackdoor path adjustment. Backdoor paths are non-causal association flows that\nconnect the \\textit{cause-entity} to the \\textit{effect-entity} through other\nvariables. Removing these paths ensures a more accurate prediction of causal\nlinks. This paper proposes CausalLPBack, a novel approach to causal link\nprediction that eliminates backdoor paths and uses knowledge graph link\nprediction methods. It extends the representation of causality in a\nneuro-symbolic framework, enabling the adoption and use of traditional causal\nAI concepts and methods. We demonstrate our approach using a causal reasoning\nbenchmark dataset of simulated videos. The evaluation involves a unique dataset\nsplitting method called the Markov-based split that's relevant for causal link\nprediction. The evaluation of the proposed approach demonstrates atleast 30\\%\nin MRR and 16\\% in Hits@K inflated performance for causal link prediction that\nis due to the bias introduced by backdoor paths for both baseline and weighted\ncausal relations.\n", "date": "2024-09-12", "categories": []}
{"id": "2410.15319", "title": "Causality for Large Language Models", "abstract": "  Recent breakthroughs in artificial intelligence have driven a paradigm shift,\nwhere large language models (LLMs) with billions or trillions of parameters are\ntrained on vast datasets, achieving unprecedented success across a series of\nlanguage tasks. However, despite these successes, LLMs still rely on\nprobabilistic modeling, which often captures spurious correlations rooted in\nlinguistic patterns and social stereotypes, rather than the true causal\nrelationships between entities and events. This limitation renders LLMs\nvulnerable to issues such as demographic biases, social stereotypes, and LLM\nhallucinations. These challenges highlight the urgent need to integrate\ncausality into LLMs, moving beyond correlation-driven paradigms to build more\nreliable and ethically aligned AI systems.\n  While many existing surveys and studies focus on utilizing prompt engineering\nto activate LLMs for causal knowledge or developing benchmarks to assess their\ncausal reasoning abilities, most of these efforts rely on human intervention to\nactivate pre-trained models. How to embed causality into the training process\nof LLMs and build more general and intelligent models remains unexplored.\nRecent research highlights that LLMs function as causal parrots, capable of\nreciting causal knowledge without truly understanding or applying it. These\nprompt-based methods are still limited to human interventional improvements.\nThis survey aims to address this gap by exploring how causality can enhance\nLLMs at every stage of their lifecycle-from token embedding learning and\nfoundation model training to fine-tuning, alignment, inference, and\nevaluation-paving the way for more interpretable, reliable, and\ncausally-informed models. Additionally, we further outline six promising future\ndirections to advance LLM development, enhance their causal reasoning\ncapabilities, and address the current limitations these models face.\n", "date": "2024-10-20", "categories": []}
{"id": "2410.15491", "title": "Structural Causality-based Generalizable Concept Discovery Models", "abstract": "  The rising need for explainable deep neural network architectures has\nutilized semantic concepts as explainable units. Several approaches utilizing\ndisentangled representation learning estimate the generative factors and\nutilize them as concepts for explaining DNNs. However, even though the\ngenerative factors for a dataset remain fixed, concepts are not fixed entities\nand vary based on downstream tasks. In this paper, we propose a disentanglement\nmechanism utilizing a variational autoencoder (VAE) for learning mutually\nindependent generative factors for a given dataset and subsequently learning\ntask-specific concepts using a structural causal model (SCM). Our method\nassumes generative factors and concepts to form a bipartite graph, with\ndirected causal edges from generative factors to concepts. Experiments are\nconducted on datasets with known generative factors: D-sprites and Shapes3D. On\nspecific downstream tasks, our proposed method successfully learns\ntask-specific concepts which are explained well by the causal edges from the\ngenerative factors. Lastly, separate from current causal concept discovery\nmethods, our methodology is generalizable to an arbitrary number of concepts\nand flexible to any downstream tasks.\n", "date": "2024-10-20", "categories": []}
{"id": "2410.15648", "title": "Linking Model Intervention to Causal Interpretation in Model Explanation", "abstract": "  Intervention intuition is often used in model explanation where the\nintervention effect of a feature on the outcome is quantified by the difference\nof a model prediction when the feature value is changed from the current value\nto the baseline value. Such a model intervention effect of a feature is\ninherently association. In this paper, we will study the conditions when an\nintuitive model intervention effect has a causal interpretation, i.e., when it\nindicates whether a feature is a direct cause of the outcome. This work links\nthe model intervention effect to the causal interpretation of a model. Such an\ninterpretation capability is important since it indicates whether a machine\nlearning model is trustworthy to domain experts. The conditions also reveal the\nlimitations of using a model intervention effect for causal interpretation in\nan environment with unobserved features. Experiments on semi-synthetic datasets\nhave been conducted to validate theorems and show the potential for using the\nmodel intervention effect for model interpretation.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.15655", "title": "Accounting for Missing Covariates in Heterogeneous Treatment Estimation", "abstract": "  Many applications of causal inference require using treatment effects\nestimated on a study population to make decisions in a separate target\npopulation. We consider the challenging setting where there are covariates that\nare observed in the target population that were not seen in the original study.\nOur goal is to estimate the tightest possible bounds on heterogeneous treatment\neffects conditioned on such newly observed covariates. We introduce a novel\npartial identification strategy based on ideas from ecological inference; the\nmain idea is that estimates of conditional treatment effects for the full\ncovariate set must marginalize correctly when restricted to only the covariates\nobserved in both populations. Furthermore, we introduce a bias-corrected\nestimator for these bounds and prove that it enjoys fast convergence rates and\nstatistical guarantees (e.g., asymptotic normality). Experimental results on\nboth real and synthetic data demonstrate that our framework can produce bounds\nthat are much tighter than would otherwise be possible.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.15706", "title": "Estimating Individual Dose-Response Curves under Unobserved Confounders\n  from Observational Data", "abstract": "  Estimating an individual's potential response to continuously varied\ntreatments is crucial for addressing causal questions across diverse domains,\nfrom healthcare to social sciences. However, existing methods are limited\neither to estimating causal effects of binary treatments, or scenarios where\nall confounding variables are measurable. In this work, we present ContiVAE, a\nnovel framework for estimating causal effects of continuous treatments,\nmeasured by individual dose-response curves, considering the presence of\nunobserved confounders using observational data. Leveraging a variational\nauto-encoder with a Tilted Gaussian prior distribution, ContiVAE models the\nhidden confounders as latent variables, and is able to predict the potential\noutcome of any treatment level for each individual while effectively capture\nthe heterogeneity among individuals. Experiments on semi-synthetic datasets\nshow that ContiVAE outperforms existing methods by up to 62%, demonstrating its\nrobustness and flexibility. Application on a real-world dataset illustrates its\npractical utility.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.15828", "title": "LLM4GRN: Discovering Causal Gene Regulatory Networks with LLMs --\n  Evaluation through Synthetic Data Generation", "abstract": "  Gene regulatory networks (GRNs) represent the causal relationships between\ntranscription factors (TFs) and target genes in single-cell RNA sequencing\n(scRNA-seq) data. Understanding these networks is crucial for uncovering\ndisease mechanisms and identifying therapeutic targets. In this work, we\ninvestigate the potential of large language models (LLMs) for GRN discovery,\nleveraging their learned biological knowledge alone or in combination with\ntraditional statistical methods. We develop a task-based evaluation strategy to\naddress the challenge of unavailable ground truth causal graphs. Specifically,\nwe use the GRNs suggested by LLMs to guide causal synthetic data generation and\ncompare the resulting data against the original dataset. Our statistical and\nbiological assessments show that LLMs can support statistical modeling and data\nsynthesis for biological research.\n", "date": "2024-10-21", "categories": []}
{"id": "2407.09632", "title": "Granger Causality in Extremes", "abstract": "  We introduce a rigorous mathematical framework for Granger causality in\nextremes, designed to identify causal links from extreme events in time series.\nGranger causality plays a pivotal role in uncovering directional relationships\namong time-varying variables. While this notion gains heightened importance\nduring extreme and highly volatile periods, state-of-the-art methods primarily\nfocus on causality within the body of the distribution, often overlooking\ncausal mechanisms that manifest only during extreme events. Our framework is\ndesigned to infer causality mainly from extreme events by leveraging the causal\ntail coefficient. We establish equivalences between causality in extremes and\nother causal concepts, including (classical) Granger causality, Sims causality,\nand structural causality. We prove other key properties of Granger causality in\nextremes and show that the framework is especially helpful under the presence\nof hidden confounders. We also propose a novel inference method for detecting\nthe presence of Granger causality in extremes from data. Our method is\nmodel-free, can handle non-linear and high-dimensional time series, outperforms\ncurrent state-of-the-art methods in all considered setups, both in performance\nand speed, and was found to uncover coherent effects when applied to financial\nand extreme weather observations.\n", "date": "2024-07-12", "categories": []}
{"id": "2410.12047", "title": "Testing Causal Explanations: A Case Study for Understanding the Effect\n  of Interventions on Chronic Kidney Disease", "abstract": "  Randomized controlled trials (RCTs) are the standard for evaluating the\neffectiveness of clinical interventions. To address the limitations of RCTs on\nreal-world populations, we developed a methodology that uses a large\nobservational electronic health record (EHR) dataset. Principles of regression\ndiscontinuity (rd) were used to derive randomized data subsets to test\nexpert-driven interventions using dynamic Bayesian Networks (DBNs)\ndo-operations. This combined method was applied to a chronic kidney disease\n(CKD) cohort of more than two million individuals and used to understand the\nassociational and causal relationships of CKD variables with respect to a\nsurrogate outcome of >=40% decline in estimated glomerular filtration rate\n(eGFR). The associational and causal analyses depicted similar findings across\nDBNs from two independent healthcare systems. The associational analysis showed\nthat the most influential variables were eGFR, urine albumin-to-creatinine\nratio, and pulse pressure, whereas the causal analysis showed eGFR as the most\ninfluential variable, followed by modifiable factors such as medications that\nmay impact kidney function over time. This methodology demonstrates how\nreal-world EHR data can be used to provide population-level insights to inform\nimproved healthcare delivery.\n", "date": "2024-10-15", "categories": []}
{"id": "2410.14127", "title": "Estimating the Causal Effects of T Cell Receptors", "abstract": "  A central question in human immunology is how a patient's repertoire of T\ncells impacts disease. Here, we introduce a method to infer the causal effects\nof T cell receptor (TCR) sequences on patient outcomes using observational TCR\nrepertoire sequencing data and clinical outcomes data. Our approach corrects\nfor unobserved confounders, such as a patient's environment and life history,\nby using the patient's immature, pre-selection TCR repertoire. The\npre-selection repertoire can be estimated from nonproductive TCR data, which is\nwidely available. It is generated by a randomized mutational process, V(D)J\nrecombination, which provides a natural experiment. We show formally how to use\nthe pre-selection repertoire to draw causal inferences, and develop a scalable\nneural-network estimator for our identification formula. Our method produces an\nestimate of the effect of interventions that add a specific TCR sequence to\npatient repertoires. As a demonstration, we use it to analyze the effects of\nTCRs on COVID-19 severity, uncovering potentially therapeutic TCRs that are (1)\nobserved in patients, (2) bind SARS-CoV-2 antigens in vitro and (3) have strong\npositive effects on clinical outcomes.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14146", "title": "CausalChat: Interactive Causal Model Development and Refinement Using\n  Large Language Models", "abstract": "  Causal networks are widely used in many fields to model the complex\nrelationships between variables. A recent approach has sought to construct\ncausal networks by leveraging the wisdom of crowds through the collective\nparticipation of humans. While this can yield detailed causal networks that\nmodel the underlying phenomena quite well, it requires a large number of\nindividuals with domain understanding. We adopt a different approach:\nleveraging the causal knowledge that large language models, such as OpenAI's\nGPT-4, have learned by ingesting massive amounts of literature. Within a\ndedicated visual analytics interface, called CausalChat, users explore single\nvariables or variable pairs recursively to identify causal relations, latent\nvariables, confounders, and mediators, constructing detailed causal networks\nthrough conversation. Each probing interaction is translated into a tailored\nGPT-4 prompt and the response is conveyed through visual representations which\nare linked to the generated text for explanations. We demonstrate the\nfunctionality of CausalChat across diverse data contexts and conduct user\nstudies involving both domain experts and laypersons.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14236", "title": "A Novel Method to Metigate Demographic and Expert Bias in ICD Coding\n  with Causal Inference", "abstract": "  ICD(International Classification of Diseases) coding involves assigning ICD\ncodes to patients visit based on their medical notes. Considering ICD coding as\na multi-label text classification task, researchers have developed\nsophisticated methods. Despite progress, these models often suffer from label\nimbalance and may develop spurious correlations with demographic factors.\nAdditionally, while human coders assign ICD codes, the inclusion of irrelevant\ninformation from unrelated experts introduces biases. To combat these issues,\nwe propose a novel method to mitigate Demographic and Expert biases in ICD\ncoding through Causal Inference (DECI). We provide a novel causality-based\ninterpretation in ICD Coding that models make predictions by three distinct\npathways. And based counterfactual reasoning, DECI mitigate demographic and\nexpert biases. Experimental results show that DECI outperforms state-of-the-art\nmodels, offering a significant advancement in accurate and unbiased ICD coding.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14436", "title": "Learning to refine domain knowledge for biological network inference", "abstract": "  Perturbation experiments allow biologists to discover causal relationships\nbetween variables of interest, but the sparsity and high dimensionality of\nthese data pose significant challenges for causal structure learning\nalgorithms. Biological knowledge graphs can bootstrap the inference of causal\nstructures in these situations, but since they compile vastly diverse\ninformation, they can bias predictions towards well-studied systems.\nAlternatively, amortized causal structure learning algorithms encode inductive\nbiases through data simulation and train supervised models to recapitulate\nthese synthetic graphs. However, realistically simulating biology is arguably\neven harder than understanding a specific system. In this work, we take\ninspiration from both strategies and propose an amortized algorithm for\nrefining domain knowledge, based on data observations. On real and synthetic\ndatasets, we show that our approach outperforms baselines in recovering ground\ntruth causal graphs and identifying errors in the prior knowledge with limited\ninterventional data.\n", "date": "2024-10-18", "categories": []}
{"id": "2301.06615", "title": "Data-Driven Estimation of Heterogeneous Treatment Effects", "abstract": "  Estimating how a treatment affects different individuals, known as\nheterogeneous treatment effect estimation, is an important problem in empirical\nsciences. In the last few years, there has been a considerable interest in\nadapting machine learning algorithms to the problem of estimating heterogeneous\neffects from observational and experimental data. However, these algorithms\noften make strong assumptions about the observed features in the data and\nignore the structure of the underlying causal model, which can lead to biased\nestimation. At the same time, the underlying causal mechanism is rarely known\nin real-world datasets, making it hard to take it into consideration. In this\nwork, we provide a survey of state-of-the-art data-driven methods for\nheterogeneous treatment effect estimation using machine learning, broadly\ncategorizing them as methods that focus on counterfactual prediction and\nmethods that directly estimate the causal effect. We also provide an overview\nof a third category of methods which rely on structural causal models and learn\nthe model structure from data. Our empirical evaluation under various\nunderlying structural model mechanisms shows the advantages and deficiencies of\nexisting estimators and of the metrics for measuring their performance.\n", "date": "2023-01-16", "categories": []}
{"id": "2310.19300", "title": "Stage-Aware Learning for Dynamic Treatments", "abstract": "  Recent advances in dynamic treatment regimes (DTRs) facilitate the search for\noptimal treatments, which are tailored to individuals' specific needs and able\nto maximize their expected clinical benefits. However, existing algorithms\nrelying on consistent trajectories, such as inverse probability weighting\nestimators (IPWEs), could suffer from insufficient sample size under optimal\ntreatments and a growing number of decision-making stages, particularly in the\ncontext of chronic diseases. To address these challenges, we propose a novel\nindividualized learning method which estimates the DTR with a focus on\nprioritizing alignment between the observed treatment trajectory and the one\nobtained by the optimal regime across decision stages. By relaxing the\nrestriction that the observed trajectory must be fully aligned with the optimal\ntreatments, our approach substantially improves the sample efficiency and\nstability of IPWE-based methods. In particular, the proposed learning scheme\nbuilds a more general framework which includes the popular outcome weighted\nlearning framework as a special case of ours. Moreover, we introduce the notion\nof stage importance scores along with an attention mechanism to explicitly\naccount for heterogeneity among decision stages. We establish the theoretical\nproperties of the proposed approach, including the Fisher consistency and\nfinite-sample performance bound. Empirically, we evaluate the proposed method\nin extensive simulated environments and a real case study for the COVID-19\npandemic.\n", "date": "2023-10-30", "categories": []}
{"id": "2405.18206", "title": "Multi-CATE: Multi-Accurate Conditional Average Treatment Effect\n  Estimation Robust to Unknown Covariate Shifts", "abstract": "  Estimating heterogeneous treatment effects is important to tailor treatments\nto those individuals who would most likely benefit. However, conditional\naverage treatment effect predictors may often be trained on one population but\npossibly deployed on different, possibly unknown populations. We use\nmethodology for learning multi-accurate predictors to post-process CATE\nT-learners (differenced regressions) to become robust to unknown covariate\nshifts at the time of deployment. The method works in general for\npseudo-outcome regression, such as the DR-learner. We show how this approach\ncan combine (large) confounded observational and (smaller) randomized datasets\nby learning a confounded predictor from the observational dataset, and auditing\nfor multi-accuracy on the randomized controlled trial. We show improvements in\nbias and mean squared error in simulations with increasingly larger covariate\nshift, and on a semi-synthetic case study of a parallel large observational\nstudy and smaller randomized controlled experiment. Overall, we establish a\nconnection between methods developed for multi-distribution learning and\nachieve appealing desiderata (e.g. external validity) in causal inference and\nmachine learning.\n", "date": "2024-05-28", "categories": []}
{"id": "2410.12799", "title": "Ads Supply Personalization via Doubly Robust Learning", "abstract": "  Ads supply personalization aims to balance the revenue and user engagement,\ntwo long-term objectives in social media ads, by tailoring the ad quantity and\ndensity. In the industry-scale system, the challenge for ads supply lies in\nmodeling the counterfactual effects of a conservative supply treatment (e.g., a\nsmall density change) over an extended duration. In this paper, we present a\nstreamlined framework for personalized ad supply. This framework optimally\nutilizes information from data collection policies through the doubly robust\nlearning. Consequently, it significantly improves the accuracy of long-term\ntreatment effect estimates. Additionally, its low-complexity design not only\nresults in computational cost savings compared to existing methods, but also\nmakes it scalable for billion-scale applications. Through both offline\nexperiments and online production tests, the framework consistently\ndemonstrated significant improvements in top-line business metrics over months.\nThe framework has been fully deployed to live traffic in one of the world's\nlargest social media platforms.\n", "date": "2024-09-29", "categories": []}
{"id": "2311.07454", "title": "Causal Discovery under Latent Class Confounding", "abstract": "  An acyclic causal structure can be described with directed acyclic graph\n(DAG), where arrows indicate the possibility of direct causation. The task of\nlearning this structure from data is known as \"causal discovery.\" Diverse\npopulations or changing environments can sometimes give rise to data that is\nheterogeneous in the following sense: each population/environment is a \"source\"\nwhich idiosyncratically determines the forms of those direct causal effects.\nFrom this perspective, the source is a latent common cause for every observed\nvariable. While some methods for causal discovery are able to work around\nlatent confounding in special cases, especially when only few observables are\nconfounded, a global confounder is a difficult challenge. The only known ways\nto deal with latent global confounding involve assumptions that limit the\nstructural equations and/or noise functions. We demonstrate that globally\nconfounded causal structures can still be identifiable with arbitrary\nstructural equations and noise functions, so long as the number of latent\nclasses remains small relative to the size and sparsity of the underlying DAG.\n", "date": "2023-11-13", "categories": []}
{"id": "2301.07427", "title": "Boosting Synthetic Data Generation with Effective Nonlinear Causal\n  Discovery", "abstract": "  Synthetic data generation has been widely adopted in software testing, data\nprivacy, imbalanced learning, and artificial intelligence explanation. In all\nsuch contexts, it is crucial to generate plausible data samples. A common\nassumption of approaches widely used for data generation is the independence of\nthe features. However, typically, the variables of a dataset depend on one\nanother, and these dependencies are not considered in data generation leading\nto the creation of implausible records. The main problem is that dependencies\namong variables are typically unknown. In this paper, we design a synthetic\ndataset generator for tabular data that can discover nonlinear causalities\namong the variables and use them at generation time. State-of-the-art methods\nfor nonlinear causal discovery are typically inefficient. We boost them by\nrestricting the causal discovery among the features appearing in the frequent\npatterns efficiently retrieved by a pattern mining algorithm. We design a\nframework for generating synthetic datasets with known causalities to validate\nour proposal. Broad experimentation on many synthetic and real datasets with\nknown causalities shows the effectiveness of the proposed method.\n", "date": "2023-01-18", "categories": []}
{"id": "2306.05082", "title": "The Importance of Time in Causal Algorithmic Recourse", "abstract": "  The application of Algorithmic Recourse in decision-making is a promising\nfield that offers practical solutions to reverse unfavorable decisions.\nHowever, the inability of these methods to consider potential dependencies\namong variables poses a significant challenge due to the assumption of feature\nindependence. Recent advancements have incorporated knowledge of causal\ndependencies, thereby enhancing the quality of the recommended recourse\nactions. Despite these improvements, the inability to incorporate the temporal\ndimension remains a significant limitation of these approaches. This is\nparticularly problematic as identifying and addressing the root causes of\nundesired outcomes requires understanding time-dependent relationships between\nvariables. In this work, we motivate the need to integrate the temporal\ndimension into causal algorithmic recourse methods to enhance recommendations'\nplausibility and reliability. The experimental evaluation highlights the\nsignificance of the role of time in this field.\n", "date": "2023-06-08", "categories": []}
{"id": "2311.05794", "title": "An Experimental Design for Anytime-Valid Causal Inference on Multi-Armed\n  Bandits", "abstract": "  Experimentation is crucial for managers to rigorously quantify the value of a\nchange and determine if it leads to a statistically significant improvement\nover the status quo. As companies increasingly mandate that all changes undergo\nexperimentation before widespread release, two challenges arise: (1) minimizing\nthe proportion of customers assigned to the inferior treatment and (2)\nincreasing experimentation velocity by enabling data-dependent stopping. This\npaper addresses both challenges by introducing the Mixture Adaptive Design\n(MAD), a new experimental design for multi-armed bandit (MAB) algorithms that\nenables anytime-valid inference on the Average Treatment Effect (ATE) for\n\\emph{any} MAB algorithm. Intuitively, MAD \"mixes\" any bandit algorithm with a\nBernoulli design, where at each time step, the probability of assigning a unit\nvia the Bernoulli design is determined by a user-specified deterministic\nsequence that can converge to zero. This sequence lets managers directly\ncontrol the trade-off between regret minimization and inferential precision.\nUnder mild conditions on the rate the sequence converges to zero, we provide a\nconfidence sequence that is asymptotically anytime-valid and guaranteed to\nshrink around the true ATE. Hence, when the true ATE converges to a non-zero\nvalue, the MAD confidence sequence is guaranteed to exclude zero in finite\ntime. Therefore, the MAD enables managers to stop experiments early while\nensuring valid inference, enhancing both the efficiency and reliability of\nadaptive experiments. Empirically, we demonstrate that the MAD achieves\nfinite-sample anytime-validity while accurately and precisely estimating the\nATE, all without incurring significant losses in reward compared to standard\nbandit designs.\n", "date": "2023-11-09", "categories": []}
{"id": "2410.11385", "title": "Do LLMs Have the Generalization Ability in Conducting Causal Inference?", "abstract": "  In causal inference, generalization capability refers to the ability to\nconduct causal inference methods on new data to estimate the causal-effect\nbetween unknown phenomenon, which is crucial for expanding the boundaries of\nknowledge. Studies have evaluated the causal inference capabilities of Large\nLanguage Models (LLMs) concerning known phenomena, yet the generalization\ncapabilities of LLMs concerning unseen phenomena remain unexplored. In this\npaper, we selected four tasks: Causal Path Discovery (CP), Backdoor Adjustment\n(BA), Factual Inference (FI), and Counterfactual Inference (CI) as\nrepresentatives of causal inference tasks. To generate evaluation questions\nabout previously unseen phenomena in new data on the four tasks, we propose a\nbenchmark generation framework, which employs randomly generated graphs and\nnode names to formulate questions within hypothetical new causal scenarios.\nBased on this framework, we compile a benchmark dataset of varying levels of\nquestion complexity. We extensively tested the generalization capabilities of\nfive leading LLMs across four tasks. Experiment results reveal that while LLMs\nexhibit good generalization performance in solving simple CP, FI, and complex\nCI questions, they encounter difficulties when tackling BA questions and face\nobvious performance fluctuations as the problem complexity changes.\nFurthermore, when the names of phenomena incorporate existing terms, even if\nthese names are entirely novel, their generalization performance can still be\nhindered by interference from familiar terms.\n", "date": "2024-10-15", "categories": []}
{"id": "2410.08976", "title": "Learning Representations of Instruments for Partial Identification of\n  Treatment Effects", "abstract": "  Reliable estimation of treatment effects from observational data is important\nin many disciplines such as medicine. However, estimation is challenging when\nunconfoundedness as a standard assumption in the causal inference literature is\nviolated. In this work, we leverage arbitrary (potentially high-dimensional)\ninstruments to estimate bounds on the conditional average treatment effect\n(CATE). Our contributions are three-fold: (1) We propose a novel approach for\npartial identification through a mapping of instruments to a discrete\nrepresentation space so that we yield valid bounds on the CATE. This is crucial\nfor reliable decision-making in real-world applications. (2) We derive a\ntwo-step procedure that learns tight bounds using a tailored neural\npartitioning of the latent instrument space. As a result, we avoid instability\nissues due to numerical approximations or adversarial training. Furthermore,\nour procedure aims to reduce the estimation variance in finite-sample settings\nto yield more reliable estimates. (3) We show theoretically that our procedure\nobtains valid bounds while reducing estimation variance. We further perform\nextensive experiments to demonstrate the effectiveness across various settings.\nOverall, our procedure offers a novel path for practitioners to make use of\npotentially high-dimensional instruments (e.g., as in Mendelian randomization).\n", "date": "2024-10-11", "categories": []}
{"id": "2410.10463", "title": "TABCF: Counterfactual Explanations for Tabular Data Using a\n  Transformer-Based VAE", "abstract": "  In the field of Explainable AI (XAI), counterfactual (CF) explanations are\none prominent method to interpret a black-box model by suggesting changes to\nthe input that would alter a prediction. In real-world applications, the input\nis predominantly in tabular form and comprised of mixed data types and complex\nfeature interdependencies. These unique data characteristics are difficult to\nmodel, and we empirically show that they lead to bias towards specific feature\ntypes when generating CFs. To overcome this issue, we introduce TABCF, a CF\nexplanation method that leverages a transformer-based Variational Autoencoder\n(VAE) tailored for modeling tabular data. Our approach uses transformers to\nlearn a continuous latent space and a novel Gumbel-Softmax detokenizer that\nenables precise categorical reconstruction while preserving end-to-end\ndifferentiability. Extensive quantitative evaluation on five financial datasets\ndemonstrates that TABCF does not exhibit bias toward specific feature types,\nand outperforms existing methods in producing effective CFs that align with\ncommon CF desiderata.\n", "date": "2024-10-14", "categories": []}
{"id": "2410.02844", "title": "CAnDOIT: Causal Discovery with Observational and Interventional Data\n  from Time-Series", "abstract": "  The study of cause-and-effect is of the utmost importance in many branches of\nscience, but also for many practical applications of intelligent systems. In\nparticular, identifying causal relationships in situations that include hidden\nfactors is a major challenge for methods that rely solely on observational data\nfor building causal models. This paper proposes CAnDOIT, a causal discovery\nmethod to reconstruct causal models using both observational and interventional\ntime-series data. The use of interventional data in the causal analysis is\ncrucial for real-world applications, such as robotics, where the scenario is\nhighly complex and observational data alone are often insufficient to uncover\nthe correct causal structure. Validation of the method is performed initially\non randomly generated synthetic models and subsequently on a well-known\nbenchmark for causal structure learning in a robotic manipulation environment.\nThe experiments demonstrate that the approach can effectively handle data from\ninterventions and exploit them to enhance the accuracy of the causal analysis.\nA Python implementation of CAnDOIT has also been developed and is publicly\navailable on GitHub: https://github.com/lcastri/causalflow.\n", "date": "2024-10-03", "categories": []}
{"id": "2410.08770", "title": "Causal machine learning for predicting treatment outcomes", "abstract": "  Causal machine learning (ML) offers flexible, data-driven methods for\npredicting treatment outcomes including efficacy and toxicity, thereby\nsupporting the assessment and safety of drugs. A key benefit of causal ML is\nthat it allows for estimating individualized treatment effects, so that\nclinical decision-making can be personalized to individual patient profiles.\nCausal ML can be used in combination with both clinical trial data and\nreal-world data, such as clinical registries and electronic health records, but\ncaution is needed to avoid biased or incorrect predictions. In this\nPerspective, we discuss the benefits of causal ML (relative to traditional\nstatistical or ML approaches) and outline the key components and steps.\nFinally, we provide recommendations for the reliable use of causal ML and\neffective translation into the clinic.\n", "date": "2024-10-11", "categories": []}
{"id": "2410.08787", "title": "Efficient Differentiable Discovery of Causal Order", "abstract": "  In the algorithm Intersort, Chevalley et al. (2024) proposed a score-based\nmethod to discover the causal order of variables in a Directed Acyclic Graph\n(DAG) model, leveraging interventional data to outperform existing methods.\nHowever, as a score-based method over the permutahedron, Intersort is\ncomputationally expensive and non-differentiable, limiting its ability to be\nutilised in problems involving large-scale datasets, such as those in genomics\nand climate models, or to be integrated into end-to-end gradient-based learning\nframeworks. We address this limitation by reformulating Intersort using\ndifferentiable sorting and ranking techniques. Our approach enables scalable\nand differentiable optimization of causal orderings, allowing the continuous\nscore function to be incorporated as a regularizer in downstream tasks.\nEmpirical results demonstrate that causal discovery algorithms benefit\nsignificantly from regularizing on the causal order, underscoring the\neffectiveness of our method. Our work opens the door to efficiently\nincorporating regularization for causal order into the training of\ndifferentiable models and thereby addresses a long-standing limitation of\npurely associational supervised learning.\n", "date": "2024-10-11", "categories": []}
{"id": "2410.08816", "title": "Uncertainty-Aware Optimal Treatment Selection for Clinical Time Series", "abstract": "  In personalized medicine, the ability to predict and optimize treatment\noutcomes across various time frames is essential. Additionally, the ability to\nselect cost-effective treatments within specific budget constraints is\ncritical. Despite recent advancements in estimating counterfactual\ntrajectories, a direct link to optimal treatment selection based on these\nestimates is missing. This paper introduces a novel method integrating\ncounterfactual estimation techniques and uncertainty quantification to\nrecommend personalized treatment plans adhering to predefined cost constraints.\nOur approach is distinctive in its handling of continuous treatment variables\nand its incorporation of uncertainty quantification to improve prediction\nreliability. We validate our method using two simulated datasets, one focused\non the cardiovascular system and the other on COVID-19. Our findings indicate\nthat our method has robust performance across different counterfactual\nestimation baselines, showing that introducing uncertainty quantification in\nthese settings helps the current baselines in finding more reliable and\naccurate treatment selection. The robustness of our method across various\nsettings highlights its potential for broad applicability in personalized\nhealthcare solutions.\n", "date": "2024-10-11", "categories": []}
{"id": "2410.08924", "title": "DiffPO: A causal diffusion model for learning distributions of potential\n  outcomes", "abstract": "  Predicting potential outcomes of interventions from observational data is\ncrucial for decision-making in medicine, but the task is challenging due to the\nfundamental problem of causal inference. Existing methods are largely limited\nto point estimates of potential outcomes with no uncertain quantification;\nthus, the full information about the distributions of potential outcomes is\ntypically ignored. In this paper, we propose a novel causal diffusion model\ncalled DiffPO, which is carefully designed for reliable inferences in medicine\nby learning the distribution of potential outcomes. In our DiffPO, we leverage\na tailored conditional denoising diffusion model to learn complex\ndistributions, where we address the selection bias through a novel orthogonal\ndiffusion loss. Another strength of our DiffPO method is that it is highly\nflexible (e.g., it can also be used to estimate different causal quantities\nsuch as CATE). Across a wide range of experiments, we show that our method\nachieves state-of-the-art performance.\n", "date": "2024-10-11", "categories": []}
{"id": "2302.00860", "title": "Modeling Causal Mechanisms with Diffusion Models for Interventional and\n  Counterfactual Queries", "abstract": "  We consider the problem of answering observational, interventional, and\ncounterfactual queries in a causally sufficient setting where only\nobservational data and the causal graph are available. Utilizing the recent\ndevelopments in diffusion models, we introduce diffusion-based causal models\n(DCM) to learn causal mechanisms, that generate unique latent encodings. These\nencodings enable us to directly sample under interventions and perform\nabduction for counterfactuals. Diffusion models are a natural fit here, since\nthey can encode each node to a latent representation that acts as a proxy for\nexogenous noise. Our empirical evaluations demonstrate significant improvements\nover existing state-of-the-art methods for answering causal queries.\nFurthermore, we provide theoretical results that offer a methodology for\nanalyzing counterfactual estimation in general encoder-decoder models, which\ncould be useful in settings beyond our proposed approach.\n", "date": "2023-02-02", "categories": []}
{"id": "2410.07673", "title": "Multimodal Clickbait Detection by De-confounding Biases Using Causal\n  Representation Inference", "abstract": "  This paper focuses on detecting clickbait posts on the Web. These posts often\nuse eye-catching disinformation in mixed modalities to mislead users to click\nfor profit. That affects the user experience and thus would be blocked by\ncontent provider. To escape detection, malicious creators use tricks to add\nsome irrelevant non-bait content into bait posts, dressing them up as legal to\nfool the detector. This content often has biased relations with non-bait\nlabels, yet traditional detectors tend to make predictions based on simple\nco-occurrence rather than grasping inherent factors that lead to malicious\nbehavior. This spurious bias would easily cause misjudgments. To address this\nproblem, we propose a new debiased method based on causal inference. We first\nemploy a set of features in multiple modalities to characterize the posts.\nConsidering these features are often mixed up with unknown biases, we then\ndisentangle three kinds of latent factors from them, including the invariant\nfactor that indicates intrinsic bait intention; the causal factor which\nreflects deceptive patterns in a certain scenario, and non-causal noise. By\neliminating the noise that causes bias, we can use invariant and causal factors\nto build a robust model with good generalization ability. Experiments on three\npopular datasets show the effectiveness of our approach.\n", "date": "2024-10-10", "categories": []}
{"id": "2410.05890", "title": "Ordering-Based Causal Discovery for Linear and Nonlinear Relations", "abstract": "  Identifying causal relations from purely observational data typically\nrequires additional assumptions on relations and/or noise. Most current methods\nrestrict their analysis to datasets that are assumed to have pure linear or\nnonlinear relations, which is often not reflective of real-world datasets that\ncontain a combination of both. This paper presents CaPS, an ordering-based\ncausal discovery algorithm that effectively handles linear and nonlinear\nrelations. CaPS introduces a novel identification criterion for topological\nordering and incorporates the concept of \"parent score\" during the\npost-processing optimization stage. These scores quantify the strength of the\naverage causal effect, helping to accelerate the pruning process and correct\ninaccurate predictions in the pruning step. Experimental results demonstrate\nthat our proposed solutions outperform state-of-the-art baselines on synthetic\ndata with varying ratios of linear and nonlinear relations. The results\nobtained from real-world data also support the competitiveness of CaPS. Code\nand datasets are available at https://github.com/E2real/CaPS.\n", "date": "2024-10-08", "categories": []}
{"id": "2410.06349", "title": "Robust Domain Generalisation with Causal Invariant Bayesian Neural\n  Networks", "abstract": "  Deep neural networks can obtain impressive performance on various tasks under\nthe assumption that their training domain is identical to their target domain.\nPerformance can drop dramatically when this assumption does not hold. One\nexplanation for this discrepancy is the presence of spurious domain-specific\ncorrelations in the training data that the network exploits. Causal mechanisms,\nin the other hand, can be made invariant under distribution changes as they\nallow disentangling the factors of distribution underlying the data generation.\nYet, learning causal mechanisms to improve out-of-distribution generalisation\nremains an under-explored area. We propose a Bayesian neural architecture that\ndisentangles the learning of the the data distribution from the inference\nprocess mechanisms. We show theoretically and experimentally that our model\napproximates reasoning under causal interventions. We demonstrate the\nperformance of our method, outperforming point estimate-counterparts, on\nout-of-distribution image recognition tasks where the data distribution acts as\nstrong adversarial confounders.\n", "date": "2024-10-08", "categories": []}
{"id": "2410.06392", "title": "Counterfactual Causal Inference in Natural Language with Large Language\n  Models", "abstract": "  Causal structure discovery methods are commonly applied to structured data\nwhere the causal variables are known and where statistical testing can be used\nto assess the causal relationships. By contrast, recovering a causal structure\nfrom unstructured natural language data such as news articles contains numerous\nchallenges due to the absence of known variables or counterfactual data to\nestimate the causal links. Large Language Models (LLMs) have shown promising\nresults in this direction but also exhibit limitations. This work investigates\nLLM's abilities to build causal graphs from text documents and perform\ncounterfactual causal inference. We propose an end-to-end causal structure\ndiscovery and causal inference method from natural language: we first use an\nLLM to extract the instantiated causal variables from text data and build a\ncausal graph. We merge causal graphs from multiple data sources to represent\nthe most exhaustive set of causes possible. We then conduct counterfactual\ninference on the estimated graph. The causal graph conditioning allows\nreduction of LLM biases and better represents the causal estimands. We use our\nmethod to show that the limitations of LLMs in counterfactual causal reasoning\ncome from prediction errors and propose directions to mitigate them. We\ndemonstrate the applicability of our method on real-world news articles.\n", "date": "2024-10-08", "categories": []}
{"id": "2410.07013", "title": "Causal Representation Learning in Temporal Data via Single-Parent\n  Decoding", "abstract": "  Scientific research often seeks to understand the causal structure underlying\nhigh-level variables in a system. For example, climate scientists study how\nphenomena, such as El Ni\\~no, affect other climate processes at remote\nlocations across the globe. However, scientists typically collect low-level\nmeasurements, such as geographically distributed temperature readings. From\nthese, one needs to learn both a mapping to causally-relevant latent variables,\nsuch as a high-level representation of the El Ni\\~no phenomenon and other\nprocesses, as well as the causal model over them. The challenge is that this\ntask, called causal representation learning, is highly underdetermined from\nobservational data alone, requiring other constraints during learning to\nresolve the indeterminacies. In this work, we consider a temporal model with a\nsparsity assumption, namely single-parent decoding: each observed low-level\nvariable is only affected by a single latent variable. Such an assumption is\nreasonable in many scientific applications that require finding groups of\nlow-level variables, such as extracting regions from geographically gridded\nmeasurement data in climate research or capturing brain regions from neural\nactivity data. We demonstrate the identifiability of the resulting model and\npropose a differentiable method, Causal Discovery with Single-parent Decoding\n(CDSD), that simultaneously learns the underlying latents and a causal graph\nover them. We assess the validity of our theoretical results using simulated\ndata and showcase the practical validity of our method in an application to\nreal-world data from the climate science field.\n", "date": "2024-10-09", "categories": []}
{"id": "2410.07135", "title": "Causal Inference with Double/Debiased Machine Learning for Evaluating\n  the Health Effects of Multiple Mismeasured Pollutants", "abstract": "  One way to quantify exposure to air pollution and its constituents in\nepidemiologic studies is to use an individual's nearest monitor. This strategy\nresults in potential inaccuracy in the actual personal exposure, introducing\nbias in estimating the health effects of air pollution and its constituents,\nespecially when evaluating the causal effects of correlated multi-pollutant\nconstituents measured with correlated error. This paper addresses estimation\nand inference for the causal effect of one constituent in the presence of other\nPM2.5 constituents, accounting for measurement error and correlations. We used\na linear regression calibration model, fitted with generalized estimating\nequations in an external validation study, and extended a double/debiased\nmachine learning (DML) approach to correct for measurement error and estimate\nthe effect of interest in the main study. We demonstrated that the DML\nestimator with regression calibration is consistent and derived its asymptotic\nvariance. Simulations showed that the proposed estimator reduced bias and\nattained nominal coverage probability across most simulation settings. We\napplied this method to assess the causal effects of PM2.5 constituents on\ncognitive function in the Nurses' Health Study and identified two PM2.5\nconstituents, Br and Mn, that showed a negative causal effect on cognitive\nfunction after measurement error correction.\n", "date": "2024-09-22", "categories": []}
{"id": "2410.05177", "title": "Are causal effect estimations enough for optimal recommendations under\n  multitreatment scenarios?", "abstract": "  When making treatment selection decisions, it is essential to include a\ncausal effect estimation analysis to compare potential outcomes under different\ntreatments or controls, assisting in optimal selection. However, merely\nestimating individual treatment effects may not suffice for truly optimal\ndecisions. Our study addressed this issue by incorporating additional criteria,\nsuch as the estimations' uncertainty, measured by the conditional\nvalue-at-risk, commonly used in portfolio and insurance management. For\ncontinuous outcomes observable before and after treatment, we incorporated a\nspecific prediction condition. We prioritized treatments that could yield\noptimal treatment effect results and lead to post-treatment outcomes more\ndesirable than pretreatment levels, with the latter condition being called the\nprediction criterion. With these considerations, we propose a comprehensive\nmethodology for multitreatment selection. Our approach ensures satisfaction of\nthe overlap assumption, crucial for comparing outcomes for treated and control\ngroups, by training propensity score models as a preliminary step before\nemploying traditional causal models. To illustrate a practical application of\nour methodology, we applied it to the credit card limit adjustment problem.\nAnalyzing a fintech company's historical data, we found that relying solely on\ncounterfactual predictions was inadequate for appropriate credit line\nmodifications. Incorporating our proposed additional criteria significantly\nenhanced policy performance.\n", "date": "2024-10-07", "categories": []}
{"id": "2410.01822", "title": "The Importance of Causality in Decision Making: A Perspective on\n  Recommender Systems", "abstract": "  Causality is receiving increasing attention in the Recommendation Systems\n(RSs) community, which has realised that RSs could greatly benefit from\ncausality to transform accurate predictions into effective and explainable\ndecisions. Indeed, the RS literature has repeatedly highlighted that, in\nreal-world scenarios, recommendation algorithms suffer many types of biases\nsince assumptions ensuring unbiasedness are likely not met. In this discussion\npaper, we formulate the RS problem in terms of causality, using potential\noutcomes and structural causal models, by giving formal definitions of the\ncausal quantities to be estimated and a general causal graph to serve as a\nreference to foster future research and development.\n", "date": "2024-09-16", "categories": []}
{"id": "2410.01221", "title": "Induced Covariance for Causal Discovery in Linear Sparse Structures", "abstract": "  Causal models seek to unravel the cause-effect relationships among variables\nfrom observed data, as opposed to mere mappings among them, as traditional\nregression models do. This paper introduces a novel causal discovery algorithm\ndesigned for settings in which variables exhibit linearly sparse relationships.\nIn such scenarios, the causal links represented by directed acyclic graphs\n(DAGs) can be encapsulated in a structural matrix. The proposed approach\nleverages the structural matrix's ability to reconstruct data and the\nstatistical properties it imposes on the data to identify the correct\nstructural matrix. This method does not rely on independence tests or graph\nfitting procedures, making it suitable for scenarios with limited training\ndata. Simulation results demonstrate that the proposed method outperforms the\nwell-known PC, GES, BIC exact search, and LINGAM-based methods in recovering\nlinearly sparse causal structures.\n", "date": "2024-10-02", "categories": []}
{"id": "2410.01227", "title": "See Me and Believe Me: Causality and Intersectionality in Testimonial\n  Injustice in Healthcare", "abstract": "  In medical settings, it is critical that all who are in need of care are\ncorrectly heard and understood. When this is not the case due to prejudices a\nlistener has, the speaker is experiencing \\emph{testimonial injustice}, which,\nbuilding upon recent work, we quantify by the presence of several categories of\nunjust vocabulary in medical notes. In this paper, we use FCI, a causal\ndiscovery method, to study the degree to which certain demographic features\ncould lead to marginalization (e.g., age, gender, and race) by way of\ncontributing to testimonial injustice. To achieve this, we review physicians'\nnotes for each patient, where we identify occurrences of unjust vocabulary,\nalong with the demographic features present, and use causal discovery to build\na Structural Causal Model (SCM) relating those demographic features to\ntestimonial injustice. We analyze and discuss the resulting SCMs to show the\ninteraction of these factors and how they influence the experience of\ninjustice. Despite the potential presence of some confounding variables, we\nobserve how one contributing feature can make a person more prone to\nexperiencing another contributor of testimonial injustice. There is no single\nroot of injustice and thus intersectionality cannot be ignored. These results\ncall for considering more than singular or equalized attributes of who a person\nis when analyzing and improving their experiences of bias and injustice. This\nwork is thus a first foray at using causal discovery to understand the nuanced\nexperiences of patients in medical settings, and its insights could be used to\nguide design principles throughout healthcare, to build trust and promote\nbetter patient care.\n", "date": "2024-10-02", "categories": []}
{"id": "2410.01658", "title": "Smaller Confidence Intervals From IPW Estimators via Data-Dependent\n  Coarsening", "abstract": "  Inverse propensity-score weighted (IPW) estimators are prevalent in causal\ninference for estimating average treatment effects in observational studies.\nUnder unconfoundedness, given accurate propensity scores and $n$ samples, the\nsize of confidence intervals of IPW estimators scales down with $n$, and,\nseveral of their variants improve the rate of scaling. However, neither IPW\nestimators nor their variants are robust to inaccuracies: even if a single\ncovariate has an $\\varepsilon>0$ additive error in the propensity score, the\nsize of confidence intervals of these estimators can increase arbitrarily.\nMoreover, even without errors, the rate with which the confidence intervals of\nthese estimators go to zero with $n$ can be arbitrarily slow in the presence of\nextreme propensity scores (those close to 0 or 1).\n  We introduce a family of Coarse IPW (CIPW) estimators that captures existing\nIPW estimators and their variants. Each CIPW estimator is an IPW estimator on a\ncoarsened covariate space, where certain covariates are merged. Under mild\nassumptions, e.g., Lipschitzness in expected outcomes and sparsity of extreme\npropensity scores, we give an efficient algorithm to find a robust estimator:\ngiven $\\varepsilon$-inaccurate propensity scores and $n$ samples, its\nconfidence interval size scales with $\\varepsilon+1/\\sqrt{n}$. In contrast,\nunder the same assumptions, existing estimators' confidence interval sizes are\n$\\Omega(1)$ irrespective of $\\varepsilon$ and $n$. Crucially, our estimator is\ndata-dependent and we show that no data-independent CIPW estimator can be\nrobust to inaccuracies.\n", "date": "2024-10-02", "categories": []}
{"id": "2301.13392", "title": "Combinatorial Causal Bandits without Graph Skeleton", "abstract": "  In combinatorial causal bandits (CCB), the learning agent chooses a subset of\nvariables in each round to intervene and collects feedback from the observed\nvariables to minimize expected regret or sample complexity. Previous works\nstudy this problem in both general causal models and binary generalized linear\nmodels (BGLMs). However, all of them require prior knowledge of causal graph\nstructure or unrealistic assumptions. This paper studies the CCB problem\nwithout the graph structure on binary general causal models and BGLMs. We first\nprovide an exponential lower bound of cumulative regrets for the CCB problem on\ngeneral causal models. To overcome the exponentially large space of parameters,\nwe then consider the CCB problem on BGLMs. We design a regret minimization\nalgorithm for BGLMs even without the graph skeleton and show that it still\nachieves $O(\\sqrt{T}\\ln T)$ expected regret, as long as the causal graph\nsatisfies a weight gap assumption. This asymptotic regret is the same as the\nstate-of-art algorithms relying on the graph structure. Moreover, we propose\nanother algorithm with $O(T^{\\frac{2}{3}}\\ln T)$ regret to remove the weight\ngap assumption.\n", "date": "2023-01-31", "categories": []}
{"id": "2402.18139", "title": "Cause and Effect: Can Large Language Models Truly Understand Causality?", "abstract": "  With the rise of Large Language Models(LLMs), it has become crucial to\nunderstand their capabilities and limitations in deciphering and explaining the\ncomplex web of causal relationships that language entails. Current methods use\neither explicit or implicit causal reasoning, yet there is a strong need for a\nunified approach combining both to tackle a wide array of causal relationships\nmore effectively. This research proposes a novel architecture called Context\nAware Reasoning Enhancement with Counterfactual Analysis(CARE CA) framework to\nenhance causal reasoning and explainability. The proposed framework\nincorporates an explicit causal detection module with ConceptNet and\ncounterfactual statements, as well as implicit causal detection through LLMs.\nOur framework goes one step further with a layer of counterfactual explanations\nto accentuate LLMs understanding of causality. The knowledge from ConceptNet\nenhances the performance of multiple causal reasoning tasks such as causal\ndiscovery, causal identification and counterfactual reasoning. The\ncounterfactual sentences add explicit knowledge of the not caused by scenarios.\nBy combining these powerful modules, our model aims to provide a deeper\nunderstanding of causal relationships, enabling enhanced interpretability.\nEvaluation of benchmark datasets shows improved performance across all metrics,\nsuch as accuracy, precision, recall, and F1 scores. We also introduce\nCausalNet, a new dataset accompanied by our code, to facilitate further\nresearch in this domain.\n", "date": "2024-02-28", "categories": []}
{"id": "2409.19060", "title": "CURATE: Scaling-up Differentially Private Causal Graph Discovery", "abstract": "  Causal Graph Discovery (CGD) is the process of estimating the underlying\nprobabilistic graphical model that represents joint distribution of features of\na dataset. CGD-algorithms are broadly classified into two categories: (i)\nConstraint-based algorithms (outcome depends on conditional independence (CI)\ntests), (ii) Score-based algorithms (outcome depends on optimized\nscore-function). Since, sensitive features of observational data is prone to\nprivacy-leakage, Differential Privacy (DP) has been adopted to ensure user\nprivacy in CGD. Adding same amount of noise in this sequential-natured\nestimation process affects the predictive performance of the algorithms. As\ninitial CI tests in constraint-based algorithms and later iterations of the\noptimization process of score-based algorithms are crucial, they need to be\nmore accurate, less noisy. Based on this key observation, we present CURATE\n(CaUsal gRaph AdapTivE privacy), a DP-CGD framework with adaptive privacy\nbudgeting. In contrast to existing DP-CGD algorithms with uniform privacy\nbudgeting across all iterations, CURATE allows adaptive privacy budgeting by\nminimizing error probability (for constraint-based), maximizing iterations of\nthe optimization problem (for score-based) while keeping the cumulative leakage\nbounded. To validate our framework, we present a comprehensive set of\nexperiments on several datasets and show that CURATE achieves higher utility\ncompared to existing DP-CGD algorithms with less privacy-leakage.\n", "date": "2024-09-27", "categories": []}
{"id": "2409.20187", "title": "Choosing DAG Models Using Markov and Minimal Edge Count in the Absence\n  of Ground Truth", "abstract": "  We give a novel nonparametric pointwise consistent statistical test (the\nMarkov Checker) of the Markov condition for directed acyclic graph (DAG) or\ncompleted partially directed acyclic graph (CPDAG) models given a dataset. We\nalso introduce the Cross-Algorithm Frugality Search (CAFS) for rejecting DAG\nmodels that either do not pass the Markov Checker test or that are not edge\nminimal. Edge minimality has been used previously by Raskutti and Uhler as a\nnonparametric simplicity criterion, though CAFS readily generalizes to other\nsimplicity conditions. Reference to the ground truth is not necessary for CAFS,\nso it is useful for finding causal structure learning algorithms and tuning\nparameter settings that output causal models that are approximately true from a\ngiven data set. We provide a software tool for this analysis that is suitable\nfor even quite large or dense models, provided a suitably fast pointwise\nconsistent test of conditional independence is available. In addition, we show\nin simulation that the CAFS procedure can pick approximately correct models\nwithout knowing the ground truth.\n", "date": "2024-09-30", "categories": []}
{"id": "2404.06349", "title": "CausalBench: A Comprehensive Benchmark for Causal Learning Capability of\n  LLMs", "abstract": "  The ability to understand causality significantly impacts the competence of\nlarge language models (LLMs) in output explanation and counterfactual\nreasoning, as causality reveals the underlying data distribution. However, the\nlack of a comprehensive benchmark currently limits the evaluation of LLMs'\ncausal learning capabilities. To fill this gap, this paper develops CausalBench\nbased on data from the causal research community, enabling comparative\nevaluations of LLMs against traditional causal learning algorithms. To provide\na comprehensive investigation, we offer three tasks of varying difficulties,\nincluding correlation, causal skeleton, and causality identification.\nEvaluations of 19 leading LLMs reveal that, while closed-source LLMs show\npotential for simple causal relationships, they significantly lag behind\ntraditional algorithms on larger-scale networks ($>50$ nodes). Specifically,\nLLMs struggle with collider structures but excel at chain structures,\nespecially at long-chain causality analogous to Chains-of-Thought techniques.\nThis supports the current prompt approaches while suggesting directions to\nenhance LLMs' causal reasoning capability. Furthermore, CausalBench\nincorporates background knowledge and training data into prompts to thoroughly\nunlock LLMs' text-comprehension ability during evaluation, whose findings\nindicate that, LLM understand causality through semantic associations with\ndistinct entities, rather than directly from contextual information or\nnumerical distributions.\n", "date": "2024-04-09", "categories": []}
{"id": "2310.13387", "title": "Assumption violations in causal discovery and the robustness of score\n  matching", "abstract": "  When domain knowledge is limited and experimentation is restricted by\nethical, financial, or time constraints, practitioners turn to observational\ncausal discovery methods to recover the causal structure, exploiting the\nstatistical properties of their data. Because causal discovery without further\nassumptions is an ill-posed problem, each algorithm comes with its own set of\nusually untestable assumptions, some of which are hard to meet in real\ndatasets. Motivated by these considerations, this paper extensively benchmarks\nthe empirical performance of recent causal discovery methods on observational\ni.i.d. data generated under different background conditions, allowing for\nviolations of the critical assumptions required by each selected approach. Our\nexperimental findings show that score matching-based methods demonstrate\nsurprising performance in the false positive and false negative rate of the\ninferred graph in these challenging scenarios, and we provide theoretical\ninsights into their performance. This work is also the first effort to\nbenchmark the stability of causal discovery algorithms with respect to the\nvalues of their hyperparameters. Finally, we hope this paper will set a new\nstandard for the evaluation of causal discovery methods and can serve as an\naccessible entry point for practitioners interested in the field, highlighting\nthe empirical implications of different algorithm choices.\n", "date": "2023-10-20", "categories": []}
{"id": "2406.05504", "title": "G-Transformer: Counterfactual Outcome Prediction under Dynamic and\n  Time-varying Treatment Regimes", "abstract": "  In the context of medical decision making, counterfactual prediction enables\nclinicians to predict treatment outcomes of interest under alternative courses\nof therapeutic actions given observed patient history. In this work, we present\nG-Transformer for counterfactual outcome prediction under dynamic and\ntime-varying treatment strategies. Our approach leverages a Transformer\narchitecture to capture complex, long-range dependencies in time-varying\ncovariates while enabling g-computation, a causal inference method for\nestimating the effects of dynamic treatment regimes. Specifically, we use a\nTransformer-based encoder architecture to estimate the conditional distribution\nof relevant covariates given covariate and treatment history at each time\npoint, then produces Monte Carlo estimates of counterfactual outcomes by\nsimulating forward patient trajectories under treatment strategies of interest.\nWe evaluate G-Transformer extensively using two simulated longitudinal datasets\nfrom mechanistic models, and a real-world sepsis ICU dataset from MIMIC-IV.\nG-Transformer outperforms both classical and state-of-the-art counterfactual\nprediction models in these settings. To the best of our knowledge, this is the\nfirst Transformer-based architecture that supports g-computation for\ncounterfactual outcome prediction under dynamic and time-varying treatment\nstrategies.\n", "date": "2024-06-08", "categories": []}
{"id": "2409.17840", "title": "Detecting and Measuring Confounding Using Causal Mechanism Shifts", "abstract": "  Detecting and measuring confounding effects from data is a key challenge in\ncausal inference. Existing methods frequently assume causal sufficiency,\ndisregarding the presence of unobserved confounding variables. Causal\nsufficiency is both unrealistic and empirically untestable. Additionally,\nexisting methods make strong parametric assumptions about the underlying causal\ngenerative process to guarantee the identifiability of confounding variables.\nRelaxing the causal sufficiency and parametric assumptions and leveraging\nrecent advancements in causal discovery and confounding analysis with\nnon-i.i.d. data, we propose a comprehensive approach for detecting and\nmeasuring confounding. We consider various definitions of confounding and\nintroduce tailored methodologies to achieve three objectives: (i) detecting and\nmeasuring confounding among a set of variables, (ii) separating observed and\nunobserved confounding effects, and (iii) understanding the relative strengths\nof confounding bias between different sets of variables. We present useful\nproperties of a confounding measure and present measures that satisfy those\nproperties. Empirical results support the theoretical analysis.\n", "date": "2024-09-26", "categories": []}
{"id": "2409.08419", "title": "Introducing CausalBench: A Flexible Benchmark Framework for Causal\n  Analysis and Machine Learning", "abstract": "  While witnessing the exceptional success of machine learning (ML)\ntechnologies in many applications, users are starting to notice a critical\nshortcoming of ML: correlation is a poor substitute for causation. The\nconventional way to discover causal relationships is to use randomized\ncontrolled experiments (RCT); in many situations, however, these are\nimpractical or sometimes unethical. Causal learning from observational data\noffers a promising alternative. While being relatively recent, causal learning\naims to go far beyond conventional machine learning, yet several major\nchallenges remain. Unfortunately, advances are hampered due to the lack of\nunified benchmark datasets, algorithms, metrics, and evaluation service\ninterfaces for causal learning. In this paper, we introduce {\\em CausalBench},\na transparent, fair, and easy-to-use evaluation platform, aiming to (a) enable\nthe advancement of research in causal learning by facilitating scientific\ncollaboration in novel algorithms, datasets, and metrics and (b) promote\nscientific objectivity, reproducibility, fairness, and awareness of bias in\ncausal learning research. CausalBench provides services for benchmarking data,\nalgorithms, models, and metrics, impacting the needs of a broad of scientific\nand engineering disciplines.\n", "date": "2024-09-12", "categories": []}
{"id": "2409.02130", "title": "From Predictive Importance to Causality: Which Machine Learning Model\n  Reflects Reality?", "abstract": "  This study analyzes the Ames Housing Dataset using CatBoost and LightGBM\nmodels to explore feature importance and causal relationships in housing price\nprediction. We examine the correlation between SHAP values and EconML\npredictions, achieving high accuracy in price forecasting. Our analysis reveals\na moderate Spearman rank correlation of 0.48 between SHAP-based feature\nimportance and causally significant features, highlighting the complexity of\naligning predictive modeling with causal understanding in housing market\nanalysis. Through extensive causal analysis, including heterogeneity\nexploration and policy tree interpretation, we provide insights into how\nspecific features like porches impact housing prices across various scenarios.\nThis work underscores the need for integrated approaches that combine\npredictive power with causal insights in real estate valuation, offering\nvaluable guidance for stakeholders in the industry.\n", "date": "2024-09-01", "categories": []}
{"id": "2404.10907", "title": "Causal Effect Estimation Using Random Hyperplane Tessellations", "abstract": "  Matching is one of the simplest approaches for estimating causal effects from\nobservational data. Matching techniques compare the observed outcomes across\npairs of individuals with similar covariate values but different treatment\nstatuses in order to estimate causal effects. However, traditional matching\ntechniques are unreliable given high-dimensional covariates due to the infamous\ncurse of dimensionality. To overcome this challenge, we propose a simple, fast,\nyet highly effective approach to matching using Random Hyperplane Tessellations\n(RHPT). First, we prove that the RHPT representation is an approximate\nbalancing score -- thus maintaining the strong ignorability assumption -- and\nprovide empirical evidence for this claim. Second, we report results of\nextensive experiments showing that matching using RHPT outperforms traditional\nmatching techniques and is competitive with state-of-the-art deep learning\nmethods for causal effect estimation. In addition, RHPT avoids the need for\ncomputationally expensive training of deep neural networks.\n", "date": "2024-04-16", "categories": []}
{"id": "2409.13210", "title": "A Unified Causal Framework for Auditing Recommender Systems for Ethical\n  Concerns", "abstract": "  As recommender systems become widely deployed in different domains, they\nincreasingly influence their users' beliefs and preferences. Auditing\nrecommender systems is crucial as it not only ensures the continuous\nimprovement of recommendation algorithms but also safeguards against potential\nissues like biases and ethical concerns. In this paper, we view recommender\nsystem auditing from a causal lens and provide a general recipe for defining\nauditing metrics. Under this general causal auditing framework, we categorize\nexisting auditing metrics and identify gaps in them -- notably, the lack of\nmetrics for auditing user agency while accounting for the multi-step dynamics\nof the recommendation process. We leverage our framework and propose two\nclasses of such metrics:future- and past-reacheability and stability, that\nmeasure the ability of a user to influence their own and other users'\nrecommendations, respectively. We provide both a gradient-based and a black-box\napproach for computing these metrics, allowing the auditor to compute them\nunder different levels of access to the recommender system. In our experiments,\nwe demonstrate the efficacy of methods for computing the proposed metrics and\ninspect the design of recommender systems through these proposed metrics.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.13423", "title": "Causal Reinforcement Learning for Optimisation of Robot Dynamics in\n  Unknown Environments", "abstract": "  Autonomous operations of robots in unknown environments are challenging due\nto the lack of knowledge of the dynamics of the interactions, such as the\nobjects' movability. This work introduces a novel Causal Reinforcement Learning\napproach to enhancing robotics operations and applies it to an urban search and\nrescue (SAR) scenario. Our proposed machine learning architecture enables\nrobots to learn the causal relationships between the visual characteristics of\nthe objects, such as texture and shape, and the objects' dynamics upon\ninteraction, such as their movability, significantly improving their\ndecision-making processes. We conducted causal discovery and RL experiments\ndemonstrating the Causal RL's superior performance, showing a notable reduction\nin learning times by over 24.5% in complex situations, compared to non-causal\nmodels.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.12203", "title": "A Simple Model to Estimate Sharing Effects in Social Networks", "abstract": "  Randomised Controlled Trials (RCTs) are the gold standard for estimating\ntreatment effects across many fields of science. Technology companies have\nadopted A/B-testing methods as a modern RCT counterpart, where end-users are\nrandomly assigned various system variants and user behaviour is tracked\ncontinuously. The objective is then to estimate the causal effect that the\ntreatment variant would have on certain metrics of interest to the business.\n  When the outcomes for randomisation units -- end-users in this case -- are\nnot statistically independent, this obfuscates identifiability of treatment\neffects, and harms decision-makers' observability of the system. Social\nnetworks exemplify this, as they are designed to promote inter-user\ninteractions. This interference by design notoriously complicates measurement\nof, e.g., the effects of sharing. In this work, we propose a simple Markov\nDecision Process (MDP)-based model describing user sharing behaviour in social\nnetworks. We derive an unbiased estimator for treatment effects under this\nmodel, and demonstrate through reproducible synthetic experiments that it\noutperforms existing methods by a significant margin.\n", "date": "2024-09-16", "categories": []}
{"id": "2402.14777", "title": "Causal Imputation for Counterfactual SCMs: Bridging Graphs and Latent\n  Factor Models", "abstract": "  We consider the task of causal imputation, where we aim to predict the\noutcomes of some set of actions across a wide range of possible contexts. As a\nrunning example, we consider predicting how different drugs affect cells from\ndifferent cell types. We study the index-only setting, where the actions and\ncontexts are categorical variables with a finite number of possible values.\nEven in this simple setting, a practical challenge arises, since often only a\nsmall subset of possible action-context pairs have been studied. Thus, models\nmust extrapolate to novel action-context pairs, which can be framed as a form\nof matrix completion with rows indexed by actions, columns indexed by contexts,\nand matrix entries corresponding to outcomes. We introduce a novel SCM-based\nmodel class, where the outcome is expressed as a counterfactual, actions are\nexpressed as interventions on an instrumental variable, and contexts are\ndefined based on the initial state of the system. We show that, under a\nlinearity assumption, this setup induces a latent factor model over the matrix\nof outcomes, with an additional fixed effect term. To perform causal prediction\nbased on this model class, we introduce simple extension to the Synthetic\nInterventions estimator (Agarwal et al., 2020). We evaluate several matrix\ncompletion approaches on the PRISM drug repurposing dataset, showing that our\nmethod outperforms all other considered matrix completion approaches.\n", "date": "2024-02-22", "categories": []}
{"id": "2405.03329", "title": "Policy Learning for Balancing Short-Term and Long-Term Rewards", "abstract": "  Empirical researchers and decision-makers spanning various domains frequently\nseek profound insights into the long-term impacts of interventions. While the\nsignificance of long-term outcomes is undeniable, an overemphasis on them may\ninadvertently overshadow short-term gains. Motivated by this, this paper\nformalizes a new framework for learning the optimal policy that effectively\nbalances both long-term and short-term rewards, where some long-term outcomes\nare allowed to be missing. In particular, we first present the identifiability\nof both rewards under mild assumptions. Next, we deduce the semiparametric\nefficiency bounds, along with the consistency and asymptotic normality of their\nestimators. We also reveal that short-term outcomes, if associated, contribute\nto improving the estimator of the long-term reward. Based on the proposed\nestimators, we develop a principled policy learning approach and further derive\nthe convergence rates of regret and estimation errors associated with the\nlearned policy. Extensive experiments are conducted to validate the\neffectiveness of the proposed method, demonstrating its practical\napplicability.\n", "date": "2024-05-06", "categories": []}
{"id": "2409.10271", "title": "Causal Discovery in Recommender Systems: Example and Discussion", "abstract": "  Causality is receiving increasing attention by the artificial intelligence\nand machine learning communities. This paper gives an example of modelling a\nrecommender system problem using causal graphs. Specifically, we approached the\ncausal discovery task to learn a causal graph by combining observational data\nfrom an open-source dataset with prior knowledge. The resulting causal graph\nshows that only a few variables effectively influence the analysed feedback\nsignals. This contrasts with the recent trend in the machine learning community\nto include more and more variables in massive models, such as neural networks.\n", "date": "2024-09-16", "categories": []}
{"id": "2408.11492", "title": "Estimating Peer Direct and Indirect Effects in Observational Network\n  Data", "abstract": "  Estimating causal effects is crucial for decision-makers in many\napplications, but it is particularly challenging with observational network\ndata due to peer interactions. Many algorithms have been proposed to estimate\ncausal effects involving network data, particularly peer effects, but they\noften overlook the variety of peer effects. To address this issue, we propose a\ngeneral setting which considers both peer direct effects and peer indirect\neffects, and the effect of an individual's own treatment, and provide\nidentification conditions of these causal effects and proofs. To estimate these\ncausal effects, we utilize attention mechanisms to distinguish the influences\nof different neighbors and explore high-order neighbor effects through\nmulti-layer graph neural networks (GNNs). Additionally, to control the\ndependency between node features and representations, we incorporate the\nHilbert-Schmidt Independence Criterion (HSIC) into the GNN, fully utilizing the\nstructural information of the graph, to enhance the robustness and accuracy of\nthe model. Extensive experiments on two semi-synthetic datasets confirm the\neffectiveness of our approach. Our theoretical findings have the potential to\nimprove intervention strategies in networked systems, with applications in\nareas such as social networks and epidemiology.\n", "date": "2024-08-21", "categories": []}
{"id": "2409.08544", "title": "Causal GNNs: A GNN-Driven Instrumental Variable Approach for Causal\n  Inference in Networks", "abstract": "  As network data applications continue to expand, causal inference within\nnetworks has garnered increasing attention. However, hidden confounders\ncomplicate the estimation of causal effects. Most methods rely on the strong\nignorability assumption, which presumes the absence of hidden confounders-an\nassumption that is both difficult to validate and often unrealistic in\npractice. To address this issue, we propose CgNN, a novel approach that\nleverages network structure as instrumental variables (IVs), combined with\ngraph neural networks (GNNs) and attention mechanisms, to mitigate hidden\nconfounder bias and improve causal effect estimation. By utilizing network\nstructure as IVs, we reduce confounder bias while preserving the correlation\nwith treatment. Our integration of attention mechanisms enhances robustness and\nimproves the identification of important nodes. Validated on two real-world\ndatasets, our results demonstrate that CgNN effectively mitigates hidden\nconfounder bias and offers a robust GNN-driven IV framework for causal\ninference in complex network data.\n", "date": "2024-09-13", "categories": []}
{"id": "2306.10614", "title": "Identifiable causal inference with noisy treatment and no side\n  information", "abstract": "  In some causal inference scenarios, the treatment variable is measured\ninaccurately, for instance in epidemiology or econometrics. Failure to correct\nfor the effect of this measurement error can lead to biased causal effect\nestimates. Previous research has not studied methods that address this issue\nfrom a causal viewpoint while allowing for complex nonlinear dependencies and\nwithout assuming access to side information. For such a scenario, this study\nproposes a model that assumes a continuous treatment variable that is\ninaccurately measured. Building on existing results for measurement error\nmodels, we prove that our model's causal effect estimates are identifiable,\neven without side information and knowledge of the measurement error variance.\nOur method relies on a deep latent variable model in which Gaussian\nconditionals are parameterized by neural networks, and we develop an amortized\nimportance-weighted variational objective for training the model. Empirical\nresults demonstrate the method's good performance with unknown measurement\nerror. More broadly, our work extends the range of applications in which\nreliable causal inference can be conducted.\n", "date": "2023-06-18", "categories": []}
{"id": "2409.07594", "title": "Automated Discovery of Pairwise Interactions from Unstructured Data", "abstract": "  Pairwise interactions between perturbations to a system can provide evidence\nfor the causal dependencies of the underlying underlying mechanisms of a\nsystem. When observations are low dimensional, hand crafted measurements,\ndetecting interactions amounts to simple statistical tests, but it is not\nobvious how to detect interactions between perturbations affecting latent\nvariables. We derive two interaction tests that are based on pairwise\ninterventions, and show how these tests can be integrated into an active\nlearning pipeline to efficiently discover pairwise interactions between\nperturbations. We illustrate the value of these tests in the context of\nbiology, where pairwise perturbation experiments are frequently used to reveal\ninteractions that are not observable from any single perturbation. Our tests\ncan be run on unstructured data, such as the pixels in an image, which enables\na more general notion of interaction than typical cell viability experiments,\nand can be run on cheaper experimental assays. We validate on several synthetic\nand real biological experiments that our tests are able to identify interacting\npairs effectively. We evaluate our approach on a real biological experiment\nwhere we knocked out 50 pairs of genes and measured the effect with microscopy\nimages. We show that we are able to recover significantly more known biological\ninteractions than random search and standard active learning baselines.\n", "date": "2024-09-11", "categories": []}
{"id": "2409.07880", "title": "Non-negative Weighted DAG Structure Learning", "abstract": "  We address the problem of learning the topology of directed acyclic graphs\n(DAGs) from nodal observations, which adhere to a linear structural equation\nmodel. Recent advances framed the combinatorial DAG structure learning task as\na continuous optimization problem, yet existing methods must contend with the\ncomplexities of non-convex optimization. To overcome this limitation, we assume\nthat the latent DAG contains only non-negative edge weights. Leveraging this\nadditional structure, we argue that cycles can be effectively characterized\n(and prevented) using a convex acyclicity function based on the log-determinant\nof the adjacency matrix. This convexity allows us to relax the task of learning\nthe non-negative weighted DAG as an abstract convex optimization problem. We\npropose a DAG recovery algorithm based on the method of multipliers, that is\nguaranteed to return a global minimizer. Furthermore, we prove that in the\ninfinite sample size regime, the convexity of our approach ensures the recovery\nof the true DAG structure. We empirically validate the performance of our\nalgorithm in several reproducible synthetic-data test cases, showing that it\noutperforms state-of-the-art alternatives.\n", "date": "2024-09-12", "categories": []}
{"id": "2310.05655", "title": "Causal structure learning with momentum: Sampling distributions over\n  Markov Equivalence Classes of DAGs", "abstract": "  In the context of inferring a Bayesian network structure (directed acyclic\ngraph, DAG for short), we devise a non-reversible continuous time Markov chain,\nthe ``Causal Zig-Zag sampler'', that targets a probability distribution over\nclasses of observationally equivalent (Markov equivalent) DAGs. The classes are\nrepresented as completed partially directed acyclic graphs (CPDAGs). The\nnon-reversible Markov chain relies on the operators used in Chickering's Greedy\nEquivalence Search (GES) and is endowed with a momentum variable, which\nimproves mixing significantly as we show empirically. The possible target\ndistributions include posterior distributions based on a prior over DAGs and a\nMarkov equivalent likelihood. We offer an efficient implementation wherein we\ndevelop new algorithms for listing, counting, uniformly sampling, and applying\npossible moves of the GES operators, all of which significantly improve upon\nthe state-of-the-art run-time.\n", "date": "2023-10-09", "categories": []}
{"id": "2402.02636", "title": "Can Large Language Models Learn Independent Causal Mechanisms?", "abstract": "  Despite impressive performance on language modelling and complex reasoning\ntasks, Large Language Models (LLMs) fall short on the same tasks in uncommon\nsettings or with distribution shifts, exhibiting a lack of generalisation\nability. By contrast, systems such as causal models, that learn abstract\nvariables and causal relationships, can demonstrate increased robustness\nagainst changes in the distribution. One reason for this success is the\nexistence and use of Independent Causal Mechanisms (ICMs) representing\nhigh-level concepts that only sparsely interact. In this work, we apply two\nconcepts from causality to learn ICMs within LLMs. We develop a new LLM\narchitecture composed of multiple sparsely interacting language modelling\nmodules. We show that such causal constraints can improve out-of-distribution\nperformance on abstract and causal reasoning tasks. We also investigate the\nlevel of independence and domain specialisation and show that LLMs rely on\npre-trained partially domain-invariant mechanisms resilient to fine-tuning.\n", "date": "2024-02-04", "categories": []}
{"id": "2406.02191", "title": "On the Recoverability of Causal Relations from Temporally Aggregated\n  I.I.D. Data", "abstract": "  We consider the effect of temporal aggregation on instantaneous\n(non-temporal) causal discovery in general setting. This is motivated by the\nobservation that the true causal time lag is often considerably shorter than\nthe observational interval. This discrepancy leads to high aggregation, causing\ntime-delay causality to vanish and instantaneous dependence to manifest.\nAlthough we expect such instantaneous dependence has consistency with the true\ncausal relation in certain sense to make the discovery results meaningful, it\nremains unclear what type of consistency we need and when will such consistency\nbe satisfied. We proposed functional consistency and conditional independence\nconsistency in formal way correspond functional causal model-based methods and\nconditional independence-based methods respectively and provide the conditions\nunder which these consistencies will hold. We show theoretically and\nexperimentally that causal discovery results may be seriously distorted by\naggregation especially in complete nonlinear case and we also find causal\nrelationship still recoverable from aggregated data if we have partial\nlinearity or appropriate prior. Our findings suggest community should take a\ncautious and meticulous approach when interpreting causal discovery results\nfrom such data and show why and when aggregation will distort the performance\nof causal discovery methods.\n", "date": "2024-06-04", "categories": []}
{"id": "2409.05872", "title": "CSRec: Rethinking Sequential Recommendation from A Causal Perspective", "abstract": "  The essence of sequential recommender systems (RecSys) lies in understanding\nhow users make decisions. Most existing approaches frame the task as sequential\nprediction based on users' historical purchase records. While effective in\ncapturing users' natural preferences, this formulation falls short in\naccurately modeling actual recommendation scenarios, particularly in accounting\nfor how unsuccessful recommendations influence future purchases. Furthermore,\nthe impact of the RecSys itself on users' decisions has not been appropriately\nisolated and quantitatively analyzed. To address these challenges, we propose a\nnovel formulation of sequential recommendation, termed Causal Sequential\nRecommendation (CSRec). Instead of predicting the next item in the sequence,\nCSRec aims to predict the probability of a recommended item's acceptance within\na sequential context and backtrack how current decisions are made. Critically,\nCSRec facilitates the isolation of various factors that affect users' final\ndecisions, especially the influence of the recommender system itself, thereby\nopening new avenues for the design of recommender systems. CSRec can be\nseamlessly integrated into existing methodologies. Experimental evaluations on\nboth synthetic and real-world datasets demonstrate that the proposed\nimplementation significantly improves upon state-of-the-art baselines.\n", "date": "2024-08-23", "categories": []}
{"id": "2409.06157", "title": "Causal Analysis of Shapley Values: Conditional vs. Marginal", "abstract": "  Shapley values, a game theoretic concept, has been one of the most popular\ntools for explaining Machine Learning (ML) models in recent years.\nUnfortunately, the two most common approaches, conditional and marginal, to\ncalculating Shapley values can lead to different results along with some\nundesirable side effects when features are correlated. This in turn has led to\nthe situation in the literature where contradictory recommendations regarding\nchoice of an approach are provided by different authors. In this paper we aim\nto resolve this controversy through the use of causal arguments. We show that\nthe differences arise from the implicit assumptions that are made within each\nmethod to deal with missing causal information. We also demonstrate that the\nconditional approach is fundamentally unsound from a causal perspective. This,\ntogether with previous work in [1], leads to the conclusion that the marginal\napproach should be preferred over the conditional one.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.06593", "title": "Advancing Causal Inference: A Nonparametric Approach to ATE and CATE\n  Estimation with Continuous Treatments", "abstract": "  This paper introduces a generalized ps-BART model for the estimation of\nAverage Treatment Effect (ATE) and Conditional Average Treatment Effect (CATE)\nin continuous treatments, addressing limitations of the Bayesian Causal Forest\n(BCF) model. The ps-BART model's nonparametric nature allows for flexibility in\ncapturing nonlinear relationships between treatment and outcome variables.\nAcross three distinct sets of Data Generating Processes (DGPs), the ps-BART\nmodel consistently outperforms the BCF model, particularly in highly nonlinear\nsettings. The ps-BART model's robustness in uncertainty estimation and accuracy\nin both point-wise and probabilistic estimation demonstrate its utility for\nreal-world applications. This research fills a crucial gap in causal inference\nliterature, providing a tool better suited for nonlinear treatment-outcome\nrelationships and opening avenues for further exploration in the domain of\ncontinuous treatment effect estimation.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.00629", "title": "Assessing the Impact of Upselling in Online Fantasy Sports", "abstract": "  This study explores the impact of upselling on user engagement. We model\nusers' deposit behaviour on the fantasy sports platform Dream11. Subsequently,\nwe develop an experimental framework to evaluate the effect of upselling using\nan intensity parameter. Our live experiments on user deposit behaviour reveal\ndecreased user recall with heightened upselling intensity. Our findings\nindicate that increased upselling intensity improves user deposit metrics and\nconcurrently diminishes user satisfaction and conversion rates. We conduct\nrobust counterfactual analysis and train causal meta-learners to personalise\nusers' upselling intensity levels to reach an optimal trade-off point.\n", "date": "2024-09-01", "categories": []}
{"id": "2409.04500", "title": "Benchmarking Estimators for Natural Experiments: A Novel Dataset and a\n  Doubly Robust Algorithm", "abstract": "  Estimating the effect of treatments from natural experiments, where\ntreatments are pre-assigned, is an important and well-studied problem. We\nintroduce a novel natural experiment dataset obtained from an early childhood\nliteracy nonprofit. Surprisingly, applying over 20 established estimators to\nthe dataset produces inconsistent results in evaluating the nonprofit's\nefficacy. To address this, we create a benchmark to evaluate estimator accuracy\nusing synthetic outcomes, whose design was guided by domain experts. The\nbenchmark extensively explores performance as real world conditions like sample\nsize, treatment correlation, and propensity score accuracy vary. Based on our\nbenchmark, we observe that the class of doubly robust treatment effect\nestimators, which are based on simple and intuitive regression adjustment,\ngenerally outperform other more complicated estimators by orders of magnitude.\nTo better support our theoretical understanding of doubly robust estimators, we\nderive a closed form expression for the variance of any such estimator that\nuses dataset splitting to obtain an unbiased estimate. This expression\nmotivates the design of a new doubly robust estimator that uses a novel loss\nfunction when fitting functions for regression adjustment. We release the\ndataset and benchmark in a Python package; the package is built in a modular\nway to facilitate new datasets and estimators.\n", "date": "2024-09-06", "categories": []}
{"id": "2409.05665", "title": "K-Fold Causal BART for CATE Estimation", "abstract": "  This research aims to propose and evaluate a novel model named K-Fold Causal\nBayesian Additive Regression Trees (K-Fold Causal BART) for improved estimation\nof Average Treatment Effects (ATE) and Conditional Average Treatment Effects\n(CATE). The study employs synthetic and semi-synthetic datasets, including the\nwidely recognized Infant Health and Development Program (IHDP) benchmark\ndataset, to validate the model's performance. Despite promising results in\nsynthetic scenarios, the IHDP dataset reveals that the proposed model is not\nstate-of-the-art for ATE and CATE estimation. Nonetheless, the research\nprovides several novel insights: 1. The ps-BART model is likely the preferred\nchoice for CATE and ATE estimation due to better generalization compared to the\nother benchmark models - including the Bayesian Causal Forest (BCF) model,\nwhich is considered by many the current best model for CATE estimation, 2. The\nBCF model's performance deteriorates significantly with increasing treatment\neffect heterogeneity, while the ps-BART model remains robust, 3. Models tend to\nbe overconfident in CATE uncertainty quantification when treatment effect\nheterogeneity is low, 4. A second K-Fold method is unnecessary for avoiding\noverfitting in CATE estimation, as it adds computational costs without\nimproving performance, 5. Detailed analysis reveals the importance of\nunderstanding dataset characteristics and using nuanced evaluation methods, 6.\nThe conclusion of Curth et al. (2021) that indirect strategies for CATE\nestimation are superior for the IHDP dataset is contradicted by the results of\nthis research. These findings challenge existing assumptions and suggest\ndirections for future research to enhance causal inference methodologies.\n", "date": "2024-09-09", "categories": []}
{"id": "2404.08839", "title": "Multiply-Robust Causal Change Attribution", "abstract": "  Comparing two samples of data, we observe a change in the distribution of an\noutcome variable. In the presence of multiple explanatory variables, how much\nof the change can be explained by each possible cause? We develop a new\nestimation strategy that, given a causal model, combines regression and\nre-weighting methods to quantify the contribution of each causal mechanism. Our\nproposed methodology is multiply robust, meaning that it still recovers the\ntarget parameter under partial misspecification. We prove that our estimator is\nconsistent and asymptotically normal. Moreover, it can be incorporated into\nexisting frameworks for causal attribution, such as Shapley values, which will\ninherit the consistency and large-sample distribution properties. Our method\ndemonstrates excellent performance in Monte Carlo simulations, and we show its\nusefulness in an empirical application. Our method is implemented as part of\nthe Python library DoWhy (arXiv:2011.04216, arXiv:2206.06821).\n", "date": "2024-04-12", "categories": []}
{"id": "2409.04174", "title": "Towards Measuring Sell Side Outcomes in Buy Side Marketplace Experiments\n  using In-Experiment Bipartite Graph", "abstract": "  In this study, we evaluate causal inference estimators for online controlled\nbipartite graph experiments in a real marketplace setting. Our novel\ncontribution is constructing a bipartite graph using in-experiment data, rather\nthan relying on prior knowledge or historical data, the common approach in the\nliterature published to date. We build the bipartite graph from various\ninteractions between buyers and sellers in the marketplace, establishing a\nnovel research direction at the intersection of bipartite experiments and\nmediation analysis. This approach is crucial for modern marketplaces aiming to\nevaluate seller-side causal effects in buyer-side experiments, or vice versa.\nWe demonstrate our method using historical buyer-side experiments conducted at\nVinted, the largest second-hand marketplace in Europe with over 80M users.\n", "date": "2024-09-06", "categories": []}
{"id": "2409.03142", "title": "Causal Temporal Representation Learning with Nonstationary Sparse\n  Transition", "abstract": "  Causal Temporal Representation Learning (Ctrl) methods aim to identify the\ntemporal causal dynamics of complex nonstationary temporal sequences. Despite\nthe success of existing Ctrl methods, they require either directly observing\nthe domain variables or assuming a Markov prior on them. Such requirements\nlimit the application of these methods in real-world scenarios when we do not\nhave such prior knowledge of the domain variables. To address this problem,\nthis work adopts a sparse transition assumption, aligned with intuitive human\nunderstanding, and presents identifiability results from a theoretical\nperspective. In particular, we explore under what conditions on the\nsignificance of the variability of the transitions we can build a model to\nidentify the distribution shifts. Based on the theoretical result, we introduce\na novel framework, Causal Temporal Representation Learning with Nonstationary\nSparse Transition (CtrlNS), designed to leverage the constraints on transition\nsparsity and conditional independence to reliably identify both distribution\nshifts and latent factors. Our experimental evaluations on synthetic and\nreal-world datasets demonstrate significant improvements over existing\nbaselines, highlighting the effectiveness of our approach.\n", "date": "2024-09-05", "categories": []}
{"id": "2407.13054", "title": "Comprehensive Review and Empirical Evaluation of Causal Discovery\n  Algorithms for Numerical Data", "abstract": "  Causal analysis has become an essential component in understanding the\nunderlying causes of phenomena across various fields. Despite its significance,\nexisting literature on causal discovery algorithms is fragmented, with\ninconsistent methodologies, i.e., there is no universal classification standard\nfor existing methods, and a lack of comprehensive evaluations, i.e., data\ncharacteristics are often ignored to be jointly analyzed when benchmarking\nalgorithms. This study addresses these gaps by conducting an exhaustive review\nand empirical evaluation for causal discovery methods on numerical data, aiming\nto provide a clearer and more structured understanding of the field. Our\nresearch begins with a comprehensive literature review spanning over two\ndecades, analyzing over 200 academic articles and identifying more than 40\nrepresentative algorithms. This extensive analysis leads to the development of\na structured taxonomy tailored to the complexities of causal discovery,\ncategorizing methods into six main types. To address the lack of comprehensive\nevaluations, our study conducts an extensive empirical assessment of 29 causal\ndiscovery algorithms on multiple synthetic and real-world datasets. We\ncategorize synthetic datasets based on size, linearity, and noise distribution,\nemploying five evaluation metrics, and summarize the top-3 algorithm\nrecommendations, providing guidelines for users in various data scenarios. Our\nresults highlight a significant impact of dataset characteristics on algorithm\nperformance. Moreover, a metadata extraction strategy with an accuracy\nexceeding 80% is developed to assist users in algorithm selection on unknown\ndatasets. Based on these insights, we offer professional and practical\nguidelines to help users choose the most suitable causal discovery methods for\ntheir specific dataset.\n", "date": "2024-07-17", "categories": []}
{"id": "2409.02332", "title": "Double Machine Learning at Scale to Predict Causal Impact of Customer\n  Actions", "abstract": "  Causal Impact (CI) of customer actions are broadly used across the industry\nto inform both short- and long-term investment decisions of various types. In\nthis paper, we apply the double machine learning (DML) methodology to estimate\nthe CI values across 100s of customer actions of business interest and 100s of\nmillions of customers. We operationalize DML through a causal ML library based\non Spark with a flexible, JSON-driven model configuration approach to estimate\nCI at scale (i.e., across hundred of actions and millions of customers). We\noutline the DML methodology and implementation, and associated benefits over\nthe traditional potential outcomes based CI model. We show population-level as\nwell as customer-level CI values along with confidence intervals. The\nvalidation metrics show a 2.2% gain over the baseline methods and a 2.5X gain\nin the computational time. Our contribution is to advance the scalable\napplication of CI, while also providing an interface that allows faster\nexperimentation, cross-platform support, ability to onboard new use cases, and\nimproves accessibility of underlying code for partner teams.\n", "date": "2024-09-03", "categories": []}
{"id": "2403.19289", "title": "Uplift Modeling Under Limited Supervision", "abstract": "  Estimating causal effects in e-commerce tends to involve costly treatment\nassignments which can be impractical in large-scale settings. Leveraging\nmachine learning to predict such treatment effects without actual intervention\nis a standard practice to diminish the risk. However, existing methods for\ntreatment effect prediction tend to rely on training sets of substantial size,\nwhich are built from real experiments and are thus inherently risky to create.\nIn this work we propose a graph neural network to diminish the required\ntraining set size, relying on graphs that are common in e-commerce data.\nSpecifically, we view the problem as node regression with a restricted number\nof labeled instances, develop a two-model neural architecture akin to previous\ncausal effect estimators, and test varying message-passing layers for encoding.\nFurthermore, as an extra step, we combine the model with an acquisition\nfunction to guide the creation of the training set in settings with extremely\nlow experimental budget. The framework is flexible since each step can be used\nseparately with other models or treatment policies. The experiments on real\nlarge-scale networks indicate a clear advantage of our methodology over the\nstate of the art, which in many cases performs close to random, underlining the\nneed for models that can generalize with limited supervision to reduce\nexperimental risks.\n", "date": "2024-03-28", "categories": []}
{"id": "2404.05746", "title": "Causality for Earth Science -- A Review on Time-series and\n  Spatiotemporal Causality Methods", "abstract": "  This survey paper covers the breadth and depth of time-series and\nspatiotemporal causality methods, and their applications in Earth Science. More\nspecifically, the paper presents an overview of causal discovery and causal\ninference, explains the underlying causal assumptions, and enlists evaluation\ntechniques and key terminologies of the domain area. The paper elicits the\nvarious state-of-the-art methods introduced for time-series and spatiotemporal\ncausal analysis along with their strengths and limitations. The paper further\ndescribes the existing applications of several methods for answering specific\nEarth Science questions such as extreme weather events, sea level rise,\nteleconnections etc. This survey paper can serve as a primer for Data Science\nresearchers interested in data-driven causal study as we share a list of\nresources, such as Earth Science datasets (synthetic, simulated and\nobservational data) and open source tools for causal analysis. It will equally\nbenefit the Earth Science community interested in taking an AI-driven approach\nto study the causality of different dynamic and thermodynamic processes as we\npresent the open challenges and opportunities in performing causality-based\nEarth Science study.\n", "date": "2024-04-03", "categories": []}
{"id": "2409.01266", "title": "Double Machine Learning meets Panel Data -- Promises, Pitfalls, and\n  Potential Solutions", "abstract": "  Estimating causal effect using machine learning (ML) algorithms can help to\nrelax functional form assumptions if used within appropriate frameworks.\nHowever, most of these frameworks assume settings with cross-sectional data,\nwhereas researchers often have access to panel data, which in traditional\nmethods helps to deal with unobserved heterogeneity between units. In this\npaper, we explore how we can adapt double/debiased machine learning (DML)\n(Chernozhukov et al., 2018) for panel data in the presence of unobserved\nheterogeneity. This adaptation is challenging because DML's cross-fitting\nprocedure assumes independent data and the unobserved heterogeneity is not\nnecessarily additively separable in settings with nonlinear observed\nconfounding. We assess the performance of several intuitively appealing\nestimators in a variety of simulations. While we find violations of the\ncross-fitting assumptions to be largely inconsequential for the accuracy of the\neffect estimates, many of the considered methods fail to adequately account for\nthe presence of unobserved heterogeneity. However, we find that using\npredictive models based on the correlated random effects approach (Mundlak,\n1978) within DML leads to accurate coefficient estimates across settings, given\na sample size that is large relative to the number of observed confounders. We\nalso show that the influence of the unobserved heterogeneity on the observed\nconfounders plays a significant role for the performance of most alternative\nmethods.\n", "date": "2024-09-02", "categories": []}
{"id": "2401.05218", "title": "Invariant Causal Prediction with Local Models", "abstract": "  We consider the task of identifying the causal parents of a target variable\namong a set of candidates from observational data. Our main assumption is that\nthe candidate variables are observed in different environments which may, under\ncertain assumptions, be regarded as interventions on the observed system. We\nassume a linear relationship between target and candidates, which can be\ndifferent in each environment with the only restriction that the causal\nstructure is invariant across environments. Within our proposed setting we\nprovide sufficient conditions for identifiability of the causal parents and\nintroduce a practical method called L-ICP ($\\textbf{L}$ocalized\n$\\textbf{I}$nvariant $\\textbf{Ca}$usal $\\textbf{P}$rediction), which is based\non a hypothesis test for parent identification using a ratio of minimum and\nmaximum statistics. We then show in a simplified setting that the statistical\npower of L-ICP converges exponentially fast in the sample size, and finally we\nanalyze the behavior of L-ICP experimentally in more general settings.\n", "date": "2024-01-10", "categories": []}
{"id": "2405.08174", "title": "Estimating Direct and Indirect Causal Effects of Spatiotemporal\n  Interventions in Presence of Spatial Interference", "abstract": "  Spatial interference (SI) occurs when the treatment at one location affects\nthe outcomes at other locations. Accounting for spatial interference in\nspatiotemporal settings poses further challenges as interference violates the\nstable unit treatment value assumption, making it infeasible for standard\ncausal inference methods to quantify the effects of time-varying treatment at\nspatially varying outcomes. In this paper, we first formalize the concept of\nspatial interference in case of time-varying treatment assignments by extending\nthe potential outcome framework under the assumption of no unmeasured\nconfounding. We then propose our deep learning based potential outcome model\nfor spatiotemporal causal inference. We utilize latent factor modeling to\nreduce the bias due to time-varying confounding while leveraging the power of\nU-Net architecture to capture global and local spatial interference in data\nover time. Our causal estimators are an extension of average treatment effect\n(ATE) for estimating direct (DATE) and indirect effects (IATE) of spatial\ninterference on treated and untreated data. Being the first of its kind deep\nlearning based spatiotemporal causal inference technique, our approach shows\nadvantages over several baseline methods based on the experiment results on two\nsynthetic datasets, with and without spatial interference. Our results on\nreal-world climate dataset also align with domain knowledge, further\ndemonstrating the effectiveness of our proposed method.\n", "date": "2024-05-13", "categories": []}
{"id": "2408.17363", "title": "Look, Learn and Leverage (L$^3$): Mitigating Visual-Domain Shift and\n  Discovering Intrinsic Relations via Symbolic Alignment", "abstract": "  Modern deep learning models have demonstrated outstanding performance on\ndiscovering the underlying mechanisms when both visual appearance and intrinsic\nrelations (e.g., causal structure) data are sufficient, such as Disentangled\nRepresentation Learning (DRL), Causal Representation Learning (CRL) and Visual\nQuestion Answering (VQA) methods. However, generalization ability of these\nmodels is challenged when the visual domain shifts and the relations data is\nabsent during finetuning. To address this challenge, we propose a novel\nlearning framework, Look, Learn and Leverage (L$^3$), which decomposes the\nlearning process into three distinct phases and systematically utilize the\nclass-agnostic segmentation masks as the common symbolic space to align visual\ndomains. Thus, a relations discovery model can be trained on the source domain,\nand when the visual domain shifts and the intrinsic relations are absent, the\npretrained relations discovery model can be directly reused and maintain a\nsatisfactory performance. Extensive performance evaluations are conducted on\nthree different tasks: DRL, CRL and VQA, and show outstanding results on all\nthree tasks, which reveals the advantages of L$^3$.\n", "date": "2024-08-30", "categories": []}
{"id": "2407.04980", "title": "Enabling Causal Discovery in Post-Nonlinear Models with Normalizing\n  Flows", "abstract": "  Post-nonlinear (PNL) causal models stand out as a versatile and adaptable\nframework for modeling intricate causal relationships. However, accurately\ncapturing the invertibility constraint required in PNL models remains\nchallenging in existing studies. To address this problem, we introduce CAF-PoNo\n(Causal discovery via Normalizing Flows for Post-Nonlinear models), harnessing\nthe power of the normalizing flows architecture to enforce the crucial\ninvertibility constraint in PNL models. Through normalizing flows, our method\nprecisely reconstructs the hidden noise, which plays a vital role in\ncause-effect identification through statistical independence testing.\nFurthermore, the proposed approach exhibits remarkable extensibility, as it can\nbe seamlessly expanded to facilitate multivariate causal discovery via causal\norder identification, empowering us to efficiently unravel complex causal\nrelationships. Extensive experimental evaluations on both simulated and real\ndatasets consistently demonstrate that the proposed method outperforms several\nstate-of-the-art approaches in both bivariate and multivariate causal discovery\ntasks.\n", "date": "2024-07-06", "categories": []}
{"id": "2407.04992", "title": "Scalable Variational Causal Discovery Unconstrained by Acyclicity", "abstract": "  Bayesian causal discovery offers the power to quantify epistemic\nuncertainties among a broad range of structurally diverse causal theories\npotentially explaining the data, represented in forms of directed acyclic\ngraphs (DAGs). However, existing methods struggle with efficient DAG sampling\ndue to the complex acyclicity constraint. In this study, we propose a scalable\nBayesian approach to effectively learn the posterior distribution over causal\ngraphs given observational data thanks to the ability to generate DAGs without\nexplicitly enforcing acyclicity. Specifically, we introduce a novel\ndifferentiable DAG sampling method that can generate a valid acyclic causal\ngraph by mapping an unconstrained distribution of implicit topological orders\nto a distribution over DAGs. Given this efficient DAG sampling scheme, we are\nable to model the posterior distribution over causal graphs using a simple\nvariational distribution over a continuous domain, which can be learned via the\nvariational inference framework. Extensive empirical experiments on both\nsimulated and real datasets demonstrate the superior performance of the\nproposed model compared to several state-of-the-art baselines.\n", "date": "2024-07-06", "categories": []}
{"id": "2308.09248", "title": "Active and Passive Causal Inference Learning", "abstract": "  This paper serves as a starting point for machine learning researchers,\nengineers and students who are interested in but not yet familiar with causal\ninference. We start by laying out an important set of assumptions that are\ncollectively needed for causal identification, such as exchangeability,\npositivity, consistency and the absence of interference. From these\nassumptions, we build out a set of important causal inference techniques, which\nwe do so by categorizing them into two buckets; active and passive approaches.\nWe describe and discuss randomized controlled trials and bandit-based\napproaches from the active category. We then describe classical approaches,\nsuch as matching and inverse probability weighting, in the passive category,\nfollowed by more recent deep learning based algorithms. By finishing the paper\nwith some of the missing aspects of causal inference from this paper, such as\ncollider biases, we expect this paper to provide readers with a diverse set of\nstarting points for further reading and research in causal inference and\ndiscovery.\n", "date": "2023-08-18", "categories": []}
{"id": "2408.13628", "title": "Enhancing Uplift Modeling in Multi-Treatment Marketing Campaigns:\n  Leveraging Score Ranking and Calibration Techniques", "abstract": "  Uplift modeling is essential for optimizing marketing strategies by selecting\nindividuals likely to respond positively to specific marketing campaigns. This\nimportance escalates in multi-treatment marketing campaigns, where diverse\ntreatment is available and we may want to assign the customers to treatment\nthat can make the most impact. While there are existing approaches with\nconvenient frameworks like Causalml, there are potential spaces to enhance the\neffect of uplift modeling in multi treatment cases. This paper introduces a\nnovel approach to uplift modeling in multi-treatment campaigns, leveraging\nscore ranking and calibration techniques to improve overall performance of the\nmarketing campaign. We review existing uplift models, including Meta Learner\nframeworks (S, T, X), and their application in real-world scenarios.\nAdditionally, we delve into insights from multi-treatment studies to highlight\nthe complexities and potential advancements in the field. Our methodology\nincorporates Meta-Learner calibration and a scoring rank-based offer selection\nstrategy. Extensive experiment results with real-world datasets demonstrate the\npractical benefits and superior performance of our approach. The findings\nunderscore the critical role of integrating score ranking and calibration\ntechniques in refining the performance and reliability of uplift predictions,\nthereby advancing predictive modeling in marketing analytics and providing\nactionable insights for practitioners seeking to optimize their campaign\nstrategies.\n", "date": "2024-08-24", "categories": []}
{"id": "2408.14101", "title": "Estimating Causal Effects from Learned Causal Networks", "abstract": "  The standard approach to answering an identifiable causal-effect query (e.g.,\n$P(Y|do(X)$) when given a causal diagram and observational data is to first\ngenerate an estimand, or probabilistic expression over the observable\nvariables, which is then evaluated using the observational data. In this paper,\nwe propose an alternative paradigm for answering causal-effect queries over\ndiscrete observable variables. We propose to instead learn the causal Bayesian\nnetwork and its confounding latent variables directly from the observational\ndata. Then, efficient probabilistic graphical model (PGM) algorithms can be\napplied to the learned model to answer queries. Perhaps surprisingly, we show\nthat this \\emph{model completion} learning approach can be more effective than\nestimand approaches, particularly for larger models in which the estimand\nexpressions become computationally difficult.\n  We illustrate our method's potential using a benchmark collection of Bayesian\nnetworks and synthetically generated causal models.\n", "date": "2024-08-26", "categories": []}
{"id": "2408.15055", "title": "Causal Rule Forest: Toward Interpretable and Precise Treatment Effect\n  Estimation", "abstract": "  Understanding and inferencing Heterogeneous Treatment Effects (HTE) and\nConditional Average Treatment Effects (CATE) are vital for developing\npersonalized treatment recommendations. Many state-of-the-art approaches\nachieve inspiring performance in estimating HTE on benchmark datasets or\nsimulation studies. However, the indirect predicting manner and complex model\narchitecture reduce the interpretability of these approaches. To mitigate the\ngap between predictive performance and heterogeneity interpretability, we\nintroduce the Causal Rule Forest (CRF), a novel approach to learning hidden\npatterns from data and transforming the patterns into interpretable multi-level\nBoolean rules. By training the other interpretable causal inference models with\ndata representation learned by CRF, we can reduce the predictive errors of\nthese models in estimating HTE and CATE, while keeping their interpretability\nfor identifying subgroups that a treatment is more effective. Our experiments\nunderscore the potential of CRF to advance personalized interventions and\npolicies, paving the way for future research to enhance its scalability and\napplication across complex causal inference challenges.\n", "date": "2024-08-27", "categories": []}
{"id": "2302.02560", "title": "Causal Estimation of Exposure Shifts with Neural Networks", "abstract": "  A fundamental task in causal inference is estimating the effect of\ndistribution shift in the treatment variable. We refer to this problem as\nshift-response function (SRF) estimation. Existing neural network methods for\ncausal inference lack theoretical guarantees and practical implementations for\nSRF estimation. In this paper, we introduce Targeted Regularization for\nExposure Shifts with Neural Networks (TRESNET), a method to estimate SRFs with\nrobustness and efficiency guarantees. Our contributions are twofold. First, we\npropose a targeted regularization loss for neural networks with theoretical\nproperties that ensure double robustness and asymptotic efficiency specific to\nSRF estimation. Second, we extend targeted regularization to support loss\nfunctions from the exponential family to accommodate non-continuous outcome\ndistributions (e.g., discrete counts). We conduct benchmark experiments\ndemonstrating TRESNET's broad applicability and competitiveness. We then apply\nour method to a key policy question in public health to estimate the causal\neffect of revising the US National Ambient Air Quality Standards (NAAQS) for PM\n2.5 from 12 ${\\mu}g/m^3$ to 9 ${\\mu}g/m^3$. This change has been recently\nproposed by the US Environmental Protection Agency (EPA). Our goal is to\nestimate the reduction in deaths that would result from this anticipated\nrevision using data consisting of 68 million individuals across the U.S.\n", "date": "2023-02-06", "categories": []}
{"id": "2306.01213", "title": "Learning Causally Disentangled Representations via the Principle of\n  Independent Causal Mechanisms", "abstract": "  Learning disentangled causal representations is a challenging problem that\nhas gained significant attention recently due to its implications for\nextracting meaningful information for downstream tasks. In this work, we define\na new notion of causal disentanglement from the perspective of independent\ncausal mechanisms. We propose ICM-VAE, a framework for learning causally\ndisentangled representations supervised by causally related observed labels. We\nmodel causal mechanisms using nonlinear learnable flow-based diffeomorphic\nfunctions to map noise variables to latent causal variables. Further, to\npromote the disentanglement of causal factors, we propose a causal\ndisentanglement prior learned from auxiliary labels and the latent causal\nstructure. We theoretically show the identifiability of causal factors and\nmechanisms up to permutation and elementwise reparameterization. We empirically\ndemonstrate that our framework induces highly disentangled causal factors,\nimproves interventional robustness, and is compatible with counterfactual\ngeneration.\n", "date": "2023-06-02", "categories": []}
{"id": "2404.11341", "title": "The Causal Chambers: Real Physical Systems as a Testbed for AI\n  Methodology", "abstract": "  In some fields of AI, machine learning and statistics, the validation of new\nmethods and algorithms is often hindered by the scarcity of suitable real-world\ndatasets. Researchers must often turn to simulated data, which yields limited\ninformation about the applicability of the proposed methods to real problems.\nAs a step forward, we have constructed two devices that allow us to quickly and\ninexpensively produce large datasets from non-trivial but well-understood\nphysical systems. The devices, which we call causal chambers, are\ncomputer-controlled laboratories that allow us to manipulate and measure an\narray of variables from these physical systems, providing a rich testbed for\nalgorithms from a variety of fields. We illustrate potential applications\nthrough a series of case studies in fields such as causal discovery,\nout-of-distribution generalization, change point detection, independent\ncomponent analysis, and symbolic regression. For applications to causal\ninference, the chambers allow us to carefully perform interventions. We also\nprovide and empirically validate a causal model of each chamber, which can be\nused as ground truth for different tasks. All hardware and software is made\nopen source, and the datasets are publicly available at causalchamber.org or\nthrough the Python package causalchamber.\n", "date": "2024-04-17", "categories": []}
{"id": "2404.17735", "title": "Causal Diffusion Autoencoders: Toward Counterfactual Generation via\n  Diffusion Probabilistic Models", "abstract": "  Diffusion probabilistic models (DPMs) have become the state-of-the-art in\nhigh-quality image generation. However, DPMs have an arbitrary noisy latent\nspace with no interpretable or controllable semantics. Although there has been\nsignificant research effort to improve image sample quality, there is little\nwork on representation-controlled generation using diffusion models.\nSpecifically, causal modeling and controllable counterfactual generation using\nDPMs is an underexplored area. In this work, we propose CausalDiffAE, a\ndiffusion-based causal representation learning framework to enable\ncounterfactual generation according to a specified causal model. Our key idea\nis to use an encoder to extract high-level semantically meaningful causal\nvariables from high-dimensional data and model stochastic variation using\nreverse diffusion. We propose a causal encoding mechanism that maps\nhigh-dimensional data to causally related latent factors and parameterize the\ncausal mechanisms among latent factors using neural networks. To enforce the\ndisentanglement of causal variables, we formulate a variational objective and\nleverage auxiliary label information in a prior to regularize the latent space.\nWe propose a DDIM-based counterfactual generation procedure subject to\ndo-interventions. Finally, to address the limited label supervision scenario,\nwe also study the application of CausalDiffAE when a part of the training data\nis unlabeled, which also enables granular control over the strength of\ninterventions in generating counterfactuals during inference. We empirically\nshow that CausalDiffAE learns a disentangled latent space and is capable of\ngenerating high-quality counterfactual images.\n", "date": "2024-04-27", "categories": []}
{"id": "2407.20003", "title": "On the Effects of Irrelevant Variables in Treatment Effect Estimation\n  with Deep Disentanglement", "abstract": "  Estimating treatment effects from observational data is paramount in\nhealthcare, education, and economics, but current deep disentanglement-based\nmethods to address selection bias are insufficiently handling irrelevant\nvariables. We demonstrate in experiments that this leads to prediction errors.\nWe disentangle pre-treatment variables with a deep embedding method and\nexplicitly identify and represent irrelevant variables, additionally to\ninstrumental, confounding and adjustment latent factors. To this end, we\nintroduce a reconstruction objective and create an embedding space for\nirrelevant variables using an attached autoencoder. Instead of relying on\nserendipitous suppression of irrelevant variables as in previous deep\ndisentanglement approaches, we explicitly force irrelevant variables into this\nembedding space and employ orthogonalization to prevent irrelevant information\nfrom leaking into the latent space representations of the other factors. Our\nexperiments with synthetic and real-world benchmark datasets show that we can\nbetter identify irrelevant variables and more precisely predict treatment\neffects than previous methods, while prediction quality degrades less when\nadditional irrelevant variables are introduced.\n", "date": "2024-07-29", "categories": []}
{"id": "2408.13498", "title": "Rethinking State Disentanglement in Causal Reinforcement Learning", "abstract": "  One of the significant challenges in reinforcement learning (RL) when dealing\nwith noise is estimating latent states from observations. Causality provides\nrigorous theoretical support for ensuring that the underlying states can be\nuniquely recovered through identifiability. Consequently, some existing work\nfocuses on establishing identifiability from a causal perspective to aid in the\ndesign of algorithms. However, these results are often derived from a purely\ncausal viewpoint, which may overlook the specific RL context. We revisit this\nresearch line and find that incorporating RL-specific context can reduce\nunnecessary assumptions in previous identifiability analyses for latent states.\nMore importantly, removing these assumptions allows algorithm design to go\nbeyond the earlier boundaries constrained by them. Leveraging these insights,\nwe propose a novel approach for general partially observable Markov Decision\nProcesses (POMDPs) by replacing the complicated structural constraints in\nprevious methods with two simple constraints for transition and reward\npreservation. With the two constraints, the proposed algorithm is guaranteed to\ndisentangle state and noise that is faithful to the underlying dynamics.\nEmpirical evidence from extensive benchmark control tasks demonstrates the\nsuperiority of our approach over existing counterparts in effectively\ndisentangling state belief from noise.\n", "date": "2024-08-24", "categories": []}
{"id": "2408.12669", "title": "Bayesian Network Modeling of Causal Influence within Cognitive Domains\n  and Clinical Dementia Severity Ratings for Western and Indian Cohorts", "abstract": "  This study investigates the causal relationships between Clinical Dementia\nRatings (CDR) and its six domain scores across two distinct aging datasets: the\nAlzheimer's Disease Neuroimaging Initiative (ADNI) and the Longitudinal Aging\nStudy of India (LASI). Using Directed Acyclic Graphs (DAGs) derived from\nBayesian network models, we analyze the dependencies among domain scores and\ntheir influence on the global CDR. Our approach leverages the PC algorithm to\nestimate the DAG structures for both datasets, revealing notable differences in\ncausal relationships and edge strengths between the Western and Indian\npopulations. The analysis highlights a stronger dependency of CDR scores on\nmemory functions in both datasets, but with significant variations in edge\nstrengths and node degrees. By contrasting these findings, we aim to elucidate\npopulation-specific differences and similarities in dementia progression,\nproviding insights that could inform targeted interventions and improve\nunderstanding of dementia across diverse demographic contexts.\n", "date": "2024-08-16", "categories": []}
{"id": "2408.12803", "title": "Multi-Treatment Multi-Task Uplift Modeling for Enhancing User Growth", "abstract": "  As a key component in boosting online user growth, uplift modeling aims to\nmeasure individual user responses (e.g., whether to play the game) to various\ntreatments, such as gaming bonuses, thereby enhancing business outcomes.\nHowever, previous research typically considers a single-task, single-treatment\nsetting, where only one treatment exists and the overall treatment effect is\nmeasured by a single type of user response. In this paper, we propose a\nMulti-Treatment Multi-Task (MTMT) uplift network to estimate treatment effects\nin a multi-task scenario. We identify the multi-treatment problem as a causal\ninference problem with a tiered response, comprising a base effect (from\noffering a treatment) and an incremental effect (from offering a specific type\nof treatment), where the base effect can be numerically much larger than the\nincremental effect. Specifically, MTMT separately encodes user features and\ntreatments. The user feature encoder uses a multi-gate mixture of experts\n(MMOE) network to encode relevant user features, explicitly learning inter-task\nrelations. The resultant embeddings are used to measure natural responses per\ntask. Furthermore, we introduce a treatment-user feature interaction module to\nmodel correlations between each treatment and user feature. Consequently, we\nseparately measure the base and incremental treatment effect for each task\nbased on the produced treatment-aware representations. Experimental results\nbased on an offline public dataset and an online proprietary dataset\ndemonstrate the effectiveness of MTMT in single/multi-treatment and\nsingle/multi-task settings. Additionally, MTMT has been deployed in our gaming\nplatform to improve user experience.\n", "date": "2024-08-23", "categories": []}
{"id": "2408.13155", "title": "Causal machine learning for sustainable agroecosystems", "abstract": "  In a changing climate, sustainable agriculture is essential for food security\nand environmental health. However, it is challenging to understand the complex\ninteractions among its biophysical, social, and economic components. Predictive\nmachine learning (ML), with its capacity to learn from data, is leveraged in\nsustainable agriculture for applications like yield prediction and weather\nforecasting. Nevertheless, it cannot explain causal mechanisms and remains\ndescriptive rather than prescriptive. To address this gap, we propose causal\nML, which merges ML's data processing with causality's ability to reason about\nchange. This facilitates quantifying intervention impacts for evidence-based\ndecision-making and enhances predictive model robustness. We showcase causal ML\nthrough eight diverse applications that benefit stakeholders across the\nagri-food chain, including farmers, policymakers, and researchers.\n", "date": "2024-08-23", "categories": []}
{"id": "2209.07111", "title": "$\\rho$-GNF: A Copula-based Sensitivity Analysis to Unobserved\n  Confounding Using Normalizing Flows", "abstract": "  We propose a novel sensitivity analysis to unobserved confounding in\nobservational studies using copulas and normalizing flows. Using the idea of\ninterventional equivalence of structural causal models, we develop $\\rho$-GNF\n($\\rho$-graphical normalizing flow), where $\\rho{\\in}[-1,+1]$ is a bounded\nsensitivity parameter. This parameter represents the back-door non-causal\nassociation due to unobserved confounding, and which is encoded with a Gaussian\ncopula. In other words, the $\\rho$-GNF enables scholars to estimate the average\ncausal effect (ACE) as a function of $\\rho$, while accounting for various\nassumed strengths of the unobserved confounding. The output of the $\\rho$-GNF\nis what we denote as the $\\rho_{curve}$ that provides the bounds for the ACE\ngiven an interval of assumed $\\rho$ values. In particular, the $\\rho_{curve}$\nenables scholars to identify the confounding strength required to nullify the\nACE, similar to other sensitivity analysis methods (e.g., the E-value).\nLeveraging on experiments from simulated and real-world data, we show the\nbenefits of $\\rho$-GNF. One benefit is that the $\\rho$-GNF uses a Gaussian\ncopula to encode the distribution of the unobserved causes, which is commonly\nused in many applied settings. This distributional assumption produces narrower\nACE bounds compared to other popular sensitivity analysis methods.\n", "date": "2022-09-15", "categories": []}
{"id": "2408.11967", "title": "Valuing an Engagement Surface using a Large Scale Dynamic Causal Model", "abstract": "  With recent rapid growth in online shopping, AI-powered Engagement Surfaces\n(ES) have become ubiquitous across retail services. These engagement surfaces\nperform an increasing range of functions, including recommending new products\nfor purchase, reminding customers of their orders and providing delivery\nnotifications. Understanding the causal effect of engagement surfaces on value\ndriven for customers and businesses remains an open scientific question. In\nthis paper, we develop a dynamic causal model at scale to disentangle value\nattributable to an ES, and to assess its effectiveness. We demonstrate the\napplication of this model to inform business decision-making by understanding\nreturns on investment in the ES, and identifying product lines and features\nwhere the ES adds the most value.\n", "date": "2024-08-21", "categories": []}
{"id": "2408.12292", "title": "Towards Deconfounded Image-Text Matching with Causal Inference", "abstract": "  Prior image-text matching methods have shown remarkable performance on many\nbenchmark datasets, but most of them overlook the bias in the dataset, which\nexists in intra-modal and inter-modal, and tend to learn the spurious\ncorrelations that extremely degrade the generalization ability of the model.\nFurthermore, these methods often incorporate biased external knowledge from\nlarge-scale datasets as prior knowledge into image-text matching model, which\nis inevitable to force model further learn biased associations. To address\nabove limitations, this paper firstly utilizes Structural Causal Models (SCMs)\nto illustrate how intra- and inter-modal confounders damage the image-text\nmatching. Then, we employ backdoor adjustment to propose an innovative\nDeconfounded Causal Inference Network (DCIN) for image-text matching task. DCIN\n(1) decomposes the intra- and inter-modal confounders and incorporates them\ninto the encoding stage of visual and textual features, effectively eliminating\nthe spurious correlations during image-text matching, and (2) uses causal\ninference to mitigate biases of external knowledge. Consequently, the model can\nlearn causality instead of spurious correlations caused by dataset bias.\nExtensive experiments on two well-known benchmark datasets, i.e., Flickr30K and\nMSCOCO, demonstrate the superiority of our proposed method.\n", "date": "2024-08-22", "categories": []}
{"id": "2408.11181", "title": "A Full DAG Score-Based Algorithm for Learning Causal Bayesian Networks\n  with Latent Confounders", "abstract": "  Causal Bayesian networks (CBN) are popular graphical probabilistic models\nthat encode causal relations among variables. Learning their graphical\nstructure from observational data has received a lot of attention in the\nliterature. When there exists no latent (unobserved) confounder, i.e., no\nunobserved direct common cause of some observed variables, learning algorithms\ncan be divided essentially into two classes: constraint-based and score-based\napproaches. The latter are often thought to be more robust than the former and\nto produce better results. However, to the best of our knowledge, when\nvariables are discrete, no score-based algorithm is capable of dealing with\nlatent confounders. This paper introduces the first fully score-based structure\nlearning algorithm searching the space of DAGs (directed acyclic graphs) that\nis capable of identifying the presence of some latent confounders. It is\njustified mathematically and experiments highlight its effectiveness.\n", "date": "2024-08-20", "categories": []}
{"id": "2208.05949", "title": "Valid Inference After Causal Discovery", "abstract": "  Causal discovery and causal effect estimation are two fundamental tasks in\ncausal inference. While many methods have been developed for each task\nindividually, statistical challenges arise when applying these methods jointly:\nestimating causal effects after running causal discovery algorithms on the same\ndata leads to \"double dipping,\" invalidating the coverage guarantees of\nclassical confidence intervals. To this end, we develop tools for valid\npost-causal-discovery inference. Across empirical studies, we show that a naive\ncombination of causal discovery and subsequent inference algorithms leads to\nhighly inflated miscoverage rates; on the other hand, applying our method\nprovides reliable coverage while achieving more accurate causal discovery than\ndata splitting.\n", "date": "2022-08-11", "categories": []}
{"id": "2305.00050", "title": "Causal Reasoning and Large Language Models: Opening a New Frontier for\n  Causality", "abstract": "  The causal capabilities of large language models (LLMs) are a matter of\nsignificant debate, with critical implications for the use of LLMs in\nsocietally impactful domains such as medicine, science, law, and policy. We\nconduct a \"behavorial\" study of LLMs to benchmark their capability in\ngenerating causal arguments. Across a wide range of tasks, we find that LLMs\ncan generate text corresponding to correct causal arguments with high\nprobability, surpassing the best-performing existing methods. Algorithms based\non GPT-3.5 and 4 outperform existing algorithms on a pairwise causal discovery\ntask (97%, 13 points gain), counterfactual reasoning task (92%, 20 points gain)\nand event causality (86% accuracy in determining necessary and sufficient\ncauses in vignettes). We perform robustness checks across tasks and show that\nthe capabilities cannot be explained by dataset memorization alone, especially\nsince LLMs generalize to novel datasets that were created after the training\ncutoff date.\n  That said, LLMs exhibit unpredictable failure modes, and we discuss the kinds\nof errors that may be improved and what are the fundamental limits of LLM-based\nanswers. Overall, by operating on the text metadata, LLMs bring capabilities so\nfar understood to be restricted to humans, such as using collected knowledge to\ngenerate causal graphs or identifying background causal context from natural\nlanguage. As a result, LLMs may be used by human domain experts to save effort\nin setting up a causal analysis, one of the biggest impediments to the\nwidespread adoption of causal methods. Given that LLMs ignore the actual data,\nour results also point to a fruitful research direction of developing\nalgorithms that combine LLMs with existing causal techniques. Code and datasets\nare available at https://github.com/py-why/pywhy-llm.\n", "date": "2023-04-28", "categories": []}
{"id": "2408.10976", "title": "Kernel-Based Differentiable Learning of Non-Parametric Directed Acyclic\n  Graphical Models", "abstract": "  Causal discovery amounts to learning a directed acyclic graph (DAG) that\nencodes a causal model. This model selection problem can be challenging due to\nits large combinatorial search space, particularly when dealing with\nnon-parametric causal models. Recent research has sought to bypass the\ncombinatorial search by reformulating causal discovery as a continuous\noptimization problem, employing constraints that ensure the acyclicity of the\ngraph. In non-parametric settings, existing approaches typically rely on\nfinite-dimensional approximations of the relationships between nodes, resulting\nin a score-based continuous optimization problem with a smooth acyclicity\nconstraint. In this work, we develop an alternative approximation method by\nutilizing reproducing kernel Hilbert spaces (RKHS) and applying general\nsparsity-inducing regularization terms based on partial derivatives. Within\nthis framework, we introduce an extended RKHS representer theorem. To enforce\nacyclicity, we advocate the log-determinant formulation of the acyclicity\nconstraint and show its stability. Finally, we assess the performance of our\nproposed RKHS-DAGMA procedure through simulations and illustrative data\nanalyses.\n", "date": "2024-08-20", "categories": []}
{"id": "2209.07028", "title": "Estimating large causal polytrees from small samples", "abstract": "  We consider the problem of estimating a large causal polytree from a\nrelatively small i.i.d. sample. This is motivated by the problem of determining\ncausal structure when the number of variables is very large compared to the\nsample size, such as in gene regulatory networks. We give an algorithm that\nrecovers the tree with high accuracy in such settings. The algorithm works\nunder essentially no distributional or modeling assumptions other than some\nmild non-degeneracy conditions.\n", "date": "2022-09-15", "categories": []}
{"id": "2304.02146", "title": "Structure Learning with Continuous Optimization: A Sober Look and Beyond", "abstract": "  This paper investigates in which cases continuous optimization for directed\nacyclic graph (DAG) structure learning can and cannot perform well and why this\nhappens, and suggests possible directions to make the search procedure more\nreliable. Reisach et al. (2021) suggested that the remarkable performance of\nseveral continuous structure learning approaches is primarily driven by a high\nagreement between the order of increasing marginal variances and the\ntopological order, and demonstrated that these approaches do not perform well\nafter data standardization. We analyze this phenomenon for continuous\napproaches assuming equal and non-equal noise variances, and show that the\nstatement may not hold in either case by providing counterexamples,\njustifications, and possible alternative explanations. We further demonstrate\nthat nonconvexity may be a main concern especially for the non-equal noise\nvariances formulation, while recent advances in continuous structure learning\nfail to achieve improvement in this case. Our findings suggest that future\nworks should take into account the non-equal noise variances formulation to\nhandle more general settings and for a more comprehensive empirical evaluation.\nLastly, we provide insights into other aspects of the search procedure,\nincluding thresholding and sparsity, and show that they play an important role\nin the final solutions.\n", "date": "2023-04-04", "categories": []}
{"id": "2405.10746", "title": "Causality in the Can: Diet Coke's Impact on Fatness", "abstract": "  Artificially sweetened beverages like Diet Coke are often considered better\nalternatives to sugary drinks, but the debate over their impact on health,\nparticularly in relation to obesity, continues. Previous research has\npredominantly used association-based methods with observational or Randomized\nControlled Trial (RCT) data, which may not accurately capture the causal\nrelationship between Diet Coke consumption and obesity, leading to potentially\nlimited conclusions. In contrast, we employed causal inference methods using\nstructural causal models, integrating both observational and RCT data.\nSpecifically, we utilized data from the National Health and Nutrition\nExamination Survey (NHANES), which includes diverse demographic information, as\nour observational data source. This data was then used to construct a causal\ngraph, and the back-door criterion, along with its adjustment formula, was\napplied to estimate the RCT data. We then calculated the counterfactual\nquantity, the Probability of Necessity and Sufficiency (PNS), using both NHANES\ndata and estimated RCT data. We propose that PNS is the essential metric for\nassessing the impact of Diet Coke on obesity. Our results indicate that between\n20 to 50 percent of individuals, especially those with poor dietary habits, are\nmore likely to gain weight from Diet Coke. Conversely, in groups like young\nfemales with healthier diets, only a small proportion experience weight gain\ndue to Diet Coke. These findings highlight the influence of individual\nlifestyle and potential hormonal factors on the varied effects of Diet Coke,\nproviding a new framework for understanding its nutritional impacts on public\nhealth.\n", "date": "2024-05-17", "categories": []}
{"id": "2402.11124", "title": "Implicit Causal Representation Learning via Switchable Mechanisms", "abstract": "  Learning causal representations from observational and interventional data in\nthe absence of known ground-truth graph structures necessitates implicit latent\ncausal representation learning. Implicit learning of causal mechanisms\ntypically involves two categories of interventional data: hard and soft\ninterventions. In real-world scenarios, soft interventions are often more\nrealistic than hard interventions, as the latter require fully controlled\nenvironments. Unlike hard interventions, which directly force changes in a\ncausal variable, soft interventions exert influence indirectly by affecting the\ncausal mechanism. However, the subtlety of soft interventions impose several\nchallenges for learning causal models. One challenge is that soft\nintervention's effects are ambiguous, since parental relations remain intact.\nIn this paper, we tackle the challenges of learning causal models using soft\ninterventions while retaining implicit modelling. We propose ICLR-SM, which\nmodels the effects of soft interventions by employing a causal mechanism switch\nvariable designed to toggle between different causal mechanisms. In our\nexperiments, we consistently observe improved learning of identifiable, causal\nrepresentations, compared to baseline approaches.\n", "date": "2024-02-16", "categories": []}
{"id": "2408.08815", "title": "An Empirical Examination of Balancing Strategy for Counterfactual\n  Estimation on Time Series", "abstract": "  Counterfactual estimation from observations represents a critical endeavor in\nnumerous application fields, such as healthcare and finance, with the primary\nchallenge being the mitigation of treatment bias. The balancing strategy aimed\nat reducing covariate disparities between different treatment groups serves as\na universal solution. However, when it comes to the time series data, the\neffectiveness of balancing strategies remains an open question, with a thorough\nanalysis of the robustness and applicability of balancing strategies still\nlacking. This paper revisits counterfactual estimation in the temporal setting\nand provides a brief overview of recent advancements in balancing strategies.\nMore importantly, we conduct a critical empirical examination for the\neffectiveness of the balancing strategies within the realm of temporal\ncounterfactual estimation in various settings on multiple datasets. Our\nfindings could be of significant interest to researchers and practitioners and\ncall for a reexamination of the balancing strategy in time series settings.\n", "date": "2024-08-16", "categories": []}
{"id": "2308.09952", "title": "Finding emergence in data by maximizing effective information", "abstract": "  Quantifying emergence and modeling emergent dynamics in a data-driven manner\nfor complex dynamical systems is challenging due to the lack of direct\nobservations at the micro-level. Thus, it's crucial to develop a framework to\nidentify emergent phenomena and capture emergent dynamics at the macro-level\nusing available data. Inspired by the theory of causal emergence (CE), this\npaper introduces a machine learning framework to learn macro-dynamics in an\nemergent latent space and quantify the degree of CE. The framework maximizes\neffective information, resulting in a macro-dynamics model with enhanced causal\neffects. Experimental results on simulated and real data demonstrate the\neffectiveness of the proposed framework. It quantifies degrees of CE\neffectively under various conditions and reveals distinct influences of\ndifferent noise types. It can learn a one-dimensional coarse-grained\nmacro-state from fMRI data, to represent complex neural activities during movie\nclip viewing. Furthermore, improved generalization to different test\nenvironments is observed across all simulation data.\n", "date": "2023-08-19", "categories": []}
{"id": "2408.07890", "title": "Local Causal Discovery with Background Knowledge", "abstract": "  Causality plays a pivotal role in various fields of study. Based on the\nframework of causal graphical models, previous works have proposed identifying\nwhether a variable is a cause or non-cause of a target in every Markov\nequivalent graph solely by learning a local structure. However, the presence of\nprior knowledge, often represented as a partially known causal graph, is common\nin many causal modeling applications. Leveraging this prior knowledge allows\nfor the further identification of causal relationships. In this paper, we first\npropose a method for learning the local structure using all types of causal\nbackground knowledge, including direct causal information, non-ancestral\ninformation and ancestral information. Then we introduce criteria for\nidentifying causal relationships based solely on the local structure in the\npresence of prior knowledge. We also apply out method to fair machine learning,\nand experiments involving local structure learning, causal relationship\nidentification, and fair machine learning demonstrate that our method is both\neffective and efficient.\n", "date": "2024-08-15", "categories": []}
{"id": "2408.08210", "title": "Does Reasoning Emerge? Examining the Probabilities of Causation in Large\n  Language Models", "abstract": "  Recent advances in AI have been significantly driven by the capabilities of\nlarge language models (LLMs) to solve complex problems in ways that resemble\nhuman thinking. However, there is an ongoing debate about the extent to which\nLLMs are capable of actual reasoning. Central to this debate are two key\nprobabilistic concepts that are essential for connecting causes to their\neffects: the probability of necessity (PN) and the probability of sufficiency\n(PS). This paper introduces a framework that is both theoretical and practical,\naimed at assessing how effectively LLMs are able to replicate real-world\nreasoning mechanisms using these probabilistic measures. By viewing LLMs as\nabstract machines that process information through a natural language\ninterface, we examine the conditions under which it is possible to compute\nsuitable approximations of PN and PS. Our research marks an important step\ntowards gaining a deeper understanding of when LLMs are capable of reasoning,\nas illustrated by a series of math examples.\n", "date": "2024-08-15", "categories": []}
{"id": "2408.07219", "title": "Causal Effect Estimation using identifiable Variational AutoEncoder with\n  Latent Confounders and Post-Treatment Variables", "abstract": "  Estimating causal effects from observational data is challenging, especially\nin the presence of latent confounders. Much work has been done on addressing\nthis challenge, but most of the existing research ignores the bias introduced\nby the post-treatment variables. In this paper, we propose a novel method of\njoint Variational AutoEncoder (VAE) and identifiable Variational AutoEncoder\n(iVAE) for learning the representations of latent confounders and latent\npost-treatment variables from their proxy variables, termed CPTiVAE, to achieve\nunbiased causal effect estimation from observational data. We further prove the\nidentifiability in terms of the representation of latent post-treatment\nvariables. Extensive experiments on synthetic and semi-synthetic datasets\ndemonstrate that the CPTiVAE outperforms the state-of-the-art methods in the\npresence of latent confounders and post-treatment variables. We further apply\nCPTiVAE to a real-world dataset to show its potential application.\n", "date": "2024-08-13", "categories": []}
{"id": "2408.07545", "title": "$\\chi$SPN: Characteristic Interventional Sum-Product Networks for Causal\n  Inference in Hybrid Domains", "abstract": "  Causal inference in hybrid domains, characterized by a mixture of discrete\nand continuous variables, presents a formidable challenge. We take a step\ntowards this direction and propose Characteristic Interventional Sum-Product\nNetwork ($\\chi$SPN) that is capable of estimating interventional distributions\nin presence of random variables drawn from mixed distributions. $\\chi$SPN uses\ncharacteristic functions in the leaves of an interventional SPN (iSPN) thereby\nproviding a unified view for discrete and continuous random variables through\nthe Fourier-Stieltjes transform of the probability measures. A neural network\nis used to estimate the parameters of the learned iSPN using the intervened\ndata. Our experiments on 3 synthetic heterogeneous datasets suggest that\n$\\chi$SPN can effectively capture the interventional distributions for both\ndiscrete and continuous variables while being expressive and causally adequate.\nWe also show that $\\chi$SPN generalize to multiple interventions while being\ntrained only on a single intervention data.\n", "date": "2024-08-14", "categories": []}
{"id": "2310.07665", "title": "Deep Backtracking Counterfactuals for Causally Compliant Explanations", "abstract": "  Counterfactuals answer questions of what would have been observed under\naltered circumstances and can therefore offer valuable insights. Whereas the\nclassical interventional interpretation of counterfactuals has been studied\nextensively, backtracking constitutes a less studied alternative where all\ncausal laws are kept intact. In the present work, we introduce a practical\nmethod called deep backtracking counterfactuals (DeepBC) for computing\nbacktracking counterfactuals in structural causal models that consist of deep\ngenerative components. We propose two distinct versions of our method--one\nutilizing Langevin Monte Carlo sampling and the other employing constrained\noptimization--to generate counterfactuals for high-dimensional data. As a\nspecial case, our formulation reduces to methods in the field of counterfactual\nexplanations. Compared to these, our approach represents a causally compliant,\nversatile and modular alternative. We demonstrate these properties\nexperimentally on a modified version of MNIST and CelebA.\n", "date": "2023-10-11", "categories": []}
{"id": "2402.05052", "title": "Causal Representation Learning from Multiple Distributions: A General\n  Setting", "abstract": "  In many problems, the measured variables (e.g., image pixels) are just\nmathematical functions of the latent causal variables (e.g., the underlying\nconcepts or objects). For the purpose of making predictions in changing\nenvironments or making proper changes to the system, it is helpful to recover\nthe latent causal variables $Z_i$ and their causal relations represented by\ngraph $\\mathcal{G}_Z$. This problem has recently been known as causal\nrepresentation learning. This paper is concerned with a general, completely\nnonparametric setting of causal representation learning from multiple\ndistributions (arising from heterogeneous data or nonstationary time series),\nwithout assuming hard interventions behind distribution changes. We aim to\ndevelop general solutions in this fundamental case; as a by product, this helps\nsee the unique benefit offered by other assumptions such as parametric causal\nmodels or hard interventions. We show that under the sparsity constraint on the\nrecovered graph over the latent variables and suitable sufficient change\nconditions on the causal influences, interestingly, one can recover the\nmoralized graph of the underlying directed acyclic graph, and the recovered\nlatent variables and their relations are related to the underlying causal model\nin a specific, nontrivial way. In some cases, most latent variables can even be\nrecovered up to component-wise transformations. Experimental results verify our\ntheoretical claims.\n", "date": "2024-02-07", "categories": []}
{"id": "2408.05584", "title": "Dynamical causality under invisible confounders", "abstract": "  Causality inference is prone to spurious causal interactions, due to the\nsubstantial confounders in a complex system. While many existing methods based\non the statistical methods or dynamical methods attempt to address\nmisidentification challenges, there remains a notable lack of effective methods\nto infer causality, in particular in the presence of invisible/unobservable\nconfounders. As a result, accurately inferring causation with invisible\nconfounders remains a largely unexplored and outstanding issue in data science\nand AI fields. In this work, we propose a method to overcome such challenges to\ninfer dynamical causality under invisible confounders (CIC method) and further\nreconstruct the invisible confounders from time-series data by developing an\northogonal decomposition theorem in a delay embedding space. The core of our\nCIC method lies in its ability to decompose the observed variables not in their\noriginal space but in their delay embedding space into the common and private\nsubspaces respectively, thereby quantifying causality between those variables\nboth theoretically and computationally. This theoretical foundation ensures the\ncausal detection for any high-dimensional system even with only two observed\nvariables under many invisible confounders, which is actually a long-standing\nproblem in the field. In addition to the invisible confounder problem, such a\ndecomposition actually makes the intertwined variables separable in the\nembedding space, thus also solving the non-separability problem of causal\ninference. Extensive validation of the CIC method is carried out using various\nreal datasets, and the experimental results demonstrates its effectiveness to\nreconstruct real biological networks even with unobserved confounders.\n", "date": "2024-08-10", "categories": []}
{"id": "2408.05647", "title": "Controlling for discrete unmeasured confounding in nonlinear causal\n  models", "abstract": "  Unmeasured confounding is a major challenge for identifying causal\nrelationships from non-experimental data. Here, we propose a method that can\naccommodate unmeasured discrete confounding. Extending recent identifiability\nresults in deep latent variable models, we show theoretically that confounding\ncan be detected and corrected under the assumption that the observed data is a\npiecewise affine transformation of a latent Gaussian mixture model and that the\nidentity of the mixture components is confounded. We provide a flow-based\nalgorithm to estimate this model and perform deconfounding. Experimental\nresults on synthetic and real-world data provide support for the effectiveness\nof our approach.\n", "date": "2024-08-10", "categories": []}
{"id": "2408.05788", "title": "Continual Learning of Nonlinear Independent Representations", "abstract": "  Identifying the causal relations between interested variables plays a pivotal\nrole in representation learning as it provides deep insights into the dataset.\nIdentifiability, as the central theme of this approach, normally hinges on\nleveraging data from multiple distributions (intervention, distribution shift,\ntime series, etc.). Despite the exciting development in this field, a practical\nbut often overlooked problem is: what if those distribution shifts happen\nsequentially? In contrast, any intelligence possesses the capacity to abstract\nand refine learned knowledge sequentially -- lifelong learning. In this paper,\nwith a particular focus on the nonlinear independent component analysis (ICA)\nframework, we move one step forward toward the question of enabling models to\nlearn meaningful (identifiable) representations in a sequential manner, termed\ncontinual causal representation learning. We theoretically demonstrate that\nmodel identifiability progresses from a subspace level to a component-wise\nlevel as the number of distributions increases. Empirically, we show that our\nmethod achieves performance comparable to nonlinear ICA methods trained jointly\non multiple offline distributions and, surprisingly, the incoming new\ndistribution does not necessarily benefit the identification of all latent\nvariables.\n", "date": "2024-08-11", "categories": []}
{"id": "2207.09768", "title": "Learning Counterfactually Invariant Predictors", "abstract": "  Notions of counterfactual invariance (CI) have proven essential for\npredictors that are fair, robust, and generalizable in the real world. We\npropose graphical criteria that yield a sufficient condition for a predictor to\nbe counterfactually invariant in terms of a conditional independence in the\nobservational distribution. In order to learn such predictors, we propose a\nmodel-agnostic framework, called Counterfactually Invariant Prediction (CIP),\nbuilding on the Hilbert-Schmidt Conditional Independence Criterion (HSCIC), a\nkernel-based conditional dependence measure. Our experimental results\ndemonstrate the effectiveness of CIP in enforcing counterfactual invariance\nacross various simulated and real-world datasets including scalar and\nmulti-variate settings.\n", "date": "2022-07-20", "categories": []}
{"id": "2408.04819", "title": "Interventional Causal Structure Discovery over Graphical Models with\n  Convergence and Optimality Guarantees", "abstract": "  Learning causal structure from sampled data is a fundamental problem with\napplications in various fields, including healthcare, machine learning and\nartificial intelligence. Traditional methods predominantly rely on\nobservational data, but there exist limits regarding the identifiability of\ncausal structures with only observational data. Interventional data, on the\nother hand, helps establish a cause-and-effect relationship by breaking the\ninfluence of confounding variables. It remains to date under-explored to\ndevelop a mathematical framework that seamlessly integrates both observational\nand interventional data in causal structure learning. Furthermore, existing\nstudies often focus on centralized approaches, necessitating the transfer of\nentire datasets to a single server, which lead to considerable communication\noverhead and heightened risks to privacy. To tackle these challenges, we\ndevelop a bilevel polynomial optimization (Bloom) framework. Bloom not only\nprovides a powerful mathematical modeling framework, underpinned by theoretical\nsupport, for causal structure discovery from both interventional and\nobservational data, but also aspires to an efficient causal discovery algorithm\nwith convergence and optimality guarantees. We further extend Bloom to a\ndistributed setting to reduce the communication overhead and mitigate data\nprivacy risks. It is seen through experiments on both synthetic and real-world\ndatasets that Bloom markedly surpasses other leading learning algorithms.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.04907", "title": "Causal Discovery of Linear Non-Gaussian Causal Models with Unobserved\n  Confounding", "abstract": "  We consider linear non-Gaussian structural equation models that involve\nlatent confounding. In this setting, the causal structure is identifiable, but,\nin general, it is not possible to identify the specific causal effects.\nInstead, a finite number of different causal effects result in the same\nobservational distribution. Most existing algorithms for identifying these\ncausal effects use overcomplete independent component analysis (ICA), which\noften suffers from convergence to local optima. Furthermore, the number of\nlatent variables must be known a priori. To address these issues, we propose an\nalgorithm that operates recursively rather than using overcomplete ICA. The\nalgorithm first infers a source, estimates the effect of the source and its\nlatent parents on their descendants, and then eliminates their influence from\nthe data. For both source identification and effect size estimation, we use\nrank conditions on matrices formed from higher-order cumulants. We prove\nasymptotic correctness under the mild assumption that locally, the number of\nlatent variables never exceeds the number of observed variables. Simulation\nstudies demonstrate that our method achieves comparable performance to\novercomplete ICA even though it does not know the number of latents in advance.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.04949", "title": "CROCODILE: Causality aids RObustness via COntrastive DIsentangled\n  LEarning", "abstract": "  Due to domain shift, deep learning image classifiers perform poorly when\napplied to a domain different from the training one. For instance, a classifier\ntrained on chest X-ray (CXR) images from one hospital may not generalize to\nimages from another hospital due to variations in scanner settings or patient\ncharacteristics. In this paper, we introduce our CROCODILE framework, showing\nhow tools from causality can foster a model's robustness to domain shift via\nfeature disentanglement, contrastive learning losses, and the injection of\nprior knowledge. This way, the model relies less on spurious correlations,\nlearns the mechanism bringing from images to prediction better, and outperforms\nbaselines on out-of-distribution (OOD) data. We apply our method to multi-label\nlung disease classification from CXRs, utilizing over 750000 images from four\ndatasets. Our bias-mitigation method improves domain generalization and\nfairness, broadening the applicability and reliability of deep learning models\nfor a safer medical image analysis. Find our code at:\nhttps://github.com/gianlucarloni/crocodile.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.04254", "title": "Generating Fine-Grained Causality in Climate Time Series Data for\n  Forecasting and Anomaly Detection", "abstract": "  Understanding the causal interaction of time series variables can contribute\nto time series data analysis for many real-world applications, such as climate\nforecasting and extreme weather alerts. However, causal relationships are\ndifficult to be fully observed in real-world complex settings, such as\nspatial-temporal data from deployed sensor networks. Therefore, to capture\nfine-grained causal relations among spatial-temporal variables for further a\nmore accurate and reliable time series analysis, we first design a conceptual\nfine-grained causal model named TBN Granger Causality, which adds\ntime-respecting Bayesian Networks to the previous time-lagged Neural Granger\nCausality to offset the instantaneous effects. Second, we propose an end-to-end\ndeep generative model called TacSas, which discovers TBN Granger Causality in a\ngenerative manner to help forecast time series data and detect possible\nanomalies during the forecast. For evaluations, besides the causality discovery\nbenchmark Lorenz-96, we also test TacSas on climate benchmark ERA5 for climate\nforecasting and the extreme weather benchmark of NOAA for extreme weather\nalerts.\n", "date": "2024-08-08", "categories": []}
{"id": "2208.03703", "title": "Granger Causality using Neural Networks", "abstract": "  Dependence between nodes in a network is an important concept that pervades\nmany areas including finance, politics, sociology, genomics and the brain\nsciences. One way to characterize dependence between components of a\nmultivariate time series data is via Granger Causality (GC). Standard\ntraditional approaches to GC estimation / inference commonly assume linear\ndynamics, however such simplification does not hold in many real-world\napplications where signals are inherently non-linear. In such cases, imposing\nlinear models such as vector autoregressive (VAR) models can lead to\nmis-characterization of true Granger Causal interactions. To overcome this\nlimitation, Tank et al (IEEE Transactions on Pattern Analysis and Machine\nLearning, 2022) proposed a solution that uses neural networks with sparse\nregularization penalties. The regularization encourages learnable weights to be\nsparse, which enables inference on GC. This paper overcomes the limitations of\ncurrent methods by leveraging advances in machine learning and deep learning\nwhich have been demonstrated to learn hidden patterns in the data. We propose\nnovel classes of models that can handle underlying non-linearity in a\ncomputationally efficient manner, simultaneously providing GC and lag order\nselection. Firstly, we present the Learned Kernel VAR (LeKVAR) model that\nlearns kernel parameterized by a shared neural net followed by penalization on\nlearnable weights to discover GC structure. Secondly, we show one can directly\ndecouple lags and individual time series importance via decoupled penalties.\nThis is important as we want to select the lag order during the process of GC\nestimation. This decoupling acts as a filtering and can be extended to any DL\nmodel including Multi-Layer Perceptrons (MLP), Recurrent Neural Networks (RNN),\nLong Short Term Memory Networks (LSTM), Transformers etc, for simultaneous GC\nestimation and lag selection.\n", "date": "2022-08-07", "categories": []}
{"id": "2407.02702", "title": "Practical Guide for Causal Pathways and Sub-group Disparity Analysis", "abstract": "  In this study, we introduce the application of causal disparity analysis to\nunveil intricate relationships and causal pathways between sensitive attributes\nand the targeted outcomes within real-world observational data. Our methodology\ninvolves employing causal decomposition analysis to quantify and examine the\ncausal interplay between sensitive attributes and outcomes. We also emphasize\nthe significance of integrating heterogeneity assessment in causal disparity\nanalysis to gain deeper insights into the impact of sensitive attributes within\nspecific sub-groups on outcomes. Our two-step investigation focuses on datasets\nwhere race serves as the sensitive attribute. The results on two datasets\nindicate the benefit of leveraging causal analysis and heterogeneity assessment\nnot only for quantifying biases in the data but also for disentangling their\ninfluences on outcomes. We demonstrate that the sub-groups identified by our\napproach to be affected the most by disparities are the ones with the largest\nML classification errors. We also show that grouping the data only based on a\nsensitive attribute is not enough, and through these analyses, we can find\nsub-groups that are directly affected by disparities. We hope that our findings\nwill encourage the adoption of such methodologies in future ethical AI\npractices and bias audits, fostering a more equitable and fair technological\nlandscape.\n", "date": "2024-07-02", "categories": []}
{"id": "2405.11250", "title": "Argumentative Causal Discovery", "abstract": "  Causal discovery amounts to unearthing causal relationships amongst features\nin data. It is a crucial companion to causal inference, necessary to build\nscientific knowledge without resorting to expensive or impossible randomised\ncontrol trials. In this paper, we explore how reasoning with symbolic\nrepresentations can support causal discovery. Specifically, we deploy\nassumption-based argumentation (ABA), a well-established and powerful knowledge\nrepresentation formalism, in combination with causality theories, to learn\ngraphs which reflect causal dependencies in the data. We prove that our method\nexhibits desirable properties, notably that, under natural conditions, it can\nretrieve ground-truth causal graphs. We also conduct experiments with an\nimplementation of our method in answer set programming (ASP) on four datasets\nfrom standard benchmarks in causal discovery, showing that our method compares\nwell against established baselines.\n", "date": "2024-05-18", "categories": []}
{"id": "2408.01445", "title": "MiranDa: Mimicking the Learning Processes of Human Doctors to Achieve\n  Causal Inference for Medication Recommendation", "abstract": "  To enhance therapeutic outcomes from a pharmacological perspective, we\npropose MiranDa, designed for medication recommendation, which is the first\nactionable model capable of providing the estimated length of stay in hospitals\n(ELOS) as counterfactual outcomes that guide clinical practice and model\ntraining. In detail, MiranDa emulates the educational trajectory of doctors\nthrough two gradient-scaling phases shifted by ELOS: an Evidence-based Training\nPhase that utilizes supervised learning and a Therapeutic Optimization Phase\ngrounds in reinforcement learning within the gradient space, explores optimal\nmedications by perturbations from ELOS. Evaluation of the Medical Information\nMart for Intensive Care III dataset and IV dataset, showcased the superior\nresults of our model across five metrics, particularly in reducing the ELOS.\nSurprisingly, our model provides structural attributes of medication\ncombinations proved in hyperbolic space and advocated \"procedure-specific\"\nmedication combinations. These findings posit that MiranDa enhanced medication\nefficacy. Notably, our paradigm can be applied to nearly all medical tasks and\nthose with information to evaluate predicted outcomes. The source code of the\nMiranDa model is available at https://github.com/azusakou/MiranDa.\n", "date": "2024-07-23", "categories": []}
{"id": "2408.01582", "title": "Conformal Diffusion Models for Individual Treatment Effect Estimation\n  and Inference", "abstract": "  Estimating treatment effects from observational data is of central interest\nacross numerous application domains. Individual treatment effect offers the\nmost granular measure of treatment effect on an individual level, and is the\nmost useful to facilitate personalized care. However, its estimation and\ninference remain underdeveloped due to several challenges. In this article, we\npropose a novel conformal diffusion model-based approach that addresses those\nintricate challenges. We integrate the highly flexible diffusion modeling, the\nmodel-free statistical inference paradigm of conformal inference, along with\npropensity score and covariate local approximation that tackle distributional\nshifts. We unbiasedly estimate the distributions of potential outcomes for\nindividual treatment effect, construct an informative confidence interval, and\nestablish rigorous theoretical guarantees. We demonstrate the competitive\nperformance of the proposed method over existing solutions through extensive\nnumerical studies.\n", "date": "2024-08-02", "categories": []}
{"id": "2408.01630", "title": "Fair Risk Minimization under Causal Path-Specific Effect Constraints", "abstract": "  This paper introduces a framework for estimating fair optimal predictions\nusing machine learning where the notion of fairness can be quantified using\npath-specific causal effects. We use a recently developed approach based on\nLagrange multipliers for infinite-dimensional functional estimation to derive\nclosed-form solutions for constrained optimization based on mean squared error\nand cross-entropy risk criteria. The theoretical forms of the solutions are\nanalyzed in detail and described as nuanced adjustments to the unconstrained\nminimizer. This analysis highlights important trade-offs between risk\nminimization and achieving fairnes. The theoretical solutions are also used as\nthe basis for construction of flexible semiparametric estimation strategies for\nthese nuisance components. We describe the robustness properties of our\nestimators in terms of achieving the optimal constrained risk, as well as in\nterms of controlling the value of the constraint. We study via simulation the\nimpact of using robust estimators of pathway-specific effects to validate our\ntheory. This work advances the discourse on algorithmic fairness by integrating\ncomplex causal considerations into model training, thus providing strategies\nfor implementing fair models in real-world applications.\n", "date": "2024-08-03", "categories": []}
{"id": "2408.02065", "title": "A Multi-class Ride-hailing Service Subsidy System Utilizing Deep Causal\n  Networks", "abstract": "  In the ride-hailing industry, subsidies are predominantly employed to\nincentivize consumers to place more orders, thereby fostering market growth.\nCausal inference techniques are employed to estimate the consumer elasticity\nwith different subsidy levels. However, the presence of confounding effects\nposes challenges in achieving an unbiased estimate of the uplift effect. We\nintroduce a consumer subsidizing system to capture relationships between\nsubsidy propensity and the treatment effect, which proves effective while\nmaintaining a lightweight online environment.\n", "date": "2024-08-04", "categories": []}
{"id": "2408.01023", "title": "Distilling interpretable causal trees from causal forests", "abstract": "  Machine learning methods for estimating treatment effect heterogeneity\npromise greater flexibility than existing methods that test a few pre-specified\nhypotheses. However, one problem these methods can have is that it can be\nchallenging to extract insights from complicated machine learning models. A\nhigh-dimensional distribution of conditional average treatment effects may give\naccurate, individual-level estimates, but it can be hard to understand the\nunderlying patterns; hard to know what the implications of the analysis are.\nThis paper proposes the Distilled Causal Tree, a method for distilling a\nsingle, interpretable causal tree from a causal forest. This compares well to\nexisting methods of extracting a single tree, particularly in noisy data or\nhigh-dimensional data where there are many correlated features. Here it even\noutperforms the base causal forest in most simulations. Its estimates are\ndoubly robust and asymptotically normal just as those of the causal forest are.\n", "date": "2024-08-02", "categories": []}
{"id": "2310.18615", "title": "Temporally Disentangled Representation Learning under Unknown\n  Nonstationarity", "abstract": "  In unsupervised causal representation learning for sequential data with\ntime-delayed latent causal influences, strong identifiability results for the\ndisentanglement of causally-related latent variables have been established in\nstationary settings by leveraging temporal structure. However, in nonstationary\nsetting, existing work only partially addressed the problem by either utilizing\nobserved auxiliary variables (e.g., class labels and/or domain indexes) as side\ninformation or assuming simplified latent causal dynamics. Both constrain the\nmethod to a limited range of scenarios. In this study, we further explored the\nMarkov Assumption under time-delayed causally related process in nonstationary\nsetting and showed that under mild conditions, the independent latent\ncomponents can be recovered from their nonlinear mixture up to a permutation\nand a component-wise transformation, without the observation of auxiliary\nvariables. We then introduce NCTRL, a principled estimation framework, to\nreconstruct time-delayed latent causal variables and identify their relations\nfrom measured sequential data only. Empirical evaluations demonstrated the\nreliable identification of time-delayed latent causal influences, with our\nmethodology substantially outperforming existing baselines that fail to exploit\nthe nonstationarity adequately and then, consequently, cannot distinguish\ndistribution shifts.\n", "date": "2023-10-28", "categories": []}
{"id": "2407.12254", "title": "COKE: Causal Discovery with Chronological Order and Expert Knowledge in\n  High Proportion of Missing Manufacturing Data", "abstract": "  Understanding causal relationships between machines is crucial for fault\ndiagnosis and optimization in manufacturing processes. Real-world datasets\nfrequently exhibit up to 90% missing data and high dimensionality from hundreds\nof sensors. These datasets also include domain-specific expert knowledge and\nchronological order information, reflecting the recording order across\ndifferent machines, which is pivotal for discerning causal relationships within\nthe manufacturing data. However, previous methods for handling missing data in\nscenarios akin to real-world conditions have not been able to effectively\nutilize expert knowledge. Conversely, prior methods that can incorporate expert\nknowledge struggle with datasets that exhibit missing values. Therefore, we\npropose COKE to construct causal graphs in manufacturing datasets by leveraging\nexpert knowledge and chronological order among sensors without imputing missing\ndata. Utilizing the characteristics of the recipe, we maximize the use of\nsamples with missing values, derive embeddings from intersections with an\ninitial graph that incorporates expert knowledge and chronological order, and\ncreate a sensor ordering graph. The graph-generating process has been optimized\nby an actor-critic architecture to obtain a final graph that has a maximum\nreward. Experimental evaluations in diverse settings of sensor quantities and\nmissing proportions demonstrate that our approach compared with the benchmark\nmethods shows an average improvement of 39.9% in the F1-score. Moreover, the\nF1-score improvement can reach 62.6% when considering the configuration similar\nto real-world datasets, and 85.0% in real-world semiconductor datasets. The\nsource code is available at https://github.com/OuTingYun/COKE.\n", "date": "2024-07-17", "categories": []}
{"id": "2408.00399", "title": "Unsupervised Pairwise Causal Discovery on Heterogeneous Data using\n  Mutual Information Measures", "abstract": "  A fundamental task in science is to determine the underlying causal relations\nbecause it is the knowledge of this functional structure what leads to the\ncorrect interpretation of an effect given the apparent associations in the\nobserved data. In this sense, Causal Discovery is a technique that tackles this\nchallenge by analyzing the statistical properties of the constituent variables.\nIn this work, we target the generalizability of the discovery method by\nfollowing a reductionist approach that only involves two variables, i.e., the\npairwise or bi-variate setting. We question the current (possibly misleading)\nbaseline results on the basis that they were obtained through supervised\nlearning, which is arguably contrary to this genuinely exploratory endeavor. In\nconsequence, we approach this problem in an unsupervised way, using robust\nMutual Information measures, and observing the impact of the different variable\ntypes, which is oftentimes ignored in the design of solutions. Thus, we provide\na novel set of standard unbiased results that can serve as a reference to guide\nfuture discovery tasks in completely unknown environments.\n", "date": "2024-08-01", "categories": []}
{"id": "2211.03374", "title": "Deep Causal Learning: Representation, Discovery and Inference", "abstract": "  Causal learning has garnered significant attention in recent years because it\nreveals the essential relationships that underpin phenomena and delineates the\nmechanisms by which the world evolves. Nevertheless, traditional causal\nlearning methods face numerous challenges and limitations, including\nhigh-dimensional, unstructured variables, combinatorial optimization problems,\nunobserved confounders, selection biases, and estimation inaccuracies. Deep\ncausal learning, which leverages deep neural networks, offers innovative\ninsights and solutions for addressing these challenges. Although numerous deep\nlearning-based methods for causal discovery and inference have been proposed,\nthere remains a dearth of reviews examining the underlying mechanisms by which\ndeep learning can enhance causal learning. In this article, we comprehensively\nreview how deep learning can contribute to causal learning by tackling\ntraditional challenges across three key dimensions: representation, discovery,\nand inference. We emphasize that deep causal learning is pivotal for advancing\nthe theoretical frontiers and broadening the practical applications of causal\nscience. We conclude by summarizing open issues and outlining potential\ndirections for future research.\n", "date": "2022-11-07", "categories": []}
{"id": "2407.18752", "title": "Knowledge Graph Structure as Prompt: Improving Small Language Models\n  Capabilities for Knowledge-based Causal Discovery", "abstract": "  Causal discovery aims to estimate causal structures among variables based on\nobservational data. Large Language Models (LLMs) offer a fresh perspective to\ntackle the causal discovery problem by reasoning on the metadata associated\nwith variables rather than their actual data values, an approach referred to as\nknowledge-based causal discovery. In this paper, we investigate the\ncapabilities of Small Language Models (SLMs, defined as LLMs with fewer than 1\nbillion parameters) with prompt-based learning for knowledge-based causal\ndiscovery. Specifically, we present KG Structure as Prompt, a novel approach\nfor integrating structural information from a knowledge graph, such as common\nneighbor nodes and metapaths, into prompt-based learning to enhance the\ncapabilities of SLMs. Experimental results on three types of biomedical and\nopen-domain datasets under few-shot settings demonstrate the effectiveness of\nour approach, surpassing most baselines and even conventional fine-tuning\napproaches trained on full datasets. Our findings further highlight the strong\ncapabilities of SLMs: in combination with knowledge graphs and prompt-based\nlearning, SLMs demonstrate the potential to surpass LLMs with larger number of\nparameters. Our code and datasets are available on GitHub.\n", "date": "2024-07-26", "categories": []}
{"id": "2407.20553", "title": "DiffusionCounterfactuals: Inferring High-dimensional Counterfactuals\n  with Guidance of Causal Representations", "abstract": "  Accurate estimation of counterfactual outcomes in high-dimensional data is\ncrucial for decision-making and understanding causal relationships and\nintervention outcomes in various domains, including healthcare, economics, and\nsocial sciences. However, existing methods often struggle to generate accurate\nand consistent counterfactuals, particularly when the causal relationships are\ncomplex. We propose a novel framework that incorporates causal mechanisms and\ndiffusion models to generate high-quality counterfactual samples guided by\ncausal representation. Our approach introduces a novel, theoretically grounded\ntraining and sampling process that enables the model to consistently generate\naccurate counterfactual high-dimensional data under multiple intervention\nsteps. Experimental results on various synthetic and real benchmarks\ndemonstrate the proposed approach outperforms state-of-the-art methods in\ngenerating accurate and high-quality counterfactuals, using different\nevaluation metrics.\n", "date": "2024-07-30", "categories": []}
{"id": "2407.20694", "title": "Detecting Causality in the Frequency Domain with Cross-Mapping Coherence", "abstract": "  Understanding causal relationships within a system is crucial for uncovering\nits underlying mechanisms. Causal discovery methods, which facilitate the\nconstruction of such models from time-series data, hold the potential to\nsignificantly advance scientific and engineering fields.\n  This study introduces the Cross-Mapping Coherence (CMC) method, designed to\nreveal causal connections in the frequency domain between time series. CMC\nbuilds upon nonlinear state-space reconstruction and extends the Convergent\nCross-Mapping algorithm to the frequency domain by utilizing coherence metrics\nfor evaluation. We tested the Cross-Mapping Coherence method using simulations\nof logistic maps, Lorenz systems, Kuramoto oscillators, and the Wilson-Cowan\nmodel of the visual cortex. CMC accurately identified the direction of causal\nconnections in all simulated scenarios. When applied to the Wilson-Cowan model,\nCMC yielded consistent results similar to spectral Granger causality.\n  Furthermore, CMC exhibits high sensitivity in detecting weak connections,\ndemonstrates sample efficiency, and maintains robustness in the presence of\nnoise.\n  In conclusion, the capability to determine directed causal influences across\ndifferent frequency bands allows CMC to provide valuable insights into the\ndynamics of complex, nonlinear systems.\n", "date": "2024-07-30", "categories": []}
{"id": "2402.06500", "title": "On the Fly Detection of Root Causes from Observed Data with Application\n  to IT Systems", "abstract": "  This paper introduces a new structural causal model tailored for representing\nthreshold-based IT systems and presents a new algorithm designed to rapidly\ndetect root causes of anomalies in such systems. When root causes are not\ncausally related, the method is proven to be correct; while an extension is\nproposed based on the intervention of an agent to relax this assumption. Our\nalgorithm and its agent-based extension leverage causal discovery from offline\ndata and engage in subgraph traversal when encountering new anomalies in online\ndata. Our extensive experiments demonstrate the superior performance of our\nmethods, even when applied to data generated from alternative structural causal\nmodels or real IT monitoring data.\n", "date": "2024-02-09", "categories": []}
{"id": "2407.19078", "title": "Practical Marketplace Optimization at Uber Using Causally-Informed\n  Machine Learning", "abstract": "  Budget allocation of marketplace levers, such as incentives for drivers and\npromotions for riders, has long been a technical and business challenge at\nUber; understanding lever budget changes' impact and estimating cost efficiency\nto achieve predefined budgets is crucial, with the goal of optimal allocations\nthat maximize business value; we introduce an end-to-end machine learning and\noptimization procedure to automate budget decision-making for cities, relying\non feature store, model training and serving, optimizers, and backtesting;\nproposing state-of-the-art deep learning (DL) estimator based on S-Learner and\na novel tensor B-Spline regression model, we solve high-dimensional\noptimization with ADMM and primal-dual interior point convex optimization,\nsubstantially improving Uber's resource allocation efficiency.\n", "date": "2024-07-26", "categories": []}
{"id": "2407.19688", "title": "Causal Interventional Prediction System for Robust and Explainable\n  Effect Forecasting", "abstract": "  Although the widespread use of AI systems in today's world is growing, many\ncurrent AI systems are found vulnerable due to hidden bias and missing\ninformation, especially in the most commonly used forecasting system. In this\nwork, we explore the robustness and explainability of AI-based forecasting\nsystems. We provide an in-depth analysis of the underlying causality involved\nin the effect prediction task and further establish a causal graph based on\ntreatment, adjustment variable, confounder, and outcome. Correspondingly, we\ndesign a causal interventional prediction system (CIPS) based on a variational\nautoencoder and fully conditional specification of multiple imputations.\nExtensive results demonstrate the superiority of our system over\nstate-of-the-art methods and show remarkable versatility and extensibility in\npractice.\n", "date": "2024-07-29", "categories": []}
{"id": "2309.03731", "title": "Using representation balancing to learn conditional-average dose\n  responses from clustered data", "abstract": "  Estimating a unit's responses to interventions with an associated dose, the\n\"conditional average dose response\" (CADR), is relevant in a variety of\ndomains, from healthcare to business, economics, and beyond. Such a response\ntypically needs to be estimated from observational data, which introduces\nseveral challenges. That is why the machine learning (ML) community has\nproposed several tailored CADR estimators. Yet, the proposal of most of these\nmethods requires strong assumptions on the distribution of data and the\nassignment of interventions, which go beyond the standard assumptions in causal\ninference. Whereas previous works have so far focused on smooth shifts in\ncovariate distributions across doses, in this work, we will study estimating\nCADR from clustered data and where different doses are assigned to different\nsegments of a population. On a novel benchmarking dataset, we show the impacts\nof clustered data on model performance and propose an estimator, CBRNet, that\nlearns cluster-agnostic and hence dose-agnostic covariate representations\nthrough representation balancing for unbiased CADR inference. We run extensive\nexperiments to illustrate the workings of our method and compare it with the\nstate of the art in ML for CADR estimation.\n", "date": "2023-09-07", "categories": []}
{"id": "2306.17557", "title": "The most likely common cause", "abstract": "  The common cause principle for two random variables $A$ and $B$ is examined\nin the case of causal insufficiency, when their common cause $C$ is known to\nexist, but only the joint probability of $A$ and $B$ is observed. As a result,\n$C$ cannot be uniquely identified (the latent confounder problem). We show that\nthe generalized maximum likelihood method can be applied to this situation and\nallows identification of $C$ that is consistent with the common cause\nprinciple. It closely relates to the maximum entropy principle. Investigation\nof the two binary symmetric variables reveals a non-analytic behavior of\nconditional probabilities reminiscent of a second-order phase transition. This\noccurs during the transition from correlation to anti-correlation in the\nobserved probability distribution. The relation between the generalized\nlikelihood approach and alternative methods, such as predictive likelihood and\nthe minimum common cause entropy, is discussed. The consideration of the common\ncause for three observed variables (and one hidden cause) uncovers causal\nstructures that defy representation through directed acyclic graphs with the\nMarkov condition.\n", "date": "2023-06-30", "categories": []}
{"id": "2407.17910", "title": "Causal Deepsets for Off-policy Evaluation under Spatial or\n  Spatio-temporal Interferences", "abstract": "  Off-policy evaluation (OPE) is widely applied in sectors such as\npharmaceuticals and e-commerce to evaluate the efficacy of novel products or\npolicies from offline datasets. This paper introduces a causal deepset\nframework that relaxes several key structural assumptions, primarily the\nmean-field assumption, prevalent in existing OPE methodologies that handle\nspatio-temporal interference. These traditional assumptions frequently prove\ninadequate in real-world settings, thereby restricting the capability of\ncurrent OPE methods to effectively address complex interference effects. In\nresponse, we advocate for the implementation of the permutation invariance (PI)\nassumption. This innovative approach enables the data-driven, adaptive learning\nof the mean-field function, offering a more flexible estimation method beyond\nconventional averaging. Furthermore, we present novel algorithms that\nincorporate the PI assumption into OPE and thoroughly examine their theoretical\nfoundations. Our numerical analyses demonstrate that this novel approach yields\nsignificantly more precise estimations than existing baseline algorithms,\nthereby substantially improving the practical applicability and effectiveness\nof OPE methodologies. A Python implementation of our proposed method is\navailable at https://github.com/BIG-S2/Causal-Deepsets.\n", "date": "2024-07-25", "categories": []}
{"id": "2407.16388", "title": "Anwendung von Causal-Discovery-Algorithmen zur Root-Cause-Analyse in der\n  Fahrzeugmontage", "abstract": "  Root Cause Analysis (RCA) is a quality management method that aims to\nsystematically investigate and identify the cause-and-effect relationships of\nproblems and their underlying causes. Traditional methods are based on the\nanalysis of problems by subject matter experts. In modern production processes,\nlarge amounts of data are collected. For this reason, increasingly\ncomputer-aided and data-driven methods are used for RCA. One of these methods\nare Causal Discovery Algorithms (CDA). This publication demonstrates the\napplication of CDA on data from the assembly of a leading automotive\nmanufacturer. The algorithms used learn the causal structure between the\ncharacteristics of the manufactured vehicles, the ergonomics and the temporal\nscope of the involved assembly processes, and quality-relevant product features\nbased on representative data. This publication compares various CDAs in terms\nof their suitability in the context of quality management. For this purpose,\nthe causal structures learned by the algorithms as well as their runtime are\ncompared. This publication provides a contribution to quality management and\ndemonstrates how CDAs can be used for RCA in assembly processes.\n", "date": "2024-07-23", "categories": []}
{"id": "2403.00957", "title": "Resolution of Simpson's paradox via the common cause principle", "abstract": "  Simpson's paradox is an obstacle to establishing a probabilistic association\nbetween two events $a_1$ and $a_2$, given the third (lurking) random variable\n$B$. We focus on scenarios when the random variables $A$ (which combines $a_1$,\n$a_2$, and their complements) and $B$ have a common cause $C$ that need not be\nobserved. Alternatively, we can assume that $C$ screens out $A$ from $B$. For\nsuch cases, the correct association between $a_1$ and $a_2$ is to be defined\nvia conditioning over $C$. This setup generalizes the original Simpson's\nparadox: now its two contradicting options refer to two particular and\ndifferent causes $C$. We show that if $B$ and $C$ are binary and $A$ is\nquaternary (the minimal and the most widespread situation for the Simpson's\nparadox), the conditioning over any binary common cause $C$ establishes the\nsame direction of association between $a_1$ and $a_2$ as the conditioning over\n$B$ in the original formulation of the paradox. Thus, for the minimal common\ncause, one should choose the option of Simpson's paradox that assumes\nconditioning over $B$ and not its marginalization. The same conclusion is\nreached when Simpson's paradox is formulated via 3 continuous Gaussian\nvariables: within the minimal formulation of the paradox (3 scalar continuous\nvariables $A_1$, $A_2$, and $B$), one should choose the option with the\nconditioning over $B$.\n", "date": "2024-03-01", "categories": []}
{"id": "2407.15259", "title": "New Rules for Causal Identification with Background Knowledge", "abstract": "  Identifying causal relations is crucial for a variety of downstream tasks. In\nadditional to observational data, background knowledge (BK), which could be\nattained from human expertise or experiments, is usually introduced for\nuncovering causal relations. This raises an open problem that in the presence\nof latent variables, what causal relations are identifiable from observational\ndata and BK. In this paper, we propose two novel rules for incorporating BK,\nwhich offer a new perspective to the open problem. In addition, we show that\nthese rules are applicable in some typical causality tasks, such as determining\nthe set of possible causal effects with observational data. Our rule-based\napproach enhances the state-of-the-art method by circumventing a process of\nenumerating block sets that would otherwise take exponential complexity.\n", "date": "2024-07-21", "categories": []}
{"id": "2207.13665", "title": "Causal foundations of bias, disparity and fairness", "abstract": "  The study of biases, such as gender or racial biases, is an important topic\nin the social and behavioural sciences. However, the literature does not always\nclearly define the concept. Definitions of bias are often ambiguous or not\nprovided at all. To study biases in a precise manner, it is important to have a\nwell-defined concept of bias. We propose to define bias as a direct causal\neffect that is unjustified. We propose to define the closely related concept of\ndisparity as a direct or indirect causal effect that includes a bias. Our\nproposed definitions can be used to study biases and disparities in a more\nrigorous and systematic way. We compare our definitions of bias and disparity\nwith various criteria of fairness introduced in the artificial intelligence\nliterature. In addition, we discuss how our definitions relate to\ndiscrimination. We illustrate our definitions of bias and disparity in two case\nstudies, focusing on gender bias in science and racial bias in police\nshootings. Our proposed definitions aim to contribute to a better appreciation\nof the causal intricacies of studies of biases and disparities. We hope that\nthis will also promote an improved understanding of the policy implications of\nsuch studies.\n", "date": "2022-07-27", "categories": []}
{"id": "2402.10877", "title": "Robust agents learn causal world models", "abstract": "  It has long been hypothesised that causal reasoning plays a fundamental role\nin robust and general intelligence. However, it is not known if agents must\nlearn causal models in order to generalise to new domains, or if other\ninductive biases are sufficient. We answer this question, showing that any\nagent capable of satisfying a regret bound under a large set of distributional\nshifts must have learned an approximate causal model of the data generating\nprocess, which converges to the true causal model for optimal agents. We\ndiscuss the implications of this result for several research areas including\ntransfer learning and causal inference.\n", "date": "2024-02-16", "categories": []}
{"id": "2407.14022", "title": "Causal Inference with Complex Treatments: A Survey", "abstract": "  Causal inference plays an important role in explanatory analysis and decision\nmaking across various fields like statistics, marketing, health care, and\neducation. Its main task is to estimate treatment effects and make intervention\npolicies. Traditionally, most of the previous works typically focus on the\nbinary treatment setting that there is only one treatment for a unit to adopt\nor not. However, in practice, the treatment can be much more complex,\nencompassing multi-valued, continuous, or bundle options. In this paper, we\nrefer to these as complex treatments and systematically and comprehensively\nreview the causal inference methods for addressing them. First, we formally\nrevisit the problem definition, the basic assumptions, and their possible\nvariations under specific conditions. Second, we sequentially review the\nrelated methods for multi-valued, continuous, and bundled treatment settings.\nIn each situation, we tentatively divide the methods into two categories: those\nconforming to the unconfoundedness assumption and those violating it.\nSubsequently, we discuss the available datasets and open-source codes. Finally,\nwe provide a brief summary of these works and suggest potential directions for\nfuture research.\n", "date": "2024-07-19", "categories": []}
{"id": "2407.14065", "title": "MSCT: Addressing Time-Varying Confounding with Marginal Structural\n  Causal Transformer for Counterfactual Post-Crash Traffic Prediction", "abstract": "  Traffic crashes profoundly impede traffic efficiency and pose economic\nchallenges. Accurate prediction of post-crash traffic status provides essential\ninformation for evaluating traffic perturbations and developing effective\nsolutions. Previous studies have established a series of deep learning models\nto predict post-crash traffic conditions, however, these correlation-based\nmethods cannot accommodate the biases caused by time-varying confounders and\nthe heterogeneous effects of crashes. The post-crash traffic prediction model\nneeds to estimate the counterfactual traffic speed response to hypothetical\ncrashes under various conditions, which demonstrates the necessity of\nunderstanding the causal relationship between traffic factors. Therefore, this\npaper presents the Marginal Structural Causal Transformer (MSCT), a novel deep\nlearning model designed for counterfactual post-crash traffic prediction. To\naddress the issue of time-varying confounding bias, MSCT incorporates a\nstructure inspired by Marginal Structural Models and introduces a balanced loss\nfunction to facilitate learning of invariant causal features. The proposed\nmodel is treatment-aware, with a specific focus on comprehending and predicting\ntraffic speed under hypothetical crash intervention strategies. In the absence\nof ground-truth data, a synthetic data generation procedure is proposed to\nemulate the causal mechanism between traffic speed, crashes, and covariates.\nThe model is validated using both synthetic and real-world data, demonstrating\nthat MSCT outperforms state-of-the-art models in multi-step-ahead prediction\nperformance. This study also systematically analyzes the impact of time-varying\nconfounding bias and dataset distribution on model performance, contributing\nvaluable insights into counterfactual prediction for intelligent transportation\nsystems.\n", "date": "2024-07-19", "categories": []}
{"id": "2407.14214", "title": "Domain Adaptation for Industrial Time-series Forecasting via\n  Counterfactual Inference", "abstract": "  Industrial time-series, as a structural data responds to production process\ninformation, can be utilized to perform data-driven decision-making for\neffective monitoring of industrial production process. However, there are some\nchallenges for time-series forecasting in industry, e.g., predicting few-shot\ncaused by data shortage, and decision-confusing caused by unknown treatment\npolicy. To cope with the problems, we propose a novel causal domain adaptation\nframework, Causal Domain Adaptation (CDA) forecaster to improve the performance\non the interested domain with limited data (target). Firstly, we analyze the\ncausality existing along with treatments, and thus ensure the shared causality\nover time. Subsequently, we propose an answer-based attention mechanism to\nachieve domain-invariant representation by the shared causality in both\ndomains. Then, a novel domain-adaptation is built to model treatments and\noutcomes jointly training on source and target domain. The main insights are\nthat our designed answer-based attention mechanism allows the target domain to\nleverage the existed causality in source time-series even with different\ntreatments, and our forecaster can predict the counterfactual outcome of\nindustrial time-series, meaning a guidance in production process. Compared with\ncommonly baselines, our method on real-world and synthetic oilfield datasets\ndemonstrates the effectiveness in across-domain prediction and the practicality\nin guiding production process\n", "date": "2024-07-19", "categories": []}
{"id": "2407.13664", "title": "Decision Focused Causal Learning for Direct Counterfactual Marketing\n  Optimization", "abstract": "  Marketing optimization plays an important role to enhance user engagement in\nonline Internet platforms. Existing studies usually formulate this problem as a\nbudget allocation problem and solve it by utilizing two fully decoupled stages,\ni.e., machine learning (ML) and operation research (OR). However, the learning\nobjective in ML does not take account of the downstream optimization task in\nOR, which causes that the prediction accuracy in ML may be not positively\nrelated to the decision quality.\n  Decision Focused Learning (DFL) integrates ML and OR into an end-to-end\nframework, which takes the objective of the downstream task as the decision\nloss function and guarantees the consistency of the optimization direction\nbetween ML and OR. However, deploying DFL in marketing is non-trivial due to\nmultiple technological challenges. Firstly, the budget allocation problem in\nmarketing is a 0-1 integer stochastic programming problem and the budget is\nuncertain and fluctuates a lot in real-world settings, which is beyond the\ngeneral problem background in DFL. Secondly, the counterfactual in marketing\ncauses that the decision loss cannot be directly computed and the optimal\nsolution can never be obtained, both of which disable the common\ngradient-estimation approaches in DFL. Thirdly, the OR solver is called\nfrequently to compute the decision loss during model training in DFL, which\nproduces huge computational cost and cannot support large-scale training data.\nIn this paper, we propose a decision focused causal learning framework (DFCL)\nfor direct counterfactual marketing optimization, which overcomes the above\ntechnological challenges. Both offline experiments and online A/B testing\ndemonstrate the effectiveness of DFCL over the state-of-the-art methods.\nCurrently, DFCL has been deployed in several marketing scenarios in Meituan,\none of the largest online food delivery platform in the world.\n", "date": "2024-07-18", "categories": []}
{"id": "2407.12437", "title": "Variable-Agnostic Causal Exploration for Reinforcement Learning", "abstract": "  Modern reinforcement learning (RL) struggles to capture real-world\ncause-and-effect dynamics, leading to inefficient exploration due to extensive\ntrial-and-error actions. While recent efforts to improve agent exploration have\nleveraged causal discovery, they often make unrealistic assumptions of causal\nvariables in the environments. In this paper, we introduce a novel framework,\nVariable-Agnostic Causal Exploration for Reinforcement Learning (VACERL),\nincorporating causal relationships to drive exploration in RL without\nspecifying environmental causal variables. Our approach automatically\nidentifies crucial observation-action steps associated with key variables using\nattention mechanisms. Subsequently, it constructs the causal graph connecting\nthese steps, which guides the agent towards observation-action pairs with\ngreater causal influence on task completion. This can be leveraged to generate\nintrinsic rewards or establish a hierarchy of subgoals to enhance exploration\nefficiency. Experimental results showcase a significant improvement in agent\nperformance in grid-world, 2d games and robotic domains, particularly in\nscenarios with sparse rewards and noisy actions, such as the notorious Noisy-TV\nenvironments.\n", "date": "2024-07-17", "categories": []}
{"id": "2407.12528", "title": "On the Complexity of Identification in Linear Structural Causal Models", "abstract": "  Learning the unknown causal parameters of a linear structural causal model is\na fundamental task in causal analysis. The task, known as the problem of\nidentification, asks to estimate the parameters of the model from a combination\nof assumptions on the graphical structure of the model and observational data,\nrepresented as a non-causal covariance matrix. In this paper, we give a new\nsound and complete algorithm for generic identification which runs in\npolynomial space. By standard simulation results, this algorithm has\nexponential running time which vastly improves the state-of-the-art double\nexponential time method using a Gr\\\"obner basis approach. The paper also\npresents evidence that parameter identification is computationally hard in\ngeneral. In particular, we prove, that the task asking whether, for a given\nfeasible correlation matrix, there are exactly one or two or more parameter\nsets explaining the observed matrix, is hard for $\\forall R$, the co-class of\nthe existential theory of the reals. In particular, this problem is\n$coNP$-hard. To our best knowledge, this is the first hardness result for some\nnotion of identifiability.\n", "date": "2024-07-17", "categories": []}
{"id": "2407.11032", "title": "Mechanisms for Data Sharing in Collaborative Causal Inference (Extended\n  Version)", "abstract": "  Collaborative causal inference (CCI) is a federated learning method for\npooling data from multiple, often self-interested, parties, to achieve a common\nlearning goal over causal structures, e.g. estimation and optimization of\ntreatment variables in a medical setting. Since obtaining data can be costly\nfor the participants and sharing unique data poses the risk of losing\ncompetitive advantages, motivating the participation of all parties through\nequitable rewards and incentives is necessary. This paper devises an evaluation\nscheme to measure the value of each party's data contribution to the common\nlearning task, tailored to causal inference's statistical demands, by comparing\ncompleted partially directed acyclic graphs (CPDAGs) inferred from\nobservational data contributed by the participants. The Data Valuation Scheme\nthus obtained can then be used to introduce mechanisms that incentivize the\nagents to contribute data. It can be leveraged to reward agents fairly,\naccording to the quality of their data, or to maximize all agents' data\ncontributions.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.11056", "title": "Industrial-Grade Time-Dependent Counterfactual Root Cause Analysis\n  through the Unanticipated Point of Incipient Failure: a Proof of Concept", "abstract": "  This paper describes the development of a counterfactual Root Cause Analysis\ndiagnosis approach for an industrial multivariate time series environment. It\ndrives the attention toward the Point of Incipient Failure, which is the moment\nin time when the anomalous behavior is first observed, and where the root cause\nis assumed to be found before the issue propagates. The paper presents the\nelementary but essential concepts of the solution and illustrates them\nexperimentally on a simulated setting. Finally, it discusses avenues of\nimprovement for the maturity of the causal technology to meet the robustness\nchallenges of increasingly complex environments in the industry.\n", "date": "2024-07-10", "categories": []}
{"id": "2407.11927", "title": "Bayesian Causal Forests for Longitudinal Data: Assessing the Impact of\n  Part-Time Work on Growth in High School Mathematics Achievement", "abstract": "  Modelling growth in student achievement is a significant challenge in the\nfield of education. Understanding how interventions or experiences such as\npart-time work can influence this growth is also important. Traditional methods\nlike difference-in-differences are effective for estimating causal effects from\nlongitudinal data. Meanwhile, Bayesian non-parametric methods have recently\nbecome popular for estimating causal effects from single time point\nobservational studies. However, there remains a scarcity of methods capable of\ncombining the strengths of these two approaches to flexibly estimate\nheterogeneous causal effects from longitudinal data. Motivated by two waves of\ndata from the High School Longitudinal Study, the NCES' most recent\nlongitudinal study which tracks a representative sample of over 20,000 students\nin the US, our study introduces a longitudinal extension of Bayesian Causal\nForests. This model allows for the flexible identification of both individual\ngrowth in mathematical ability and the effects of participation in part-time\nwork. Simulation studies demonstrate the predictive performance and reliable\nuncertainty quantification of the proposed model. Results reveal the negative\nimpact of part time work for most students, but hint at potential benefits for\nthose students with an initially low sense of school belonging. Clear signs of\na widening achievement gap between students with high and low academic\nachievement are also identified. Potential policy implications are discussed,\nalong with promising areas for future research.\n", "date": "2024-07-16", "categories": []}
{"id": "2305.15742", "title": "Counterfactual Generative Models for Time-Varying Treatments", "abstract": "  Estimating the counterfactual outcome of treatment is essential for\ndecision-making in public health and clinical science, among others. Often,\ntreatments are administered in a sequential, time-varying manner, leading to an\nexponentially increased number of possible counterfactual outcomes.\nFurthermore, in modern applications, the outcomes are high-dimensional and\nconventional average treatment effect estimation fails to capture disparities\nin individuals. To tackle these challenges, we propose a novel conditional\ngenerative framework capable of producing counterfactual samples under\ntime-varying treatment, without the need for explicit density estimation. Our\nmethod carefully addresses the distribution mismatch between the observed and\ncounterfactual distributions via a loss function based on inverse probability\nre-weighting, and supports integration with state-of-the-art conditional\ngenerative models such as the guided diffusion and conditional variational\nautoencoder. We present a thorough evaluation of our method using both\nsynthetic and real-world data. Our results demonstrate that our method is\ncapable of generating high-quality counterfactual samples and outperforms the\nstate-of-the-art baselines.\n", "date": "2023-05-25", "categories": []}
{"id": "2407.10132", "title": "Optimal Kernel Choice for Score Function-based Causal Discovery", "abstract": "  Score-based methods have demonstrated their effectiveness in discovering\ncausal relationships by scoring different causal structures based on their\ngoodness of fit to the data. Recently, Huang et al. proposed a generalized\nscore function that can handle general data distributions and causal\nrelationships by modeling the relations in reproducing kernel Hilbert space\n(RKHS). The selection of an appropriate kernel within this score function is\ncrucial for accurately characterizing causal relationships and ensuring precise\ncausal discovery. However, the current method involves manual heuristic\nselection of kernel parameters, making the process tedious and less likely to\nensure optimality. In this paper, we propose a kernel selection method within\nthe generalized score function that automatically selects the optimal kernel\nthat best fits the data. Specifically, we model the generative process of the\nvariables involved in each step of the causal graph search procedure as a\nmixture of independent noise variables. Based on this model, we derive an\nautomatic kernel selection method by maximizing the marginal likelihood of the\nvariables involved in each search step. We conduct experiments on both\nsynthetic data and real-world benchmarks, and the results demonstrate that our\nproposed method outperforms heuristic kernel selection methods.\n", "date": "2024-07-14", "categories": []}
{"id": "2407.10779", "title": "The Missing Link: Allocation Performance in Causal Machine Learning", "abstract": "  Automated decision-making (ADM) systems are being deployed across a diverse\nrange of critical problem areas such as social welfare and healthcare. Recent\nwork highlights the importance of causal ML models in ADM systems, but\nimplementing them in complex social environments poses significant challenges.\nResearch on how these challenges impact the performance in specific downstream\ndecision-making tasks is limited. Addressing this gap, we make use of a\ncomprehensive real-world dataset of jobseekers to illustrate how the\nperformance of a single CATE model can vary significantly across different\ndecision-making scenarios and highlight the differential influence of\nchallenges such as distribution shifts on predictions and allocations.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.10878", "title": "Deep Causal Learning to Explain and Quantify The Geo-Tension's Impact on\n  Natural Gas Market", "abstract": "  Natural gas demand is a crucial factor for predicting natural gas prices and\nthus has a direct influence on the power system. However, existing methods face\nchallenges in assessing the impact of shocks, such as the outbreak of the\nRussian-Ukrainian war. In this context, we apply deep neural network-based\nGranger causality to identify important drivers of natural gas demand.\nFurthermore, the resulting dependencies are used to construct a counterfactual\ncase without the outbreak of the war, providing a quantifiable estimate of the\noverall effect of the shock on various German energy sectors. The code and\ndataset are available at https://github.com/bonaldli/CausalEnergy.\n", "date": "2024-07-15", "categories": []}
{"id": "2307.12625", "title": "De-confounding Representation Learning for Counterfactual Inference on\n  Continuous Treatment via Generative Adversarial Network", "abstract": "  Counterfactual inference for continuous rather than binary treatment\nvariables is more common in real-world causal inference tasks. While there are\nalready some sample reweighting methods based on Marginal Structural Model for\neliminating the confounding bias, they generally focus on removing the\ntreatment's linear dependence on confounders and rely on the accuracy of the\nassumed parametric models, which are usually unverifiable. In this paper, we\npropose a de-confounding representation learning (DRL) framework for\ncounterfactual outcome estimation of continuous treatment by generating the\nrepresentations of covariates disentangled with the treatment variables. The\nDRL is a non-parametric model that eliminates both linear and nonlinear\ndependence between treatment and covariates. Specifically, we train the\ncorrelations between the de-confounded representations and the treatment\nvariables against the correlations between the covariate representations and\nthe treatment variables to eliminate confounding bias. Further, a\ncounterfactual inference network is embedded into the framework to make the\nlearned representations serve both de-confounding and trusted inference.\nExtensive experiments on synthetic datasets show that the DRL model performs\nsuperiorly in learning de-confounding representations and outperforms\nstate-of-the-art counterfactual inference models for continuous treatment\nvariables. In addition, we apply the DRL model to a real-world medical dataset\nMIMIC and demonstrate a detailed causal relationship between red cell width\ndistribution and mortality.\n", "date": "2023-07-24", "categories": []}
{"id": "2403.18717", "title": "Semi-Supervised Learning for Deep Causal Generative Models", "abstract": "  Developing models that are capable of answering questions of the form \"How\nwould x change if y had been z?'\" is fundamental to advancing medical image\nanalysis. Training causal generative models that address such counterfactual\nquestions, though, currently requires that all relevant variables have been\nobserved and that the corresponding labels are available in the training data.\nHowever, clinical data may not have complete records for all patients and state\nof the art causal generative models are unable to take full advantage of this.\nWe thus develop, for the first time, a semi-supervised deep causal generative\nmodel that exploits the causal relationships between variables to maximise the\nuse of all available data. We explore this in the setting where each sample is\neither fully labelled or fully unlabelled, as well as the more clinically\nrealistic case of having different labels missing for each sample. We leverage\ntechniques from causal inference to infer missing values and subsequently\ngenerate realistic counterfactuals, even for samples with incomplete labels.\n", "date": "2024-03-27", "categories": []}
{"id": "2407.05869", "title": "PORCA: Root Cause Analysis with Partially Observed Data", "abstract": "  Root Cause Analysis (RCA) aims at identifying the underlying causes of system\nfaults by uncovering and analyzing the causal structure from complex systems.\nIt has been widely used in many application domains. Reliable diagnostic\nconclusions are of great importance in mitigating system failures and financial\nlosses. However, previous studies implicitly assume a full observation of the\nsystem, which neglect the effect of partial observation (i.e., missing nodes\nand latent malfunction). As a result, they fail in deriving reliable RCA\nresults. In this paper, we unveil the issues of unobserved confounders and\nheterogeneity in partial observation and come up with a new problem of root\ncause analysis with partially observed data. To achieve this, we propose PORCA,\na novel RCA framework which can explore reliable root causes under both\nunobserved confounders and unobserved heterogeneity. PORCA leverages magnified\nscore-based causal discovery to efficiently optimize acyclic directed mixed\ngraph under unobserved confounders. In addition, we also develop a\nheterogeneity-aware scheduling strategy to provide adaptive sample weights.\nExtensive experimental results on one synthetic and two real-world datasets\ndemonstrate the effectiveness and superiority of the proposed framework.\n", "date": "2024-07-08", "categories": []}
{"id": "2407.07933", "title": "Identification and Estimation of the Bi-Directional MR with Some Invalid\n  Instruments", "abstract": "  We consider the challenging problem of estimating causal effects from purely\nobservational data in the bi-directional Mendelian randomization (MR), where\nsome invalid instruments, as well as unmeasured confounding, usually exist. To\naddress this problem, most existing methods attempt to find proper valid\ninstrumental variables (IVs) for the target causal effect by expert knowledge\nor by assuming that the causal model is a one-directional MR model. As such, in\nthis paper, we first theoretically investigate the identification of the\nbi-directional MR from observational data. In particular, we provide necessary\nand sufficient conditions under which valid IV sets are correctly identified\nsuch that the bi-directional MR model is identifiable, including the causal\ndirections of a pair of phenotypes (i.e., the treatment and outcome). Moreover,\nbased on the identification theory, we develop a cluster fusion-like method to\ndiscover valid IV sets and estimate the causal effects of interest. We\ntheoretically demonstrate the correctness of the proposed algorithm.\nExperimental results show the effectiveness of our method for estimating causal\neffects in bi-directional MR.\n", "date": "2024-07-10", "categories": []}
{"id": "2402.08616", "title": "Adjustment Identification Distance: A gadjid for Causal Structure\n  Learning", "abstract": "  Evaluating graphs learned by causal discovery algorithms is difficult: The\nnumber of edges that differ between two graphs does not reflect how the graphs\ndiffer with respect to the identifying formulas they suggest for causal\neffects. We introduce a framework for developing causal distances between\ngraphs which includes the structural intervention distance for directed acyclic\ngraphs as a special case. We use this framework to develop improved\nadjustment-based distances as well as extensions to completed partially\ndirected acyclic graphs and causal orders. We develop new reachability\nalgorithms to compute the distances efficiently and to prove their low\npolynomial time complexity. In our package gadjid (open source at\nhttps://github.com/CausalDisco/gadjid), we provide implementations of our\ndistances; they are orders of magnitude faster with proven lower time\ncomplexity than the structural intervention distance and thereby provide a\nsuccess metric for causal discovery that scales to graph sizes that were\npreviously prohibitive.\n", "date": "2024-02-13", "categories": []}
{"id": "2403.04221", "title": "Why Online Reinforcement Learning is Causal", "abstract": "  Reinforcement learning (RL) and causal modelling naturally complement each\nother. The goal of causal modelling is to predict the effects of interventions\nin an environment, while the goal of reinforcement learning is to select\ninterventions that maximize the rewards the agent receives from the\nenvironment. Reinforcement learning includes the two most powerful sources of\ninformation for estimating causal relationships: temporal ordering and the\nability to act on an environment. This paper examines which reinforcement\nlearning settings we can expect to benefit from causal modelling, and how. In\nonline learning, the agent has the ability to interact directly with their\nenvironment, and learn from exploring it. Our main argument is that in online\nlearning, conditional probabilities are causal, and therefore offline RL is the\nsetting where causal learning has the most potential to make a difference.\nEssentially, the reason is that when an agent learns from their {\\em own}\nexperience, there are no unobserved confounders that influence both the agent's\nown exploratory actions and the rewards they receive. Our paper formalizes this\nargument. For offline RL, where an agent may and typically does learn from the\nexperience of {\\em others}, we describe previous and new methods for leveraging\na causal model, including support for counterfactual queries.\n", "date": "2024-03-07", "categories": []}
{"id": "2407.08560", "title": "Causal inference through multi-stage learning and doubly robust deep\n  neural networks", "abstract": "  Deep neural networks (DNNs) have demonstrated remarkable empirical\nperformance in large-scale supervised learning problems, particularly in\nscenarios where both the sample size $n$ and the dimension of covariates $p$\nare large. This study delves into the application of DNNs across a wide\nspectrum of intricate causal inference tasks, where direct estimation falls\nshort and necessitates multi-stage learning. Examples include estimating the\nconditional average treatment effect and dynamic treatment effect. In this\nframework, DNNs are constructed sequentially, with subsequent stages building\nupon preceding ones. To mitigate the impact of estimation errors from early\nstages on subsequent ones, we integrate DNNs in a doubly robust manner. In\ncontrast to previous research, our study offers theoretical assurances\nregarding the effectiveness of DNNs in settings where the dimensionality $p$\nexpands with the sample size. These findings are significant independently and\nextend to degenerate single-stage learning problems.\n", "date": "2024-07-11", "categories": []}
{"id": "2407.07291", "title": "Causal Discovery in Semi-Stationary Time Series", "abstract": "  Discovering causal relations from observational time series without making\nthe stationary assumption is a significant challenge. In practice, this\nchallenge is common in many areas, such as retail sales, transportation\nsystems, and medical science. Here, we consider this problem for a class of\nnon-stationary time series. The structural causal model (SCM) of this type of\ntime series, called the semi-stationary time series, exhibits that a finite\nnumber of different causal mechanisms occur sequentially and periodically\nacross time. This model holds considerable practical utility because it can\nrepresent periodicity, including common occurrences such as seasonality and\ndiurnal variation. We propose a constraint-based, non-parametric algorithm for\ndiscovering causal relations in this setting. The resulting algorithm,\nPCMCI$_{\\Omega}$, can capture the alternating and recurring changes in the\ncausal mechanisms and then identify the underlying causal graph with\nconditional independence (CI) tests. We show that this algorithm is sound in\nidentifying causal relations on discrete time series. We validate the algorithm\nwith extensive experiments on continuous and discrete simulated data. We also\napply our algorithm to a real-world climate dataset.\n", "date": "2024-07-10", "categories": []}
{"id": "2304.05493", "title": "Optimizing Data-driven Causal Discovery Using Knowledge-guided Search", "abstract": "  Learning causal relationships solely from observational data often fails to\nreveal the underlying causal mechanisms due to the vast search space of\npossible causal graphs, which can grow exponentially, especially for greedy\nalgorithms using score-based approaches. Leveraging prior causal information,\nsuch as the presence or absence of causal edges, can help restrict and guide\nthe score-based discovery process, leading to a more accurate search. In the\nhealthcare domain, prior knowledge is abundant from sources like medical\njournals, electronic health records (EHRs), and clinical intervention outcomes.\nThis study introduces a knowledge-guided causal structure search (KGS) approach\nthat utilizes observational data and structural priors (such as causal edges)\nas constraints to learn the causal graph. KGS leverages prior edge information\nbetween variables, including the presence of a directed edge, the absence of an\nedge, and the presence of an undirected edge. We extensively evaluate KGS in\nmultiple settings using synthetic and benchmark real-world datasets, as well as\nin a real-life healthcare application related to oxygen therapy treatment. To\nobtain causal priors, we use GPT-4 to retrieve relevant literature information.\nOur results show that structural priors of any type and amount enhance the\nsearch process, improving performance and optimizing causal discovery. This\nguided strategy ensures that the discovered edges align with established causal\nknowledge, enhancing the trustworthiness of findings while expediting the\nsearch process. It also enables a more focused exploration of causal\nmechanisms, potentially leading to more effective and personalized healthcare\nsolutions.\n", "date": "2023-04-11", "categories": []}
{"id": "2406.00853", "title": "A Tutorial on Doubly Robust Learning for Causal Inference", "abstract": "  Doubly robust learning offers a robust framework for causal inference from\nobservational data by integrating propensity score and outcome modeling.\nDespite its theoretical appeal, practical adoption remains limited due to\nperceived complexity and inaccessible software. This tutorial aims to demystify\ndoubly robust methods and demonstrate their application using the EconML\npackage. We provide an introduction to causal inference, discuss the principles\nof outcome modeling and propensity scores, and illustrate the doubly robust\napproach through simulated case studies. By simplifying the methodology and\noffering practical coding examples, we intend to make doubly robust learning\naccessible to researchers and practitioners in data science and statistics.\n", "date": "2024-06-02", "categories": []}
{"id": "2407.05330", "title": "Fast Proxy Experiment Design for Causal Effect Identification", "abstract": "  Identifying causal effects is a key problem of interest across many\ndisciplines. The two long-standing approaches to estimate causal effects are\nobservational and experimental (randomized) studies. Observational studies can\nsuffer from unmeasured confounding, which may render the causal effects\nunidentifiable. On the other hand, direct experiments on the target variable\nmay be too costly or even infeasible to conduct. A middle ground between these\ntwo approaches is to estimate the causal effect of interest through proxy\nexperiments, which are conducted on variables with a lower cost to intervene on\ncompared to the main target. Akbari et al. [2022] studied this setting and\ndemonstrated that the problem of designing the optimal (minimum-cost)\nexperiment for causal effect identification is NP-complete and provided a naive\nalgorithm that may require solving exponentially many NP-hard problems as a\nsub-routine in the worst case. In this work, we provide a few reformulations of\nthe problem that allow for designing significantly more efficient algorithms to\nsolve it as witnessed by our extensive simulations. Additionally, we study the\nclosely-related problem of designing experiments that enable us to identify a\ngiven effect through valid adjustments sets.\n", "date": "2024-07-07", "categories": []}
{"id": "2407.05732", "title": "FairPFN: Transformers Can do Counterfactual Fairness", "abstract": "  Machine Learning systems are increasingly prevalent across healthcare, law\nenforcement, and finance but often operate on historical data, which may carry\nbiases against certain demographic groups. Causal and counterfactual fairness\nprovides an intuitive way to define fairness that closely aligns with legal\nstandards. Despite its theoretical benefits, counterfactual fairness comes with\nseveral practical limitations, largely related to the reliance on domain\nknowledge and approximate causal discovery techniques in constructing a causal\nmodel. In this study, we take a fresh perspective on counterfactually fair\nprediction, building upon recent work in in context learning (ICL) and prior\nfitted networks (PFNs) to learn a transformer called FairPFN. This model is\npretrained using synthetic fairness data to eliminate the causal effects of\nprotected attributes directly from observational data, removing the requirement\nof access to the correct causal model in practice. In our experiments, we\nthoroughly assess the effectiveness of FairPFN in eliminating the causal impact\nof protected attributes on a series of synthetic case studies and real world\ndatasets. Our findings pave the way for a new and promising research area:\ntransformers for causal and counterfactual fairness.\n", "date": "2024-07-08", "categories": []}
{"id": "2407.06015", "title": "Simulation-based Benchmarking for Causal Structure Learning in Gene\n  Perturbation Experiments", "abstract": "  Causal structure learning (CSL) refers to the task of learning causal\nrelationships from data. Advances in CSL now allow learning of causal graphs in\ndiverse application domains, which has the potential to facilitate data-driven\ncausal decision-making. Real-world CSL performance depends on a number of\n$\\textit{context-specific}$ factors, including context-specific data\ndistributions and non-linear dependencies, that are important in practical\nuse-cases. However, our understanding of how to assess and select CSL methods\nin specific contexts remains limited. To address this gap, we present\n$\\textit{CausalRegNet}$, a multiplicative effect structural causal model that\nallows for generating observational and interventional data incorporating\ncontext-specific properties, with a focus on the setting of gene perturbation\nexperiments. Using real-world gene perturbation data, we show that CausalRegNet\ngenerates accurate distributions and scales far better than current simulation\nframeworks. We illustrate the use of CausalRegNet in assessing CSL methods in\nthe context of interventional experiments in biology.\n", "date": "2024-07-08", "categories": []}
{"id": "2308.01375", "title": "CausalOps -- Towards an Industrial Lifecycle for Causal Probabilistic\n  Graphical Models", "abstract": "  Causal probabilistic graph-based models have gained widespread utility,\nenabling the modeling of cause-and-effect relationships across diverse domains.\nWith their rising adoption in new areas, such as automotive system safety and\nmachine learning, the need for an integrated lifecycle framework akin to DevOps\nand MLOps has emerged. Currently, a process reference for organizations\ninterested in employing causal engineering is missing. To address this gap and\nfoster widespread industrial adoption, we propose CausalOps, a novel lifecycle\nframework for causal model development and application. By defining key\nentities, dependencies, and intermediate artifacts generated during causal\nengineering, we establish a consistent vocabulary and workflow model. This work\ncontextualizes causal model usage across different stages and stakeholders,\noutlining a holistic view of creating and maintaining them. CausalOps' aim is\nto drive the adoption of causal methods in practical applications within\ninterested organizations and the causality community.\n", "date": "2023-08-02", "categories": []}
{"id": "2407.04605", "title": "Linear causal disentanglement via higher-order cumulants", "abstract": "  Linear causal disentanglement is a recent method in causal representation\nlearning to describe a collection of observed variables via latent variables\nwith causal dependencies between them. It can be viewed as a generalization of\nboth independent component analysis and linear structural equation models. We\nstudy the identifiability of linear causal disentanglement, assuming access to\ndata under multiple contexts, each given by an intervention on a latent\nvariable. We show that one perfect intervention on each latent variable is\nsufficient and in the worst case necessary to recover parameters under perfect\ninterventions, generalizing previous work to allow more latent than observed\nvariables. We give a constructive proof that computes parameters via a coupled\ntensor decomposition. For soft interventions, we find the equivalence class of\nlatent graphs and parameters that are consistent with observed data, via the\nstudy of a system of polynomial equations. Our results hold assuming the\nexistence of non-zero higher-order cumulants, which implies non-Gaussianity of\nvariables.\n", "date": "2024-07-05", "categories": []}
{"id": "2407.02825", "title": "Representation learning with CGAN for casual inference", "abstract": "  Conditional Generative Adversarial Nets (CGAN) is often used to improve\nconditional image generation performance. However, there is little research on\nRepresentation learning with CGAN for causal inference. This paper proposes a\nnew method for finding representation learning functions by adopting the\nadversarial idea. We apply the pattern of CGAN and theoretically emonstrate the\nfeasibility of finding a suitable representation function in the context of two\ndistributions being balanced. The theoretical result shows that when two\ndistributions are balanced, the ideal representation function can be found and\nthus can be used to further research.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.03082", "title": "Stable Heterogeneous Treatment Effect Estimation across\n  Out-of-Distribution Populations", "abstract": "  Heterogeneous treatment effect (HTE) estimation is vital for understanding\nthe change of treatment effect across individuals or subgroups. Most existing\nHTE estimation methods focus on addressing selection bias induced by imbalanced\ndistributions of confounders between treated and control units, but ignore\ndistribution shifts across populations. Thereby, their applicability has been\nlimited to the in-distribution (ID) population, which shares a similar\ndistribution with the training dataset. In real-world applications, where\npopulation distributions are subject to continuous changes, there is an urgent\nneed for stable HTE estimation across out-of-distribution (OOD) populations,\nwhich, however, remains an open problem. As pioneers in resolving this problem,\nwe propose a novel Stable Balanced Representation Learning with\nHierarchical-Attention Paradigm (SBRL-HAP) framework, which consists of 1)\nBalancing Regularizer for eliminating selection bias, 2) Independence\nRegularizer for addressing the distribution shift issue, 3)\nHierarchical-Attention Paradigm for coordination between balance and\nindependence. In this way, SBRL-HAP regresses counterfactual outcomes using ID\ndata, while ensuring the resulting HTE estimation can be successfully\ngeneralized to out-of-distribution scenarios, thereby enhancing the model's\napplicability in real-world settings. Extensive experiments conducted on\nsynthetic and real-world datasets demonstrate the effectiveness of our SBRL-HAP\nin achieving stable HTE estimation across OOD populations, with an average 10%\nreduction in the error metric PEHE and 11% decrease in the ATE bias, compared\nto the SOTA methods.\n", "date": "2024-07-03", "categories": []}
{"id": "2306.15328", "title": "Simulating counterfactuals", "abstract": "  Counterfactual inference considers a hypothetical intervention in a parallel\nworld that shares some evidence with the factual world. If the evidence\nspecifies a conditional distribution on a manifold, counterfactuals may be\nanalytically intractable. We present an algorithm for simulating values from a\ncounterfactual distribution where conditions can be set on both discrete and\ncontinuous variables. We show that the proposed algorithm can be presented as a\nparticle filter leading to asymptotically valid inference. The algorithm is\napplied to fairness analysis in credit-scoring.\n", "date": "2023-06-27", "categories": []}
{"id": "2406.00549", "title": "Zero Inflation as a Missing Data Problem: a Proxy-based Approach", "abstract": "  A common type of zero-inflated data has certain true values incorrectly\nreplaced by zeros due to data recording conventions (rare outcomes assumed to\nbe absent) or details of data recording equipment (e.g. artificial zeros in\ngene expression data).\n  Existing methods for zero-inflated data either fit the observed data\nlikelihood via parametric mixture models that explicitly represent excess\nzeros, or aim to replace excess zeros by imputed values. If the goal of the\nanalysis relies on knowing true data realizations, a particular challenge with\nzero-inflated data is identifiability, since it is difficult to correctly\ndetermine which observed zeros are real and which are inflated.\n  This paper views zero-inflated data as a general type of missing data\nproblem, where the observability indicator for a potentially censored variable\nis itself unobserved whenever a zero is recorded. We show that, without\nadditional assumptions, target parameters involving a zero-inflated variable\nare not identified. However, if a proxy of the missingness indicator is\nobserved, a modification of the effect restoration approach of Kuroki and Pearl\nallows identification and estimation, given the proxy-indicator relationship is\nknown.\n  If this relationship is unknown, our approach yields a partial identification\nstrategy for sensitivity analysis. Specifically, we show that only certain\nproxy-indicator relationships are compatible with the observed data\ndistribution. We give an analytic bound for this relationship in cases with a\ncategorical outcome, which is sharp in certain models. For more complex cases,\nsharp numerical bounds may be computed using methods in Duarte et al.[2023].\n  We illustrate our method via simulation studies and a data application on\ncentral line-associated bloodstream infections (CLABSIs).\n", "date": "2024-06-01", "categories": []}
{"id": "2407.01686", "title": "Everything that can be learned about a causal structure with latent\n  variables by observational and interventional probing schemes", "abstract": "  What types of differences among causal structures with latent variables are\nimpossible to distinguish by statistical data obtained by probing each visible\nvariable? If the probing scheme is simply passive observation, then it is\nwell-known that many different causal structures can realize the same joint\nprobability distributions. Even for the simplest case of two visible variables,\nfor instance, one cannot distinguish between one variable being a causal parent\nof the other and the two variables sharing a latent common cause. However, it\nis possible to distinguish between these two causal structures if we have\nrecourse to more powerful probing schemes, such as the possibility of\nintervening on one of the variables and observing the other. Herein, we address\nthe question of which causal structures remain indistinguishable even given the\nmost informative types of probing schemes on the visible variables. We find\nthat two causal structures remain indistinguishable if and only if they are\nboth associated with the same mDAG structure (as defined by Evans (2016)). We\nalso consider the question of when one causal structure dominates another in\nthe sense that it can realize all of the joint probability distributions that\ncan be realized by the other using a given probing scheme. (Equivalence of\ncausal structures is the special case of mutual dominance.) Finally, we\ninvestigate to what extent one can weaken the probing schemes implemented on\nthe visible variables and still have the same discrimination power as a\nmaximally informative probing scheme.\n", "date": "2024-07-01", "categories": []}
{"id": "2308.09790", "title": "A Two-Part Machine Learning Approach to Characterizing Network\n  Interference in A/B Testing", "abstract": "  The reliability of controlled experiments, commonly referred to as \"A/B\ntests,\" is often compromised by network interference, where the outcomes of\nindividual units are influenced by interactions with others. Significant\nchallenges in this domain include the lack of accounting for complex social\nnetwork structures and the difficulty in suitably characterizing network\ninterference. To address these challenges, we propose a machine learning-based\nmethod. We introduce \"causal network motifs\" and utilize transparent machine\nlearning models to characterize network interference patterns underlying an A/B\ntest on networks. Our method's performance has been demonstrated through\nsimulations on both a synthetic experiment and a large-scale test on Instagram.\nOur experiments show that our approach outperforms conventional methods such as\ndesign-based cluster randomization and conventional analysis-based neighborhood\nexposure mapping. Our approach provides a comprehensive and automated solution\nto address network interference for A/B testing practitioners. This aids in\ninforming strategic business decisions in areas such as marketing effectiveness\nand product customization.\n", "date": "2023-08-18", "categories": []}
{"id": "2407.00529", "title": "Detecting and Identifying Selection Structure in Sequential Data", "abstract": "  We argue that the selective inclusion of data points based on latent\nobjectives is common in practical situations, such as music sequences. Since\nthis selection process often distorts statistical analysis, previous work\nprimarily views it as a bias to be corrected and proposes various methods to\nmitigate its effect. However, while controlling this bias is crucial, selection\nalso offers an opportunity to provide a deeper insight into the hidden\ngeneration process, as it is a fundamental mechanism underlying what we\nobserve. In particular, overlooking selection in sequential data can lead to an\nincomplete or overcomplicated inductive bias in modeling, such as assuming a\nuniversal autoregressive structure for all dependencies. Therefore, rather than\nmerely viewing it as a bias, we explore the causal structure of selection in\nsequential data to delve deeper into the complete causal process. Specifically,\nwe show that selection structure is identifiable without any parametric\nassumptions or interventional experiments. Moreover, even in cases where\nselection variables coexist with latent confounders, we still establish the\nnonparametric identifiability under appropriate structural conditions.\nMeanwhile, we also propose a provably correct algorithm to detect and identify\nselection structures as well as other types of dependencies. The framework has\nbeen validated empirically on both synthetic data and real-world music.\n", "date": "2024-06-29", "categories": []}
{"id": "2407.00744", "title": "Disentangled Representations for Causal Cognition", "abstract": "  Complex adaptive agents consistently achieve their goals by solving problems\nthat seem to require an understanding of causal information, information\npertaining to the causal relationships that exist among elements of combined\nagent-environment systems. Causal cognition studies and describes the main\ncharacteristics of causal learning and reasoning in human and non-human\nanimals, offering a conceptual framework to discuss cognitive performances\nbased on the level of apparent causal understanding of a task. Despite the use\nof formal intervention-based models of causality, including causal Bayesian\nnetworks, psychological and behavioural research on causal cognition does not\nyet offer a computational account that operationalises how agents acquire a\ncausal understanding of the world. Machine and reinforcement learning research\non causality, especially involving disentanglement as a candidate process to\nbuild causal representations, represent on the one hand a concrete attempt at\ndesigning causal artificial agents that can shed light on the inner workings of\nnatural causal cognition. In this work, we connect these two areas of research\nto build a unifying framework for causal cognition that will offer a\ncomputational perspective on studies of animal cognition, and provide insights\nin the development of new algorithms for causal reinforcement learning in AI.\n", "date": "2024-06-30", "categories": []}
{"id": "2407.00950", "title": "Causal Bandits: The Pareto Optimal Frontier of Adaptivity, a Reduction\n  to Linear Bandits, and Limitations around Unknown Marginals", "abstract": "  In this work, we investigate the problem of adapting to the presence or\nabsence of causal structure in multi-armed bandit problems. In addition to the\nusual reward signal, we assume the learner has access to additional variables,\nobserved in each round after acting. When these variables $d$-separate the\naction from the reward, existing work in causal bandits demonstrates that one\ncan achieve strictly better (minimax) rates of regret (Lu et al., 2020). Our\ngoal is to adapt to this favorable \"conditionally benign\" structure, if it is\npresent in the environment, while simultaneously recovering worst-case minimax\nregret, if it is not. Notably, the learner has no prior knowledge of whether\nthe favorable structure holds. In this paper, we establish the Pareto optimal\nfrontier of adaptive rates. We prove upper and matching lower bounds on the\npossible trade-offs in the performance of learning in conditionally benign and\narbitrary environments, resolving an open question raised by Bilodeau et al.\n(2022). Furthermore, we are the first to obtain instance-dependent bounds for\ncausal bandits, by reducing the problem to the linear bandit setting. Finally,\nwe examine the common assumption that the marginal distributions of the\npost-action contexts are known and show that a nontrivial estimate is necessary\nfor better-than-worst-case minimax rates.\n", "date": "2024-07-01", "categories": []}
{"id": "2407.01004", "title": "CURLS: Causal Rule Learning for Subgroups with Significant Treatment\n  Effect", "abstract": "  In causal inference, estimating heterogeneous treatment effects (HTE) is\ncritical for identifying how different subgroups respond to interventions, with\nbroad applications in fields such as precision medicine and personalized\nadvertising. Although HTE estimation methods aim to improve accuracy, how to\nprovide explicit subgroup descriptions remains unclear, hindering data\ninterpretation and strategic intervention management. In this paper, we propose\nCURLS, a novel rule learning method leveraging HTE, which can effectively\ndescribe subgroups with significant treatment effects. Specifically, we frame\ncausal rule learning as a discrete optimization problem, finely balancing\ntreatment effect with variance and considering the rule interpretability. We\ndesign an iterative procedure based on the minorize-maximization algorithm and\nsolve a submodular lower bound as an approximation for the original.\nQuantitative experiments and qualitative case studies verify that compared with\nstate-of-the-art methods, CURLS can find subgroups where the estimated and true\neffects are 16.1% and 13.8% higher and the variance is 12.0% smaller, while\nmaintaining similar or better estimation accuracy and rule interpretability.\nCode is available at https://osf.io/zwp2k/.\n", "date": "2024-07-01", "categories": []}
{"id": "2407.01154", "title": "Wind Estimation in Unmanned Aerial Vehicles with Causal Machine Learning", "abstract": "  In this work we demonstrate the possibility of estimating the wind\nenvironment of a UAV without specialised sensors, using only the UAV's\ntrajectory, applying a causal machine learning approach. We implement the\ncausal curiosity method which combines machine learning times series\nclassification and clustering with a causal framework. We analyse three\ndistinct wind environments: constant wind, shear wind, and turbulence, and\nexplore different optimisation strategies for optimal UAV manoeuvres to\nestimate the wind conditions. The proposed approach can be used to design\noptimal trajectories in challenging weather conditions, and to avoid\nspecialised sensors that add to the UAV's weight and compromise its\nfunctionality.\n", "date": "2024-07-01", "categories": []}
{"id": "2311.10263", "title": "Stable Differentiable Causal Discovery", "abstract": "  Inferring causal relationships as directed acyclic graphs (DAGs) is an\nimportant but challenging problem. Differentiable Causal Discovery (DCD) is a\npromising approach to this problem, framing the search as a continuous\noptimization. But existing DCD methods are numerically unstable, with poor\nperformance beyond tens of variables. In this paper, we propose Stable\nDifferentiable Causal Discovery (SDCD), a new method that improves previous DCD\nmethods in two ways: (1) It employs an alternative constraint for acyclicity;\nthis constraint is more stable, both theoretically and empirically, and fast to\ncompute. (2) It uses a training procedure tailored for sparse causal graphs,\nwhich are common in real-world scenarios. We first derive SDCD and prove its\nstability and correctness. We then evaluate it with both observational and\ninterventional data and on both small-scale and large-scale settings. We find\nthat SDCD outperforms existing methods in both convergence speed and accuracy\nand can scale to thousands of variables. We provide code at\nhttps://github.com/azizilab/sdcd.\n", "date": "2023-11-17", "categories": []}
{"id": "2401.03183", "title": "Exploring Defeasibility in Causal Reasoning", "abstract": "  Defeasibility in causal reasoning implies that the causal relationship\nbetween cause and effect can be strengthened or weakened. Namely, the causal\nstrength between cause and effect should increase or decrease with the\nincorporation of strengthening arguments (supporters) or weakening arguments\n(defeaters), respectively. However, existing works ignore defeasibility in\ncausal reasoning and fail to evaluate existing causal strength metrics in\ndefeasible settings. In this work, we present $\\delta$-CAUSAL, the first\nbenchmark dataset for studying defeasibility in causal reasoning.\n$\\delta$-CAUSAL includes around 11K events spanning ten domains, featuring\ndefeasible causality pairs, i.e., cause-effect pairs accompanied by supporters\nand defeaters. We further show current causal strength metrics fail to reflect\nthe change of causal strength with the incorporation of supporters or defeaters\nin $\\delta$-CAUSAL. To this end, we propose CESAR (Causal Embedding aSsociation\nwith Attention Rating), a metric that measures causal strength based on\ntoken-level causal relationships. CESAR achieves a significant 69.7% relative\nimprovement over existing metrics, increasing from 47.2% to 80.1% in capturing\nthe causal strength change brought by supporters and defeaters. We further\ndemonstrate even Large Language Models (LLMs) like GPT-3.5 still lag 4.5 and\n10.7 points behind humans in generating supporters and defeaters, emphasizing\nthe challenge posed by $\\delta$-CAUSAL.\n", "date": "2024-01-06", "categories": []}
{"id": "2406.17228", "title": "Greedy equivalence search for nonparametric graphical models", "abstract": "  One of the hallmark achievements of the theory of graphical models and\nBayesian model selection is the celebrated greedy equivalence search (GES)\nalgorithm due to Chickering and Meek. GES is known to consistently estimate the\nstructure of directed acyclic graph (DAG) models in various special cases\nincluding Gaussian and discrete models, which are in particular curved\nexponential families. A general theory that covers general nonparametric DAG\nmodels, however, is missing. Here, we establish the consistency of greedy\nequivalence search for general families of DAG models that satisfy smoothness\nconditions on the Markov factorization, and hence may not be curved exponential\nfamilies, or even parametric. The proof leverages recent advances in\nnonparametric Bayes to construct a test for comparing misspecified DAG models\nthat avoids arguments based on the Laplace approximation. Nonetheless, when the\nLaplace approximation is valid and a consistent scoring function exists, we\nrecover the classical result. As a result, we obtain a general consistency\ntheorem for GES applied to general DAG models.\n", "date": "2024-06-25", "categories": []}
{"id": "2406.17698", "title": "Identifying Nonstationary Causal Structures with High-Order Markov\n  Switching Models", "abstract": "  Causal discovery in time series is a rapidly evolving field with a wide\nvariety of applications in other areas such as climate science and\nneuroscience. Traditional approaches assume a stationary causal graph, which\ncan be adapted to nonstationary time series with time-dependent effects or\nheterogeneous noise. In this work we address nonstationarity via\nregime-dependent causal structures. We first establish identifiability for\nhigh-order Markov Switching Models, which provide the foundations for\nidentifiable regime-dependent causal discovery. Our empirical studies\ndemonstrate the scalability of our proposed approach for high-order\nregime-dependent structure estimation, and we illustrate its applicability on\nbrain activity data.\n", "date": "2024-06-25", "categories": []}
{"id": "2307.12797", "title": "Causal Fair Machine Learning via Rank-Preserving Interventional\n  Distributions", "abstract": "  A decision can be defined as fair if equal individuals are treated equally\nand unequals unequally. Adopting this definition, the task of designing machine\nlearning (ML) models that mitigate unfairness in automated decision-making\nsystems must include causal thinking when introducing protected attributes:\nFollowing a recent proposal, we define individuals as being normatively equal\nif they are equal in a fictitious, normatively desired (FiND) world, where the\nprotected attributes have no (direct or indirect) causal effect on the target.\nWe propose rank-preserving interventional distributions to define a specific\nFiND world in which this holds and a warping method for estimation. Evaluation\ncriteria for both the method and the resulting ML model are presented and\nvalidated through simulations. Experiments on empirical data showcase the\npractical application of our method and compare results with \"fairadapt\"\n(Ple\\v{c}ko and Meinshausen, 2020), a different approach for mitigating\nunfairness by causally preprocessing data that uses quantile regression\nforests. With this, we show that our warping approach effectively identifies\nthe most discriminated individuals and mitigates unfairness.\n", "date": "2023-07-24", "categories": []}
{"id": "2310.06312", "title": "Discovering Mixtures of Structural Causal Models from Time Series Data", "abstract": "  Discovering causal relationships from time series data is significant in\nfields such as finance, climate science, and neuroscience. However,\ncontemporary techniques rely on the simplifying assumption that data originates\nfrom the same causal model, while in practice, data is heterogeneous and can\nstem from different causal models. In this work, we relax this assumption and\nperform causal discovery from time series data originating from a mixture of\ncausal models. We propose a general variational inference-based framework\ncalled MCD to infer the underlying causal models as well as the mixing\nprobability of each sample. Our approach employs an end-to-end training process\nthat maximizes an evidence-lower bound for the data likelihood. We present two\nvariants: MCD-Linear for linear relationships and independent noise, and\nMCD-Nonlinear for nonlinear causal relationships and history-dependent noise.\nWe demonstrate that our method surpasses state-of-the-art benchmarks in causal\ndiscovery tasks through extensive experimentation on synthetic and real-world\ndatasets, particularly when the data emanates from diverse underlying causal\ngraphs. Theoretically, we prove the identifiability of such a model under some\nmild assumptions.\n", "date": "2023-10-10", "categories": []}
{"id": "2405.11548", "title": "Adaptive Online Experimental Design for Causal Discovery", "abstract": "  Causal discovery aims to uncover cause-and-effect relationships encoded in\ncausal graphs by leveraging observational, interventional data, or their\ncombination. The majority of existing causal discovery methods are developed\nassuming infinite interventional data. We focus on data interventional\nefficiency and formalize causal discovery from the perspective of online\nlearning, inspired by pure exploration in bandit problems. A graph separating\nsystem, consisting of interventions that cut every edge of the graph at least\nonce, is sufficient for learning causal graphs when infinite interventional\ndata is available, even in the worst case. We propose a track-and-stop causal\ndiscovery algorithm that adaptively selects interventions from the graph\nseparating system via allocation matching and learns the causal graph based on\nsampling history. Given any desired confidence value, the algorithm determines\na termination condition and runs until it is met. We analyze the algorithm to\nestablish a problem-dependent upper bound on the expected number of required\ninterventional samples. Our proposed algorithm outperforms existing methods in\nsimulations across various randomly generated causal graphs. It achieves higher\naccuracy, measured by the structural hamming distance (SHD) between the learned\ncausal graph and the ground truth, with significantly fewer samples.\n", "date": "2024-05-19", "categories": []}
{"id": "2406.16605", "title": "CLEAR: Can Language Models Really Understand Causal Graphs?", "abstract": "  Causal reasoning is a cornerstone of how humans interpret the world. To model\nand reason about causality, causal graphs offer a concise yet effective\nsolution. Given the impressive advancements in language models, a crucial\nquestion arises: can they really understand causal graphs? To this end, we\npioneer an investigation into language models' understanding of causal graphs.\nSpecifically, we develop a framework to define causal graph understanding, by\nassessing language models' behaviors through four practical criteria derived\nfrom diverse disciplines (e.g., philosophy and psychology). We then develop\nCLEAR, a novel benchmark that defines three complexity levels and encompasses\n20 causal graph-based tasks across these levels. Finally, based on our\nframework and benchmark, we conduct extensive experiments on six leading\nlanguage models and summarize five empirical findings. Our results indicate\nthat while language models demonstrate a preliminary understanding of causal\ngraphs, significant potential for improvement remains. Our project website is\nat https://github.com/OpenCausaLab/CLEAR.\n", "date": "2024-06-24", "categories": []}
{"id": "2406.16708", "title": "CausalFormer: An Interpretable Transformer for Temporal Causal Discovery", "abstract": "  Temporal causal discovery is a crucial task aimed at uncovering the causal\nrelations within time series data. The latest temporal causal discovery methods\nusually train deep learning models on prediction tasks to uncover the causality\nbetween time series. They capture causal relations by analyzing the parameters\nof some components of the trained models, e.g., attention weights and\nconvolution weights. However, this is an incomplete mapping process from the\nmodel parameters to the causality and fails to investigate the other\ncomponents, e.g., fully connected layers and activation functions, that are\nalso significant for causal discovery. To facilitate the utilization of the\nwhole deep learning models in temporal causal discovery, we proposed an\ninterpretable transformer-based causal discovery model termed CausalFormer,\nwhich consists of the causality-aware transformer and the decomposition-based\ncausality detector. The causality-aware transformer learns the causal\nrepresentation of time series data using a prediction task with the designed\nmulti-kernel causal convolution which aggregates each input time series along\nthe temporal dimension under the temporal priority constraint. Then, the\ndecomposition-based causality detector interprets the global structure of the\ntrained causality-aware transformer with the proposed regression relevance\npropagation to identify potential causal relations and finally construct the\ncausal graph. Experiments on synthetic, simulated, and real datasets\ndemonstrate the state-of-the-art performance of CausalFormer on discovering\ntemporal causality. Our code is available at\nhttps://github.com/lingbai-kong/CausalFormer.\n", "date": "2024-06-24", "categories": []}
{"id": "2406.16728", "title": "CausalMMM: Learning Causal Structure for Marketing Mix Modeling", "abstract": "  In online advertising, marketing mix modeling (MMM) is employed to predict\nthe gross merchandise volume (GMV) of brand shops and help decision-makers to\nadjust the budget allocation of various advertising channels. Traditional MMM\nmethods leveraging regression techniques can fail in handling the complexity of\nmarketing. Although some efforts try to encode the causal structures for better\nprediction, they have the strict restriction that causal structures are\nprior-known and unchangeable. In this paper, we define a new causal MMM problem\nthat automatically discovers the interpretable causal structures from data and\nyields better GMV predictions. To achieve causal MMM, two essential challenges\nshould be addressed: (1) Causal Heterogeneity. The causal structures of\ndifferent kinds of shops vary a lot. (2) Marketing Response Patterns. Various\nmarketing response patterns i.e., carryover effect and shape effect, have been\nvalidated in practice. We argue that causal MMM needs dynamically discover\nspecific causal structures for different shops and the predictions should\ncomply with the prior known marketing response patterns. Thus, we propose\nCausalMMM that integrates Granger causality in a variational inference\nframework to measure the causal relationships between different channels and\npredict the GMV with the regularization of both temporal and saturation\nmarketing response patterns. Extensive experiments show that CausalMMM can not\nonly achieve superior performance of causal structure learning on synthetic\ndatasets with improvements of 5.7%\\sim 7.1%, but also enhance the GMV\nprediction results on a representative E-commerce platform.\n", "date": "2024-06-24", "categories": []}
{"id": "2305.15936", "title": "Learning DAGs from Data with Few Root Causes", "abstract": "  We present a novel perspective and algorithm for learning directed acyclic\ngraphs (DAGs) from data generated by a linear structural equation model (SEM).\nFirst, we show that a linear SEM can be viewed as a linear transform that, in\nprior work, computes the data from a dense input vector of random valued root\ncauses (as we will call them) associated with the nodes. Instead, we consider\nthe case of (approximately) few root causes and also introduce noise in the\nmeasurement of the data. Intuitively, this means that the DAG data is produced\nby few data-generating events whose effect percolates through the DAG. We prove\nidentifiability in this new setting and show that the true DAG is the global\nminimizer of the $L^0$-norm of the vector of root causes. For data with few\nroot causes, with and without noise, we show superior performance compared to\nprior DAG learning methods.\n", "date": "2023-05-25", "categories": []}
{"id": "2403.03589", "title": "Active Adaptive Experimental Design for Treatment Effect Estimation with\n  Covariate Choices", "abstract": "  This study designs an adaptive experiment for efficiently estimating average\ntreatment effects (ATEs). In each round of our adaptive experiment, an\nexperimenter sequentially samples an experimental unit, assigns a treatment,\nand observes the corresponding outcome immediately. At the end of the\nexperiment, the experimenter estimates an ATE using the gathered samples. The\nobjective is to estimate the ATE with a smaller asymptotic variance. Existing\nstudies have designed experiments that adaptively optimize the propensity score\n(treatment-assignment probability). As a generalization of such an approach, we\npropose optimizing the covariate density as well as the propensity score.\nFirst, we derive the efficient covariate density and propensity score that\nminimize the semiparametric efficiency bound and find that optimizing both\ncovariate density and propensity score minimizes the semiparametric efficiency\nbound more effectively than optimizing only the propensity score. Next, we\ndesign an adaptive experiment using the efficient covariate density and\npropensity score sequentially estimated during the experiment. Lastly, we\npropose an ATE estimator whose asymptotic variance aligns with the minimized\nsemiparametric efficiency bound.\n", "date": "2024-03-06", "categories": []}
{"id": "2406.11308", "title": "Management Decisions in Manufacturing using Causal Machine Learning --\n  To Rework, or not to Rework?", "abstract": "  In this paper, we present a data-driven model for estimating optimal rework\npolicies in manufacturing systems. We consider a single production stage within\na multistage, lot-based system that allows for optional rework steps. While the\nrework decision depends on an intermediate state of the lot and system, the\nfinal product inspection, and thus the assessment of the actual yield, is\ndelayed until production is complete. Repair steps are applied uniformly to the\nlot, potentially improving some of the individual items while degrading others.\nThe challenge is thus to balance potential yield improvement with the rework\ncosts incurred. Given the inherently causal nature of this decision problem, we\npropose a causal model to estimate yield improvement. We apply methods from\ncausal machine learning, in particular double/debiased machine learning (DML)\ntechniques, to estimate conditional treatment effects from data and derive\npolicies for rework decisions. We validate our decision model using real-world\ndata from opto-electronic semiconductor manufacturing, achieving a yield\nimprovement of 2 - 3% during the color-conversion process of white\nlight-emitting diodes (LEDs).\n", "date": "2024-06-17", "categories": []}
{"id": "2406.13371", "title": "Identifiable Causal Representation Learning: Unsupervised, Multi-View,\n  and Multi-Environment", "abstract": "  Causal models provide rich descriptions of complex systems as sets of\nmechanisms by which each variable is influenced by its direct causes. They\nsupport reasoning about manipulating parts of the system and thus hold promise\nfor addressing some of the open challenges of artificial intelligence (AI),\nsuch as planning, transferring knowledge in changing environments, or\nrobustness to distribution shifts. However, a key obstacle to more widespread\nuse of causal models in AI is the requirement that the relevant variables be\nspecified a priori, which is typically not the case for the high-dimensional,\nunstructured data processed by modern AI systems. At the same time, machine\nlearning (ML) has proven quite successful at automatically extracting useful\nand compact representations of such complex data. Causal representation\nlearning (CRL) aims to combine the core strengths of ML and causality by\nlearning representations in the form of latent variables endowed with causal\nmodel semantics.\n  In this thesis, we study and present new results for different CRL settings.\nA central theme is the question of identifiability: Given infinite data, when\nare representations satisfying the same learning objective guaranteed to be\nequivalent? This is an important prerequisite for CRL, as it formally\ncharacterises if and when a learning task is, at least in principle, feasible.\nSince learning causal models, even without a representation learning component,\nis notoriously difficult, we require additional assumptions on the model class\nor rich data beyond the classical i.i.d. setting. By partially characterising\nidentifiability for different settings, this thesis investigates what is\npossible for CRL without direct supervision, and thus contributes to its\ntheoretical foundations. Ideally, the developed insights can help inform data\ncollection practices or inspire the design of new practical estimation methods.\n", "date": "2024-06-19", "categories": []}
{"id": "2406.13731", "title": "Integrating Fuzzy Logic with Causal Inference: Enhancing the Pearl and\n  Neyman-Rubin Methodologies", "abstract": "  In this paper, we generalize the Pearl and Neyman-Rubin methodologies in\ncausal inference by introducing a generalized approach that incorporates fuzzy\nlogic. Indeed, we introduce a fuzzy causal inference approach that consider\nboth the vagueness and imprecision inherent in data, as well as the subjective\nhuman perspective characterized by fuzzy terms such as 'high', 'medium', and\n'low'. To do so, we introduce two fuzzy causal effect formulas: the Fuzzy\nAverage Treatment Effect (FATE) and the Generalized Fuzzy Average Treatment\nEffect (GFATE), together with their normalized versions: NFATE and NGFATE. When\ndealing with a binary treatment variable, our fuzzy causal effect formulas\ncoincide with classical Average Treatment Effect (ATE) formula, that is a\nwell-established and popular metric in causal inference. In FATE, all values of\nthe treatment variable are considered equally important. In contrast, GFATE\ntakes into account the rarity and frequency of these values. We show that for\nlinear Structural Equation Models (SEMs), the normalized versions of our\nformulas, NFATE and NGFATE, are equivalent to ATE. Further, we provide\nidentifiability criteria for these formulas and show their stability with\nrespect to minor variations in the fuzzy subsets and the probability\ndistributions involved. This ensures the robustness of our approach in handling\nsmall perturbations in the data. Finally, we provide several experimental\nexamples to empirically validate and demonstrate the practical application of\nour proposed fuzzy causal inference methods.\n", "date": "2024-06-19", "categories": []}
{"id": "2406.13966", "title": "Causal Inference with Latent Variables: Recent Advances and Future\n  Prospectives", "abstract": "  Causality lays the foundation for the trajectory of our world. Causal\ninference (CI), which aims to infer intrinsic causal relations among variables\nof interest, has emerged as a crucial research topic. Nevertheless, the lack of\nobservation of important variables (e.g., confounders, mediators, exogenous\nvariables, etc.) severely compromises the reliability of CI methods. The issue\nmay arise from the inherent difficulty in measuring the variables.\nAdditionally, in observational studies where variables are passively recorded,\ncertain covariates might be inadvertently omitted by the experimenter.\nDepending on the type of unobserved variables and the specific CI task, various\nconsequences can be incurred if these latent variables are carelessly handled,\nsuch as biased estimation of causal effects, incomplete understanding of causal\nmechanisms, lack of individual-level causal consideration, etc. In this survey,\nwe provide a comprehensive review of recent developments in CI with latent\nvariables. We start by discussing traditional CI techniques when variables of\ninterest are assumed to be fully observed. Afterward, under the taxonomy of\ncircumvention and inference-based methods, we provide an in-depth discussion of\nvarious CI strategies to handle latent variables, covering the tasks of causal\neffect estimation, mediation analysis, counterfactual reasoning, and causal\ndiscovery. Furthermore, we generalize the discussion to graph data where\ninterference among units may exist. Finally, we offer fresh aspects for further\nadvancement of CI with latent variables, especially new opportunities in the\nera of large language models (LLMs).\n", "date": "2024-06-20", "categories": []}
{"id": "2402.15301", "title": "Causal Graph Discovery with Retrieval-Augmented Generation based Large\n  Language Models", "abstract": "  Causal graph recovery is traditionally done using statistical\nestimation-based methods or based on individual's knowledge about variables of\ninterests. They often suffer from data collection biases and limitations of\nindividuals' knowledge. The advance of large language models (LLMs) provides\nopportunities to address these problems. We propose a novel method that\nleverages LLMs to deduce causal relationships in general causal graph recovery\ntasks. This method leverages knowledge compressed in LLMs and knowledge LLMs\nextracted from scientific publication database as well as experiment data about\nfactors of interest to achieve this goal. Our method gives a prompting strategy\nto extract associational relationships among those factors and a mechanism to\nperform causality verification for these associations. Comparing to other\nLLM-based methods that directly instruct LLMs to do the highly complex causal\nreasoning, our method shows clear advantage on causal graph quality on\nbenchmark datasets. More importantly, as causality among some factors may\nchange as new research results emerge, our method show sensitivity to new\nevidence in the literature and can provide useful information for updating\ncausal graphs accordingly.\n", "date": "2024-02-23", "categories": []}
{"id": "2406.09177", "title": "Scalable and Flexible Causal Discovery with an Efficient Test for\n  Adjacency", "abstract": "  To make accurate predictions, understand mechanisms, and design interventions\nin systems of many variables, we wish to learn causal graphs from large scale\ndata. Unfortunately the space of all possible causal graphs is enormous so\nscalably and accurately searching for the best fit to the data is a challenge.\nIn principle we could substantially decrease the search space, or learn the\ngraph entirely, by testing the conditional independence of variables. However,\ndeciding if two variables are adjacent in a causal graph may require an\nexponential number of tests. Here we build a scalable and flexible method to\nevaluate if two variables are adjacent in a causal graph, the Differentiable\nAdjacency Test (DAT). DAT replaces an exponential number of tests with a\nprovably equivalent relaxed problem. It then solves this problem by training\ntwo neural networks. We build a graph learning method based on DAT, DAT-Graph,\nthat can also learn from data with interventions. DAT-Graph can learn graphs of\n1000 variables with state of the art accuracy. Using the graph learned by\nDAT-Graph, we also build models that make much more accurate predictions of the\neffects of interventions on large scale RNA sequencing data.\n", "date": "2024-06-13", "categories": []}
{"id": "2406.10807", "title": "Bayesian Networks and Machine Learning for COVID-19 Severity Explanation\n  and Demographic Symptom Classification", "abstract": "  With the prevailing efforts to combat the coronavirus disease 2019 (COVID-19)\npandemic, there are still uncertainties that are yet to be discovered about its\nspread, future impact, and resurgence. In this paper, we present a three-stage\ndata-driven approach to distill the hidden information about COVID-19. The\nfirst stage employs a Bayesian network structure learning method to identify\nthe causal relationships among COVID-19 symptoms and their intrinsic\ndemographic variables. As a second stage, the output from the Bayesian network\nstructure learning, serves as a useful guide to train an unsupervised machine\nlearning (ML) algorithm that uncovers the similarities in patients' symptoms\nthrough clustering. The final stage then leverages the labels obtained from\nclustering to train a demographic symptom identification (DSID) model which\npredicts a patient's symptom class and the corresponding demographic\nprobability distribution. We applied our method on the COVID-19 dataset\nobtained from the Centers for Disease Control and Prevention (CDC) in the\nUnited States. Results from the experiments show a testing accuracy of 99.99%,\nas against the 41.15% accuracy of a heuristic ML method. This strongly reveals\nthe viability of our Bayesian network and ML approach in understanding the\nrelationship between the virus symptoms, and providing insights on patients'\nstratification towards reducing the severity of the virus.\n", "date": "2024-06-16", "categories": []}
{"id": "2406.11501", "title": "Teleporter Theory: A General and Simple Approach for Modeling\n  Cross-World Counterfactual Causality", "abstract": "  Leveraging the development of structural causal model (SCM), researchers can\nestablish graphical models for exploring the causal mechanisms behind machine\nlearning techniques. As the complexity of machine learning applications rises,\nsingle-world interventionism causal analysis encounters theoretical adaptation\nlimitations. Accordingly, cross-world counterfactual approach extends our\nunderstanding of causality beyond observed data, enabling hypothetical\nreasoning about alternative scenarios. However, the joint involvement of\ncross-world variables, encompassing counterfactual variables and real-world\nvariables, challenges the construction of the graphical model. Twin network is\na subtle attempt, establishing a symbiotic relationship, to bridge the gap\nbetween graphical modeling and the introduction of counterfactuals albeit with\nroom for improvement in generalization. In this regard, we demonstrate the\ntheoretical breakdowns of twin networks in certain cross-world counterfactual\nscenarios. To this end, we propose a novel teleporter theory to establish a\ngeneral and simple graphical representation of counterfactuals, which provides\ncriteria for determining teleporter variables to connect multiple worlds. In\ntheoretical application, we determine that introducing the proposed teleporter\ntheory can directly obtain the conditional independence between counterfactual\nvariables and real-world variables from the cross-world SCM without requiring\ncomplex algebraic derivations. Accordingly, we can further identify\ncounterfactual causal effects through cross-world symbolic derivation. We\ndemonstrate the generality of the teleporter theory to the practical\napplication. Adhering to the proposed theory, we build a plug-and-play module,\nand the effectiveness of which are substantiated by experiments on benchmarks.\n", "date": "2024-06-17", "categories": []}
{"id": "2406.12158", "title": "LLMs Are Prone to Fallacies in Causal Inference", "abstract": "  Recent work shows that causal facts can be effectively extracted from LLMs\nthrough prompting, facilitating the creation of causal graphs for causal\ninference tasks. However, it is unclear if this success is limited to\nexplicitly-mentioned causal facts in the pretraining data which the model can\nmemorize. Thus, this work investigates: Can LLMs infer causal relations from\nother relational data in text? To disentangle the role of memorized causal\nfacts vs inferred causal relations, we finetune LLMs on synthetic data\ncontaining temporal, spatial and counterfactual relations, and measure whether\nthe LLM can then infer causal relations. We find that: (a) LLMs are susceptible\nto inferring causal relations from the order of two entity mentions in text\n(e.g. X mentioned before Y implies X causes Y); (b) if the order is randomized,\nLLMs still suffer from the post hoc fallacy, i.e. X occurs before Y (temporal\nrelation) implies X causes Y. We also find that while LLMs can correctly deduce\nthe absence of causal relations from temporal and spatial relations, they have\ndifficulty inferring causal relations from counterfactuals, questioning their\nunderstanding of causality.\n", "date": "2024-06-18", "categories": []}
{"id": "2401.16986", "title": "Causal Machine Learning for Cost-Effective Allocation of Development Aid", "abstract": "  The Sustainable Development Goals (SDGs) of the United Nations provide a\nblueprint of a better future by 'leaving no one behind', and, to achieve the\nSDGs by 2030, poor countries require immense volumes of development aid. In\nthis paper, we develop a causal machine learning framework for predicting\nheterogeneous treatment effects of aid disbursements to inform effective aid\nallocation. Specifically, our framework comprises three components: (i) a\nbalancing autoencoder that uses representation learning to embed\nhigh-dimensional country characteristics while addressing treatment selection\nbias; (ii) a counterfactual generator to compute counterfactual outcomes for\nvarying aid volumes to address small sample-size settings; and (iii) an\ninference model that is used to predict heterogeneous treatment-response\ncurves. We demonstrate the effectiveness of our framework using data with\nofficial development aid earmarked to end HIV/AIDS in 105 countries, amounting\nto more than USD 5.2 billion. For this, we first show that our framework\nsuccessfully computes heterogeneous treatment-response curves using\nsemi-synthetic data. Then, we demonstrate our framework using real-world HIV\ndata. Our framework points to large opportunities for a more effective aid\nallocation, suggesting that the total number of new HIV infections could be\nreduced by up to 3.3% (~50,000 cases) compared to the current allocation\npractice.\n", "date": "2024-01-30", "categories": []}
{"id": "2403.08335", "title": "A Sparsity Principle for Partially Observable Causal Representation\n  Learning", "abstract": "  Causal representation learning aims at identifying high-level causal\nvariables from perceptual data. Most methods assume that all latent causal\nvariables are captured in the high-dimensional observations. We instead\nconsider a partially observed setting, in which each measurement only provides\ninformation about a subset of the underlying causal state. Prior work has\nstudied this setting with multiple domains or views, each depending on a fixed\nsubset of latents. Here, we focus on learning from unpaired observations from a\ndataset with an instance-dependent partial observability pattern. Our main\ncontribution is to establish two identifiability results for this setting: one\nfor linear mixing functions without parametric assumptions on the underlying\ncausal model, and one for piecewise linear mixing functions with Gaussian\nlatent causal variables. Based on these insights, we propose two methods for\nestimating the underlying causal variables by enforcing sparsity in the\ninferred representation. Experiments on different simulated datasets and\nestablished benchmarks highlight the effectiveness of our approach in\nrecovering the ground-truth latents.\n", "date": "2024-03-13", "categories": []}
{"id": "2406.10419", "title": "Learning Flexible Time-windowed Granger Causality Integrating\n  Heterogeneous Interventional Time Series Data", "abstract": "  Granger causality, commonly used for inferring causal structures from time\nseries data, has been adopted in widespread applications across various fields\ndue to its intuitive explainability and high compatibility with emerging deep\nneural network prediction models. To alleviate challenges in better deciphering\ncausal structures unambiguously from time series, the use of interventional\ndata has become a practical approach. However, existing methods have yet to be\nexplored in the context of imperfect interventions with unknown targets, which\nare more common and often more beneficial in a wide range of real-world\napplications. Additionally, the identifiability issues of Granger causality\nwith unknown interventional targets in complex network models remain unsolved.\nOur work presents a theoretically-grounded method that infers Granger causal\nstructure and identifies unknown targets by leveraging heterogeneous\ninterventional time series data. We further illustrate that learning Granger\ncausal structure and recovering interventional targets can mutually promote\neach other. Comparative experiments demonstrate that our method outperforms\nseveral robust baseline methods in learning Granger causal structure from\ninterventional time series data.\n", "date": "2024-06-14", "categories": []}
{"id": "2406.10537", "title": "Scalable Differentiable Causal Discovery in the Presence of Latent\n  Confounders with Skeleton Posterior (Extended Version)", "abstract": "  Differentiable causal discovery has made significant advancements in the\nlearning of directed acyclic graphs. However, its application to real-world\ndatasets remains restricted due to the ubiquity of latent confounders and the\nrequirement to learn maximal ancestral graphs (MAGs). To date, existing\ndifferentiable MAG learning algorithms have been limited to small datasets and\nfailed to scale to larger ones (e.g., with more than 50 variables).\n  The key insight in this paper is that the causal skeleton, which is the\nundirected version of the causal graph, has potential for improving accuracy\nand reducing the search space of the optimization procedure, thereby enhancing\nthe performance of differentiable causal discovery. Therefore, we seek to\naddress a two-fold challenge to harness the potential of the causal skeleton\nfor differentiable causal discovery in the presence of latent confounders: (1)\nscalable and accurate estimation of skeleton and (2) universal integration of\nskeleton estimation with differentiable causal discovery.\n  To this end, we propose SPOT (Skeleton Posterior-guided OpTimization), a\ntwo-phase framework that harnesses skeleton posterior for differentiable causal\ndiscovery in the presence of latent confounders. On the contrary to a\n``point-estimation'', SPOT seeks to estimate the posterior distribution of\nskeletons given the dataset. It first formulates the posterior inference as an\ninstance of amortized inference problem and concretizes it with a supervised\ncausal learning (SCL)-enabled solution to estimate the skeleton posterior. To\nincorporate the skeleton posterior with differentiable causal discovery, SPOT\nthen features a skeleton posterior-guided stochastic optimization procedure to\nguide the optimization of MAGs. [abridged due to length limit]\n", "date": "2024-06-15", "categories": []}
{"id": "2406.10917", "title": "Bayesian Intervention Optimization for Causal Discovery", "abstract": "  Causal discovery is crucial for understanding complex systems and informing\ndecisions. While observational data can uncover causal relationships under\ncertain assumptions, it often falls short, making active interventions\nnecessary. Current methods, such as Bayesian and graph-theoretical approaches,\ndo not prioritize decision-making and often rely on ideal conditions or\ninformation gain, which is not directly related to hypothesis testing. We\npropose a novel Bayesian optimization-based method inspired by Bayes factors\nthat aims to maximize the probability of obtaining decisive and correct\nevidence. Our approach uses observational data to estimate causal models under\ndifferent hypotheses, evaluates potential interventions pre-experimentally, and\niteratively updates priors to refine interventions. We demonstrate the\neffectiveness of our method through various experiments. Our contributions\nprovide a robust framework for efficient causal discovery through active\ninterventions, enhancing the practical application of theoretical advancements.\n", "date": "2024-06-16", "categories": []}
{"id": "2406.05855", "title": "Self-Distilled Disentangled Learning for Counterfactual Prediction", "abstract": "  The advancements in disentangled representation learning significantly\nenhance the accuracy of counterfactual predictions by granting precise control\nover instrumental variables, confounders, and adjustable variables. An\nappealing method for achieving the independent separation of these factors is\nmutual information minimization, a task that presents challenges in numerous\nmachine learning scenarios, especially within high-dimensional spaces. To\ncircumvent this challenge, we propose the Self-Distilled Disentanglement\nframework, referred to as $SD^2$. Grounded in information theory, it ensures\ntheoretically sound independent disentangled representations without intricate\nmutual information estimator designs for high-dimensional representations. Our\ncomprehensive experiments, conducted on both synthetic and real-world datasets,\nconfirms the effectiveness of our approach in facilitating counterfactual\ninference in the presence of both observed and unobserved confounders.\n", "date": "2024-06-09", "categories": []}
{"id": "2406.08851", "title": "Inverse Probability of Treatment Weighting with Deep Sequence Models\n  Enables Accurate treatment effect Estimation from Electronic Health Records", "abstract": "  Observational data have been actively used to estimate treatment effect,\ndriven by the growing availability of electronic health records (EHRs).\nHowever, EHRs typically consist of longitudinal records, often introducing\ntime-dependent confoundings that hinder the unbiased estimation of treatment\neffect. Inverse probability of treatment weighting (IPTW) is a widely used\npropensity score method since it provides unbiased treatment effect estimation\nand its derivation is straightforward. In this study, we aim to utilize IPTW to\nestimate treatment effect in the presence of time-dependent confounding using\nclaims records. Previous studies have utilized propensity score methods with\nfeatures derived from claims records through feature processing, which\ngenerally requires domain knowledge and additional resources to extract\ninformation to accurately estimate propensity scores. Deep sequence models,\nparticularly recurrent neural networks and self-attention-based architectures,\nhave demonstrated good performance in modeling EHRs for various downstream\ntasks. We propose that these deep sequence models can provide accurate IPTW\nestimation of treatment effect by directly estimating the propensity scores\nfrom claims records without the need for feature processing. We empirically\ndemonstrate this by conducting comprehensive evaluations using synthetic and\nsemi-synthetic datasets.\n", "date": "2024-06-13", "categories": []}
{"id": "2406.09049", "title": "Efficiently Deciding Algebraic Equivalence of Bow-Free Acyclic Path\n  Diagrams", "abstract": "  For causal discovery in the presence of latent confounders, constraints\nbeyond conditional independences exist that can enable causal discovery\nalgorithms to distinguish more pairs of graphs. Such constraints are not\nwell-understood yet. In the setting of linear structural equation models\nwithout bows, we study algebraic constraints and argue that these provide the\nmost fine-grained resolution achievable. We propose efficient algorithms that\ndecide whether two graphs impose the same algebraic constraints, or whether the\nconstraints imposed by one graph are a subset of those imposed by another\ngraph.\n", "date": "2024-06-10", "categories": []}
{"id": "2406.09318", "title": "Characterising Interventions in Causal Games", "abstract": "  Causal games are probabilistic graphical models that enable causal queries to\nbe answered in multi-agent settings. They extend causal Bayesian networks by\nspecifying decision and utility variables to represent the agents' degrees of\nfreedom and objectives. In multi-agent settings, whether each agent decides on\ntheir policy before or after knowing the causal intervention is important as\nthis affects whether they can respond to the intervention by adapting their\npolicy. Consequently, previous work in causal games imposed chronological\nconstraints on permissible interventions. We relax this by outlining a sound\nand complete set of primitive causal interventions so the effect of any\narbitrarily complex interventional query can be studied in multi-agent\nsettings. We also demonstrate applications to the design of safe AI systems by\nconsidering causal mechanism design and commitment.\n", "date": "2024-06-13", "categories": []}
{"id": "2406.09338", "title": "Learning the Influence Graph of a High-Dimensional Markov Process with\n  Memory", "abstract": "  Motivated by multiple applications in social networks, nervous systems, and\nfinancial risk analysis, we consider the problem of learning the underlying\n(directed) influence graph or causal graph of a high-dimensional multivariate\ndiscrete-time Markov process with memory. At any discrete time instant, each\nobserved variable of the multivariate process is a binary string of random\nlength, which is parameterized by an unobservable or hidden [0,1]-valued\nscalar. The hidden scalars corresponding to the variables evolve according to\ndiscrete-time linear stochastic dynamics dictated by the underlying influence\ngraph whose nodes are the variables. We extend an existing algorithm for\nlearning i.i.d. graphical models to this Markovian setting with memory and\nprove that it can learn the influence graph based on the binary observations\nusing logarithmic (in number of variables or nodes) samples when the degree of\nthe influence graph is bounded. The crucial analytical contribution of this\nwork is the derivation of the sample complexity result by upper and lower\nbounding the rate of convergence of the observed Markov process with memory to\nits stationary distribution in terms of the parameters of the influence graph.\n", "date": "2024-06-13", "categories": []}
{"id": "2404.00848", "title": "Predictive Performance Comparison of Decision Policies Under Confounding", "abstract": "  Predictive models are often introduced to decision-making tasks under the\nrationale that they improve performance over an existing decision-making\npolicy. However, it is challenging to compare predictive performance against an\nexisting decision-making policy that is generally under-specified and dependent\non unobservable factors. These sources of uncertainty are often addressed in\npractice by making strong assumptions about the data-generating mechanism. In\nthis work, we propose a method to compare the predictive performance of\ndecision policies under a variety of modern identification approaches from the\ncausal inference and off-policy evaluation literatures (e.g., instrumental\nvariable, marginal sensitivity model, proximal variable). Key to our method is\nthe insight that there are regions of uncertainty that we can safely ignore in\nthe policy comparison. We develop a practical approach for finite-sample\nestimation of regret intervals under no assumptions on the parametric form of\nthe status quo policy. We verify our framework theoretically and via synthetic\ndata experiments. We conclude with a real-world application using our framework\nto support a pre-deployment evaluation of a proposed modification to a\nhealthcare enrollment policy.\n", "date": "2024-04-01", "categories": []}
{"id": "2405.15301", "title": "Rankability-enhanced Revenue Uplift Modeling Framework for Online\n  Marketing", "abstract": "  Uplift modeling has been widely employed in online marketing by predicting\nthe response difference between the treatment and control groups, so as to\nidentify the sensitive individuals toward interventions like coupons or\ndiscounts. Compared with traditional \\textit{conversion uplift modeling},\n\\textit{revenue uplift modeling} exhibits higher potential due to its direct\nconnection with the corporate income. However, previous works can hardly handle\nthe continuous long-tail response distribution in revenue uplift modeling.\nMoreover, they have neglected to optimize the uplift ranking among different\nindividuals, which is actually the core of uplift modeling. To address such\nissues, in this paper, we first utilize the zero-inflated lognormal (ZILN) loss\nto regress the responses and customize the corresponding modeling network,\nwhich can be adapted to different existing uplift models. Then, we study the\nranking-related uplift modeling error from the theoretical perspective and\npropose two tighter error bounds as the additional loss terms to the\nconventional response regression loss. Finally, we directly model the uplift\nranking error for the entire population with a listwise uplift ranking loss.\nThe experiment results on offline public and industrial datasets validate the\neffectiveness of our method for revenue uplift modeling. Furthermore, we\nconduct large-scale experiments on a prominent online fintech marketing\nplatform, Tencent FiT, which further demonstrates the superiority of our method\nin real-world applications.\n", "date": "2024-05-24", "categories": []}
{"id": "2406.07940", "title": "Simple yet Sharp Sensitivity Analysis for Any Contrast Under Unmeasured\n  Confounding", "abstract": "  We extend our previous work on sensitivity analysis for the risk ratio and\ndifference contrasts under unmeasured confounding to any contrast. We prove\nthat the bounds produced are still arbitrarily sharp, i.e. practically\nattainable. We illustrate the usability of the bounds with real data.\n", "date": "2024-06-12", "categories": []}
{"id": "2406.08106", "title": "Counterfactual-based Root Cause Analysis for Dynamical Systems", "abstract": "  Identifying the underlying reason for a failing dynamic process or otherwise\nanomalous observation is a fundamental challenge, yet has numerous industrial\napplications. Identifying the failure-causing sub-system using causal\ninference, one can ask the question: \"Would the observed failure also occur, if\nwe had replaced the behaviour of a sub-system at a certain point in time with\nits normal behaviour?\" To this end, a formal description of behaviour of the\nfull system is needed in which such counterfactual questions can be answered.\nHowever, existing causal methods for root cause identification are typically\nlimited to static settings and focusing on additive external influences causing\nfailures rather than structural influences. In this paper, we address these\nproblems by modelling the dynamic causal system using a Residual Neural Network\nand deriving corresponding counterfactual distributions over trajectories. We\nshow quantitatively that more root causes are identified when an intervention\nis performed on the structural equation and the external influence, compared to\nan intervention on the external influence only. By employing an efficient\napproximation to a corresponding Shapley value, we also obtain a ranking\nbetween the different subsystems at different points in time being responsible\nfor an observed failure, which is applicable in settings with large number of\nvariables. We illustrate the effectiveness of the proposed method on a\nbenchmark dynamic system as well as on a real world river dataset.\n", "date": "2024-06-12", "categories": []}
{"id": "2406.08206", "title": "Sources of Gain: Decomposing Performance in Conditional Average Dose\n  Response Estimation", "abstract": "  Estimating conditional average dose responses (CADR) is an important but\nchallenging problem. Estimators must correctly model the potentially complex\nrelationships between covariates, interventions, doses, and outcomes. In recent\nyears, the machine learning community has shown great interest in developing\ntailored CADR estimators that target specific challenges. Their performance is\ntypically evaluated against other methods on (semi-) synthetic benchmark\ndatasets. Our paper analyses this practice and shows that using popular\nbenchmark datasets without further analysis is insufficient to judge model\nperformance. Established benchmarks entail multiple challenges, whose impacts\nmust be disentangled. Therefore, we propose a novel decomposition scheme that\nallows the evaluation of the impact of five distinct components contributing to\nCADR estimator performance. We apply this scheme to eight popular CADR\nestimators on four widely-used benchmark datasets, running nearly 1,500\nindividual experiments. Our results reveal that most established benchmarks are\nchallenging for reasons different from their creators' claims. Notably,\nconfounding, the key challenge tackled by most estimators, is not an issue in\nany of the considered datasets. We discuss the major implications of our\nfindings and present directions for future research.\n", "date": "2024-06-12", "categories": []}
{"id": "2406.03575", "title": "Reconciling Heterogeneous Effects in Causal Inference", "abstract": "  In this position and problem pitch paper, we offer a solution to the\nreference class problem in causal inference. We apply the Reconcile algorithm\nfor model multiplicity in machine learning to reconcile heterogeneous effects\nin causal inference. Discrepancy between conditional average treatment effect\n(CATE) estimators of heterogeneous effects poses the reference class problem,\nwhere estimates for individual predictions differ by choice of reference class.\nBy adopting the individual to group framework for interpreting probability, we\ncan recognize that the reference class problem -- which appears across fields\nsuch as philosophy of science and causal inference -- is equivalent to the\nmodel multiplicity problem in computer science. We then apply the Reconcile\nAlgorithm to reconcile differences in estimates of individual probability among\nCATE estimators. Because the reference class problem manifests in contexts of\nindividual probability prediction using group-based evidence, our results have\ntangible implications for ensuring fair outcomes in high-stakes such as\nhealthcare, insurance, and housing, especially for marginalized communities. By\nhighlighting the importance of mitigating disparities in predictive modeling,\nour work invites further exploration into interdisciplinary strategies that\ncombine technical rigor with a keen awareness of social implications.\nUltimately, our findings advocate for a holistic approach to algorithmic\nfairness, underscoring the critical role of thoughtful, well-rounded solutions\nin achieving the broader goals of equity and access.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.06768", "title": "Data-Driven Switchback Experiments: Theoretical Tradeoffs and Empirical\n  Bayes Designs", "abstract": "  We study the design and analysis of switchback experiments conducted on a\nsingle aggregate unit. The design problem is to partition the continuous time\nspace into intervals and switch treatments between intervals, in order to\nminimize the estimation error of the treatment effect. We show that the\nestimation error depends on four factors: carryover effects, periodicity,\nserially correlated outcomes, and impacts from simultaneous experiments. We\nderive a rigorous bias-variance decomposition and show the tradeoffs of the\nestimation error from these factors. The decomposition provides three new\ninsights in choosing a design: First, balancing the periodicity between treated\nand control intervals reduces the variance; second, switching less frequently\nreduces the bias from carryover effects while increasing the variance from\ncorrelated outcomes, and vice versa; third, randomizing interval start and end\npoints reduces both bias and variance from simultaneous experiments. Combining\nthese insights, we propose a new empirical Bayes design approach. This approach\nuses prior data and experiments for designing future experiments. We illustrate\nthis approach using real data from a ride-sharing platform, yielding a design\nthat reduces MSE by 33% compared to the status quo design used on the platform.\n", "date": "2024-06-10", "categories": []}
{"id": "2406.06829", "title": "Personalized Binomial DAGs Learning with Network Structured Covariates", "abstract": "  The causal dependence in data is often characterized by Directed Acyclic\nGraphical (DAG) models, widely used in many areas. Causal discovery aims to\nrecover the DAG structure using observational data. This paper focuses on\ncausal discovery with multi-variate count data. We are motivated by real-world\nweb visit data, recording individual user visits to multiple websites. Building\na causal diagram can help understand user behavior in transitioning between\nwebsites, inspiring operational strategy. A challenge in modeling is user\nheterogeneity, as users with different backgrounds exhibit varied behaviors.\nAdditionally, social network connections can result in similar behaviors among\nfriends. We introduce personalized Binomial DAG models to address heterogeneity\nand network dependency between observations, which are common in real-world\napplications. To learn the proposed DAG model, we develop an algorithm that\nembeds the network structure into a dimension-reduced covariate, learns each\nnode's neighborhood to reduce the DAG search space, and explores the\nvariance-mean relation to determine the ordering. Simulations show our\nalgorithm outperforms state-of-the-art competitors in heterogeneous data. We\ndemonstrate its practical usefulness on a real-world web visit dataset.\n", "date": "2024-06-10", "categories": []}
{"id": "2406.07020", "title": "Learning Discrete Latent Variable Structures with Tensor Rank Conditions", "abstract": "  Unobserved discrete data are ubiquitous in many scientific disciplines, and\nhow to learn the causal structure of these latent variables is crucial for\nuncovering data patterns. Most studies focus on the linear latent variable\nmodel or impose strict constraints on latent structures, which fail to address\ncases in discrete data involving non-linear relationships or complex latent\nstructures. To achieve this, we explore a tensor rank condition on contingency\ntables for an observed variable set $\\mathbf{X}_p$, showing that the rank is\ndetermined by the minimum support of a specific conditional set (not necessary\nin $\\mathbf{X}_p$) that d-separates all variables in $\\mathbf{X}_p$. By this,\none can locate the latent variable through probing the rank on different\nobserved variables set, and further identify the latent causal structure under\nsome structure assumptions. We present the corresponding identification\nalgorithm and conduct simulated experiments to verify the effectiveness of our\nmethod. In general, our results elegantly extend the identification boundary\nfor causal discovery with discrete latent variables and expand the application\nscope of causal discovery with latent variables.\n", "date": "2024-06-11", "categories": []}
{"id": "2406.07378", "title": "Large Language Models for Constrained-Based Causal Discovery", "abstract": "  Causality is essential for understanding complex systems, such as the\neconomy, the brain, and the climate. Constructing causal graphs often relies on\neither data-driven or expert-driven approaches, both fraught with challenges.\nThe former methods, like the celebrated PC algorithm, face issues with data\nrequirements and assumptions of causal sufficiency, while the latter demand\nsubstantial time and domain knowledge. This work explores the capabilities of\nLarge Language Models (LLMs) as an alternative to domain experts for causal\ngraph generation. We frame conditional independence queries as prompts to LLMs\nand employ the PC algorithm with the answers. The performance of the LLM-based\nconditional independence oracle on systems with known causal graphs shows a\nhigh degree of variability. We improve the performance through a proposed\nstatistical-inspired voting schema that allows some control over false-positive\nand false-negative rates. Inspecting the chain-of-thought argumentation, we\nfind causal reasoning to justify its answer to a probabilistic query. We show\nevidence that knowledge-based CIT could eventually become a complementary tool\nfor data-driven causal discovery.\n", "date": "2024-06-11", "categories": []}
{"id": "2308.06718", "title": "Generalized Independent Noise Condition for Estimating Causal Structure\n  with Latent Variables", "abstract": "  We investigate the task of learning causal structure in the presence of\nlatent variables, including locating latent variables and determining their\nquantity, and identifying causal relationships among both latent and observed\nvariables. To this end, we propose a Generalized Independent Noise (GIN)\ncondition for linear non-Gaussian acyclic causal models that incorporate latent\nvariables, which establishes the independence between a linear combination of\ncertain measured variables and some other measured variables. Specifically, for\ntwo observed random vectors $\\bf{Y}$ and $\\bf{Z}$, GIN holds if and only if\n$\\omega^{\\intercal}\\mathbf{Y}$ and $\\mathbf{Z}$ are independent, where $\\omega$\nis a non-zero parameter vector determined by the cross-covariance between\n$\\mathbf{Y}$ and $\\mathbf{Z}$. We then give necessary and sufficient graphical\ncriteria of the GIN condition in linear non-Gaussian acyclic models. Roughly\nspeaking, GIN implies the existence of a set $\\mathcal{S}$ such that\n$\\mathcal{S}$ is causally earlier (w.r.t. the causal ordering) than\n$\\mathbf{Y}$, and that every active (collider-free) path between $\\mathbf{Y}$\nand $\\mathbf{Z}$ must contain a node from $\\mathcal{S}$. Interestingly, we find\nthat the independent noise condition (i.e., if there is no confounder, causes\nare independent of the residual derived from regressing the effect on the\ncauses) can be seen as a special case of GIN. With such a connection between\nGIN and latent causal structures, we further leverage the proposed GIN\ncondition, together with a well-designed search procedure, to efficiently\nestimate Linear, Non-Gaussian Latent Hierarchical Models (LiNGLaHs), where\nlatent confounders may also be causally related and may even follow a\nhierarchical structure. We show that the causal structure of a LiNGLaH is\nidentifiable in light of GIN conditions. Experimental results show the\neffectiveness of the proposed method.\n", "date": "2023-08-13", "categories": []}
{"id": "2310.11334", "title": "Agent-Specific Effects: A Causal Effect Propagation Analysis in\n  Multi-Agent MDPs", "abstract": "  Establishing causal relationships between actions and outcomes is fundamental\nfor accountable multi-agent decision-making. However, interpreting and\nquantifying agents' contributions to such relationships pose significant\nchallenges. These challenges are particularly prominent in the context of\nmulti-agent sequential decision-making, where the causal effect of an agent's\naction on the outcome depends on how other agents respond to that action. In\nthis paper, our objective is to present a systematic approach for attributing\nthe causal effects of agents' actions to the influence they exert on other\nagents. Focusing on multi-agent Markov decision processes, we introduce\nagent-specific effects (ASE), a novel causal quantity that measures the effect\nof an agent's action on the outcome that propagates through other agents. We\nthen turn to the counterfactual counterpart of ASE (cf-ASE), provide a\nsufficient set of conditions for identifying cf-ASE, and propose a practical\nsampling-based algorithm for estimating it. Finally, we experimentally evaluate\nthe utility of cf-ASE through a simulation-based testbed, which includes a\nsepsis management environment.\n", "date": "2023-10-17", "categories": []}
{"id": "2312.12844", "title": "Effective Causal Discovery under Identifiable Heteroscedastic Noise\n  Model", "abstract": "  Capturing the underlying structural causal relations represented by Directed\nAcyclic Graphs (DAGs) has been a fundamental task in various AI disciplines.\nCausal DAG learning via the continuous optimization framework has recently\nachieved promising performance in terms of both accuracy and efficiency.\nHowever, most methods make strong assumptions of homoscedastic noise, i.e.,\nexogenous noises have equal variances across variables, observations, or even\nboth. The noises in real data usually violate both assumptions due to the\nbiases introduced by different data collection processes. To address the issue\nof heteroscedastic noise, we introduce relaxed and implementable sufficient\nconditions, proving the identifiability of a general class of SEM subject to\nthese conditions. Based on the identifiable general SEM, we propose a novel\nformulation for DAG learning that accounts for the variation in noise variance\nacross variables and observations. We then propose an effective two-phase\niterative DAG learning algorithm to address the increasing optimization\ndifficulties and to learn a causal DAG from data with heteroscedastic variable\nnoise under varying variance. We show significant empirical gains of the\nproposed approaches over state-of-the-art methods on both synthetic data and\nreal data.\n", "date": "2023-12-20", "categories": []}
{"id": "2406.05633", "title": "Heterogeneous Treatment Effects in Panel Data", "abstract": "  We address a core problem in causal inference: estimating heterogeneous\ntreatment effects using panel data with general treatment patterns. Many\nexisting methods either do not utilize the potential underlying structure in\npanel data or have limitations in the allowable treatment patterns. In this\nwork, we propose and evaluate a new method that first partitions observations\ninto disjoint clusters with similar treatment effects using a regression tree,\nand then leverages the (assumed) low-rank structure of the panel data to\nestimate the average treatment effect for each cluster. Our theoretical results\nestablish the convergence of the resulting estimates to the true treatment\neffects. Computation experiments with semi-synthetic data show that our method\nachieves superior accuracy compared to alternative approaches, using a\nregression tree with no more than 40 leaves. Hence, our method provides more\naccurate and interpretable estimates than alternative methods.\n", "date": "2024-06-09", "categories": []}
{"id": "2406.05746", "title": "Methodology and Real-World Applications of Dynamic Uncertain Causality\n  Graph for Clinical Diagnosis with Explainability and Invariance", "abstract": "  AI-aided clinical diagnosis is desired in medical care. Existing deep\nlearning models lack explainability and mainly focus on image analysis. The\nrecently developed Dynamic Uncertain Causality Graph (DUCG) approach is\ncausality-driven, explainable, and invariant across different application\nscenarios, without problems of data collection, labeling, fitting, privacy,\nbias, generalization, high cost and high energy consumption. Through close\ncollaboration between clinical experts and DUCG technicians, 46 DUCG models\ncovering 54 chief complaints were constructed. Over 1,000 diseases can be\ndiagnosed without triage. Before being applied in real-world, the 46 DUCG\nmodels were retrospectively verified by third-party hospitals. The verified\ndiagnostic precisions were no less than 95%, in which the diagnostic precision\nfor every disease including uncommon ones was no less than 80%. After\nverifications, the 46 DUCG models were applied in the real-world in China. Over\none million real diagnosis cases have been performed, with only 17 incorrect\ndiagnoses identified. Due to DUCG's transparency, the mistakes causing the\nincorrect diagnoses were found and corrected. The diagnostic abilities of the\nclinicians who applied DUCG frequently were improved significantly. Following\nthe introduction to the earlier presented DUCG methodology, the recommendation\nalgorithm for potential medical checks is presented and the key idea of DUCG is\nextracted.\n", "date": "2024-06-09", "categories": []}
{"id": "2310.15709", "title": "Causal Representation Learning Made Identifiable by Grouping of\n  Observational Variables", "abstract": "  A topic of great current interest is Causal Representation Learning (CRL),\nwhose goal is to learn a causal model for hidden features in a data-driven\nmanner. Unfortunately, CRL is severely ill-posed since it is a combination of\nthe two notoriously ill-posed problems of representation learning and causal\ndiscovery. Yet, finding practical identifiability conditions that guarantee a\nunique solution is crucial for its practical applicability. Most approaches so\nfar have been based on assumptions on the latent causal mechanisms, such as\ntemporal causality, or existence of supervision or interventions; these can be\ntoo restrictive in actual applications. Here, we show identifiability based on\nnovel, weak constraints, which requires no temporal structure, intervention,\nnor weak supervision. The approach is based on assuming the observational\nmixing exhibits a suitable grouping of the observational variables. We also\npropose a novel self-supervised estimation framework consistent with the model,\nprove its statistical consistency, and experimentally show its superior CRL\nperformances compared to the state-of-the-art baselines. We further demonstrate\nits robustness against latent confounders and causal cycles.\n", "date": "2023-10-24", "categories": []}
{"id": "2406.04598", "title": "OCDB: Revisiting Causal Discovery with a Comprehensive Benchmark and\n  Evaluation Framework", "abstract": "  Large language models (LLMs) have excelled in various natural language\nprocessing tasks, but challenges in interpretability and trustworthiness\npersist, limiting their use in high-stakes fields. Causal discovery offers a\npromising approach to improve transparency and reliability. However, current\nevaluations are often one-sided and lack assessments focused on\ninterpretability performance. Additionally, these evaluations rely on synthetic\ndata and lack comprehensive assessments of real-world datasets. These lead to\npromising methods potentially being overlooked. To address these issues, we\npropose a flexible evaluation framework with metrics for evaluating differences\nin causal structures and causal effects, which are crucial attributes that help\nimprove the interpretability of LLMs. We introduce the Open Causal Discovery\nBenchmark (OCDB), based on real data, to promote fair comparisons and drive\noptimization of algorithms. Additionally, our new metrics account for\nundirected edges, enabling fair comparisons between Directed Acyclic Graphs\n(DAGs) and Completed Partially Directed Acyclic Graphs (CPDAGs). Experimental\nresults show significant shortcomings in existing algorithms' generalization\ncapabilities on real data, highlighting the potential for performance\nimprovement and the importance of our framework in advancing causal discovery\ntechniques.\n", "date": "2024-06-07", "categories": []}
{"id": "2206.06821", "title": "DoWhy-GCM: An extension of DoWhy for causal inference in graphical\n  causal models", "abstract": "  We present DoWhy-GCM, an extension of the DoWhy Python library, which\nleverages graphical causal models. Unlike existing causality libraries, which\nmainly focus on effect estimation, DoWhy-GCM addresses diverse causal queries,\nsuch as identifying the root causes of outliers and distributional changes,\nattributing causal influences to the data generating process of each node, or\ndiagnosis of causal structures. With DoWhy-GCM, users typically specify\ncause-effect relations via a causal graph, fit causal mechanisms, and pose\ncausal queries -- all with just a few lines of code. The general documentation\nis available at https://www.pywhy.org/dowhy and the DoWhy-GCM specific code at\nhttps://github.com/py-why/dowhy/tree/main/dowhy/gcm.\n", "date": "2022-06-14", "categories": []}
{"id": "2304.14545", "title": "Augmented balancing weights as linear regression", "abstract": "  We provide a novel characterization of augmented balancing weights, also\nknown as automatic debiased machine learning (AutoDML). These popular doubly\nrobust or de-biased machine learning estimators combine outcome modeling with\nbalancing weights - weights that achieve covariate balance directly in lieu of\nestimating and inverting the propensity score. When the outcome and weighting\nmodels are both linear in some (possibly infinite) basis, we show that the\naugmented estimator is equivalent to a single linear model with coefficients\nthat combine the coefficients from the original outcome model and coefficients\nfrom an unpenalized ordinary least squares (OLS) fit on the same data. We see\nthat, under certain choices of regularization parameters, the augmented\nestimator often collapses to the OLS estimator alone; this occurs for example\nin a re-analysis of the Lalonde 1986 dataset. We then extend these results to\nspecific choices of outcome and weighting models. We first show that the\naugmented estimator that uses (kernel) ridge regression for both outcome and\nweighting models is equivalent to a single, undersmoothed (kernel) ridge\nregression. This holds numerically in finite samples and lays the groundwork\nfor a novel analysis of undersmoothing and asymptotic rates of convergence.\nWhen the weighting model is instead lasso-penalized regression, we give\nclosed-form expressions for special cases and demonstrate a ``double\nselection'' property. Our framework opens the black box on this increasingly\npopular class of estimators, bridges the gap between existing results on the\nsemiparametric efficiency of undersmoothed and doubly robust estimators, and\nprovides new insights into the performance of augmented balancing weights.\n", "date": "2023-04-27", "categories": []}
{"id": "2305.17139", "title": "A Measure-Theoretic Axiomatisation of Causality", "abstract": "  Causality is a central concept in a wide range of research areas, yet there\nis still no universally agreed axiomatisation of causality. We view causality\nboth as an extension of probability theory and as a study of \\textit{what\nhappens when one intervenes on a system}, and argue in favour of taking\nKolmogorov's measure-theoretic axiomatisation of probability as the starting\npoint towards an axiomatisation of causality. To that end, we propose the\nnotion of a \\textit{causal space}, consisting of a probability space along with\na collection of transition probability kernels, called \\textit{causal kernels},\nthat encode the causal information of the space. Our proposed framework is not\nonly rigorously grounded in measure theory, but it also sheds light on\nlong-standing limitations of existing frameworks including, for example,\ncycles, latent variables and stochastic processes.\n", "date": "2023-05-19", "categories": []}
{"id": "2402.14979", "title": "Optimizing Language Models for Human Preferences is a Causal Inference\n  Problem", "abstract": "  As large language models (LLMs) see greater use in academic and commercial\nsettings, there is increasing interest in methods that allow language models to\ngenerate texts aligned with human preferences. In this paper, we present an\ninitial exploration of language model optimization for human preferences from\ndirect outcome datasets, where each sample consists of a text and an associated\nnumerical outcome measuring the reader's response. We first propose that\nlanguage model optimization should be viewed as a causal problem to ensure that\nthe model correctly learns the relationship between the text and the outcome.\nWe formalize this causal language optimization problem, and we develop a\nmethod--causal preference optimization (CPO)--that solves an unbiased surrogate\nobjective for the problem. We further extend CPO with doubly robust CPO\n(DR-CPO), which reduces the variance of the surrogate objective while retaining\nprovably strong guarantees on bias. Finally, we empirically demonstrate the\neffectiveness of (DR-)CPO in optimizing state-of-the-art LLMs for human\npreferences on direct outcome data, and we validate the robustness of DR-CPO\nunder difficult confounding conditions.\n", "date": "2024-02-22", "categories": []}
{"id": "2405.16225", "title": "Local Causal Structure Learning in the Presence of Latent Variables", "abstract": "  Discovering causal relationships from observational data, particularly in the\npresence of latent variables, poses a challenging problem. While current local\nstructure learning methods have proven effective and efficient when the focus\nlies solely on the local relationships of a target variable, they operate under\nthe assumption of causal sufficiency. This assumption implies that all the\ncommon causes of the measured variables are observed, leaving no room for\nlatent variables. Such a premise can be easily violated in various real-world\napplications, resulting in inaccurate structures that may adversely impact\ndownstream tasks. In light of this, our paper delves into the primary\ninvestigation of locally identifying potential parents and children of a target\nfrom observational data that may include latent variables. Specifically, we\nharness the causal information from m-separation and V-structures to derive\ntheoretical consistency results, effectively bridging the gap between global\nand local structure learning. Together with the newly developed stop rules, we\npresent a principled method for determining whether a variable is a direct\ncause or effect of a target. Further, we theoretically demonstrate the\ncorrectness of our approach under the standard causal Markov and faithfulness\nconditions, with infinite samples. Experimental results on both synthetic and\nreal-world data validate the effectiveness and efficiency of our approach.\n", "date": "2024-05-25", "categories": []}
{"id": "2406.03864", "title": "PairNet: Training with Observed Pairs to Estimate Individual Treatment\n  Effect", "abstract": "  Given a dataset of individuals each described by a covariate vector, a\ntreatment, and an observed outcome on the treatment, the goal of the individual\ntreatment effect (ITE) estimation task is to predict outcome changes resulting\nfrom a change in treatment. A fundamental challenge is that in the\nobservational data, a covariate's outcome is observed only under one treatment,\nwhereas we need to infer the difference in outcomes under two different\ntreatments. Several existing approaches address this issue through training\nwith inferred pseudo-outcomes, but their success relies on the quality of these\npseudo-outcomes. We propose PairNet, a novel ITE estimation training strategy\nthat minimizes losses over pairs of examples based on their factual observed\noutcomes. Theoretical analysis for binary treatments reveals that PairNet is a\nconsistent estimator of ITE risk, and achieves smaller generalization error\nthan baseline models. Empirical comparison with thirteen existing methods\nacross eight benchmarks, covering both discrete and continuous treatments,\nshows that PairNet achieves significantly lower ITE error compared to the\nbaselines. Also, it is model-agnostic and easy to implement.\n", "date": "2024-06-06", "categories": []}
{"id": "2306.00382", "title": "Calibrated and Conformal Propensity Scores for Causal Effect Estimation", "abstract": "  Propensity scores are commonly used to estimate treatment effects from\nobservational data. We argue that the probabilistic output of a learned\npropensity score model should be calibrated -- i.e., a predictive treatment\nprobability of 90% should correspond to 90% of individuals being assigned the\ntreatment group -- and we propose simple recalibration techniques to ensure\nthis property. We prove that calibration is a necessary condition for unbiased\ntreatment effect estimation when using popular inverse propensity weighted and\ndoubly robust estimators. We derive error bounds on causal effect estimates\nthat directly relate to the quality of uncertainties provided by the\nprobabilistic propensity score model and show that calibration strictly\nimproves this error bound while also avoiding extreme propensity weights. We\ndemonstrate improved causal effect estimation with calibrated propensity scores\nin several tasks including high-dimensional image covariates and genome-wide\nassociation studies (GWASs). Calibrated propensity scores improve the speed of\nGWAS analysis by more than two-fold by enabling the use of simpler models that\nare faster to train.\n", "date": "2023-06-01", "categories": []}
{"id": "2406.02873", "title": "Prediction-powered Generalization of Causal Inferences", "abstract": "  Causal inferences from a randomized controlled trial (RCT) may not pertain to\na target population where some effect modifiers have a different distribution.\nPrior work studies generalizing the results of a trial to a target population\nwith no outcome but covariate data available. We show how the limited size of\ntrials makes generalization a statistically infeasible task, as it requires\nestimating complex nuisance functions. We develop generalization algorithms\nthat supplement the trial data with a prediction model learned from an\nadditional observational study (OS), without making any assumptions on the OS.\nWe theoretically and empirically show that our methods facilitate better\ngeneralization when the OS is high-quality, and remain robust when it is not,\nand e.g., have unmeasured confounding.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.03209", "title": "Challenges and Considerations in the Evaluation of Bayesian Causal\n  Discovery", "abstract": "  Representing uncertainty in causal discovery is a crucial component for\nexperimental design, and more broadly, for safe and reliable causal decision\nmaking. Bayesian Causal Discovery (BCD) offers a principled approach to\nencapsulating this uncertainty. Unlike non-Bayesian causal discovery, which\nrelies on a single estimated causal graph and model parameters for assessment,\nevaluating BCD presents challenges due to the nature of its inferred quantity -\nthe posterior distribution. As a result, the research community has proposed\nvarious metrics to assess the quality of the approximate posterior. However,\nthere is, to date, no consensus on the most suitable metric(s) for evaluation.\nIn this work, we reexamine this question by dissecting various metrics and\nunderstanding their limitations. Through extensive empirical evaluation, we\nfind that many existing metrics fail to exhibit a strong correlation with the\nquality of approximation to the true posterior, especially in scenarios with\nlow sample sizes where BCD is most desirable. We highlight the suitability (or\nlack thereof) of these metrics under two distinct factors: the identifiability\nof the underlying causal model and the quantity of available data. Both factors\naffect the entropy of the true posterior, indicating that the current metrics\nare less fitting in settings of higher entropy. Our findings underline the\nimportance of a more nuanced evaluation of new methods by taking into account\nthe nature of the true posterior, as well as guide and motivate the development\nof new evaluation procedures for this challenge.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.03212", "title": "Inferring the time-varying coupling of dynamical systems with temporal\n  convolutional autoencoders", "abstract": "  Most approaches for assessing causality in complex dynamical systems fail\nwhen the interactions between variables are inherently non-linear and\nnon-stationary. Here we introduce Temporal Autoencoders for Causal Inference\n(TACI), a methodology that combines a new surrogate data metric for assessing\ncausal interactions with a novel two-headed machine learning architecture to\nidentify and measure the direction and strength of time-varying causal\ninteractions. Through tests on both synthetic and real-world datasets, we\ndemonstrate TACI's ability to accurately quantify dynamic causal interactions\nacross a variety of systems. Our findings display the method's effectiveness\ncompared to existing approaches and also highlight our approach's potential to\nbuild a deeper understanding of the mechanisms that underlie time-varying\ninteractions in physical and biological systems.\n", "date": "2024-06-05", "categories": []}
{"id": "2310.00809", "title": "Towards Causal Foundation Model: on Duality between Causal Inference and\n  Attention", "abstract": "  Foundation models have brought changes to the landscape of machine learning,\ndemonstrating sparks of human-level intelligence across a diverse array of\ntasks. However, a gap persists in complex tasks such as causal inference,\nprimarily due to challenges associated with intricate reasoning steps and high\nnumerical precision requirements. In this work, we take a first step towards\nbuilding causally-aware foundation models for treatment effect estimations. We\npropose a novel, theoretically justified method called Causal Inference with\nAttention (CInA), which utilizes multiple unlabeled datasets to perform\nself-supervised causal learning, and subsequently enables zero-shot causal\ninference on unseen tasks with new data. This is based on our theoretical\nresults that demonstrate the primal-dual connection between optimal covariate\nbalancing and self-attention, facilitating zero-shot causal inference through\nthe final layer of a trained transformer-type architecture. We demonstrate\nempirically that CInA effectively generalizes to out-of-distribution datasets\nand various real-world datasets, matching or even surpassing traditional\nper-dataset methodologies. These results provide compelling evidence that our\nmethod has the potential to serve as a stepping stone for the development of\ncausal foundation models.\n", "date": "2023-10-01", "categories": []}
{"id": "2402.08845", "title": "Feature Attribution with Necessity and Sufficiency via Dual-stage\n  Perturbation Test for Causal Explanation", "abstract": "  We investigate the problem of explainability for machine learning models,\nfocusing on Feature Attribution Methods (FAMs) that evaluate feature importance\nthrough perturbation tests. Despite their utility, FAMs struggle to distinguish\nthe contributions of different features, when their prediction changes are\nsimilar after perturbation. To enhance FAMs' discriminative power, we introduce\nFeature Attribution with Necessity and Sufficiency (FANS), which find a\nneighborhood of the input such that perturbing samples within this neighborhood\nhave a high Probability of being Necessity and Sufficiency (PNS) cause for the\nchange in predictions, and use this PNS as the importance of the feature.\nSpecifically, FANS compute this PNS via a heuristic strategy for estimating the\nneighborhood and a perturbation test involving two stages (factual and\ninterventional) for counterfactual reasoning. To generate counterfactual\nsamples, we use a resampling-based approach on the observed samples to\napproximate the required conditional distribution. We demonstrate that FANS\noutperforms existing attribution methods on six benchmarks. Please refer to the\nsource code via \\url{https://github.com/DMIRLAB-Group/FANS}.\n", "date": "2024-02-13", "categories": []}
{"id": "2406.01823", "title": "Causal Discovery with Fewer Conditional Independence Tests", "abstract": "  Many questions in science center around the fundamental problem of\nunderstanding causal relationships. However, most constraint-based causal\ndiscovery algorithms, including the well-celebrated PC algorithm, often incur\nan exponential number of conditional independence (CI) tests, posing\nlimitations in various applications. Addressing this, our work focuses on\ncharacterizing what can be learned about the underlying causal graph with a\nreduced number of CI tests. We show that it is possible to a learn a coarser\nrepresentation of the hidden causal graph with a polynomial number of tests.\nThis coarser representation, named Causal Consistent Partition Graph (CCPG),\ncomprises of a partition of the vertices and a directed graph defined over its\ncomponents. CCPG satisfies consistency of orientations and additional\nconstraints which favor finer partitions. Furthermore, it reduces to the\nunderlying causal graph when the causal graph is identifiable. As a\nconsequence, our results offer the first efficient algorithm for recovering the\ntrue causal graph with a polynomial number of tests, in special cases where the\ncausal graph is fully identifiable through observational data and potentially\nadditional interventions.\n", "date": "2024-06-03", "categories": []}
{"id": "2406.02049", "title": "Causal Effect Identification in LiNGAM Models with Latent Confounders", "abstract": "  We study the generic identifiability of causal effects in linear non-Gaussian\nacyclic models (LiNGAM) with latent variables. We consider the problem in two\nmain settings: When the causal graph is known a priori, and when it is unknown.\nIn both settings, we provide a complete graphical characterization of the\nidentifiable direct or total causal effects among observed variables. Moreover,\nwe propose efficient algorithms to certify the graphical conditions. Finally,\nwe propose an adaptation of the reconstruction independent component analysis\n(RICA) algorithm that estimates the causal effects from the observational data\ngiven the causal graph. Experimental results show the effectiveness of the\nproposed method in estimating the causal effects.\n", "date": "2024-06-04", "categories": []}
{"id": "2406.02310", "title": "Disentangled Representation via Variational AutoEncoder for Continuous\n  Treatment Effect Estimation", "abstract": "  Continuous treatment effect estimation holds significant practical importance\nacross various decision-making and assessment domains, such as healthcare and\nthe military. However, current methods for estimating dose-response curves\nhinge on balancing the entire representation by treating all covariates as\nconfounding variables. Although various approaches disentangle covariates into\ndifferent factors for treatment effect estimation, they are confined to binary\ntreatment settings. Moreover, observational data are often tainted with\nnon-causal noise information that is imperceptible to the human. Hence, in this\npaper, we propose a novel Dose-Response curve estimator via Variational\nAutoEncoder (DRVAE) disentangled covariates representation. Our model is\ndedicated to disentangling covariates into instrumental factors, confounding\nfactors, adjustment factors, and external noise factors, thereby facilitating\nthe estimation of treatment effects under continuous treatment settings by\nbalancing the disentangled confounding factors. Extensive results on synthetic\nand semi-synthetic datasets demonstrate that our model outperforms the current\nstate-of-the-art methods.\n", "date": "2024-06-04", "categories": []}
{"id": "2406.02464", "title": "Meta-Learners for Partially-Identified Treatment Effects Across Multiple\n  Environments", "abstract": "  Estimating the conditional average treatment effect (CATE) from observational\ndata is relevant for many applications such as personalized medicine. Here, we\nfocus on the widespread setting where the observational data come from multiple\nenvironments, such as different hospitals, physicians, or countries.\nFurthermore, we allow for violations of standard causal assumptions, namely,\noverlap within the environments and unconfoundedness. To this end, we move away\nfrom point identification and focus on partial identification. Specifically, we\nshow that current assumptions from the literature on multiple environments\nallow us to interpret the environment as an instrumental variable (IV). This\nallows us to adapt bounds from the IV literature for partial identification of\nCATE by leveraging treatment assignment mechanisms across environments. Then,\nwe propose different model-agnostic learners (so-called meta-learners) to\nestimate the bounds that can be used in combination with arbitrary machine\nlearning models. We further demonstrate the effectiveness of our meta-learners\nacross various experiments using both simulated and real-world data. Finally,\nwe discuss the applicability of our meta-learners to partial identification in\ninstrumental variable settings, such as randomized controlled trials with\nnon-compliance.\n", "date": "2024-06-04", "categories": []}
{"id": "2211.02763", "title": "Bayesian learning of Causal Structure and Mechanisms with GFlowNets and\n  Variational Bayes", "abstract": "  Bayesian causal structure learning aims to learn a posterior distribution\nover directed acyclic graphs (DAGs), and the mechanisms that define the\nrelationship between parent and child variables. By taking a Bayesian approach,\nit is possible to reason about the uncertainty of the causal model. The notion\nof modelling the uncertainty over models is particularly crucial for causal\nstructure learning since the model could be unidentifiable when given only a\nfinite amount of observational data. In this paper, we introduce a novel method\nto jointly learn the structure and mechanisms of the causal model using\nVariational Bayes, which we call Variational Bayes-DAG-GFlowNet (VBG). We\nextend the method of Bayesian causal structure learning using GFlowNets to\nlearn not only the posterior distribution over the structure, but also the\nparameters of a linear-Gaussian model. Our results on simulated data suggest\nthat VBG is competitive against several baselines in modelling the posterior\nover DAGs and mechanisms, while offering several advantages over existing\nmethods, including the guarantee to sample acyclic graphs, and the flexibility\nto generalize to non-linear causal mechanisms.\n", "date": "2022-11-04", "categories": []}
{"id": "2306.15479", "title": "Predictive Coding beyond Correlations", "abstract": "  Recently, there has been extensive research on the capabilities of\nbiologically plausible algorithms. In this work, we show how one of such\nalgorithms, called predictive coding, is able to perform causal inference\ntasks. First, we show how a simple change in the inference process of\npredictive coding enables to compute interventions without the need to mutilate\nor redefine a causal graph. Then, we explore applications in cases where the\ngraph is unknown, and has to be inferred from observational data. Empirically,\nwe show how such findings can be used to improve the performance of predictive\ncoding in image classification tasks, and conclude that such models are able to\nperform simple end-to-end causal inference tasks.\n", "date": "2023-06-27", "categories": []}
{"id": "2310.08672", "title": "Machine Learning Who to Nudge: Causal vs Predictive Targeting in a Field\n  Experiment on Student Financial Aid Renewal", "abstract": "  In many settings, interventions may be more effective for some individuals\nthan others, so that targeting interventions may be beneficial. We analyze the\nvalue of targeting in the context of a large-scale field experiment with over\n53,000 college students, where the goal was to use \"nudges\" to encourage\nstudents to renew their financial-aid applications before a non-binding\ndeadline. We begin with baseline approaches to targeting. First, we target\nbased on a causal forest that estimates heterogeneous treatment effects and\nthen assigns students to treatment according to those estimated to have the\nhighest treatment effects. Next, we evaluate two alternative targeting\npolicies, one targeting students with low predicted probability of renewing\nfinancial aid in the absence of the treatment, the other targeting those with\nhigh probability. The predicted baseline outcome is not the ideal criterion for\ntargeting, nor is it a priori clear whether to prioritize low, high, or\nintermediate predicted probability. Nonetheless, targeting on low baseline\noutcomes is common in practice, for example because the relationship between\nindividual characteristics and treatment effects is often difficult or\nimpossible to estimate with historical data. We propose hybrid approaches that\nincorporate the strengths of both predictive approaches (accurate estimation)\nand causal approaches (correct criterion); we show that targeting intermediate\nbaseline outcomes is most effective in our specific application, while\ntargeting based on low baseline outcomes is detrimental. In one year of the\nexperiment, nudging all students improved early filing by an average of 6.4\npercentage points over a baseline average of 37% filing, and we estimate that\ntargeting half of the students using our preferred policy attains around 75% of\nthis benefit.\n", "date": "2023-10-12", "categories": []}
{"id": "2310.17816", "title": "Local Discovery by Partitioning: Polynomial-Time Causal Discovery Around\n  Exposure-Outcome Pairs", "abstract": "  Causal discovery is crucial for causal inference in observational studies, as\nit can enable the identification of valid adjustment sets (VAS) for unbiased\neffect estimation. However, global causal discovery is notoriously hard in the\nnonparametric setting, with exponential time and sample complexity in the worst\ncase. To address this, we propose local discovery by partitioning (LDP): a\nlocal causal discovery method that is tailored for downstream inference tasks\nwithout requiring parametric and pretreatment assumptions. LDP is a\nconstraint-based procedure that returns a VAS for an exposure-outcome pair\nunder latent confounding, given sufficient conditions. The total number of\nindependence tests performed is worst-case quadratic with respect to the\ncardinality of the variable set. Asymptotic theoretical guarantees are\nnumerically validated on synthetic graphs. Adjustment sets from LDP yield less\nbiased and more precise average treatment effect estimates than baseline\ndiscovery algorithms, with LDP outperforming on confounder recall, runtime, and\ntest count for VAS discovery. Notably, LDP ran at least 1300x faster than\nbaselines on a benchmark.\n", "date": "2023-10-25", "categories": []}
{"id": "2311.18639", "title": "Targeted Reduction of Causal Models", "abstract": "  Why does a phenomenon occur? Addressing this question is central to most\nscientific inquiries and often relies on simulations of scientific models. As\nmodels become more intricate, deciphering the causes behind phenomena in\nhigh-dimensional spaces of interconnected variables becomes increasingly\nchallenging. Causal Representation Learning (CRL) offers a promising avenue to\nuncover interpretable causal patterns within these simulations through an\ninterventional lens. However, developing general CRL frameworks suitable for\npractical applications remains an open challenge. We introduce Targeted Causal\nReduction (TCR), a method for condensing complex intervenable models into a\nconcise set of causal factors that explain a specific target phenomenon. We\npropose an information theoretic objective to learn TCR from interventional\ndata of simulations, establish identifiability for continuous variables under\nshift interventions and present a practical algorithm for learning TCRs. Its\nability to generate interpretable high-level explanations from complex models\nis demonstrated on toy and mechanical systems, illustrating its potential to\nassist scientists in the study of complex phenomena in a broad range of\ndisciplines.\n", "date": "2023-11-30", "categories": []}
{"id": "2402.15255", "title": "Optimal Transport for Structure Learning Under Missing Data", "abstract": "  Causal discovery in the presence of missing data introduces a chicken-and-egg\ndilemma. While the goal is to recover the true causal structure, robust\nimputation requires considering the dependencies or, preferably, causal\nrelations among variables. Merely filling in missing values with existing\nimputation methods and subsequently applying structure learning on the complete\ndata is empirically shown to be sub-optimal. To address this problem, we\npropose a score-based algorithm for learning causal structures from missing\ndata based on optimal transport. This optimal transport viewpoint diverges from\nexisting score-based approaches that are dominantly based on expectation\nmaximization. We formulate structure learning as a density fitting problem,\nwhere the goal is to find the causal model that induces a distribution of\nminimum Wasserstein distance with the observed data distribution. Our framework\nis shown to recover the true causal graphs more effectively than competing\nmethods in most simulations and real-data settings. Empirical evidence also\nshows the superior scalability of our approach, along with the flexibility to\nincorporate any off-the-shelf causal discovery methods for complete data.\n", "date": "2024-02-23", "categories": []}
{"id": "2404.17483", "title": "Differentiable Pareto-Smoothed Weighting for High-Dimensional\n  Heterogeneous Treatment Effect Estimation", "abstract": "  There is a growing interest in estimating heterogeneous treatment effects\nacross individuals using their high-dimensional feature attributes. Achieving\nhigh performance in such high-dimensional heterogeneous treatment effect\nestimation is challenging because in this setup, it is usual that some features\ninduce sample selection bias while others do not but are predictive of\npotential outcomes. To avoid losing such predictive feature information,\nexisting methods learn separate feature representations using inverse\nprobability weighting (IPW). However, due to their numerically unstable IPW\nweights, these methods suffer from estimation bias under a finite sample setup.\nTo develop a numerically robust estimator by weighted representation learning,\nwe propose a differentiable Pareto-smoothed weighting framework that replaces\nextreme weight values in an end-to-end fashion. Our experimental results show\nthat by effectively correcting the weight values, our proposed method\noutperforms the existing ones, including traditional weighting schemes. Our\ncode is available at https://github.com/ychika/DPSW.\n", "date": "2024-04-26", "categories": []}
{"id": "2405.18626", "title": "Causal Contextual Bandits with Adaptive Context", "abstract": "  We study a variant of causal contextual bandits where the context is chosen\nbased on an initial intervention chosen by the learner. At the beginning of\neach round, the learner selects an initial action, depending on which a\nstochastic context is revealed by the environment. Following this, the learner\nthen selects a final action and receives a reward. Given $T$ rounds of\ninteractions with the environment, the objective of the learner is to learn a\npolicy (of selecting the initial and the final action) with maximum expected\nreward. In this paper we study the specific situation where every action\ncorresponds to intervening on a node in some known causal graph. We extend\nprior work from the deterministic context setting to obtain simple regret\nminimization guarantees. This is achieved through an instance-dependent causal\nparameter, $\\lambda$, which characterizes our upper bound. Furthermore, we\nprove that our simple regret is essentially tight for a large class of\ninstances. A key feature of our work is that we use convex optimization to\naddress the bandit exploration problem. We also conduct experiments to validate\nour theoretical results, and release our code at our project GitHub repository:\nhttps://github.com/adaptiveContextualCausalBandits/aCCB.\n", "date": "2024-05-28", "categories": []}
{"id": "2406.00317", "title": "Combining Experimental and Historical Data for Policy Evaluation", "abstract": "  This paper studies policy evaluation with multiple data sources, especially\nin scenarios that involve one experimental dataset with two arms, complemented\nby a historical dataset generated under a single control arm. We propose novel\ndata integration methods that linearly integrate base policy value estimators\nconstructed based on the experimental and historical data, with weights\noptimized to minimize the mean square error (MSE) of the resulting combined\nestimator. We further apply the pessimistic principle to obtain more robust\nestimators, and extend these developments to sequential decision making.\nTheoretically, we establish non-asymptotic error bounds for the MSEs of our\nproposed estimators, and derive their oracle, efficiency and robustness\nproperties across a broad spectrum of reward shift scenarios. Numerical\nexperiments and real-data-based analyses from a ridesharing company demonstrate\nthe superior performance of the proposed estimators.\n", "date": "2024-06-01", "categories": []}
{"id": "2406.00394", "title": "Learning Causal Abstractions of Linear Structural Causal Models", "abstract": "  The need for modelling causal knowledge at different levels of granularity\narises in several settings. Causal Abstraction provides a framework for\nformalizing this problem by relating two Structural Causal Models at different\nlevels of detail. Despite increasing interest in applying causal abstraction,\ne.g. in the interpretability of large machine learning models, the graphical\nand parametrical conditions under which a causal model can abstract another are\nnot known. Furthermore, learning causal abstractions from data is still an open\nproblem. In this work, we tackle both issues for linear causal models with\nlinear abstraction functions. First, we characterize how the low-level\ncoefficients and the abstraction function determine the high-level coefficients\nand how the high-level model constrains the causal ordering of low-level\nvariables. Then, we apply our theoretical results to learn high-level and\nlow-level causal models and their abstraction function from observational data.\nIn particular, we introduce Abs-LiNGAM, a method that leverages the constraints\ninduced by the learned high-level model and the abstraction function to speedup\nthe recovery of the larger low-level model, under the assumption of\nnon-Gaussian noise terms. In simulated settings, we show the effectiveness of\nlearning causal abstractions from data and the potential of our method in\nimproving scalability of causal discovery.\n", "date": "2024-06-01", "categories": []}
{"id": "2406.00983", "title": "Take its Essence, Discard its Dross! Debiasing for Toxic Language\n  Detection via Counterfactual Causal Effect", "abstract": "  Current methods of toxic language detection (TLD) typically rely on specific\ntokens to conduct decisions, which makes them suffer from lexical bias, leading\nto inferior performance and generalization. Lexical bias has both \"useful\" and\n\"misleading\" impacts on understanding toxicity. Unfortunately, instead of\ndistinguishing between these impacts, current debiasing methods typically\neliminate them indiscriminately, resulting in a degradation in the detection\naccuracy of the model. To this end, we propose a Counterfactual Causal\nDebiasing Framework (CCDF) to mitigate lexical bias in TLD. It preserves the\n\"useful impact\" of lexical bias and eliminates the \"misleading impact\".\nSpecifically, we first represent the total effect of the original sentence and\nbiased tokens on decisions from a causal view. We then conduct counterfactual\ninference to exclude the direct causal effect of lexical bias from the total\neffect. Empirical evaluations demonstrate that the debiased TLD model\nincorporating CCDF achieves state-of-the-art performance in both accuracy and\nfairness compared to competitive baselines applied on several vanilla models.\nThe generalization capability of our model outperforms current debiased models\nfor out-of-distribution data.\n", "date": "2024-06-03", "categories": []}
{"id": "2401.11130", "title": "Identification and Estimation of Conditional Average Partial Causal\n  Effects via Instrumental Variable", "abstract": "  There has been considerable recent interest in estimating heterogeneous\ncausal effects. In this paper, we study conditional average partial causal\neffects (CAPCE) to reveal the heterogeneity of causal effects with continuous\ntreatment. We provide conditions for identifying CAPCE in an instrumental\nvariable setting. Notably, CAPCE is identifiable under a weaker assumption than\nrequired by a commonly used measure for estimating heterogeneous causal effects\nof continuous treatment. We develop three families of CAPCE estimators: sieve,\nparametric, and reproducing kernel Hilbert space (RKHS)-based, and analyze\ntheir statistical properties. We illustrate the proposed CAPCE estimators on\nsynthetic and real-world data.\n", "date": "2024-01-20", "categories": []}
{"id": "2405.20487", "title": "Probabilities of Causation for Continuous and Vector Variables", "abstract": "  Probabilities of causation (PoC) are valuable concepts for explainable\nartificial intelligence and practical decision-making. PoC are originally\ndefined for scalar binary variables. In this paper, we extend the concept of\nPoC to continuous treatment and outcome variables, and further generalize PoC\nto capture causal effects between multiple treatments and multiple outcomes. In\naddition, we consider PoC for a sub-population and PoC with multi-hypothetical\nterms to capture more sophisticated counterfactual information useful for\ndecision-making. We provide a nonparametric identification theorem for each\ntype of PoC we introduce. Finally, we illustrate the application of our results\non a real-world dataset about education.\n", "date": "2024-05-30", "categories": []}
{"id": "2308.13047", "title": "Federated Causal Inference from Observational Data", "abstract": "  Decentralized data sources are prevalent in real-world applications, posing a\nformidable challenge for causal inference. These sources cannot be consolidated\ninto a single entity owing to privacy constraints. The presence of dissimilar\ndata distributions and missing values within them can potentially introduce\nbias to the causal estimands. In this article, we propose a framework to\nestimate causal effects from decentralized data sources. The proposed framework\navoid exchanging raw data among the sources, thus contributing towards\nprivacy-preserving causal learning. Three instances of the proposed framework\nare introduced to estimate causal effects across a wide range of diverse\nscenarios within a federated setting. (1) FedCI: a Bayesian framework based on\nGaussian processes for estimating causal effects from federated observational\ndata sources. It estimates the posterior distributions of the causal effects to\ncompute the higher-order statistics that capture the uncertainty. (2)\nCausalRFF: an adaptive transfer algorithm that learns the similarities among\nthe data sources by utilizing Random Fourier Features to disentangle the loss\nfunction into multiple components, each of which is associated with a data\nsource. It estimates the similarities among the sources through transfer\ncoefficients, and hence requiring no prior information about the similarity\nmeasures. (3) CausalFI: a new approach for federated causal inference from\nincomplete data, enabling the estimation of causal effects from multiple\ndecentralized and incomplete data sources. It accounts for the missing data\nunder the missing at random assumption, while also estimating higher-order\nstatistics of the causal estimands. The proposed federated framework and its\ninstances are an important step towards a privacy-preserving causal learning\nmodel.\n", "date": "2023-08-24", "categories": []}
{"id": "2401.14535", "title": "CaRiNG: Learning Temporal Causal Representation under Non-Invertible\n  Generation Process", "abstract": "  Identifying the underlying time-delayed latent causal processes in sequential\ndata is vital for grasping temporal dynamics and making downstream reasoning.\nWhile some recent methods can robustly identify these latent causal variables,\nthey rely on strict assumptions about the invertible generation process from\nlatent variables to observed data. However, these assumptions are often hard to\nsatisfy in real-world applications containing information loss. For instance,\nthe visual perception process translates a 3D space into 2D images, or the\nphenomenon of persistence of vision incorporates historical data into current\nperceptions. To address this challenge, we establish an identifiability theory\nthat allows for the recovery of independent latent components even when they\ncome from a nonlinear and non-invertible mix. Using this theory as a\nfoundation, we propose a principled approach, CaRiNG, to learn the CAusal\nRepresentatIon of Non-invertible Generative temporal data with identifiability\nguarantees. Specifically, we utilize temporal context to recover lost latent\ninformation and apply the conditions in our theory to guide the training\nprocess. Through experiments conducted on synthetic datasets, we validate that\nour CaRiNG method reliably identifies the causal process, even when the\ngeneration process is non-invertible. Moreover, we demonstrate that our\napproach considerably improves temporal understanding and reasoning in\npractical applications.\n", "date": "2024-01-25", "categories": []}
{"id": "2405.19463", "title": "Stochastic Optimization Algorithms for Instrumental Variable Regression\n  with Streaming Data", "abstract": "  We develop and analyze algorithms for instrumental variable regression by\nviewing the problem as a conditional stochastic optimization problem. In the\ncontext of least-squares instrumental variable regression, our algorithms\nneither require matrix inversions nor mini-batches and provides a fully online\napproach for performing instrumental variable regression with streaming data.\nWhen the true model is linear, we derive rates of convergence in expectation,\nthat are of order $\\mathcal{O}(\\log T/T)$ and $\\mathcal{O}(1/T^{1-\\iota})$ for\nany $\\iota>0$, respectively under the availability of two-sample and one-sample\noracles, respectively, where $T$ is the number of iterations. Importantly,\nunder the availability of the two-sample oracle, our procedure avoids\nexplicitly modeling and estimating the relationship between confounder and the\ninstrumental variables, demonstrating the benefit of the proposed approach over\nrecent works based on reformulating the problem as minimax optimization\nproblems. Numerical experiments are provided to corroborate the theoretical\nresults.\n", "date": "2024-05-29", "categories": []}
{"id": "2405.06902", "title": "Causal Inference from Slowly Varying Nonstationary Processes", "abstract": "  Causal inference from observational data following the restricted structural\ncausal models (SCM) framework hinges largely on the asymmetry between cause and\neffect from the data generating mechanisms, such as non-Gaussianity or\nnon-linearity. This methodology can be adapted to stationary time series, yet\ninferring causal relationships from nonstationary time series remains a\nchallenging task. In this work, we propose a new class of restricted SCM, via a\ntime-varying filter and stationary noise, and exploit the asymmetry from\nnonstationarity for causal identification in both bivariate and network\nsettings. We propose efficient procedures by leveraging powerful estimates of\nthe bivariate evolutionary spectra for slowly varying processes. Various\nsynthetic and real datasets that involve high-order and non-smooth filters are\nevaluated to demonstrate the effectiveness of our proposed methodology.\n", "date": "2024-05-11", "categories": []}
{"id": "2405.18836", "title": "Do Finetti: On Causal Effects for Exchangeable Data", "abstract": "  We study causal effect estimation in a setting where the data are not i.i.d.\n(independent and identically distributed). We focus on exchangeable data\nsatisfying an assumption of independent causal mechanisms. Traditional causal\neffect estimation frameworks, e.g., relying on structural causal models and\ndo-calculus, are typically limited to i.i.d. data and do not extend to more\ngeneral exchangeable generative processes, which naturally arise in\nmulti-environment data. To address this gap, we develop a generalized framework\nfor exchangeable data and introduce a truncated factorization formula that\nfacilitates both the identification and estimation of causal effects in our\nsetting. To illustrate potential applications, we introduce a causal P\\'olya\nurn model and demonstrate how intervention propagates effects in exchangeable\ndata settings. Finally, we develop an algorithm that performs simultaneous\ncausal discovery and effect estimation given multi-environment data.\n", "date": "2024-05-29", "categories": []}
{"id": "2405.18875", "title": "Counterfactual Metarules for Local and Global Recourse", "abstract": "  We introduce T-CREx, a novel model-agnostic method for local and global\ncounterfactual explanation (CE), which summarises recourse options for both\nindividuals and groups in the form of human-readable rules. It leverages\ntree-based surrogate models to learn the counterfactual rules, alongside\n'metarules' denoting their regions of optimality, providing both a global\nanalysis of model behaviour and diverse recourse options for users. Experiments\nindicate that T-CREx achieves superior aggregate performance over existing\nrule-based baselines on a range of CE desiderata, while being orders of\nmagnitude faster to run.\n", "date": "2024-05-29", "categories": []}
{"id": "2306.02931", "title": "Bivariate Causal Discovery using Bayesian Model Selection", "abstract": "  Much of the causal discovery literature prioritises guaranteeing the\nidentifiability of causal direction in statistical models. For structures\nwithin a Markov equivalence class, this requires strong assumptions which may\nnot hold in real-world datasets, ultimately limiting the usability of these\nmethods. Building on previous attempts, we show how to incorporate causal\nassumptions within the Bayesian framework. Identifying causal direction then\nbecomes a Bayesian model selection problem. This enables us to construct models\nwith realistic assumptions, and consequently allows for the differentiation\nbetween Markov equivalent causal structures. We analyse why Bayesian model\nselection works in situations where methods based on maximum likelihood fail.\nTo demonstrate our approach, we construct a Bayesian non-parametric model that\ncan flexibly model the joint distribution. We then outperform previous methods\non a wide range of benchmark datasets with varying data generating assumptions.\n", "date": "2023-06-05", "categories": []}
{"id": "2311.02790", "title": "CausalCite: A Causal Formulation of Paper Citations", "abstract": "  Citation count of a paper is a commonly used proxy for evaluating the\nsignificance of a paper in the scientific community. Yet citation measures are\nwidely criticized for failing to accurately reflect the true impact of a paper.\nThus, we propose CausalCite, a new way to measure the significance of a paper\nby assessing the causal impact of the paper on its follow-up papers. CausalCite\nis based on a novel causal inference method, TextMatch, which adapts the\ntraditional matching framework to high-dimensional text embeddings. TextMatch\nencodes each paper using text embeddings from large language models (LLMs),\nextracts similar samples by cosine similarity, and synthesizes a counterfactual\nsample as the weighted average of similar papers according to their similarity\nvalues. We demonstrate the effectiveness of CausalCite on various criteria,\nsuch as high correlation with paper impact as reported by scientific experts on\na previous dataset of 1K papers, (test-of-time) awards for past papers, and its\nstability across various subfields of AI. We also provide a set of findings\nthat can serve as suggested ways for future researchers to use our metric for a\nbetter understanding of the quality of a paper. Our code is available at\nhttps://github.com/causalNLP/causal-cite.\n", "date": "2023-11-05", "categories": []}
{"id": "2402.17042", "title": "Towards Generalizing Inferences from Trials to Target Populations", "abstract": "  Randomized Controlled Trials (RCTs) are pivotal in generating internally\nvalid estimates with minimal assumptions, serving as a cornerstone for\nresearchers dedicated to advancing causal inference methods. However, extending\nthese findings beyond the experimental cohort to achieve externally valid\nestimates is crucial for broader scientific inquiry. This paper delves into the\nforefront of addressing these external validity challenges, encapsulating the\nessence of a multidisciplinary workshop held at the Institute for Computational\nand Experimental Research in Mathematics (ICERM), Brown University, in Fall\n2023. The workshop congregated experts from diverse fields including social\nscience, medicine, public health, statistics, computer science, and education,\nto tackle the unique obstacles each discipline faces in extrapolating\nexperimental findings. Our study presents three key contributions: we integrate\nongoing efforts, highlighting methodological synergies across fields; provide\nan exhaustive review of generalizability and transportability based on the\nworkshop's discourse; and identify persistent hurdles while suggesting avenues\nfor future research. By doing so, this paper aims to enhance the collective\nunderstanding of the generalizability and transportability of causal effects,\nfostering cross-disciplinary collaboration and offering valuable insights for\nresearchers working on refining and applying causal inference methods.\n", "date": "2024-02-26", "categories": []}
{"id": "2404.14786", "title": "RealTCD: Temporal Causal Discovery from Interventional Data with Large\n  Language Model", "abstract": "  In the field of Artificial Intelligence for Information Technology\nOperations, causal discovery is pivotal for operation and maintenance of graph\nconstruction, facilitating downstream industrial tasks such as root cause\nanalysis. Temporal causal discovery, as an emerging method, aims to identify\ntemporal causal relationships between variables directly from observations by\nutilizing interventional data. However, existing methods mainly focus on\nsynthetic datasets with heavy reliance on intervention targets and ignore the\ntextual information hidden in real-world systems, failing to conduct causal\ndiscovery for real industrial scenarios. To tackle this problem, in this paper\nwe propose to investigate temporal causal discovery in industrial scenarios,\nwhich faces two critical challenges: 1) how to discover causal relationships\nwithout the interventional targets that are costly to obtain in practice, and\n2) how to discover causal relations via leveraging the textual information in\nsystems which can be complex yet abundant in industrial contexts. To address\nthese challenges, we propose the RealTCD framework, which is able to leverage\ndomain knowledge to discover temporal causal relationships without\ninterventional targets. Specifically, we first develop a score-based temporal\ncausal discovery method capable of discovering causal relations for root cause\nanalysis without relying on interventional targets through strategic masking\nand regularization. Furthermore, by employing Large Language Models (LLMs) to\nhandle texts and integrate domain knowledge, we introduce LLM-guided\nmeta-initialization to extract the meta-knowledge from textual information\nhidden in systems to boost the quality of discovery. We conduct extensive\nexperiments on simulation and real-world datasets to show the superiority of\nour proposed RealTCD framework over existing baselines in discovering temporal\ncausal structures.\n", "date": "2024-04-23", "categories": []}
{"id": "2405.16130", "title": "Automating the Selection of Proxy Variables of Unmeasured Confounders", "abstract": "  Recently, interest has grown in the use of proxy variables of unobserved\nconfounding for inferring the causal effect in the presence of unmeasured\nconfounders from observational data. One difficulty inhibiting the practical\nuse is finding valid proxy variables of unobserved confounding to a target\ncausal effect of interest. These proxy variables are typically justified by\nbackground knowledge. In this paper, we investigate the estimation of causal\neffects among multiple treatments and a single outcome, all of which are\naffected by unmeasured confounders, within a linear causal model, without prior\nknowledge of the validity of proxy variables. To be more specific, we first\nextend the existing proxy variable estimator, originally addressing a single\nunmeasured confounder, to accommodate scenarios where multiple unmeasured\nconfounders exist between the treatments and the outcome. Subsequently, we\npresent two different sets of precise identifiability conditions for selecting\nvalid proxy variables of unmeasured confounders, based on the second-order\nstatistics and higher-order statistics of the data, respectively. Moreover, we\npropose two data-driven methods for the selection of proxy variables and for\nthe unbiased estimation of causal effects. Theoretical analysis demonstrates\nthe correctness of our proposed algorithms. Experimental results on both\nsynthetic and real-world data show the effectiveness of the proposed approach.\n", "date": "2024-05-25", "categories": []}
{"id": "2405.16718", "title": "Amortized Active Causal Induction with Deep Reinforcement Learning", "abstract": "  We present Causal Amortized Active Structure Learning (CAASL), an active\nintervention design policy that can select interventions that are adaptive,\nreal-time and that does not require access to the likelihood. This policy, an\namortized network based on the transformer, is trained with reinforcement\nlearning on a simulator of the design environment, and a reward function that\nmeasures how close the true causal graph is to a causal graph posterior\ninferred from the gathered data. On synthetic data and a single-cell gene\nexpression simulator, we demonstrate empirically that the data acquired through\nour policy results in a better estimate of the underlying causal graph than\nalternative strategies. Our design policy successfully achieves amortized\nintervention design on the distribution of the training environment while also\ngeneralizing well to distribution shifts in test-time design environments.\nFurther, our policy also demonstrates excellent zero-shot generalization to\ndesign environments with dimensionality higher than that during training, and\nto intervention types that it has not been trained on.\n", "date": "2024-05-26", "categories": []}
{"id": "2311.07259", "title": "Towards Bounding Causal Effects under Markov Equivalence", "abstract": "  Predicting the effect of unseen interventions is a fundamental research\nquestion across the data sciences. It is well established that in general such\nquestions cannot be answered definitively from observational data. This\nrealization has fuelled a growing literature introducing various identifying\nassumptions, for example in the form of a causal diagram among relevant\nvariables. In practice, this paradigm is still too rigid for many practical\napplications as it is generally not possible to confidently delineate the true\ncausal diagram. In this paper, we consider the derivation of bounds on causal\neffects given only observational data. We propose to take as input a less\ninformative structure known as a Partial Ancestral Graph, which represents a\nMarkov equivalence class of causal diagrams and is learnable from data. In this\nmore ``data-driven'' setting, we provide a systematic algorithm to derive\nbounds on causal effects that exploit the invariant properties of the\nequivalence class, and that can be computed analytically. We demonstrate our\nmethod with synthetic and real data examples.\n", "date": "2023-11-13", "categories": []}
{"id": "2402.03614", "title": "Bayesian Vector AutoRegression with Factorised Granger-Causal Graphs", "abstract": "  We study the problem of automatically discovering Granger causal relations\nfrom observational multivariate time-series data.Vector autoregressive (VAR)\nmodels have been time-tested for this problem, including Bayesian variants and\nmore recent developments using deep neural networks. Most existing VAR methods\nfor Granger causality use sparsity-inducing penalties/priors or post-hoc\nthresholds to interpret their coefficients as Granger causal graphs. Instead,\nwe propose a new Bayesian VAR model with a hierarchical factorised prior\ndistribution over binary Granger causal graphs, separately from the VAR\ncoefficients. We develop an efficient algorithm to infer the posterior over\nbinary Granger causal graphs. Comprehensive experiments on synthetic,\nsemi-synthetic, and climate data show that our method is more uncertainty\naware, has less hyperparameters, and achieves better performance than competing\napproaches, especially in low-data regimes where there are less observations.\n", "date": "2024-02-06", "categories": []}
{"id": "2405.15358", "title": "Coordinated Multi-Neighborhood Learning on a Directed Acyclic Graph", "abstract": "  Learning the structure of causal directed acyclic graphs (DAGs) is useful in\nmany areas of machine learning and artificial intelligence, with wide\napplications. However, in the high-dimensional setting, it is challenging to\nobtain good empirical and theoretical results without strong and often\nrestrictive assumptions. Additionally, it is questionable whether all of the\nvariables purported to be included in the network are observable. It is of\ninterest then to restrict consideration to a subset of the variables for\nrelevant and reliable inferences. In fact, researchers in various disciplines\ncan usually select a set of target nodes in the network for causal discovery.\nThis paper develops a new constraint-based method for estimating the local\nstructure around multiple user-specified target nodes, enabling coordination in\nstructure learning between neighborhoods. Our method facilitates causal\ndiscovery without learning the entire DAG structure. We establish consistency\nresults for our algorithm with respect to the local neighborhood structure of\nthe target nodes in the true graph. Experimental results on synthetic and\nreal-world data show that our algorithm is more accurate in learning the\nneighborhood structures with much less computational cost than standard methods\nthat estimate the entire DAG. An R package implementing our methods may be\naccessed at https://github.com/stephenvsmith/CML.\n", "date": "2024-05-24", "categories": []}
{"id": "2405.15505", "title": "Revisiting Counterfactual Regression through the Lens of\n  Gromov-Wasserstein Information Bottleneck", "abstract": "  As a promising individualized treatment effect (ITE) estimation method,\ncounterfactual regression (CFR) maps individuals' covariates to a latent space\nand predicts their counterfactual outcomes. However, the selection bias between\ncontrol and treatment groups often imbalances the two groups' latent\ndistributions and negatively impacts this method's performance. In this study,\nwe revisit counterfactual regression through the lens of information bottleneck\nand propose a novel learning paradigm called Gromov-Wasserstein information\nbottleneck (GWIB). In this paradigm, we learn CFR by maximizing the mutual\ninformation between covariates' latent representations and outcomes while\npenalizing the kernelized mutual information between the latent representations\nand the covariates. We demonstrate that the upper bound of the penalty term can\nbe implemented as a new regularizer consisting of $i)$ the fused\nGromov-Wasserstein distance between the latent representations of different\ngroups and $ii)$ the gap between the transport cost generated by the model and\nthe cross-group Gromov-Wasserstein distance between the latent representations\nand the covariates. GWIB effectively learns the CFR model through alternating\noptimization, suppressing selection bias while avoiding trivial latent\ndistributions. Experiments on ITE estimation tasks show that GWIB consistently\noutperforms state-of-the-art CFR methods. To promote the research community, we\nrelease our project at https://github.com/peteryang1031/Causal-GWIB.\n", "date": "2024-05-24", "categories": []}
{"id": "2209.08860", "title": "A Survey of Deep Causal Models and Their Industrial Applications", "abstract": "  The notion of causality assumes a paramount position within the realm of\nhuman cognition. Over the past few decades, there has been significant\nadvancement in the domain of causal effect estimation across various\ndisciplines, including but not limited to computer science, medicine,\neconomics, and industrial applications. Given the continous advancements in\ndeep learning methodologies, there has been a notable surge in its utilization\nfor the estimation of causal effects using counterfactual data. Typically, deep\ncausal models map the characteristics of covariates to a representation space\nand then design various objective functions to estimate counterfactual data\nunbiasedly. Different from the existing surveys on causal models in machine\nlearning, this review mainly focuses on the overview of the deep causal models\nbased on neural networks, and its core contributions are as follows: 1) we cast\ninsight on a comprehensive overview of deep causal models from both timeline of\ndevelopment and method classification perspectives; 2) we outline some typical\napplications of causal effect estimation to industry; 3) we also endeavor to\npresent a detailed categorization and analysis on relevant datasets, source\ncodes and experiments.\n", "date": "2022-09-19", "categories": []}
{"id": "2306.11697", "title": "Treatment Effects in Extreme Regimes", "abstract": "  Understanding treatment effects in extreme regimes is important for\ncharacterizing risks associated with different interventions. This is hindered\nby the unavailability of counterfactual outcomes and the rarity and difficulty\nof collecting extreme data in practice. To address this issue, we propose a new\nframework based on extreme value theory for estimating treatment effects in\nextreme regimes. We quantify these effects using variations in tail decay rates\nof potential outcomes in the presence and absence of treatments. We establish\nalgorithms for calculating these quantities and develop related theoretical\nresults. We demonstrate the efficacy of our approach on various standard\nsynthetic and semi-synthetic datasets.\n", "date": "2023-06-20", "categories": []}
{"id": "2310.11011", "title": "From Identifiable Causal Representations to Controllable Counterfactual\n  Generation: A Survey on Causal Generative Modeling", "abstract": "  Deep generative models have shown tremendous capability in data density\nestimation and data generation from finite samples. While these models have\nshown impressive performance by learning correlations among features in the\ndata, some fundamental shortcomings are their lack of explainability, tendency\nto induce spurious correlations, and poor out-of-distribution extrapolation. To\nremedy such challenges, recent work has proposed a shift toward causal\ngenerative models. Causal models offer several beneficial properties to deep\ngenerative models, such as distribution shift robustness, fairness, and\ninterpretability. Structural causal models (SCMs) describe data-generating\nprocesses and model complex causal relationships and mechanisms among variables\nin a system. Thus, SCMs can naturally be combined with deep generative models.\nWe provide a technical survey on causal generative modeling categorized into\ncausal representation learning and controllable counterfactual generation\nmethods. We focus on fundamental theory, methodology, drawbacks, datasets, and\nmetrics. Then, we cover applications of causal generative models in fairness,\nprivacy, out-of-distribution generalization, precision medicine, and biological\nsciences. Lastly, we discuss open problems and fruitful research directions for\nfuture work in the field.\n", "date": "2023-10-17", "categories": []}
{"id": "2310.11287", "title": "Assessing the Causal Impact of Humanitarian Aid on Food Security", "abstract": "  In the face of climate change-induced droughts, vulnerable regions encounter\nsevere threats to food security, demanding urgent humanitarian assistance. This\npaper introduces a causal inference framework for the Horn of Africa, aiming to\nassess the impact of cash-based interventions on food crises. Our contributions\ninclude identifying causal relationships within the food security system,\nharmonizing a comprehensive database including socio-economic, weather and\nremote sensing data, and estimating the causal effect of humanitarian\ninterventions on malnutrition. On a country level, our results revealed no\nsignificant effects, likely due to limited sample size, suboptimal data\nquality, and an imperfect causal graph resulting from our limited understanding\nof multidisciplinary systems like food security. Instead, on a district level,\nresults revealed significant effects, further implying the context-specific\nnature of the system. This underscores the need to enhance data collection and\nrefine causal models with domain experts for more effective future\ninterventions and policies, improving transparency and accountability in\nhumanitarian aid.\n", "date": "2023-10-17", "categories": []}
{"id": "2402.00168", "title": "Continuous Treatment Effects with Surrogate Outcomes", "abstract": "  In many real-world causal inference applications, the primary outcomes\n(labels) are often partially missing, especially if they are expensive or\ndifficult to collect. If the missingness depends on covariates (i.e.,\nmissingness is not completely at random), analyses based on fully observed\nsamples alone may be biased. Incorporating surrogates, which are fully observed\npost-treatment variables related to the primary outcome, can improve estimation\nin this case. In this paper, we study the role of surrogates in estimating\ncontinuous treatment effects and propose a doubly robust method to efficiently\nincorporate surrogates in the analysis, which uses both labeled and unlabeled\ndata and does not suffer from the above selection bias problem. Importantly, we\nestablish the asymptotic normality of the proposed estimator and show possible\nimprovements on the variance compared with methods that solely use labeled\ndata. Extensive simulations show our methods enjoy appealing empirical\nperformance.\n", "date": "2024-01-31", "categories": []}
{"id": "2403.04919", "title": "Identifying Causal Effects Under Functional Dependencies", "abstract": "  We study the identification of causal effects, motivated by two improvements\nto identifiability which can be attained if one knows that some variables in a\ncausal graph are functionally determined by their parents (without needing to\nknow the specific functions). First, an unidentifiable causal effect may become\nidentifiable when certain variables are functional. Second, certain functional\nvariables can be excluded from being observed without affecting the\nidentifiability of a causal effect, which may significantly reduce the number\nof needed variables in observational data. Our results are largely based on an\nelimination procedure which removes functional variables from a causal graph\nwhile preserving key properties in the resulting causal graph, including the\nidentifiability of causal effects.\n", "date": "2024-03-07", "categories": []}
{"id": "2405.13092", "title": "CausalPlayground: Addressing Data-Generation Requirements in\n  Cutting-Edge Causality Research", "abstract": "  Research on causal effects often relies on synthetic data due to the scarcity\nof real-world datasets with ground-truth effects. Since current data-generating\ntools do not always meet all requirements for state-of-the-art research, ad-hoc\nmethods are often employed. This leads to heterogeneity among datasets and\ndelays research progress. We address the shortcomings of current\ndata-generating libraries by introducing CausalPlayground, a Python library\nthat provides a standardized platform for generating, sampling, and sharing\nstructural causal models (SCMs). CausalPlayground offers fine-grained control\nover SCMs, interventions, and the generation of datasets of SCMs for learning\nand quantitative research. Furthermore, by integrating with Gymnasium, the\nstandard framework for reinforcement learning (RL) environments, we enable\nonline interaction with the SCMs. Overall, by introducing CausalPlayground we\naim to foster more efficient and comparable research in the field. All code and\nAPI documentation is available at https://github.com/sa-and/CausalPlayground.\n", "date": "2024-05-21", "categories": []}
{"id": "2405.13100", "title": "Better Simulations for Validating Causal Discovery with the\n  DAG-Adaptation of the Onion Method", "abstract": "  The number of artificial intelligence algorithms for learning causal models\nfrom data is growing rapidly. Most ``causal discovery'' or ``causal structure\nlearning'' algorithms are primarily validated through simulation studies.\nHowever, no widely accepted simulation standards exist and publications often\nreport conflicting performance statistics -- even when only considering\npublications that simulate data from linear models. In response, several\nmanuscripts have criticized a popular simulation design for validating\nalgorithms in the linear case.\n  We propose a new simulation design for generating linear models for directed\nacyclic graphs (DAGs): the DAG-adaptation of the Onion (DaO) method. DaO\nsimulations are fundamentally different from existing simulations because they\nprioritize the distribution of correlation matrices rather than the\ndistribution of linear effects. Specifically, the DaO method uniformly samples\nthe space of all correlation matrices consistent with (i.e. Markov to) a DAG.\nWe also discuss how to sample DAGs and present methods for generating DAGs with\nscale-free in-degree or out-degree. We compare the DaO method against two\nalternative simulation designs and provide implementations of the DaO method in\nPython and R: https://github.com/bja43/DaO_simulation. We advocate for others\nto adopt DaO simulations as a fair universal benchmark.\n", "date": "2024-05-21", "categories": []}
{"id": "2405.13551", "title": "Large Language Models are Effective Priors for Causal Graph Discovery", "abstract": "  Causal structure discovery from observations can be improved by integrating\nbackground knowledge provided by an expert to reduce the hypothesis space.\nRecently, Large Language Models (LLMs) have begun to be considered as sources\nof prior information given the low cost of querying them relative to a human\nexpert. In this work, firstly, we propose a set of metrics for assessing LLM\njudgments for causal graph discovery independently of the downstream algorithm.\nSecondly, we systematically study a set of prompting designs that allows the\nmodel to specify priors about the structure of the causal graph. Finally, we\npresent a general methodology for the integration of LLM priors in graph\ndiscovery algorithms, finding that they help improve performance on\ncommon-sense benchmarks and especially when used for assessing edge\ndirectionality. Our work highlights the potential as well as the shortcomings\nof the use of LLMs in this problem space.\n", "date": "2024-05-22", "categories": []}
{"id": "2405.14728", "title": "Intervention and Conditioning in Causal Bayesian Networks", "abstract": "  Causal models are crucial for understanding complex systems and identifying\ncausal relationships among variables. Even though causal models are extremely\npopular, conditional probability calculation of formulas involving\ninterventions pose significant challenges. In case of Causal Bayesian Networks\n(CBNs), Pearl assumes autonomy of mechanisms that determine interventions to\ncalculate a range of probabilities. We show that by making simple yet often\nrealistic independence assumptions, it is possible to uniquely estimate the\nprobability of an interventional formula (including the well-studied notions of\nprobability of sufficiency and necessity). We discuss when these assumptions\nare appropriate. Importantly, in many cases of interest, when the assumptions\nare appropriate, these probability estimates can be evaluated using\nobservational data, which carries immense significance in scenarios where\nconducting experiments is impractical or unfeasible.\n", "date": "2024-05-23", "categories": []}
{"id": "2405.14743", "title": "Iterative Causal Segmentation: Filling the Gap between Market\n  Segmentation and Marketing Strategy", "abstract": "  The field of causal Machine Learning (ML) has made significant strides in\nrecent years. Notable breakthroughs include methods such as meta learners\n(arXiv:1706.03461v6) and heterogeneous doubly robust estimators\n(arXiv:2004.14497) introduced in the last five years. Despite these\nadvancements, the field still faces challenges, particularly in managing\ntightly coupled systems where both the causal treatment variable and a\nconfounding covariate must serve as key decision-making indicators. This\nscenario is common in applications of causal ML for marketing, such as\nmarketing segmentation and incremental marketing uplift. In this work, we\npresent our formally proven algorithm, iterative causal segmentation, to\naddress this issue.\n", "date": "2024-05-23", "categories": []}
{"id": "2307.03687", "title": "Leveraging text data for causal inference using electronic health\n  records", "abstract": "  In studies that rely on data from electronic health records (EHRs),\nunstructured text data such as clinical progress notes offer a rich source of\ninformation about patient characteristics and care that may be missing from\nstructured data. Despite the prevalence of text in clinical research, these\ndata are often ignored for the purposes of quantitative analysis due their\ncomplexity. This paper presents a unified framework for leveraging text data to\nsupport causal inference with electronic health data at multiple stages of\nanalysis. In particular, we consider how natural language processing and\nstatistical text analysis can be combined with standard inferential techniques\nto address common challenges due to missing data, confounding bias, and\ntreatment effect heterogeneity. Through an application to a recent EHR study\ninvestigating the effects of a non-randomized medical intervention on patient\noutcomes, we show how incorporating text data in a traditional matching\nanalysis can help strengthen the validity of an estimated treatment effect and\nidentify patient subgroups that may benefit most from treatment. We believe\nthese methods have the potential to expand the scope of secondary analysis of\nclinical data to domains where structured EHR data is limited, such as in\ndeveloping countries. To this end, we provide code and open-source replication\nmaterials to encourage adoption and broader exploration of these techniques in\nclinical research.\n", "date": "2023-06-09", "categories": []}
{"id": "2405.12340", "title": "Cascade-based Randomization for Inferring Causal Effects under Diffusion\n  Interference", "abstract": "  The presence of interference, where the outcome of an individual may depend\non the treatment assignment and behavior of neighboring nodes, can lead to\nbiased causal effect estimation. Current approaches to network experiment\ndesign focus on limiting interference through cluster-based randomization, in\nwhich clusters are identified using graph clustering, and cluster randomization\ndictates the node assignment to treatment and control. However, cluster-based\nrandomization approaches perform poorly when interference propagates in\ncascades, whereby the response of individuals to treatment propagates to their\nmulti-hop neighbors. When we have knowledge of the cascade seed nodes, we can\nleverage this interference structure to mitigate the resulting causal effect\nestimation bias. With this goal, we propose a cascade-based network experiment\ndesign that initiates treatment assignment from the cascade seed node and\npropagates the assignment to their multi-hop neighbors to limit interference\nduring cascade growth and thereby reduce the overall causal effect estimation\nerror. Our extensive experiments on real-world and synthetic datasets\ndemonstrate that our proposed framework outperforms the existing\nstate-of-the-art approaches in estimating causal effects in network data.\n", "date": "2024-05-20", "categories": []}
{"id": "2405.12387", "title": "Conformal Counterfactual Inference under Hidden Confounding", "abstract": "  Personalized decision making requires the knowledge of potential outcomes\nunder different treatments, and confidence intervals about the potential\noutcomes further enrich this decision-making process and improve its\nreliability in high-stakes scenarios. Predicting potential outcomes along with\nits uncertainty in a counterfactual world poses the foundamental challenge in\ncausal inference. Existing methods that construct confidence intervals for\ncounterfactuals either rely on the assumption of strong ignorability, or need\naccess to un-identifiable lower and upper bounds that characterize the\ndifference between observational and interventional distributions. To overcome\nthese limitations, we first propose a novel approach wTCP-DR based on\ntransductive weighted conformal prediction, which provides confidence intervals\nfor counterfactual outcomes with marginal converage guarantees, even under\nhidden confounding. With less restrictive assumptions, our approach requires\naccess to a fraction of interventional data (from randomized controlled trials)\nto account for the covariate shift from observational distributoin to\ninterventional distribution. Theoretical results explicitly demonstrate the\nconditions under which our algorithm is strictly advantageous to the naive\nmethod that only uses interventional data. After ensuring valid intervals on\ncounterfactuals, it is straightforward to construct intervals for individual\ntreatment effects (ITEs). We demonstrate our method across synthetic and\nreal-world data, including recommendation systems, to verify the superiority of\nour methods compared against state-of-the-art baselines in terms of both\ncoverage and efficiency\n", "date": "2024-05-20", "categories": []}
{"id": "2405.12615", "title": "Learning Causal Dynamics Models in Object-Oriented Environments", "abstract": "  Causal dynamics models (CDMs) have demonstrated significant potential in\naddressing various challenges in reinforcement learning. To learn CDMs, recent\nstudies have performed causal discovery to capture the causal dependencies\namong environmental variables. However, the learning of CDMs is still confined\nto small-scale environments due to computational complexity and sample\nefficiency constraints. This paper aims to extend CDMs to large-scale\nobject-oriented environments, which consist of a multitude of objects\nclassified into different categories. We introduce the Object-Oriented CDM\n(OOCDM) that shares causalities and parameters among objects belonging to the\nsame class. Furthermore, we propose a learning method for OOCDM that enables it\nto adapt to a varying number of objects. Experiments on large-scale tasks\nindicate that OOCDM outperforms existing CDMs in terms of causal discovery,\nprediction accuracy, generalization, and computational efficiency.\n", "date": "2024-05-21", "categories": []}
{"id": "2405.03069", "title": "On Probabilistic and Causal Reasoning with Summation Operators", "abstract": "  Ibeling et al. (2023). axiomatize increasingly expressive languages of\ncausation and probability, and Mosse et al. (2024) show that reasoning\n(specifically the satisfiability problem) in each causal language is as\ndifficult, from a computational complexity perspective, as reasoning in its\nmerely probabilistic or \"correlational\" counterpart. Introducing a summation\noperator to capture common devices that appear in applications -- such as the\n$do$-calculus of Pearl (2009) for causal inference, which makes ample use of\nmarginalization -- van der Zander et al. (2023) partially extend these earlier\ncomplexity results to causal and probabilistic languages with marginalization.\nWe complete this extension, fully characterizing the complexity of\nprobabilistic and causal reasoning with summation, demonstrating that these\nagain remain equally difficult. Surprisingly, allowing free variables for\nrandom variable values results in a system that is undecidable, so long as the\nranges of these random variables are unrestricted. We finally axiomatize these\nlanguages featuring marginalization (or more generally summation), resolving\nopen questions posed by Ibeling et al. (2023).\n", "date": "2024-05-05", "categories": []}
{"id": "2405.11377", "title": "Causal Customer Churn Analysis with Low-rank Tensor Block Hazard Model", "abstract": "  This study introduces an innovative method for analyzing the impact of\nvarious interventions on customer churn, using the potential outcomes\nframework. We present a new causal model, the tensorized latent factor block\nhazard model, which incorporates tensor completion methods for a principled\ncausal analysis of customer churn. A crucial element of our approach is the\nformulation of a 1-bit tensor completion for the parameter tensor. This\ncaptures hidden customer characteristics and temporal elements from churn\nrecords, effectively addressing the binary nature of churn data and its\ntime-monotonic trends. Our model also uniquely categorizes interventions by\ntheir similar impacts, enhancing the precision and practicality of implementing\ncustomer retention strategies. For computational efficiency, we apply a\nprojected gradient descent algorithm combined with spectral clustering. We lay\ndown the theoretical groundwork for our model, including its non-asymptotic\nproperties. The efficacy and superiority of our model are further validated\nthrough comprehensive experiments on both simulated and real-world\napplications.\n", "date": "2024-05-18", "categories": []}
{"id": "2405.11633", "title": "Geometry-Aware Instrumental Variable Regression", "abstract": "  Instrumental variable (IV) regression can be approached through its\nformulation in terms of conditional moment restrictions (CMR). Building on\nvariants of the generalized method of moments, most CMR estimators are\nimplicitly based on approximating the population data distribution via\nreweightings of the empirical sample. While for large sample sizes, in the\nindependent identically distributed (IID) setting, reweightings can provide\nsufficient flexibility, they might fail to capture the relevant information in\npresence of corrupted data or data prone to adversarial attacks. To address\nthese shortcomings, we propose the Sinkhorn Method of Moments, an optimal\ntransport-based IV estimator that takes into account the geometry of the data\nmanifold through data-derivative information. We provide a simple plug-and-play\nimplementation of our method that performs on par with related estimators in\nstandard settings but improves robustness against data corruption and\nadversarial attacks.\n", "date": "2024-05-19", "categories": []}
{"id": "2405.09190", "title": "Advancing Explainable AI with Causal Analysis in Large-Scale Fuzzy\n  Cognitive Maps", "abstract": "  In the quest for accurate and interpretable AI models, eXplainable AI (XAI)\nhas become crucial. Fuzzy Cognitive Maps (FCMs) stand out as an advanced XAI\nmethod because of their ability to synergistically combine and exploit both\nexpert knowledge and data-driven insights, providing transparency and intrinsic\ninterpretability. This letter introduces and investigates the \"Total Causal\nEffect Calculation for FCMs\" (TCEC-FCM) algorithm, an innovative approach that,\nfor the first time, enables the efficient calculation of total causal effects\namong concepts in large-scale FCMs by leveraging binary search and graph\ntraversal techniques, thereby overcoming the challenge of exhaustive causal\npath exploration that hinder existing methods. We evaluate the proposed method\nacross various synthetic FCMs that demonstrate TCEC-FCM's superior performance\nover exhaustive methods, marking a significant advancement in causal effect\nanalysis within FCMs, thus broadening their usability for modern complex XAI\napplications.\n", "date": "2024-05-15", "categories": []}
{"id": "2405.09288", "title": "DeCoDEx: Confounder Detector Guidance for Improved Diffusion-based\n  Counterfactual Explanations", "abstract": "  Deep learning classifiers are prone to latching onto dominant confounders\npresent in a dataset rather than on the causal markers associated with the\ntarget class, leading to poor generalization and biased predictions. Although\nexplainability via counterfactual image generation has been successful at\nexposing the problem, bias mitigation strategies that permit accurate\nexplainability in the presence of dominant and diverse artifacts remain\nunsolved. In this work, we propose the DeCoDEx framework and show how an\nexternal, pre-trained binary artifact detector can be leveraged during\ninference to guide a diffusion-based counterfactual image generator towards\naccurate explainability. Experiments on the CheXpert dataset, using both\nsynthetic artifacts and real visual artifacts (support devices), show that the\nproposed method successfully synthesizes the counterfactual images that change\nthe causal pathology markers associated with Pleural Effusion while preserving\nor ignoring the visual artifacts. Augmentation of ERM and Group-DRO classifiers\nwith the DeCoDEx generated images substantially improves the results across\nunderrepresented groups that are out of distribution for each class. The code\nis made publicly available at https://github.com/NimaFathi/DeCoDEx.\n", "date": "2024-05-15", "categories": []}
{"id": "2405.09516", "title": "Generalization Bounds for Causal Regression: Insights, Guarantees and\n  Sensitivity Analysis", "abstract": "  Many algorithms have been recently proposed for causal machine learning. Yet,\nthere is little to no theory on their quality, especially considering finite\nsamples. In this work, we propose a theory based on generalization bounds that\nprovides such guarantees. By introducing a novel change-of-measure inequality,\nwe are able to tightly bound the model loss in terms of the deviation of the\ntreatment propensities over the population, which we show can be empirically\nlimited. Our theory is fully rigorous and holds even in the face of hidden\nconfounding and violations of positivity. We demonstrate our bounds on\nsemi-synthetic and real data, showcasing their remarkable tightness and\npractical utility.\n", "date": "2024-05-15", "categories": []}
{"id": "2405.08699", "title": "Weakly-supervised causal discovery based on fuzzy knowledge and complex\n  data complementarity", "abstract": "  Causal discovery based on observational data is important for deciphering the\ncausal mechanism behind complex systems. However, the effectiveness of existing\ncausal discovery methods is limited due to inferior prior knowledge, domain\ninconsistencies, and the challenges of high-dimensional datasets with small\nsample sizes. To address this gap, we propose a novel weakly-supervised fuzzy\nknowledge and data co-driven causal discovery method named KEEL. KEEL adopts a\nfuzzy causal knowledge schema to encapsulate diverse types of fuzzy knowledge,\nand forms corresponding weakened constraints. This schema not only lessens the\ndependency on expertise but also allows various types of limited and\nerror-prone fuzzy knowledge to guide causal discovery. It can enhance the\ngeneralization and robustness of causal discovery, especially in\nhigh-dimensional and small-sample scenarios. In addition, we integrate the\nextended linear causal model (ELCM) into KEEL for dealing with the\nmulti-distribution and incomplete data. Extensive experiments with different\ndatasets demonstrate the superiority of KEEL over several state-of-the-art\nmethods in accuracy, robustness and computational efficiency. For causal\ndiscovery in real protein signal transduction processes, KEEL outperforms the\nbenchmark method with limited data. In summary, KEEL is effective to tackle the\ncausal discovery tasks with higher accuracy while alleviating the requirement\nfor extensive domain expertise.\n", "date": "2024-05-14", "categories": []}
{"id": "2405.08779", "title": "Jacobian Regularizer-based Neural Granger Causality", "abstract": "  With the advancement of neural networks, diverse methods for neural Granger\ncausality have emerged, which demonstrate proficiency in handling complex data,\nand nonlinear relationships. However, the existing framework of neural Granger\ncausality has several limitations. It requires the construction of separate\npredictive models for each target variable, and the relationship depends on the\nsparsity on the weights of the first layer, resulting in challenges in\neffectively modeling complex relationships between variables as well as\nunsatisfied estimation accuracy of Granger causality. Moreover, most of them\ncannot grasp full-time Granger causality. To address these drawbacks, we\npropose a Jacobian Regularizer-based Neural Granger Causality (JRNGC) approach,\na straightforward yet highly effective method for learning multivariate summary\nGranger causality and full-time Granger causality by constructing a single\nmodel for all target variables. Specifically, our method eliminates the\nsparsity constraints of weights by leveraging an input-output Jacobian matrix\nregularizer, which can be subsequently represented as the weighted causal\nmatrix in the post-hoc analysis. Extensive experiments show that our proposed\napproach achieves competitive performance with the state-of-the-art methods for\nlearning summary Granger causality and full-time Granger causality while\nmaintaining lower model complexity and high scalability.\n", "date": "2024-05-14", "categories": []}
{"id": "2405.08793", "title": "A Brief Introduction to Causal Inference in Machine Learning", "abstract": "  This is a lecture note produced for DS-GA 3001.003 \"Special Topics in DS -\nCausal Inference in Machine Learning\" at the Center for Data Science, New York\nUniversity in Spring, 2024. This course was created to target master's and PhD\nlevel students with basic background in machine learning but who were not\nexposed to causal inference or causal reasoning in general previously. In\nparticular, this course focuses on introducing such students to expand their\nview and knowledge of machine learning to incorporate causal reasoning, as this\naspect is at the core of so-called out-of-distribution generalization (or lack\nthereof.)\n", "date": "2024-05-14", "categories": []}
{"id": "2311.08605", "title": "Exploring the Jungle of Bias: Political Bias Attribution in Language\n  Models via Dependency Analysis", "abstract": "  The rapid advancement of Large Language Models (LLMs) has sparked intense\ndebate regarding the prevalence of bias in these models and its mitigation.\nYet, as exemplified by both results on debiasing methods in the literature and\nreports of alignment-related defects from the wider community, bias remains a\npoorly understood topic despite its practical relevance. To enhance the\nunderstanding of the internal causes of bias, we analyse LLM bias through the\nlens of causal fairness analysis, which enables us to both comprehend the\norigins of bias and reason about its downstream consequences and mitigation. To\noperationalize this framework, we propose a prompt-based method for the\nextraction of confounding and mediating attributes which contribute to the LLM\ndecision process. By applying Activity Dependency Networks (ADNs), we then\nanalyse how these attributes influence an LLM's decision process. We apply our\nmethod to LLM ratings of argument quality in political debates. We find that\nthe observed disparate treatment can at least in part be attributed to\nconfounding and mitigating attributes and model misalignment, and discuss the\nconsequences of our findings for human-AI alignment and bias mitigation. Our\ncode and data are at https://github.com/david-jenny/LLM-Political-Study.\n", "date": "2023-11-15", "categories": []}
{"id": "2405.07220", "title": "On Discovery of Local Independence over Continuous Variables via Neural\n  Contextual Decomposition", "abstract": "  Conditional independence provides a way to understand causal relationships\namong the variables of interest. An underlying system may exhibit more\nfine-grained causal relationships especially between a variable and its\nparents, which will be called the local independence relationships. One of the\nmost widely studied local relationships is Context-Specific Independence (CSI),\nwhich holds in a specific assignment of conditioned variables. However, its\napplicability is often limited since it does not allow continuous variables:\ndata conditioned to the specific value of a continuous variable contains few\ninstances, if not none, making it infeasible to test independence. In this\nwork, we define and characterize the local independence relationship that holds\nin a specific set of joint assignments of parental variables, which we call\ncontext-set specific independence (CSSI). We then provide a canonical\nrepresentation of CSSI and prove its fundamental properties. Based on our\ntheoretical findings, we cast the problem of discovering multiple CSSI\nrelationships in a system as finding a partition of the joint outcome space.\nFinally, we propose a novel method, coined neural contextual decomposition\n(NCD), which learns such partition by imposing each set to induce CSSI via\nmodeling a conditional distribution. We empirically demonstrate that the\nproposed method successfully discovers the ground truth local independence\nrelationships in both synthetic dataset and complex system reflecting the\nreal-world physical dynamics.\n", "date": "2024-05-12", "categories": []}
{"id": "2405.07795", "title": "Improved Bound for Robust Causal Bandits with Linear Models", "abstract": "  This paper investigates the robustness of causal bandits (CBs) in the face of\ntemporal model fluctuations. This setting deviates from the existing\nliterature's widely-adopted assumption of constant causal models. The focus is\non causal systems with linear structural equation models (SEMs). The SEMs and\nthe time-varying pre- and post-interventional statistical models are all\nunknown and subject to variations over time. The goal is to design a sequence\nof interventions that incur the smallest cumulative regret compared to an\noracle aware of the entire causal model and its fluctuations. A robust CB\nalgorithm is proposed, and its cumulative regret is analyzed by establishing\nboth upper and lower bounds on the regret. It is shown that in a graph with\nmaximum in-degree $d$, length of the largest causal path $L$, and an aggregate\nmodel deviation $C$, the regret is upper bounded by\n$\\tilde{\\mathcal{O}}(d^{L-\\frac{1}{2}}(\\sqrt{T} + C))$ and lower bounded by\n$\\Omega(d^{\\frac{L}{2}-2}\\max\\{\\sqrt{T}\\; ,\\; d^2C\\})$. The proposed algorithm\nachieves nearly optimal $\\tilde{\\mathcal{O}}(\\sqrt{T})$ regret when $C$ is\n$o(\\sqrt{T})$, maintaining sub-linear regret for a broad range of $C$.\n", "date": "2024-05-13", "categories": []}
{"id": "2404.04446", "title": "Bounding Causal Effects with Leaky Instruments", "abstract": "  Instrumental variables (IVs) are a popular and powerful tool for estimating\ncausal effects in the presence of unobserved confounding. However, classical\napproaches rely on strong assumptions such as the $\\textit{exclusion\ncriterion}$, which states that instrumental effects must be entirely mediated\nby treatments. This assumption often fails in practice. When IV methods are\nimproperly applied to data that do not meet the exclusion criterion, estimated\ncausal effects may be badly biased. In this work, we propose a novel solution\nthat provides $\\textit{partial}$ identification in linear systems given a set\nof $\\textit{leaky instruments}$, which are allowed to violate the exclusion\ncriterion to some limited degree. We derive a convex optimization objective\nthat provides provably sharp bounds on the average treatment effect under some\ncommon forms of information leakage, and implement inference procedures to\nquantify the uncertainty of resulting estimates. We demonstrate our method in a\nset of experiments with simulated data, where it performs favorably against the\nstate of the art. An accompanying $\\texttt{R}$ package, $\\texttt{leakyIV}$, is\navailable from $\\texttt{CRAN}$.\n", "date": "2024-04-05", "categories": []}
{"id": "2311.03309", "title": "Neural Structure Learning with Stochastic Differential Equations", "abstract": "  Discovering the underlying relationships among variables from temporal\nobservations has been a longstanding challenge in numerous scientific\ndisciplines, including biology, finance, and climate science. The dynamics of\nsuch systems are often best described using continuous-time stochastic\nprocesses. Unfortunately, most existing structure learning approaches assume\nthat the underlying process evolves in discrete-time and/or observations occur\nat regular time intervals. These mismatched assumptions can often lead to\nincorrect learned structures and models. In this work, we introduce a novel\nstructure learning method, SCOTCH, which combines neural stochastic\ndifferential equations (SDE) with variational inference to infer a posterior\ndistribution over possible structures. This continuous-time approach can\nnaturally handle both learning from and predicting observations at arbitrary\ntime points. Theoretically, we establish sufficient conditions for an SDE and\nSCOTCH to be structurally identifiable, and prove its consistency under\ninfinite data limits. Empirically, we demonstrate that our approach leads to\nimproved structure learning performance on both synthetic and real-world\ndatasets compared to relevant baselines under regular and irregular sampling\nintervals.\n", "date": "2023-11-06", "categories": []}
{"id": "2405.03130", "title": "Deep Learning for Causal Inference: A Comparison of Architectures for\n  Heterogeneous Treatment Effect Estimation", "abstract": "  Causal inference has gained much popularity in recent years, with interests\nranging from academic, to industrial, to educational, and all in between.\nConcurrently, the study and usage of neural networks has also grown profoundly\n(albeit at a far faster rate). What we aim to do in this blog write-up is\ndemonstrate a Neural Network causal inference architecture. We develop a fully\nconnected neural network implementation of the popular Bayesian Causal Forest\nalgorithm, a state of the art tree based method for estimating heterogeneous\ntreatment effects. We compare our implementation to existing neural network\ncausal inference methodologies, showing improvements in performance in\nsimulation settings. We apply our method to a dataset examining the effect of\nstress on sleep.\n", "date": "2024-05-06", "categories": []}
{"id": "2311.04037", "title": "Causal Discovery Under Local Privacy", "abstract": "  Differential privacy is a widely adopted framework designed to safeguard the\nsensitive information of data providers within a data set. It is based on the\napplication of controlled noise at the interface between the server that stores\nand processes the data, and the data consumers. Local differential privacy is a\nvariant that allows data providers to apply the privatization mechanism\nthemselves on their data individually. Therefore it provides protection also in\ncontexts in which the server, or even the data collector, cannot be trusted.\nThe introduction of noise, however, inevitably affects the utility of the data,\nparticularly by distorting the correlations between individual data components.\nThis distortion can prove detrimental to tasks such as causal discovery. In\nthis paper, we consider various well-known locally differentially private\nmechanisms and compare the trade-off between the privacy they provide, and the\naccuracy of the causal structure produced by algorithms for causal learning\nwhen applied to data obfuscated by these mechanisms. Our analysis yields\nvaluable insights for selecting appropriate local differentially private\nprotocols for causal discovery tasks. We foresee that our findings will aid\nresearchers and practitioners in conducting locally private causal discovery.\n", "date": "2023-11-07", "categories": []}
{"id": "2405.01708", "title": "A deep causal inference model for fully-interpretable travel behaviour\n  analysis", "abstract": "  Transport policy assessment often involves causal questions, yet the causal\ninference capabilities of traditional travel behavioural models are at best\nlimited. We present the deep CAusal infeRence mOdel for traveL behavIour\naNAlysis (CAROLINA), a framework that explicitly models causality in travel\nbehaviour, enhances predictive accuracy, and maintains interpretability by\nleveraging causal inference, deep learning, and traditional discrete choice\nmodelling. Within this framework, we introduce a Generative Counterfactual\nmodel for forecasting human behaviour by adapting the Normalizing Flow method.\nThrough the case studies of virtual reality-based pedestrian crossing\nbehaviour, revealed preference travel behaviour from London, and synthetic\ndata, we demonstrate the effectiveness of our proposed models in uncovering\ncausal relationships, prediction accuracy, and assessing policy interventions.\nOur results show that intervention mechanisms that can reduce pedestrian stress\nlevels lead to a 38.5% increase in individuals experiencing shorter waiting\ntimes. Reducing the travel distances in London results in a 47% increase in\nsustainable travel modes.\n", "date": "2024-05-02", "categories": []}
{"id": "2405.02183", "title": "Metalearners for Ranking Treatment Effects", "abstract": "  Efficiently allocating treatments with a budget constraint constitutes an\nimportant challenge across various domains. In marketing, for example, the use\nof promotions to target potential customers and boost conversions is limited by\nthe available budget. While much research focuses on estimating causal effects,\nthere is relatively limited work on learning to allocate treatments while\nconsidering the operational context. Existing methods for uplift modeling or\ncausal inference primarily estimate treatment effects, without considering how\nthis relates to a profit maximizing allocation policy that respects budget\nconstraints. The potential downside of using these methods is that the\nresulting predictive model is not aligned with the operational context.\nTherefore, prediction errors are propagated to the optimization of the budget\nallocation problem, subsequently leading to a suboptimal allocation policy. We\npropose an alternative approach based on learning to rank. Our proposed\nmethodology directly learns an allocation policy by prioritizing instances in\nterms of their incremental profit. We propose an efficient sampling procedure\nfor the optimization of the ranking model to scale our methodology to\nlarge-scale data sets. Theoretically, we show how learning to rank can maximize\nthe area under a policy's incremental profit curve. Empirically, we validate\nour methodology and show its effectiveness in practice through a series of\nexperiments on both synthetic and real-world data.\n", "date": "2024-05-03", "categories": []}
{"id": "2305.05281", "title": "Causal Discovery via Conditional Independence Testing with Proxy\n  Variables", "abstract": "  Distinguishing causal connections from correlations is important in many\nscenarios. However, the presence of unobserved variables, such as the latent\nconfounder, can introduce bias in conditional independence testing commonly\nemployed in constraint-based causal discovery for identifying causal relations.\nTo address this issue, existing methods introduced proxy variables to adjust\nfor the bias caused by unobserveness. However, these methods were either\nlimited to categorical variables or relied on strong parametric assumptions for\nidentification. In this paper, we propose a novel hypothesis-testing procedure\nthat can effectively examine the existence of the causal relationship over\ncontinuous variables, without any parametric constraint. Our procedure is based\non discretization, which under completeness conditions, is able to\nasymptotically establish a linear equation whose coefficient vector is\nidentifiable under the causal null hypothesis. Based on this, we introduce our\ntest statistic and demonstrate its asymptotic level and power. We validate the\neffectiveness of our procedure using both synthetic and real-world data.\n", "date": "2023-05-09", "categories": []}
{"id": "2306.08765", "title": "Causal Discovery from Time Series with Hybrids of Constraint-Based and\n  Noise-Based Algorithms", "abstract": "  Constraint-based methods and noise-based methods are two distinct families of\nmethods proposed for uncovering causal graphs from observational data. However,\nboth operate under strong assumptions that may be challenging to validate or\ncould be violated in real-world scenarios. In response to these challenges,\nthere is a growing interest in hybrid methods that amalgamate principles from\nboth methods, showing robustness to assumption violations. This paper\nintroduces a novel comprehensive framework for hybridizing constraint-based and\nnoise-based methods designed to uncover causal graphs from observational time\nseries. The framework is structured into two classes. The first class employs a\nnoise-based strategy to identify a super graph, containing the true graph,\nfollowed by a constraint-based strategy to eliminate unnecessary edges. In the\nsecond class, a constraint-based strategy is applied to identify a skeleton,\nwhich is then oriented using a noise-based strategy. The paper provides\ntheoretical guarantees for each class under the condition that all assumptions\nare satisfied, and it outlines some properties when assumptions are violated.\nTo validate the efficacy of the framework, two algorithms from each class are\nexperimentally tested on simulated data, realistic ecological data, and real\ndatasets sourced from diverse applications. Notably, two novel datasets related\nto Information Technology monitoring are introduced within the set of\nconsidered real datasets. The experimental results underscore the robustness\nand effectiveness of the hybrid approaches across a broad spectrum of datasets.\n", "date": "2023-06-14", "categories": []}
{"id": "2403.14385", "title": "Estimating Causal Effects with Double Machine Learning -- A Method\n  Evaluation", "abstract": "  The estimation of causal effects with observational data continues to be a\nvery active research area. In recent years, researchers have developed new\nframeworks which use machine learning to relax classical assumptions necessary\nfor the estimation of causal effects. In this paper, we review one of the most\nprominent methods - \"double/debiased machine learning\" (DML) - and empirically\nevaluate it by comparing its performance on simulated data relative to more\ntraditional statistical methods, before applying it to real-world data. Our\nfindings indicate that the application of a suitably flexible machine learning\nalgorithm within DML improves the adjustment for various nonlinear confounding\nrelationships. This advantage enables a departure from traditional functional\nform assumptions typically necessary in causal effect estimation. However, we\ndemonstrate that the method continues to critically depend on standard\nassumptions about causal structure and identification. When estimating the\neffects of air pollution on housing prices in our application, we find that DML\nestimates are consistently larger than estimates of less flexible methods. From\nour overall results, we provide actionable recommendations for specific choices\nresearchers must make when applying DML in practice.\n", "date": "2024-03-21", "categories": []}
{"id": "2404.19596", "title": "Debiased Collaborative Filtering with Kernel-Based Causal Balancing", "abstract": "  Debiased collaborative filtering aims to learn an unbiased prediction model\nby removing different biases in observational datasets. To solve this problem,\none of the simple and effective methods is based on the propensity score, which\nadjusts the observational sample distribution to the target one by reweighting\nobserved instances. Ideally, propensity scores should be learned with causal\nbalancing constraints. However, existing methods usually ignore such\nconstraints or implement them with unreasonable approximations, which may\naffect the accuracy of the learned propensity scores. To bridge this gap, in\nthis paper, we first analyze the gaps between the causal balancing requirements\nand existing methods such as learning the propensity with cross-entropy loss or\nmanually selecting functions to balance. Inspired by these gaps, we propose to\napproximate the balancing functions in reproducing kernel Hilbert space and\ndemonstrate that, based on the universal property and representer theorem of\nkernel functions, the causal balancing constraints can be better satisfied.\nMeanwhile, we propose an algorithm that adaptively balances the kernel function\nand theoretically analyze the generalization error bound of our methods. We\nconduct extensive experiments to demonstrate the effectiveness of our methods,\nand to promote this research direction, we have released our project at\nhttps://github.com/haoxuanli-pku/ICLR24-Kernel-Balancing.\n", "date": "2024-04-30", "categories": []}
{"id": "2404.19620", "title": "Be Aware of the Neighborhood Effect: Modeling Selection Bias under\n  Interference", "abstract": "  Selection bias in recommender system arises from the recommendation process\nof system filtering and the interactive process of user selection. Many\nprevious studies have focused on addressing selection bias to achieve unbiased\nlearning of the prediction model, but ignore the fact that potential outcomes\nfor a given user-item pair may vary with the treatments assigned to other\nuser-item pairs, named neighborhood effect. To fill the gap, this paper\nformally formulates the neighborhood effect as an interference problem from the\nperspective of causal inference and introduces a treatment representation to\ncapture the neighborhood effect. On this basis, we propose a novel ideal loss\nthat can be used to deal with selection bias in the presence of neighborhood\neffect. We further develop two new estimators for estimating the proposed ideal\nloss. We theoretically establish the connection between the proposed and\nprevious debiasing methods ignoring the neighborhood effect, showing that the\nproposed methods can achieve unbiased learning when both selection bias and\nneighborhood effect are present, while the existing methods are biased.\nExtensive semi-synthetic and real-world experiments are conducted to\ndemonstrate the effectiveness of the proposed methods.\n", "date": "2024-04-30", "categories": []}
{"id": "2211.01939", "title": "Empirical Analysis of Model Selection for Heterogeneous Causal Effect\n  Estimation", "abstract": "  We study the problem of model selection in causal inference, specifically for\nconditional average treatment effect (CATE) estimation. Unlike machine\nlearning, there is no perfect analogue of cross-validation for model selection\nas we do not observe the counterfactual potential outcomes. Towards this, a\nvariety of surrogate metrics have been proposed for CATE model selection that\nuse only observed data. However, we do not have a good understanding regarding\ntheir effectiveness due to limited comparisons in prior studies. We conduct an\nextensive empirical analysis to benchmark the surrogate model selection metrics\nintroduced in the literature, as well as the novel ones introduced in this\nwork. We ensure a fair comparison by tuning the hyperparameters associated with\nthese metrics via AutoML, and provide more detailed trends by incorporating\nrealistic datasets via generative modeling. Our analysis suggests novel model\nselection strategies based on careful hyperparameter selection of CATE\nestimators and causal ensembling.\n", "date": "2022-11-03", "categories": []}
{"id": "2404.18197", "title": "A General Causal Inference Framework for Cross-Sectional Observational\n  Data", "abstract": "  Causal inference methods for observational data are highly regarded due to\ntheir wide applicability. While there are already numerous methods available\nfor de-confounding bias, these methods generally assume that covariates consist\nsolely of confounders or make naive assumptions about the covariates. Such\nassumptions face challenges in both theory and practice, particularly when\ndealing with high-dimensional covariates. Relaxing these naive assumptions and\nidentifying the confounding covariates that truly require correction can\neffectively enhance the practical significance of these methods. Therefore,\nthis paper proposes a General Causal Inference (GCI) framework specifically\ndesigned for cross-sectional observational data, which precisely identifies the\nkey confounding covariates and provides corresponding identification algorithm.\nSpecifically, based on progressive derivations of the Markov property on\nDirected Acyclic Graph, we conclude that the key confounding covariates are\nequivalent to the common root ancestors of the treatment and the outcome\nvariable. Building upon this conclusion, the GCI framework is composed of a\nnovel Ancestor Set Identification (ASI) algorithm and de-confounding inference\nmethods. Firstly, the ASI algorithm is theoretically supported by the\nconditional independence properties and causal asymmetry between variables,\nenabling the identification of key confounding covariates. Subsequently, the\nidentified confounding covariates are used in the de-confounding inference\nmethods to obtain unbiased causal effect estimation, which can support informed\ndecision-making. Extensive experiments on synthetic datasets demonstrate that\nthe GCI framework can effectively identify the critical confounding covariates\nand significantly improve the precision, stability, and interpretability of\ncausal inference in observational studies.\n", "date": "2024-04-28", "categories": []}
{"id": "2404.18490", "title": "Reduced-Rank Multi-objective Policy Learning and Optimization", "abstract": "  Evaluating the causal impacts of possible interventions is crucial for\ninforming decision-making, especially towards improving access to opportunity.\nHowever, if causal effects are heterogeneous and predictable from covariates,\npersonalized treatment decisions can improve individual outcomes and contribute\nto both efficiency and equity. In practice, however, causal researchers do not\nhave a single outcome in mind a priori and often collect multiple outcomes of\ninterest that are noisy estimates of the true target of interest. For example,\nin government-assisted social benefit programs, policymakers collect many\noutcomes to understand the multidimensional nature of poverty. The ultimate\ngoal is to learn an optimal treatment policy that in some sense maximizes\nmultiple outcomes simultaneously. To address such issues, we present a\ndata-driven dimensionality-reduction methodology for multiple outcomes in the\ncontext of optimal policy learning with multiple objectives. We learn a\nlow-dimensional representation of the true outcome from the observed outcomes\nusing reduced rank regression. We develop a suite of estimates that use the\nmodel to denoise observed outcomes, including commonly-used index weightings.\nThese methods improve estimation error in policy evaluation and optimization,\nincluding on a case study of real-world cash transfer and social intervention\ndata. Reducing the variance of noisy social outcomes can improve the\nperformance of algorithmic allocations.\n", "date": "2024-04-29", "categories": []}
{"id": "2209.07397", "title": "From algorithms to action: improving patient care requires causality", "abstract": "  In cancer research there is much interest in building and validating outcome\npredicting outcomes to support treatment decisions. However, because most\noutcome prediction models are developed and validated without regard to the\ncausal aspects of treatment decision making, many published outcome prediction\nmodels may cause harm when used for decision making, despite being found\naccurate in validation studies. Guidelines on prediction model validation and\nthe checklist for risk model endorsement by the American Joint Committee on\nCancer do not protect against prediction models that are accurate during\ndevelopment and validation but harmful when used for decision making. We\nexplain why this is the case and how to build and validate models that are\nuseful for decision making.\n", "date": "2022-09-15", "categories": []}
{"id": "2404.17019", "title": "Neyman Meets Causal Machine Learning: Experimental Evaluation of\n  Individualized Treatment Rules", "abstract": "  A century ago, Neyman showed how to evaluate the efficacy of treatment using\na randomized experiment under a minimal set of assumptions. This classical\nrepeated sampling framework serves as a basis of routine experimental analyses\nconducted by today's scientists across disciplines. In this paper, we\ndemonstrate that Neyman's methodology can also be used to experimentally\nevaluate the efficacy of individualized treatment rules (ITRs), which are\nderived by modern causal machine learning algorithms. In particular, we show\nhow to account for additional uncertainty resulting from a training process\nbased on cross-fitting. The primary advantage of Neyman's approach is that it\ncan be applied to any ITR regardless of the properties of machine learning\nalgorithms that are used to derive the ITR. We also show, somewhat\nsurprisingly, that for certain metrics, it is more efficient to conduct this\nex-post experimental evaluation of an ITR than to conduct an ex-ante\nexperimental evaluation that randomly assigns some units to the ITR. Our\nanalysis demonstrates that Neyman's repeated sampling framework is as relevant\nfor causal inference today as it has been since its inception.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.16277", "title": "Causally Inspired Regularization Enables Domain General Representations", "abstract": "  Given a causal graph representing the data-generating process shared across\ndifferent domains/distributions, enforcing sufficient graph-implied conditional\nindependencies can identify domain-general (non-spurious) feature\nrepresentations. For the standard input-output predictive setting, we\ncategorize the set of graphs considered in the literature into two distinct\ngroups: (i) those in which the empirical risk minimizer across training domains\ngives domain-general representations and (ii) those where it does not. For the\nlatter case (ii), we propose a novel framework with regularizations, which we\ndemonstrate are sufficient for identifying domain-general feature\nrepresentations without a priori knowledge (or proxies) of the spurious\nfeatures. Empirically, our proposed method is effective for both (semi)\nsynthetic and real-world data, outperforming other state-of-the-art methods in\naverage and worst-domain transfer accuracy.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.16326", "title": "NeuroKoopman Dynamic Causal Discovery", "abstract": "  In many real-world applications where the system dynamics has an underlying\ninterdependency among its variables (such as power grid, economics,\nneuroscience, omics networks, environmental ecosystems, and others), one is\noften interested in knowing whether the past values of one time series\ninfluences the future of another, known as Granger causality, and the\nassociated underlying dynamics. This paper introduces a Koopman-inspired\nframework that leverages neural networks for data-driven learning of the\nKoopman bases, termed NeuroKoopman Dynamic Causal Discovery (NKDCD), for\nreliably inferring the Granger causality along with the underlying nonlinear\ndynamics. NKDCD employs an autoencoder architecture that lifts the nonlinear\ndynamics to a higher dimension using data-learned bases, where the lifted time\nseries can be reliably modeled linearly. The lifting function, the linear\nGranger causality lag matrices, and the projection function (from lifted space\nto base space) are all represented as multilayer perceptrons and are all\nlearned simultaneously in one go. NKDCD also utilizes sparsity-inducing\npenalties on the weights of the lag matrices, encouraging the model to select\nonly the needed causal dependencies within the data. Through extensive testing\non practically applicable datasets, it is shown that the NKDCD outperforms the\nexisting nonlinear Granger causality discovery approaches.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.16361", "title": "Evolutionary Causal Discovery with Relative Impact Stratification for\n  Interpretable Data Analysis", "abstract": "  This study proposes Evolutionary Causal Discovery (ECD) for causal discovery\nthat tailors response variables, predictor variables, and corresponding\noperators to research datasets. Utilizing genetic programming for variable\nrelationship parsing, the method proceeds with the Relative Impact\nStratification (RIS) algorithm to assess the relative impact of predictor\nvariables on the response variable, facilitating expression simplification and\nenhancing the interpretability of variable relationships. ECD proposes an\nexpression tree to visualize the RIS results, offering a differentiated\ndepiction of unknown causal relationships compared to conventional causal\ndiscovery. The ECD method represents an evolution and augmentation of existing\ncausal discovery methods, providing an interpretable approach for analyzing\nvariable relationships in complex systems, particularly in healthcare settings\nwith Electronic Health Record (EHR) data. Experiments on both synthetic and\nreal-world EHR datasets demonstrate the efficacy of ECD in uncovering patterns\nand mechanisms among variables, maintaining high accuracy and stability across\ndifferent noise levels. On the real-world EHR dataset, ECD reveals the\nintricate relationships between the response variable and other predictive\nvariables, aligning with the results of structural equation modeling and\nshapley additive explanations analyses.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.15746", "title": "Collaborative Heterogeneous Causal Inference Beyond Meta-analysis", "abstract": "  Collaboration between different data centers is often challenged by\nheterogeneity across sites. To account for the heterogeneity, the\nstate-of-the-art method is to re-weight the covariate distributions in each\nsite to match the distribution of the target population. Nevertheless, this\nmethod could easily fail when a certain site couldn't cover the entire\npopulation. Moreover, it still relies on the concept of traditional\nmeta-analysis after adjusting for the distribution shift.\n  In this work, we propose a collaborative inverse propensity score weighting\nestimator for causal inference with heterogeneous data. Instead of adjusting\nthe distribution shift separately, we use weighted propensity score models to\ncollaboratively adjust for the distribution shift. Our method shows significant\nimprovements over the methods based on meta-analysis when heterogeneity\nincreases. To account for the vulnerable density estimation, we further discuss\nthe double machine method and show the possibility of using nonparametric\ndensity estimation with d<8 and a flexible machine learning method to guarantee\nasymptotic normality. We propose a federated learning algorithm to\ncollaboratively train the outcome model while preserving privacy. Using\nsynthetic and real datasets, we demonstrate the advantages of our method.\n", "date": "2024-04-24", "categories": []}
{"id": "2404.14460", "title": "Inference of Causal Networks using a Topological Threshold", "abstract": "  We propose a constraint-based algorithm, which automatically determines\ncausal relevance thresholds, to infer causal networks from data. We call these\ntopological thresholds. We present two methods for determining the threshold:\nthe first seeks a set of edges that leaves no disconnected nodes in the\nnetwork; the second seeks a causal large connected component in the data.\n  We tested these methods both for discrete synthetic and real data, and\ncompared the results with those obtained for the PC algorithm, which we took as\nthe benchmark. We show that this novel algorithm is generally faster and more\naccurate than the PC algorithm.\n  The algorithm for determining the thresholds requires choosing a measure of\ncausality. We tested our methods for Fisher Correlations, commonly used in PC\nalgorithm (for instance in \\cite{kalisch2005}), and further proposed a discrete\nand asymmetric measure of causality, that we called Net Influence, which\nprovided very good results when inferring causal networks from discrete data.\nThis metric allows for inferring directionality of the edges in the process of\napplying the thresholds, speeding up the inference of causal DAGs.\n", "date": "2024-04-21", "categories": []}
{"id": "2404.12238", "title": "Neural Networks with Causal Graph Constraints: A New Approach for\n  Treatment Effects Estimation", "abstract": "  In recent years, there has been a growing interest in using machine learning\ntechniques for the estimation of treatment effects. Most of the best-performing\nmethods rely on representation learning strategies that encourage shared\nbehavior among potential outcomes to increase the precision of treatment effect\nestimates. In this paper we discuss and classify these models in terms of their\nalgorithmic inductive biases and present a new model, NN-CGC, that considers\nadditional information from the causal graph. NN-CGC tackles bias resulting\nfrom spurious variable interactions by implementing novel constraints on\nmodels, and it can be integrated with other representation learning methods. We\ntest the effectiveness of our method using three different base models on\ncommon benchmarks. Our results indicate that our model constraints lead to\nsignificant improvements, achieving new state-of-the-art results in treatment\neffects estimation. We also show that our method is robust to imperfect causal\ngraphs and that using partial causal information is preferable to ignoring it.\n", "date": "2024-04-18", "categories": []}
{"id": "2209.04589", "title": "Causal Intervention for Fairness in Multi-behavior Recommendation", "abstract": "  Recommender systems usually learn user interests from various user behaviors,\nincluding clicks and post-click behaviors (e.g., like and favorite). However,\nthese behaviors inevitably exhibit popularity bias, leading to some unfairness\nissues: 1) for items with similar quality, more popular ones get more exposure;\nand 2) even worse the popular items with lower popularity might receive more\nexposure. Existing work on mitigating popularity bias blindly eliminates the\nbias and usually ignores the effect of item quality. We argue that the\nrelationships between different user behaviors (e.g., conversion rate) actually\nreflect the item quality. Therefore, to handle the unfairness issues, we\npropose to mitigate the popularity bias by considering multiple user behaviors.\n  In this work, we examine causal relationships behind the interaction\ngeneration procedure in multi-behavior recommendation. Specifically, we find\nthat: 1) item popularity is a confounder between the exposed items and users'\npost-click interactions, leading to the first unfairness; and 2) some hidden\nconfounders (e.g., the reputation of item producers) affect both item\npopularity and quality, resulting in the second unfairness. To alleviate these\nconfounding issues, we propose a causal framework to estimate the causal\neffect, which leverages backdoor adjustment to block the backdoor paths caused\nby the confounders. In the inference stage, we remove the negative effect of\npopularity and utilize the good effect of quality for recommendation.\nExperiments on two real-world datasets validate the effectiveness of our\nproposed framework, which enhances fairness without sacrificing recommendation\naccuracy.\n", "date": "2022-09-10", "categories": []}
{"id": "2404.10883", "title": "Automated Discovery of Functional Actual Causes in Complex Environments", "abstract": "  Reinforcement learning (RL) algorithms often struggle to learn policies that\ngeneralize to novel situations due to issues such as causal confusion,\noverfitting to irrelevant factors, and failure to isolate control of state\nfactors. These issues stem from a common source: a failure to accurately\nidentify and exploit state-specific causal relationships in the environment.\nWhile some prior works in RL aim to identify these relationships explicitly,\nthey rely on informal domain-specific heuristics such as spatial and temporal\nproximity. Actual causality offers a principled and general framework for\ndetermining the causes of particular events. However, existing definitions of\nactual cause often attribute causality to a large number of events, even if\nmany of them rarely influence the outcome. Prior work on actual causality\nproposes normality as a solution to this problem, but its existing\nimplementations are challenging to scale to complex and continuous-valued RL\nenvironments. This paper introduces functional actual cause (FAC), a framework\nthat uses context-specific independencies in the environment to restrict the\nset of actual causes. We additionally introduce Joint Optimization for Actual\nCause Inference (JACI), an algorithm that learns from observational data to\ninfer functional actual causes. We demonstrate empirically that FAC agrees with\nknown results on a suite of examples from the actual causality literature, and\nJACI identifies actual causes with significantly higher accuracy than existing\nheuristic methods in a set of complex, continuous-valued environments.\n", "date": "2024-04-16", "categories": []}
{"id": "2404.10547", "title": "A/B testing under Interference with Partial Network Information", "abstract": "  A/B tests are often required to be conducted on subjects that might have\nsocial connections. For e.g., experiments on social media, or medical and\nsocial interventions to control the spread of an epidemic. In such settings,\nthe SUTVA assumption for randomized-controlled trials is violated due to\nnetwork interference, or spill-over effects, as treatments to group A can\npotentially also affect the control group B. When the underlying social network\nis known exactly, prior works have demonstrated how to conduct A/B tests\nadequately to estimate the global average treatment effect (GATE). However, in\npractice, it is often impossible to obtain knowledge about the exact underlying\nnetwork. In this paper, we present UNITE: a novel estimator that relax this\nassumption and can identify GATE while only relying on knowledge of the\nsuperset of neighbors for any subject in the graph. Through theoretical\nanalysis and extensive experiments, we show that the proposed approach performs\nbetter in comparison to standard estimators.\n", "date": "2024-04-16", "categories": []}
{"id": "2306.11281", "title": "Towards Characterizing Domain Counterfactuals For Invertible Latent\n  Causal Models", "abstract": "  Answering counterfactual queries has important applications such as\nexplainability, robustness, and fairness but is challenging when the causal\nvariables are unobserved and the observations are non-linear mixtures of these\nlatent variables, such as pixels in images. One approach is to recover the\nlatent Structural Causal Model (SCM), which may be infeasible in practice due\nto requiring strong assumptions, e.g., linearity of the causal mechanisms or\nperfect atomic interventions. Meanwhile, more practical ML-based approaches\nusing naive domain translation models to generate counterfactual samples lack\ntheoretical grounding and may construct invalid counterfactuals. In this work,\nwe strive to strike a balance between practicality and theoretical guarantees\nby analyzing a specific type of causal query called domain counterfactuals,\nwhich hypothesizes what a sample would have looked like if it had been\ngenerated in a different domain (or environment). We show that recovering the\nlatent SCM is unnecessary for estimating domain counterfactuals, thereby\nsidestepping some of the theoretic challenges. By assuming invertibility and\nsparsity of intervention, we prove domain counterfactual estimation error can\nbe bounded by a data fit term and intervention sparsity term. Building upon our\ntheoretical results, we develop a theoretically grounded practical algorithm\nthat simplifies the modeling process to generative model estimation under\nautoregressive and shared parameter constraints that enforce intervention\nsparsity. Finally, we show an improvement in counterfactual estimation over\nbaseline methods through extensive simulated and image-based experiments.\n", "date": "2023-06-20", "categories": []}
{"id": "2307.16405", "title": "Causal-learn: Causal Discovery in Python", "abstract": "  Causal discovery aims at revealing causal relations from observational data,\nwhich is a fundamental task in science and engineering. We describe\n$\\textit{causal-learn}$, an open-source Python library for causal discovery.\nThis library focuses on bringing a comprehensive collection of causal discovery\nmethods to both practitioners and researchers. It provides easy-to-use APIs for\nnon-specialists, modular building blocks for developers, detailed documentation\nfor learners, and comprehensive methods for all. Different from previous\npackages in R or Java, $\\textit{causal-learn}$ is fully developed in Python,\nwhich could be more in tune with the recent preference shift in programming\nlanguages within related communities. The library is available at\nhttps://github.com/py-why/causal-learn.\n", "date": "2023-07-31", "categories": []}
{"id": "2309.02027", "title": "Granger Causal Inference in Multivariate Hawkes Processes by Minimum\n  Message Length", "abstract": "  Multivariate Hawkes processes (MHPs) are versatile probabilistic tools used\nto model various real-life phenomena: earthquakes, operations on stock markets,\nneuronal activity, virus propagation and many others. In this paper, we focus\non MHPs with exponential decay kernels and estimate connectivity graphs, which\nrepresent the Granger causal relations between their components. We approach\nthis inference problem by proposing an optimization criterion and model\nselection algorithm based on the minimum message length (MML) principle. MML\ncompares Granger causal models using the Occam's razor principle in the\nfollowing way: even when models have a comparable goodness-of-fit to the\nobserved data, the one generating the most concise explanation of the data is\npreferred. While most of the state-of-art methods using lasso-type penalization\ntend to overfitting in scenarios with short time horizons, the proposed\nMML-based method achieves high F1 scores in these settings. We conduct a\nnumerical study comparing the proposed algorithm to other related classical and\nstate-of-art methods, where we achieve the highest F1 scores in specific sparse\ngraph settings. We illustrate the proposed method also on G7 sovereign bond\ndata and obtain causal connections, which are in agreement with the expert\nknowledge available in the literature.\n", "date": "2023-09-05", "categories": []}
{"id": "2404.04979", "title": "CAVIAR: Categorical-Variable Embeddings for Accurate and Robust\n  Inference", "abstract": "  Social science research often hinges on the relationship between categorical\nvariables and outcomes. We introduce CAVIAR, a novel method for embedding\ncategorical variables that assume values in a high-dimensional ambient space\nbut are sampled from an underlying manifold. Our theoretical and numerical\nanalyses outline challenges posed by such categorical variables in causal\ninference. Specifically, dynamically varying and sparse levels can lead to\nviolations of the Donsker conditions and a failure of the estimation\nfunctionals to converge to a tight Gaussian process. Traditional approaches,\nincluding the exclusion of rare categorical levels and principled variable\nselection models like LASSO, fall short. CAVIAR embeds the data into a\nlower-dimensional global coordinate system. The mapping can be derived from\nboth structured and unstructured data, and ensures stable and robust estimates\nthrough dimensionality reduction. In a dataset of direct-to-consumer apparel\nsales, we illustrate how high-dimensional categorical variables, such as zip\ncodes, can be succinctly represented, facilitating inference and analysis.\n", "date": "2024-04-07", "categories": []}
{"id": "2404.07225", "title": "Unveiling the Impact of Macroeconomic Policies: A Double Machine\n  Learning Approach to Analyzing Interest Rate Effects on Financial Markets", "abstract": "  This study examines the effects of macroeconomic policies on financial\nmarkets using a novel approach that combines Machine Learning (ML) techniques\nand causal inference. It focuses on the effect of interest rate changes made by\nthe US Federal Reserve System (FRS) on the returns of fixed income and equity\nfunds between January 1986 and December 2021. The analysis makes a distinction\nbetween actively and passively managed funds, hypothesizing that the latter are\nless susceptible to changes in interest rates. The study contrasts gradient\nboosting and linear regression models using the Double Machine Learning (DML)\nframework, which supports a variety of statistical learning techniques. Results\nindicate that gradient boosting is a useful tool for predicting fund returns;\nfor example, a 1% increase in interest rates causes an actively managed fund's\nreturn to decrease by -11.97%. This understanding of the relationship between\ninterest rates and fund performance provides opportunities for additional\nresearch and insightful, data-driven advice for fund managers and investors\n", "date": "2024-03-31", "categories": []}
{"id": "2302.08070", "title": "Local Causal Discovery for Estimating Causal Effects", "abstract": "  Even when the causal graph underlying our data is unknown, we can use\nobservational data to narrow down the possible values that an average treatment\neffect (ATE) can take by (1) identifying the graph up to a Markov equivalence\nclass; and (2) estimating that ATE for each graph in the class. While the PC\nalgorithm can identify this class under strong faithfulness assumptions, it can\nbe computationally prohibitive. Fortunately, only the local graph structure\naround the treatment is required to identify the set of possible ATE values, a\nfact exploited by local discovery algorithms to improve computational\nefficiency. In this paper, we introduce Local Discovery using Eager Collider\nChecks (LDECC), a new local causal discovery algorithm that leverages\nunshielded colliders to orient the treatment's parents differently from\nexisting methods. We show that there exist graphs where LDECC exponentially\noutperforms existing local discovery algorithms and vice versa. Moreover, we\nshow that LDECC and existing algorithms rely on different faithfulness\nassumptions, leveraging this insight to weaken the assumptions for identifying\nthe set of possible ATE values.\n", "date": "2023-02-16", "categories": []}
{"id": "2404.06681", "title": "Causal Unit Selection using Tractable Arithmetic Circuits", "abstract": "  The unit selection problem aims to find objects, called units, that optimize\na causal objective function which describes the objects' behavior in a causal\ncontext (e.g., selecting customers who are about to churn but would most likely\nchange their mind if encouraged). While early studies focused mainly on\nbounding a specific class of counterfactual objective functions using data,\nmore recent work allows one to find optimal units exactly by reducing the\ncausal objective to a classical objective on a meta-model, and then applying a\nvariant of the classical Variable Elimination (VE) algorithm to the meta-model\n-- assuming a fully specified causal model is available. In practice, however,\nfinding optimal units using this approach can be very expensive because the\nused VE algorithm must be exponential in the constrained treewidth of the\nmeta-model, which is larger and denser than the original model. We address this\ncomputational challenge by introducing a new approach for unit selection that\nis not necessarily limited by the constrained treewidth. This is done through\ncompiling the meta-model into a special class of tractable arithmetic circuits\nthat allows the computation of optimal units in time linear in the circuit\nsize. We finally present empirical results on random causal models that show\norder-of-magnitude speedups based on the proposed method for solving unit\nselection.\n", "date": "2024-04-10", "categories": []}
{"id": "2311.16026", "title": "A Neural Framework for Generalized Causal Sensitivity Analysis", "abstract": "  Unobserved confounding is common in many applications, making causal\ninference from observational data challenging. As a remedy, causal sensitivity\nanalysis is an important tool to draw causal conclusions under unobserved\nconfounding with mathematical guarantees. In this paper, we propose NeuralCSA,\na neural framework for generalized causal sensitivity analysis. Unlike previous\nwork, our framework is compatible with (i) a large class of sensitivity models,\nincluding the marginal sensitivity model, f-sensitivity models, and Rosenbaum's\nsensitivity model; (ii) different treatment types (i.e., binary and\ncontinuous); and (iii) different causal queries, including (conditional)\naverage treatment effects and simultaneous effects on multiple outcomes. The\ngenerality of NeuralCSA is achieved by learning a latent distribution shift\nthat corresponds to a treatment intervention using two conditional normalizing\nflows. We provide theoretical guarantees that NeuralCSA is able to infer valid\nbounds on the causal query of interest and also demonstrate this empirically\nusing both simulated and real-world data.\n", "date": "2023-11-27", "categories": []}
{"id": "2403.00886", "title": "Evaluating and Correcting Performative Effects of Decision Support\n  Systems via Causal Domain Shift", "abstract": "  When predicting a target variable $Y$ from features $X$, the prediction\n$\\hat{Y}$ can be performative: an agent might act on this prediction, affecting\nthe value of $Y$ that we eventually observe. Performative predictions are\ndeliberately prevalent in algorithmic decision support, where a Decision\nSupport System (DSS) provides a prediction for an agent to affect the value of\nthe target variable. When deploying a DSS in high-stakes settings (e.g.\nhealthcare, law, predictive policing, or child welfare screening) it is\nimperative to carefully assess the performative effects of the DSS. In the case\nthat the DSS serves as an alarm for a predicted negative outcome, naive\nretraining of the prediction model is bound to result in a model that\nunderestimates the risk, due to effective workings of the previous model. In\nthis work, we propose to model the deployment of a DSS as causal domain shift\nand provide novel cross-domain identification results for the conditional\nexpectation $E[Y | X]$, allowing for pre- and post-hoc assessment of the\ndeployment of the DSS, and for retraining of a model that assesses the risk\nunder a baseline policy where the DSS is not deployed. Using a running example,\nwe empirically show that a repeated regression procedure provides a practical\nframework for estimating these quantities, even when the data is affected by\nsample selection bias and selective labelling, offering for a practical,\nunified solution for multiple forms of target variable bias.\n", "date": "2024-03-01", "categories": []}
{"id": "2404.04224", "title": "Active Causal Learning for Decoding Chemical Complexities with Targeted\n  Interventions", "abstract": "  Predicting and enhancing inherent properties based on molecular structures is\nparamount to design tasks in medicine, materials science, and environmental\nmanagement. Most of the current machine learning and deep learning approaches\nhave become standard for predictions, but they face challenges when applied\nacross different datasets due to reliance on correlations between molecular\nrepresentation and target properties. These approaches typically depend on\nlarge datasets to capture the diversity within the chemical space, facilitating\na more accurate approximation, interpolation, or extrapolation of the chemical\nbehavior of molecules. In our research, we introduce an active learning\napproach that discerns underlying cause-effect relationships through strategic\nsampling with the use of a graph loss function. This method identifies the\nsmallest subset of the dataset capable of encoding the most information\nrepresentative of a much larger chemical space. The identified causal relations\nare then leveraged to conduct systematic interventions, optimizing the design\ntask within a chemical space that the models have not encountered previously.\nWhile our implementation focused on the QM9 quantum-chemical dataset for a\nspecific design task-finding molecules with a large dipole moment-our active\ncausal learning approach, driven by intelligent sampling and interventions,\nholds potential for broader applications in molecular, materials design and\ndiscovery.\n", "date": "2024-04-05", "categories": []}
{"id": "2402.13332", "title": "Causal hybrid modeling with double machine learning", "abstract": "  Hybrid modeling integrates machine learning with scientific knowledge to\nenhance interpretability, generalization, and adherence to natural laws.\nNevertheless, equifinality and regularization biases pose challenges in hybrid\nmodeling to achieve these purposes. This paper introduces a novel approach to\nestimating hybrid models via a causal inference framework, specifically\nemploying Double Machine Learning (DML) to estimate causal effects. We showcase\nits use for the Earth sciences on two problems related to carbon dioxide\nfluxes. In the $Q_{10}$ model, we demonstrate that DML-based hybrid modeling is\nsuperior in estimating causal parameters over end-to-end deep neural network\n(DNN) approaches, proving efficiency, robustness to bias from regularization\nmethods, and circumventing equifinality. Our approach, applied to carbon flux\npartitioning, exhibits flexibility in accommodating heterogeneous causal\neffects. The study emphasizes the necessity of explicitly defining causal\ngraphs and relationships, advocating for this as a general best practice. We\nencourage the continued exploration of causality in hybrid models for more\ninterpretable and trustworthy results in knowledge-guided machine learning.\n", "date": "2024-02-20", "categories": []}
{"id": "2211.13715", "title": "Trust Your $\\nabla$: Gradient-based Intervention Targeting for Causal\n  Discovery", "abstract": "  Inferring causal structure from data is a challenging task of fundamental\nimportance in science. Observational data are often insufficient to identify a\nsystem's causal structure uniquely. While conducting interventions (i.e.,\nexperiments) can improve the identifiability, such samples are usually\nchallenging and expensive to obtain. Hence, experimental design approaches for\ncausal discovery aim to minimize the number of interventions by estimating the\nmost informative intervention target. In this work, we propose a novel\nGradient-based Intervention Targeting method, abbreviated GIT, that 'trusts'\nthe gradient estimator of a gradient-based causal discovery framework to\nprovide signals for the intervention acquisition function. We provide extensive\nexperiments in simulated and real-world datasets and demonstrate that GIT\nperforms on par with competitive baselines, surpassing them in the low-data\nregime.\n", "date": "2022-11-24", "categories": []}
{"id": "2310.17463", "title": "Bayesian Neural Controlled Differential Equations for Treatment Effect\n  Estimation", "abstract": "  Treatment effect estimation in continuous time is crucial for personalized\nmedicine. However, existing methods for this task are limited to point\nestimates of the potential outcomes, whereas uncertainty estimates have been\nignored. Needless to say, uncertainty quantification is crucial for reliable\ndecision-making in medical applications. To fill this gap, we propose a novel\nBayesian neural controlled differential equation (BNCDE) for treatment effect\nestimation in continuous time. In our BNCDE, the time dimension is modeled\nthrough a coupled system of neural controlled differential equations and neural\nstochastic differential equations, where the neural stochastic differential\nequations allow for tractable variational Bayesian inference. Thereby, for an\nassigned sequence of treatments, our BNCDE provides meaningful posterior\npredictive distributions of the potential outcomes. To the best of our\nknowledge, ours is the first tailored neural method to provide uncertainty\nestimates of treatment effects in continuous time. As such, our method is of\ndirect practical value for promoting reliable decision-making in medicine.\n", "date": "2023-10-26", "categories": []}
{"id": "2308.16859", "title": "Information Theoretically Optimal Sample Complexity of Learning\n  Dynamical Directed Acyclic Graphs", "abstract": "  In this article, the optimal sample complexity of learning the underlying\ninteractions or dependencies of a Linear Dynamical System (LDS) over a Directed\nAcyclic Graph (DAG) is studied. We call such a DAG underlying an LDS as\ndynamical DAG (DDAG). In particular, we consider a DDAG where the nodal\ndynamics are driven by unobserved exogenous noise sources that are wide-sense\nstationary (WSS) in time but are mutually uncorrelated, and have the same\n{power spectral density (PSD)}. Inspired by the static DAG setting, a metric\nand an algorithm based on the PSD matrix of the observed time series are\nproposed to reconstruct the DDAG. It is shown that the optimal sample\ncomplexity (or length of state trajectory) needed to learn the DDAG is\n$n=\\Theta(q\\log(p/q))$, where $p$ is the number of nodes and $q$ is the maximum\nnumber of parents per node. To prove the sample complexity upper bound, a\nconcentration bound for the PSD estimation is derived, under two different\nsampling strategies. A matching min-max lower bound using generalized Fano's\ninequality also is provided, thus showing the order optimality of the proposed\nalgorithm.\n", "date": "2023-08-31", "categories": []}
{"id": "2310.15333", "title": "Safe and Interpretable Estimation of Optimal Treatment Regimes", "abstract": "  Recent statistical and reinforcement learning methods have significantly\nadvanced patient care strategies. However, these approaches face substantial\nchallenges in high-stakes contexts, including missing data, inherent\nstochasticity, and the critical requirements for interpretability and patient\nsafety. Our work operationalizes a safe and interpretable framework to identify\noptimal treatment regimes. This approach involves matching patients with\nsimilar medical and pharmacological characteristics, allowing us to construct\nan optimal policy via interpolation. We perform a comprehensive simulation\nstudy to demonstrate the framework's ability to identify optimal policies even\nin complex settings. Ultimately, we operationalize our approach to study\nregimes for treating seizures in critically ill patients. Our findings strongly\nsupport personalized treatment strategies based on a patient's medical history\nand pharmacological features. Notably, we identify that reducing medication\ndoses for patients with mild and brief seizure episodes while adopting\naggressive treatment for patients in intensive care unit experiencing intense\nseizures leads to more favorable outcomes.\n", "date": "2023-10-23", "categories": []}
{"id": "2404.00207", "title": "Causal Inference for Human-Language Model Collaboration", "abstract": "  In this paper, we examine the collaborative dynamics between humans and\nlanguage models (LMs), where the interactions typically involve LMs proposing\ntext segments and humans editing or responding to these proposals. Productive\nengagement with LMs in such scenarios necessitates that humans discern\neffective text-based interaction strategies, such as editing and response\nstyles, from historical human-LM interactions. This objective is inherently\ncausal, driven by the counterfactual `what-if' question: how would the outcome\nof collaboration change if humans employed a different text editing/refinement\nstrategy? A key challenge in answering this causal inference question is\nformulating an appropriate causal estimand: the conventional average treatment\neffect (ATE) estimand is inapplicable to text-based treatments due to their\nhigh dimensionality. To address this concern, we introduce a new causal\nestimand -- Incremental Stylistic Effect (ISE) -- which characterizes the\naverage impact of infinitesimally shifting a text towards a specific style,\nsuch as increasing formality. We establish the conditions for the\nnon-parametric identification of ISE. Building on this, we develop\nCausalCollab, an algorithm designed to estimate the ISE of various interaction\nstrategies in dynamic human-LM collaborations. Our empirical investigations\nacross three distinct human-LM collaboration scenarios reveal that CausalCollab\neffectively reduces confounding and significantly improves counterfactual\nestimation over a set of competitive baselines.\n", "date": "2024-03-30", "categories": []}
{"id": "2404.00751", "title": "C-XGBoost: A tree boosting model for causal effect estimation", "abstract": "  Causal effect estimation aims at estimating the Average Treatment Effect as\nwell as the Conditional Average Treatment Effect of a treatment to an outcome\nfrom the available data. This knowledge is important in many safety-critical\ndomains, where it often needs to be extracted from observational data. In this\nwork, we propose a new causal inference model, named C-XGBoost, for the\nprediction of potential outcomes. The motivation of our approach is to exploit\nthe superiority of tree-based models for handling tabular data together with\nthe notable property of causal inference neural network-based models to learn\nrepresentations that are useful for estimating the outcome for both the\ntreatment and non-treatment cases. The proposed model also inherits the\nconsiderable advantages of XGBoost model such as efficiently handling features\nwith missing values requiring minimum preprocessing effort, as well as it is\nequipped with regularization techniques to avoid overfitting/bias. Furthermore,\nwe propose a new loss function for efficiently training the proposed causal\ninference model. The experimental analysis, which is based on the performance\nprofiles of Dolan and Mor{\\'e} as well as on post-hoc and non-parametric\nstatistical tests, provide strong evidence about the effectiveness of the\nproposed approach.\n", "date": "2024-03-31", "categories": []}
{"id": "2310.13240", "title": "Transparency challenges in policy evaluation with causal machine\n  learning -- improving usability and accountability", "abstract": "  Causal machine learning tools are beginning to see use in real-world policy\nevaluation tasks to flexibly estimate treatment effects. One issue with these\nmethods is that the machine learning models used are generally black boxes,\ni.e., there is no globally interpretable way to understand how a model makes\nestimates. This is a clear problem in policy evaluation applications,\nparticularly in government, because it is difficult to understand whether such\nmodels are functioning in ways that are fair, based on the correct\ninterpretation of evidence and transparent enough to allow for accountability\nif things go wrong. However, there has been little discussion of transparency\nproblems in the causal machine learning literature and how these might be\novercome. This paper explores why transparency issues are a problem for causal\nmachine learning in public policy evaluation applications and considers ways\nthese problems might be addressed through explainable AI tools and by\nsimplifying models in line with interpretable AI principles. It then applies\nthese ideas to a case-study using a causal forest model to estimate conditional\naverage treatment effects for a hypothetical change in the school leaving age\nin Australia. It shows that existing tools for understanding black-box\npredictive models are poorly suited to causal machine learning and that\nsimplifying the model to make it interpretable leads to an unacceptable\nincrease in error (in this application). It concludes that new tools are needed\nto properly understand causal machine learning models and the algorithms that\nfit them.\n", "date": "2023-10-20", "categories": []}
{"id": "2403.20250", "title": "Optimal Policy Learning with Observational Data in Multi-Action\n  Scenarios: Estimation, Risk Preference, and Potential Failures", "abstract": "  This paper deals with optimal policy learning (OPL) with observational data,\ni.e. data-driven optimal decision-making, in multi-action (or multi-arm)\nsettings, where a finite set of decision options is available. It is organized\nin three parts, where I discuss respectively: estimation, risk preference, and\npotential failures. The first part provides a brief review of the key\napproaches to estimating the reward (or value) function and optimal policy\nwithin this context of analysis. Here, I delineate the identification\nassumptions and statistical properties related to offline optimal policy\nlearning estimators. In the second part, I delve into the analysis of decision\nrisk. This analysis reveals that the optimal choice can be influenced by the\ndecision maker's attitude towards risks, specifically in terms of the trade-off\nbetween reward conditional mean and conditional variance. Here, I present an\napplication of the proposed model to real data, illustrating that the average\nregret of a policy with multi-valued treatment is contingent on the\ndecision-maker's attitude towards risk. The third part of the paper discusses\nthe limitations of optimal data-driven decision-making by highlighting\nconditions under which decision-making can falter. This aspect is linked to the\nfailure of the two fundamental assumptions essential for identifying the\noptimal choice: (i) overlapping, and (ii) unconfoundedness. Some conclusions\nend the paper.\n", "date": "2024-03-29", "categories": []}
{"id": "2403.18994", "title": "Causal-StoNet: Causal Inference for High-Dimensional Complex Data", "abstract": "  With the advancement of data science, the collection of increasingly complex\ndatasets has become commonplace. In such datasets, the data dimension can be\nextremely high, and the underlying data generation process can be unknown and\nhighly nonlinear. As a result, the task of making causal inference with\nhigh-dimensional complex data has become a fundamental problem in many\ndisciplines, such as medicine, econometrics, and social science. However, the\nexisting methods for causal inference are frequently developed under the\nassumption that the data dimension is low or that the underlying data\ngeneration process is linear or approximately linear. To address these\nchallenges, this paper proposes a novel causal inference approach for dealing\nwith high-dimensional complex data. The proposed approach is based on deep\nlearning techniques, including sparse deep learning theory and stochastic\nneural networks, that have been developed in recent literature. By using these\ntechniques, the proposed approach can address both the high dimensionality and\nunknown data generation process in a coherent way. Furthermore, the proposed\napproach can also be used when missing values are present in the datasets.\nExtensive numerical studies indicate that the proposed approach outperforms\nexisting ones.\n", "date": "2024-03-27", "categories": []}
{"id": "2311.02695", "title": "Identifying Linearly-Mixed Causal Representations from Multi-Node\n  Interventions", "abstract": "  The task of inferring high-level causal variables from low-level\nobservations, commonly referred to as causal representation learning, is\nfundamentally underconstrained. As such, recent works to address this problem\nfocus on various assumptions that lead to identifiability of the underlying\nlatent causal variables. A large corpus of these preceding approaches consider\nmulti-environment data collected under different interventions on the causal\nmodel. What is common to virtually all of these works is the restrictive\nassumption that in each environment, only a single variable is intervened on.\nIn this work, we relax this assumption and provide the first identifiability\nresult for causal representation learning that allows for multiple variables to\nbe targeted by an intervention within one environment. Our approach hinges on a\ngeneral assumption on the coverage and diversity of interventions across\nenvironments, which also includes the shared assumption of single-node\ninterventions of previous works. The main idea behind our approach is to\nexploit the trace that interventions leave on the variance of the ground truth\ncausal variables and regularizing for a specific notion of sparsity with\nrespect to this trace. In addition to and inspired by our theoretical\ncontributions, we present a practical algorithm to learn causal representations\nfrom multi-node interventional data and provide empirical evidence that\nvalidates our identifiability results.\n", "date": "2023-11-05", "categories": []}
{"id": "2311.02760", "title": "Causal Question Answering with Reinforcement Learning", "abstract": "  Causal questions inquire about causal relationships between different events\nor phenomena. They are important for a variety of use cases, including virtual\nassistants and search engines. However, many current approaches to causal\nquestion answering cannot provide explanations or evidence for their answers.\nHence, in this paper, we aim to answer causal questions with a causality graph,\na large-scale dataset of causal relations between noun phrases along with the\nrelations' provenance data. Inspired by recent, successful applications of\nreinforcement learning to knowledge graph tasks, such as link prediction and\nfact-checking, we explore the application of reinforcement learning on a\ncausality graph for causal question answering. We introduce an\nActor-Critic-based agent which learns to search through the graph to answer\ncausal questions. We bootstrap the agent with a supervised learning procedure\nto deal with large action spaces and sparse rewards. Our evaluation shows that\nthe agent successfully prunes the search space to answer binary causal\nquestions by visiting less than 30 nodes per question compared to over 3,000\nnodes by a naive breadth-first search. Our ablation study indicates that our\nsupervised learning strategy provides a strong foundation upon which our\nreinforcement learning agent improves. The paths returned by our agent explain\nthe mechanisms by which a cause produces an effect. Moreover, for each edge on\na path, our causality graph provides its original source allowing for easy\nverification of paths.\n", "date": "2023-11-05", "categories": []}
{"id": "2401.15911", "title": "Distribution-consistency Structural Causal Models", "abstract": "  In the field of causal modeling, potential outcomes (PO) and structural\ncausal models (SCMs) stand as the predominant frameworks. However, these\nframeworks face notable challenges in practically modeling counterfactuals,\nformalized as parameters of the joint distribution of potential outcomes.\nCounterfactual reasoning holds paramount importance in contemporary\ndecision-making processes, especially in scenarios that demand personalized\nincentives based on the joint values of $(Y(0), Y(1))$. This paper begins with\nan investigation of the PO and SCM frameworks for modeling counterfactuals.\nThrough the analysis, we identify an inherent model capacity limitation, termed\nas the ``degenerative counterfactual problem'', emerging from the consistency\nrule that is the cornerstone of both frameworks. To address this limitation, we\nintroduce a novel \\textit{distribution-consistency} assumption, and in\nalignment with it, we propose the Distribution-consistency Structural Causal\nModels (DiscoSCMs) offering enhanced capabilities to model counterfactuals. To\nconcretely reveal the enhanced model capacity, we introduce a new identifiable\ncausal parameter, \\textit{the probability of consistency}, which holds\npractical significance within DiscoSCM alone, showcased with a personalized\nincentive example. Furthermore, we provide a comprehensive set of theoretical\nresults about the ``Ladder of Causation'' within the DiscoSCM framework. We\nhope it opens new avenues for future research of counterfactual modeling,\nultimately enhancing our understanding of causality and its real-world\napplications.\n", "date": "2024-01-29", "categories": []}
{"id": "2403.15499", "title": "A Causal Analysis of CO2 Reduction Strategies in Electricity Markets\n  Through Machine Learning-Driven Metalearners", "abstract": "  This study employs the Causal Machine Learning (CausalML) statistical method\nto analyze the influence of electricity pricing policies on carbon dioxide\n(CO2) levels in the household sector. Investigating the causality between\npotential outcomes and treatment effects, where changes in pricing policies are\nthe treatment, our analysis challenges the conventional wisdom surrounding\nincentive-based electricity pricing. The study's findings suggest that adopting\nsuch policies may inadvertently increase CO2 intensity. Additionally, we\nintegrate a machine learning-based meta-algorithm, reflecting a contemporary\nstatistical approach, to enhance the depth of our causal analysis. The study\nconducts a comparative analysis of learners X, T, S, and R to ascertain the\noptimal methods based on the defined question's specified goals and contextual\nnuances. This research contributes valuable insights to the ongoing dialogue on\nsustainable development practices, emphasizing the importance of considering\nunintended consequences in policy formulation.\n", "date": "2024-03-21", "categories": []}
{"id": "2403.15500", "title": "Gene Regulatory Network Inference in the Presence of Dropouts: a Causal\n  View", "abstract": "  Gene regulatory network inference (GRNI) is a challenging problem,\nparticularly owing to the presence of zeros in single-cell RNA sequencing data:\nsome are biological zeros representing no gene expression, while some others\nare technical zeros arising from the sequencing procedure (aka dropouts), which\nmay bias GRNI by distorting the joint distribution of the measured gene\nexpressions. Existing approaches typically handle dropout error via imputation,\nwhich may introduce spurious relations as the true joint distribution is\ngenerally unidentifiable. To tackle this issue, we introduce a causal graphical\nmodel to characterize the dropout mechanism, namely, Causal Dropout Model. We\nprovide a simple yet effective theoretical result: interestingly, the\nconditional independence (CI) relations in the data with dropouts, after\ndeleting the samples with zero values (regardless if technical or not) for the\nconditioned variables, are asymptotically identical to the CI relations in the\noriginal data without dropouts. This particular test-wise deletion procedure,\nin which we perform CI tests on the samples without zeros for the conditioned\nvariables, can be seamlessly integrated with existing structure learning\napproaches including constraint-based and greedy score-based methods, thus\ngiving rise to a principled framework for GRNI in the presence of dropouts. We\nfurther show that the causal dropout model can be validated from data, and many\nexisting statistical models to handle dropouts fit into our model as specific\nparametric instances. Empirical evaluation on synthetic, curated, and\nreal-world experimental transcriptomic data comprehensively demonstrate the\nefficacy of our method.\n", "date": "2024-03-21", "categories": []}
{"id": "2403.16031", "title": "Learning Directed Acyclic Graphs from Partial Orderings", "abstract": "  Directed acyclic graphs (DAGs) are commonly used to model causal\nrelationships among random variables. In general, learning the DAG structure is\nboth computationally and statistically challenging. Moreover, without\nadditional information, the direction of edges may not be estimable from\nobservational data. In contrast, given a complete causal ordering of the\nvariables, the problem can be solved efficiently, even in high dimensions. In\nthis paper, we consider the intermediate problem of learning DAGs when a\npartial causal ordering of variables is available. We propose a general\nestimation framework for leveraging the partial ordering and present efficient\nestimation algorithms for low- and high-dimensional problems. The advantages of\nthe proposed framework are illustrated via numerical studies.\n", "date": "2024-03-24", "categories": []}
{"id": "2403.16523", "title": "Causal Discovery from Poisson Branching Structural Causal Model Using\n  High-Order Cumulant with Path Analysis", "abstract": "  Count data naturally arise in many fields, such as finance, neuroscience, and\nepidemiology, and discovering causal structure among count data is a crucial\ntask in various scientific and industrial scenarios. One of the most common\ncharacteristics of count data is the inherent branching structure described by\na binomial thinning operator and an independent Poisson distribution that\ncaptures both branching and noise. For instance, in a population count\nscenario, mortality and immigration contribute to the count, where survival\nfollows a Bernoulli distribution, and immigration follows a Poisson\ndistribution. However, causal discovery from such data is challenging due to\nthe non-identifiability issue: a single causal pair is Markov equivalent, i.e.,\n$X\\rightarrow Y$ and $Y\\rightarrow X$ are distributed equivalent. Fortunately,\nin this work, we found that the causal order from $X$ to its child $Y$ is\nidentifiable if $X$ is a root vertex and has at least two directed paths to\n$Y$, or the ancestor of $X$ with the most directed path to $X$ has a directed\npath to $Y$ without passing $X$. Specifically, we propose a Poisson Branching\nStructure Causal Model (PB-SCM) and perform a path analysis on PB-SCM using\nhigh-order cumulants. Theoretical results establish the connection between the\npath and cumulant and demonstrate that the path information can be obtained\nfrom the cumulant. With the path information, causal order is identifiable\nunder some graphical conditions. A practical algorithm for learning causal\nstructure under PB-SCM is proposed and the experiments demonstrate and verify\nthe effectiveness of the proposed method.\n", "date": "2024-03-25", "categories": []}
{"id": "2403.14843", "title": "Local Causal Discovery with Linear non-Gaussian Cyclic Models", "abstract": "  Local causal discovery is of great practical significance, as there are often\nsituations where the discovery of the global causal structure is unnecessary,\nand the interest lies solely on a single target variable. Most existing local\nmethods utilize conditional independence relations, providing only a partially\ndirected graph, and assume acyclicity for the ground-truth structure, even\nthough real-world scenarios often involve cycles like feedback mechanisms. In\nthis work, we present a general, unified local causal discovery method with\nlinear non-Gaussian models, whether they are cyclic or acyclic. We extend the\napplication of independent component analysis from the global context to\nindependent subspace analysis, enabling the exact identification of the\nequivalent local directed structures and causal strengths from the Markov\nblanket of the target variable. We also propose an alternative regression-based\nmethod in the particular acyclic scenarios. Our identifiability results are\nempirically validated using both synthetic and real-world datasets.\n", "date": "2024-03-21", "categories": []}
{"id": "2312.10569", "title": "Interpretable Causal Inference for Analyzing Wearable, Sensor, and\n  Distributional Data", "abstract": "  Many modern causal questions ask how treatments affect complex outcomes that\nare measured using wearable devices and sensors. Current analysis approaches\nrequire summarizing these data into scalar statistics (e.g., the mean), but\nthese summaries can be misleading. For example, disparate distributions can\nhave the same means, variances, and other statistics. Researchers can overcome\nthe loss of information by instead representing the data as distributions. We\ndevelop an interpretable method for distributional data analysis that ensures\ntrustworthy and robust decision-making: Analyzing Distributional Data via\nMatching After Learning to Stretch (ADD MALTS). We (i) provide analytical\nguarantees of the correctness of our estimation strategy, (ii) demonstrate via\nsimulation that ADD MALTS outperforms other distributional data analysis\nmethods at estimating treatment effects, and (iii) illustrate ADD MALTS'\nability to verify whether there is enough cohesion between treatment and\ncontrol units within subpopulations to trustworthily estimate treatment\neffects. We demonstrate ADD MALTS' utility by studying the effectiveness of\ncontinuous glucose monitors in mitigating diabetes risks.\n", "date": "2023-12-17", "categories": []}
{"id": "2403.14228", "title": "Recovering Latent Confounders from High-dimensional Proxy Variables", "abstract": "  Detecting latent confounders from proxy variables is an essential problem in\ncausal effect estimation. Previous approaches are limited to low-dimensional\nproxies, sorted proxies, and binary treatments. We remove these assumptions and\npresent a novel Proxy Confounder Factorization (PCF) framework for continuous\ntreatment effect estimation when latent confounders manifest through\nhigh-dimensional, mixed proxy variables. For specific sample sizes, our\ntwo-step PCF implementation, using Independent Component Analysis (ICA-PCF),\nand the end-to-end implementation, using Gradient Descent (GD-PCF), achieve\nhigh correlation with the latent confounder and low absolute error in causal\neffect estimation with synthetic datasets in the high sample size regime. Even\nwhen faced with climate data, ICA-PCF recovers four components that explain\n$75.9\\%$ of the variance in the North Atlantic Oscillation, a known confounder\nof precipitation patterns in Europe. Code for our PCF implementations and\nexperiments can be found here: https://github.com/IPL-UV/confound_it. The\nproposed methodology constitutes a stepping stone towards discovering latent\nconfounders and can be applied to many problems in disciplines dealing with\nhigh-dimensional observed proxies, e.g., spatiotemporal fields.\n", "date": "2024-03-21", "categories": []}
{"id": "2403.14232", "title": "Contrastive Balancing Representation Learning for Heterogeneous\n  Dose-Response Curves Estimation", "abstract": "  Estimating the individuals' potential response to varying treatment doses is\ncrucial for decision-making in areas such as precision medicine and management\nscience. Most recent studies predict counterfactual outcomes by learning a\ncovariate representation that is independent of the treatment variable.\nHowever, such independence constraints neglect much of the covariate\ninformation that is useful for counterfactual prediction, especially when the\ntreatment variables are continuous. To tackle the above issue, in this paper,\nwe first theoretically demonstrate the importance of the balancing and\nprognostic representations for unbiased estimation of the heterogeneous\ndose-response curves, that is, the learned representations are constrained to\nsatisfy the conditional independence between the covariates and both of the\ntreatment variables and the potential responses. Based on this, we propose a\nnovel Contrastive balancing Representation learning Network using a partial\ndistance measure, called CRNet, for estimating the heterogeneous dose-response\ncurves without losing the continuity of treatments. Extensive experiments are\nconducted on synthetic and real-world datasets demonstrating that our proposal\nsignificantly outperforms previous methods.\n", "date": "2024-03-21", "categories": []}
{"id": "2403.14327", "title": "Investigating the validity of structure learning algorithms in\n  identifying risk factors for intervention in patients with diabetes", "abstract": "  Diabetes, a pervasive and enduring health challenge, imposes significant\nglobal implications on health, financial healthcare systems, and societal\nwell-being. This study undertakes a comprehensive exploration of various\nstructural learning algorithms to discern causal pathways amongst potential\nrisk factors influencing diabetes progression. The methodology involves the\napplication of these algorithms to relevant diabetes data, followed by the\nconversion of their output graphs into Causal Bayesian Networks (CBNs),\nenabling predictive analysis and the evaluation of discrepancies in the effect\nof hypothetical interventions within our context-specific case study.\n  This study highlights the substantial impact of algorithm selection on\nintervention outcomes. To consolidate insights from diverse algorithms, we\nemploy a model-averaging technique that helps us obtain a unique causal model\nfor diabetes derived from a varied set of structural learning algorithms. We\nalso investigate how each of those individual graphs, as well as the average\ngraph, compare to the structures elicited by a domain expert who categorised\ngraph edges into high confidence, moderate, and low confidence types, leading\ninto three individual graphs corresponding to the three levels of confidence.\n  The resulting causal model and data are made available online, and serve as a\nvaluable resource and a guide for informed decision-making by healthcare\npractitioners, offering a comprehensive understanding of the interactions\nbetween relevant risk factors and the effect of hypothetical interventions.\nTherefore, this research not only contributes to the academic discussion on\ndiabetes, but also provides practical guidance for healthcare professionals in\ndeveloping efficient intervention and risk management strategies.\n", "date": "2024-03-21", "categories": []}
{"id": "2311.00118", "title": "Extracting the Multiscale Causal Backbone of Brain Dynamics", "abstract": "  The bulk of the research effort on brain connectivity revolves around\nstatistical associations among brain regions, which do not directly relate to\nthe causal mechanisms governing brain dynamics. Here we propose the multiscale\ncausal backbone (MCB) of brain dynamics, shared by a set of individuals across\nmultiple temporal scales, and devise a principled methodology to extract it.\n  Our approach leverages recent advances in multiscale causal structure\nlearning and optimizes the trade-off between the model fit and its complexity.\nEmpirical assessment on synthetic data shows the superiority of our methodology\nover a baseline based on canonical functional connectivity networks. When\napplied to resting-state fMRI data, we find sparse MCBs for both the left and\nright brain hemispheres. Thanks to its multiscale nature, our approach shows\nthat at low-frequency bands, causal dynamics are driven by brain regions\nassociated with high-level cognitive functions; at higher frequencies instead,\nnodes related to sensory processing play a crucial role. Finally, our analysis\nof individual multiscale causal structures confirms the existence of a causal\nfingerprint of brain connectivity, thus supporting the existing extensive\nresearch in brain connectivity fingerprinting from a causal perspective.\n", "date": "2023-10-31", "categories": []}
{"id": "2402.05946", "title": "Unveiling Latent Causal Rules: A Temporal Point Process Approach for\n  Abnormal Event Explanation", "abstract": "  In high-stakes systems such as healthcare, it is critical to understand the\ncausal reasons behind unusual events, such as sudden changes in patient's\nhealth. Unveiling the causal reasons helps with quick diagnoses and precise\ntreatment planning. In this paper, we propose an automated method for\nuncovering \"if-then\" logic rules to explain observational events. We introduce\ntemporal point processes to model the events of interest, and discover the set\nof latent rules to explain the occurrence of events. To achieve this, we employ\nan Expectation-Maximization (EM) algorithm. In the E-step, we calculate the\nlikelihood of each event being explained by each discovered rule. In the\nM-step, we update both the rule set and model parameters to enhance the\nlikelihood function's lower bound. Notably, we optimize the rule set in a\ndifferential manner. Our approach demonstrates accurate performance in both\ndiscovering rules and identifying root causes. We showcase its promising\nresults using synthetic and real healthcare datasets.\n", "date": "2024-02-03", "categories": []}
{"id": "2209.14568", "title": "Local and Regional Counterfactual Rules: Summarized and Robust Recourses", "abstract": "  Counterfactual Explanations (CE) face several unresolved challenges, such as\nensuring stability, synthesizing multiple CEs, and providing plausibility and\nsparsity guarantees. From a more practical point of view, recent studies\n[Pawelczyk et al., 2022] show that the prescribed counterfactual recourses are\noften not implemented exactly by individuals and demonstrate that most\nstate-of-the-art CE algorithms are very likely to fail in this noisy\nenvironment. To address these issues, we propose a probabilistic framework that\ngives a sparse local counterfactual rule for each observation, providing rules\nthat give a range of values capable of changing decisions with high\nprobability. These rules serve as a summary of diverse counterfactual\nexplanations and yield robust recourses. We further aggregate these local rules\ninto a regional counterfactual rule, identifying shared recourses for subgroups\nof the data. Our local and regional rules are derived from the Random Forest\nalgorithm, which offers statistical guarantees and fidelity to data\ndistribution by selecting recourses in high-density regions. Moreover, our\nrules are sparse as we first select the smallest set of variables having a high\nprobability of changing the decision. We have conducted experiments to validate\nthe effectiveness of our counterfactual rules in comparison to standard CE and\nrecent similar attempts. Our methods are available as a Python package.\n", "date": "2022-09-29", "categories": []}
{"id": "2310.17405", "title": "Causal Modeling with Stationary Diffusions", "abstract": "  We develop a novel approach towards causal inference. Rather than structural\nequations over a causal graph, we learn stochastic differential equations\n(SDEs) whose stationary densities model a system's behavior under\ninterventions. These stationary diffusion models do not require the formalism\nof causal graphs, let alone the common assumption of acyclicity. We show that\nin several cases, they generalize to unseen interventions on their variables,\noften better than classical approaches. Our inference method is based on a new\ntheoretical result that expresses a stationarity condition on the diffusion's\ngenerator in a reproducing kernel Hilbert space. The resulting kernel deviation\nfrom stationarity (KDS) is an objective function of independent interest.\n", "date": "2023-10-26", "categories": []}
{"id": "2403.10618", "title": "Limits of Approximating the Median Treatment Effect", "abstract": "  Average Treatment Effect (ATE) estimation is a well-studied problem in causal\ninference. However, it does not necessarily capture the heterogeneity in the\ndata, and several approaches have been proposed to tackle the issue, including\nestimating the Quantile Treatment Effects. In the finite population setting\ncontaining $n$ individuals, with treatment and control values denoted by the\npotential outcome vectors $\\mathbf{a}, \\mathbf{b}$, much of the prior work\nfocused on estimating median$(\\mathbf{a}) -$ median$(\\mathbf{b})$, where\nmedian($\\mathbf x$) denotes the median value in the sorted ordering of all the\nvalues in vector $\\mathbf x$. It is known that estimating the difference of\nmedians is easier than the desired estimand of median$(\\mathbf{a-b})$, called\nthe Median Treatment Effect (MTE). The fundamental problem of causal inference\n-- for every individual $i$, we can only observe one of the potential outcome\nvalues, i.e., either the value $a_i$ or $b_i$, but not both, makes estimating\nMTE particularly challenging. In this work, we argue that MTE is not estimable\nand detail a novel notion of approximation that relies on the sorted order of\nthe values in $\\mathbf{a-b}$. Next, we identify a quantity called variability\nthat exactly captures the complexity of MTE estimation. By drawing connections\nto instance-optimality studied in theoretical computer science, we show that\nevery algorithm for estimating the MTE obtains an approximation error that is\nno better than the error of an algorithm that computes variability. Finally, we\nprovide a simple linear time algorithm for computing the variability exactly.\nUnlike much prior work, a particular highlight of our work is that we make no\nassumptions about how the potential outcome vectors are generated or how they\nare correlated, except that the potential outcome values are $k$-ary, i.e.,\ntake one of $k$ discrete values.\n", "date": "2024-03-15", "categories": []}
{"id": "2403.10766", "title": "ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference", "abstract": "  Inferring unbiased treatment effects has received widespread attention in the\nmachine learning community. In recent years, our community has proposed\nnumerous solutions in standard settings, high-dimensional treatment settings,\nand even longitudinal settings. While very diverse, the solution has mostly\nrelied on neural networks for inference and simultaneous correction of\nassignment bias. New approaches typically build on top of previous approaches\nby proposing new (or refined) architectures and learning algorithms. However,\nthe end result -- a neural-network-based inference machine -- remains\nunchallenged. In this paper, we introduce a different type of solution in the\nlongitudinal setting: a closed-form ordinary differential equation (ODE). While\nwe still rely on continuous optimization to learn an ODE, the resulting\ninference machine is no longer a neural network. Doing so yields several\nadvantages such as interpretability, irregular sampling, and a different set of\nidentification assumptions. Above all, we consider the introduction of a\ncompletely new type of solution to be our most important contribution as it may\nspark entirely new innovations in treatment effects in general. We facilitate\nthis by formulating our contribution as a framework that can transform any ODE\ndiscovery method into a treatment effects method.\n", "date": "2024-03-16", "categories": []}
{"id": "2403.11217", "title": "Research on Personal Credit Risk Assessment Methods Based on Causal\n  Inference", "abstract": "  The discussion on causality in human history dates back to ancient Greece,\nyet to this day, there is still no consensus. Fundamentally, this stems from\nthe nature of human cognition, as understanding causality requires abstract\ntools to transcend the limitations of human cognition. In recent decades, the\nrapid development of mathematical and computational tools has provided new\ntheoretical and technical means for exploring causality, creating more avenues\nfor investigation.\n  Based on this, this paper introduces a new definition of causality using\ncategory theory, proposed by Samuel Eilenberg and Saunders Mac Lane in 1945 to\navoid the self-referential contradictions in set theory, notably the Russell\nparadox. Within this framework, the feasibility of indicator synthesis in\ncausal inference is demonstrated. Due to the limitations in the development of\ncategory theory-related technical tools, this paper adopts the widely-used\nprobabilistic causal graph tool proposed by Judea Pearl in 1995 to study the\napplication of causal inference in personal credit risk management. The\nspecific work includes: research on the construction method of causal inference\nindex system, definition of causality and feasibility proof of indicator\nsynthesis causal inference within this framework, application methods of causal\ngraph model and intervention alternative criteria in personal credit risk\nmanagement, and so on.\n", "date": "2024-03-17", "categories": []}
{"id": "2403.11219", "title": "Causality from Bottom to Top: A Survey", "abstract": "  Causality has become a fundamental approach for explaining the relationships\nbetween events, phenomena, and outcomes in various fields of study. It has\ninvaded various fields and applications, such as medicine, healthcare,\neconomics, finance, fraud detection, cybersecurity, education, public policy,\nrecommender systems, anomaly detection, robotics, control, sociology,\nmarketing, and advertising. In this paper, we survey its development over the\npast five decades, shedding light on the differences between causality and\nother approaches, as well as the preconditions for using it. Furthermore, the\npaper illustrates how causality interacts with new approaches such as\nArtificial Intelligence (AI), Generative AI (GAI), Machine and Deep Learning,\nReinforcement Learning (RL), and Fuzzy Logic. We study the impact of causality\non various fields, its contribution, and its interaction with state-of-the-art\napproaches. Additionally, the paper exemplifies the trustworthiness and\nexplainability of causality models. We offer several ways to evaluate causality\nmodels and discuss future directions.\n", "date": "2024-03-17", "categories": []}
{"id": "2403.11841", "title": "Pessimistic Causal Reinforcement Learning with Mediators for Confounded\n  Offline Data", "abstract": "  In real-world scenarios, datasets collected from randomized experiments are\noften constrained by size, due to limitations in time and budget. As a result,\nleveraging large observational datasets becomes a more attractive option for\nachieving high-quality policy learning. However, most existing offline\nreinforcement learning (RL) methods depend on two key\nassumptions--unconfoundedness and positivity--which frequently do not hold in\nobservational data contexts. Recognizing these challenges, we propose a novel\npolicy learning algorithm, PESsimistic CAusal Learning (PESCAL). We utilize the\nmediator variable based on front-door criterion to remove the confounding bias;\nadditionally, we adopt the pessimistic principle to address the distributional\nshift between the action distributions induced by candidate policies, and the\nbehavior policy that generates the observational data. Our key observation is\nthat, by incorporating auxiliary variables that mediate the effect of actions\non system dynamics, it is sufficient to learn a lower bound of the mediator\ndistribution function, instead of the Q-function, to partially mitigate the\nissue of distributional shift. This insight significantly simplifies our\nalgorithm, by circumventing the challenging task of sequential uncertainty\nquantification for the estimated Q-function. Moreover, we provide theoretical\nguarantees for the algorithms we propose, and demonstrate their efficacy\nthrough simulations, as well as real-world experiments utilizing offline\ndatasets from a leading ride-hailing platform.\n", "date": "2024-03-18", "categories": []}
{"id": "2307.09552", "title": "Self-Compatibility: Evaluating Causal Discovery without Ground Truth", "abstract": "  As causal ground truth is incredibly rare, causal discovery algorithms are\ncommonly only evaluated on simulated data. This is concerning, given that\nsimulations reflect preconceptions about generating processes regarding noise\ndistributions, model classes, and more. In this work, we propose a novel method\nfor falsifying the output of a causal discovery algorithm in the absence of\nground truth. Our key insight is that while statistical learning seeks\nstability across subsets of data points, causal learning should seek stability\nacross subsets of variables. Motivated by this insight, our method relies on a\nnotion of compatibility between causal graphs learned on different subsets of\nvariables. We prove that detecting incompatibilities can falsify wrongly\ninferred causal relations due to violation of assumptions or errors from finite\nsample effects. Although passing such compatibility tests is only a necessary\ncriterion for good performance, we argue that it provides strong evidence for\nthe causal models whenever compatibility entails strong implications for the\njoint distribution. We also demonstrate experimentally that detection of\nincompatibilities can aid in causal model selection.\n", "date": "2023-07-18", "categories": []}
{"id": "2403.05407", "title": "Algorithmic Identification of Essential Exogenous Nodes for Causal\n  Sufficiency in Brain Networks", "abstract": "  In the investigation of any causal mechanisms, such as the brain's causal\nnetworks, the assumption of causal sufficiency plays a critical role. Notably,\nneglecting this assumption can result in significant errors, a fact that is\noften disregarded in the causal analysis of brain networks. In this study, we\npropose an algorithmic identification approach for determining essential\nexogenous nodes that satisfy the critical need for causal sufficiency to adhere\nto it in such inquiries. Our approach consists of three main steps: First, by\ncapturing the essence of the Peter-Clark (PC) algorithm, we conduct\nindependence tests for pairs of regions within a network, as well as for the\nsame pairs conditioned on nodes from other networks. Next, we distinguish\ncandidate confounders by analyzing the differences between the conditional and\nunconditional results, using the Kolmogorov-Smirnov test. Subsequently, we\nutilize Non-Factorized identifiable Variational Autoencoders (NF-iVAE) along\nwith the Correlation Coefficient index (CCI) metric to identify the confounding\nvariables within these candidate nodes. Applying our method to the Human\nConnectome Projects (HCP) movie-watching task data, we demonstrate that while\ninteractions exist between dorsal and ventral regions, only dorsal regions\nserve as confounders for the visual networks, and vice versa. These findings\nalign consistently with those resulting from the neuroscientific perspective.\nFinally, we show the reliability of our results by testing 30 independent runs\nfor NF-iVAE initialization.\n", "date": "2024-03-08", "categories": []}
{"id": "2403.09683", "title": "Counterfactual Image Editing", "abstract": "  Counterfactual image editing is an important task in generative AI, which\nasks how an image would look if certain features were different. The current\nliterature on the topic focuses primarily on changing individual features while\nremaining silent about the causal relationships between these features, as\npresent in the real world. In this paper, we formalize the counterfactual image\nediting task using formal language, modeling the causal relationships between\nlatent generative factors and images through a special type of model called\naugmented structural causal models (ASCMs). Second, we show two fundamental\nimpossibility results: (1) counterfactual editing is impossible from i.i.d.\nimage samples and their corresponding labels alone; (2) even when the causal\nrelationships between the latent generative factors and images are available,\nno guarantees regarding the output of the model can be provided. Third, we\npropose a relaxation for this challenging problem by approximating\nnon-identifiable counterfactual distributions with a new family of\ncounterfactual-consistent estimators. This family exhibits the desirable\nproperty of preserving features that the user cares about across both factual\nand counterfactual worlds. Finally, we develop an efficient algorithm to\ngenerate counterfactual images by leveraging neural causal models.\n", "date": "2024-02-07", "categories": []}
{"id": "2403.09830", "title": "Towards the Reusability and Compositionality of Causal Representations", "abstract": "  Causal Representation Learning (CRL) aims at identifying high-level causal\nfactors and their relationships from high-dimensional observations, e.g.,\nimages. While most CRL works focus on learning causal representations in a\nsingle environment, in this work we instead propose a first step towards\nlearning causal representations from temporal sequences of images that can be\nadapted in a new environment, or composed across multiple related environments.\nIn particular, we introduce DECAF, a framework that detects which causal\nfactors can be reused and which need to be adapted from previously learned\ncausal representations. Our approach is based on the availability of\nintervention targets, that indicate which variables are perturbed at each time\nstep. Experiments on three benchmark datasets show that integrating our\nframework with four state-of-the-art CRL approaches leads to accurate\nrepresentations in a new environment with only a few samples.\n", "date": "2024-03-14", "categories": []}
{"id": "2403.10184", "title": "Lifted Causal Inference in Relational Domains", "abstract": "  Lifted inference exploits symmetries in probabilistic graphical models by\nusing a representative for indistinguishable objects, thereby speeding up query\nanswering while maintaining exact answers. Even though lifting is a\nwell-established technique for the task of probabilistic inference in\nrelational domains, it has not yet been applied to the task of causal\ninference. In this paper, we show how lifting can be applied to efficiently\ncompute causal effects in relational domains. More specifically, we introduce\nparametric causal factor graphs as an extension of parametric factor graphs\nincorporating causal knowledge and give a formal semantics of interventions\ntherein. We further present the lifted causal inference algorithm to compute\ncausal effects on a lifted level, thereby drastically speeding up causal\ninference compared to propositional inference, e.g., in causal Bayesian\nnetworks. In our empirical evaluation, we demonstrate the effectiveness of our\napproach.\n", "date": "2024-03-15", "categories": []}
{"id": "2403.09300", "title": "Recursive Causal Discovery", "abstract": "  Causal discovery, i.e., learning the causal graph from data, is often the\nfirst step toward the identification and estimation of causal effects, a key\nrequirement in numerous scientific domains. Causal discovery is hampered by two\nmain challenges: limited data results in errors in statistical testing and the\ncomputational complexity of the learning task is daunting. This paper builds\nupon and extends four of our prior publications (Mokhtarian et al., 2021;\nAkbari et al., 2021; Mokhtarian et al., 2022, 2023a). These works introduced\nthe concept of removable variables, which are the only variables that can be\nremoved recursively for the purpose of causal discovery. Presence and\nidentification of removable variables allow recursive approaches for causal\ndiscovery, a promising solution that helps to address the aforementioned\nchallenges by reducing the problem size successively. This reduction not only\nminimizes conditioning sets in each conditional independence (CI) test, leading\nto fewer errors but also significantly decreases the number of required CI\ntests. The worst-case performances of these methods nearly match the lower\nbound. In this paper, we present a unified framework for the proposed\nalgorithms, refined with additional details and enhancements for a coherent\npresentation. A comprehensive literature review is also included, comparing the\ncomputational complexity of our methods with existing approaches, showcasing\ntheir state-of-the-art efficiency. Another contribution of this paper is the\nrelease of RCD, a Python package that efficiently implements these algorithms.\nThis package is designed for practitioners and researchers interested in\napplying these methods in practical scenarios. The package is available at\ngithub.com/ban-epfl/rcd, with comprehensive documentation provided at\nrcdpackage.com.\n", "date": "2024-03-14", "categories": []}
{"id": "2403.09422", "title": "Mitigating attribute amplification in counterfactual image generation", "abstract": "  Causal generative modelling is gaining interest in medical imaging due to its\nability to answer interventional and counterfactual queries. Most work focuses\non generating counterfactual images that look plausible, using auxiliary\nclassifiers to enforce effectiveness of simulated interventions. We investigate\npitfalls in this approach, discovering the issue of attribute amplification,\nwhere unrelated attributes are spuriously affected during interventions,\nleading to biases across protected characteristics and disease status. We show\nthat attribute amplification is caused by the use of hard labels in the\ncounterfactual training process and propose soft counterfactual fine-tuning to\nmitigate this issue. Our method substantially reduces the amplification effect\nwhile maintaining effectiveness of generated images, demonstrated on a large\nchest X-ray dataset. Our work makes an important advancement towards more\nfaithful and unbiased causal modelling in medical imaging.\n", "date": "2024-03-14", "categories": []}
{"id": "2303.15027", "title": "A Survey on Causal Discovery Methods for I.I.D. and Time Series Data", "abstract": "  The ability to understand causality from data is one of the major milestones\nof human-level intelligence. Causal Discovery (CD) algorithms can identify the\ncause-effect relationships among the variables of a system from related\nobservational data with certain assumptions. Over the years, several methods\nhave been developed primarily based on the statistical properties of data to\nuncover the underlying causal mechanism. In this study, we present an extensive\ndiscussion on the methods designed to perform causal discovery from both\nindependent and identically distributed (I.I.D.) data and time series data. For\nthis purpose, we first introduce the common terminologies used in causal\ndiscovery literature and then provide a comprehensive discussion of the\nalgorithms designed to identify causal relations in different settings. We\nfurther discuss some of the benchmark datasets available for evaluating the\nalgorithmic performance, off-the-shelf tools or software packages to perform\ncausal discovery readily, and the common metrics used to evaluate these\nmethods. We also evaluate some widely used causal discovery algorithms on\nmultiple benchmark datasets and compare their performances. Finally, we\nconclude by discussing the research challenges and the applications of causal\ndiscovery algorithms in multiple areas of interest.\n", "date": "2023-03-27", "categories": []}
{"id": "2310.02895", "title": "CoLiDE: Concomitant Linear DAG Estimation", "abstract": "  We deal with the combinatorial problem of learning directed acyclic graph\n(DAG) structure from observational data adhering to a linear structural\nequation model (SEM). Leveraging advances in differentiable, nonconvex\ncharacterizations of acyclicity, recent efforts have advocated a continuous\nconstrained optimization paradigm to efficiently explore the space of DAGs.\nMost existing methods employ lasso-type score functions to guide this search,\nwhich (i) require expensive penalty parameter retuning when the\n$\\textit{unknown}$ SEM noise variances change across problem instances; and\n(ii) implicitly rely on limiting homoscedasticity assumptions. In this work, we\npropose a new convex score function for sparsity-aware learning of linear DAGs,\nwhich incorporates concomitant estimation of scale and thus effectively\ndecouples the sparsity parameter from the exogenous noise levels.\nRegularization via a smooth, nonconvex acyclicity penalty term yields CoLiDE\n($\\textbf{Co}$ncomitant $\\textbf{Li}$near $\\textbf{D}$AG\n$\\textbf{E}$stimation), a regression-based criterion amenable to efficient\ngradient computation and closed-form estimation of noise variances in\nheteroscedastic scenarios. Our algorithm outperforms state-of-the-art methods\nwithout incurring added complexity, especially when the DAGs are larger and the\nnoise level profile is heterogeneous. We also find CoLiDE exhibits enhanced\nstability manifested via reduced standard deviations in several domain-specific\nmetrics, underscoring the robustness of our novel linear DAG estimator.\n", "date": "2023-10-04", "categories": []}
{"id": "2403.08414", "title": "Causal Graph Neural Networks for Wildfire Danger Prediction", "abstract": "  Wildfire forecasting is notoriously hard due to the complex interplay of\ndifferent factors such as weather conditions, vegetation types and human\nactivities. Deep learning models show promise in dealing with this complexity\nby learning directly from data. However, to inform critical decision making, we\nargue that we need models that are right for the right reasons; that is, the\nimplicit rules learned should be grounded by the underlying processes driving\nwildfires. In that direction, we propose integrating causality with Graph\nNeural Networks (GNNs) that explicitly model the causal mechanism among complex\nvariables via graph learning. The causal adjacency matrix considers the\nsynergistic effect among variables and removes the spurious links from highly\ncorrelated impacts. Our methodology's effectiveness is demonstrated through\nsuperior performance forecasting wildfire patterns in the European boreal and\nmediterranean biome. The gain is especially prominent in a highly imbalanced\ndataset, showcasing an enhanced robustness of the model to adapt to regime\nshifts in functional relationships. Furthermore, SHAP values from our trained\nmodel further enhance our understanding of the model's inner workings.\n", "date": "2024-03-13", "categories": []}
{"id": "2403.08572", "title": "Caformer: Rethinking Time Series Analysis from Causal Perspective", "abstract": "  Time series analysis is a vital task with broad applications in various\ndomains. However, effectively capturing cross-dimension and cross-time\ndependencies in non-stationary time series poses significant challenges,\nparticularly in the context of environmental factors. The spurious correlation\ninduced by the environment confounds the causal relationships between\ncross-dimension and cross-time dependencies. In this paper, we introduce a\nnovel framework called Caformer (\\underline{\\textbf{Ca}}usal\nTrans\\underline{\\textbf{former}}) for time series analysis from a causal\nperspective. Specifically, our framework comprises three components: Dynamic\nLearner, Environment Learner, and Dependency Learner. The Dynamic Learner\nunveils dynamic interactions among dimensions, the Environment Learner\nmitigates spurious correlations caused by environment with a back-door\nadjustment, and the Dependency Learner aims to infer robust interactions across\nboth time and dimensions. Our Caformer demonstrates consistent state-of-the-art\nperformance across five mainstream time series analysis tasks, including long-\nand short-term forecasting, imputation, classification, and anomaly detection,\nwith proper interpretability.\n", "date": "2024-03-13", "categories": []}
{"id": "2403.08638", "title": "Disparate Effect Of Missing Mediators On Transportability of Causal\n  Effects", "abstract": "  Transported mediation effects provide an avenue to understand how upstream\ninterventions (such as improved neighborhood conditions like green spaces)\nwould work differently when applied to different populations as a result of\nfactors that mediate the effects. However, when mediators are missing in the\npopulation where the effect is to be transported, these estimates could be\nbiased. We study this issue of missing mediators, motivated by challenges in\npublic health, wherein mediators can be missing, not at random. We propose a\nsensitivity analysis framework that quantifies the impact of missing mediator\ndata on transported mediation effects. This framework enables us to identify\nthe settings under which the conditional transported mediation effect is\nrendered insignificant for the subgroup with missing mediator data.\nSpecifically, we provide the bounds on the transported mediation effect as a\nfunction of missingness. We then apply the framework to longitudinal data from\nthe Moving to Opportunity Study, a large-scale housing voucher experiment, to\nquantify the effect of missing mediators on transport effect estimates of\nvoucher receipt, an upstream intervention on living location, in childhood on\nsubsequent risk of mental health or substance use disorder mediated through\nparental health across sites. Our findings provide a tangible understanding of\nhow much missing data can be withstood for unbiased effect estimates.\n", "date": "2024-03-13", "categories": []}
{"id": "2306.14114", "title": "TNPAR: Topological Neural Poisson Auto-Regressive Model for Learning\n  Granger Causal Structure from Event Sequences", "abstract": "  Learning Granger causality from event sequences is a challenging but\nessential task across various applications. Most existing methods rely on the\nassumption that event sequences are independent and identically distributed\n(i.i.d.). However, this i.i.d. assumption is often violated due to the inherent\ndependencies among the event sequences. Fortunately, in practice, we find these\ndependencies can be modeled by a topological network, suggesting a potential\nsolution to the non-i.i.d. problem by introducing the prior topological network\ninto Granger causal discovery. This observation prompts us to tackle two\nensuing challenges: 1) how to model the event sequences while incorporating\nboth the prior topological network and the latent Granger causal structure, and\n2) how to learn the Granger causal structure. To this end, we devise a unified\ntopological neural Poisson auto-regressive model with two processes. In the\ngeneration process, we employ a variant of the neural Poisson process to model\nthe event sequences, considering influences from both the topological network\nand the Granger causal structure. In the inference process, we formulate an\namortized inference algorithm to infer the latent Granger causal structure. We\nencapsulate these two processes within a unified likelihood function, providing\nan end-to-end framework for this task. Experiments on simulated and real-world\ndata demonstrate the effectiveness of our approach.\n", "date": "2023-06-25", "categories": []}
{"id": "2403.02624", "title": "Pareto-Optimal Estimation and Policy Learning on Short-term and\n  Long-term Treatment Effects", "abstract": "  This paper focuses on developing Pareto-optimal estimation and policy\nlearning to identify the most effective treatment that maximizes the total\nreward from both short-term and long-term effects, which might conflict with\neach other. For example, a higher dosage of medication might increase the speed\nof a patient's recovery (short-term) but could also result in severe long-term\nside effects. Although recent works have investigated the problems about\nshort-term or long-term effects or the both, how to trade-off between them to\nachieve optimal treatment remains an open challenge. Moreover, when multiple\nobjectives are directly estimated using conventional causal representation\nlearning, the optimization directions among various tasks can conflict as well.\nIn this paper, we systematically investigate these issues and introduce a\nPareto-Efficient algorithm, comprising Pareto-Optimal Estimation (POE) and\nPareto-Optimal Policy Learning (POPL), to tackle them. POE incorporates a\ncontinuous Pareto module with representation balancing, enhancing estimation\nefficiency across multiple tasks. As for POPL, it involves deriving short-term\nand long-term outcomes linked with various treatment levels, facilitating an\nexploration of the Pareto frontier emanating from these outcomes. Results on\nboth the synthetic and real-world datasets demonstrate the superiority of our\nmethod.\n", "date": "2024-03-05", "categories": []}
{"id": "2403.07442", "title": "Proxy Methods for Domain Adaptation", "abstract": "  We study the problem of domain adaptation under distribution shift, where the\nshift is due to a change in the distribution of an unobserved, latent variable\nthat confounds both the covariates and the labels. In this setting, neither the\ncovariate shift nor the label shift assumptions apply. Our approach to\nadaptation employs proximal causal learning, a technique for estimating causal\neffects in settings where proxies of unobserved confounders are available. We\ndemonstrate that proxy variables allow for adaptation to distribution shift\nwithout explicitly recovering or modeling latent variables. We consider two\nsettings, (i) Concept Bottleneck: an additional ''concept'' variable is\nobserved that mediates the relationship between the covariates and labels; (ii)\nMulti-domain: training data from multiple source domains is available, where\neach source domain exhibits a different distribution over the latent\nconfounder. We develop a two-stage kernel estimation approach to adapt to\ncomplex distribution shifts in both settings. In our experiments, we show that\nour approach outperforms other methods, notably those which explicitly recover\nthe latent confounder.\n", "date": "2024-03-12", "categories": []}
{"id": "2403.07745", "title": "Probabilistic Easy Variational Causal Effect", "abstract": "  Let $X$ and $Z$ be random vectors, and $Y=g(X,Z)$. In this paper, on the one\nhand, for the case that $X$ and $Z$ are continuous, by using the ideas from the\ntotal variation and the flux of $g$, we develop a point of view in causal\ninference capable of dealing with a broad domain of causal problems. Indeed, we\nfocus on a function, called Probabilistic Easy Variational Causal Effect\n(PEACE), which can measure the direct causal effect of $X$ on $Y$ with respect\nto continuously and interventionally changing the values of $X$ while keeping\nthe value of $Z$ constant. PEACE is a function of $d\\ge 0$, which is a degree\nmanaging the strengths of probability density values $f(x|z)$. On the other\nhand, we generalize the above idea for the discrete case and show its\ncompatibility with the continuous case. Further, we investigate some properties\nof PEACE using measure theoretical concepts. Furthermore, we provide some\nidentifiability criteria and several examples showing the generic capability of\nPEACE. We note that PEACE can deal with the causal problems for which\nmicro-level or just macro-level changes in the value of the input variables are\nimportant. Finally, PEACE is stable under small changes in $\\partial\ng_{in}/\\partial x$ and the joint distribution of $X$ and $Z$, where $g_{in}$ is\nobtained from $g$ by removing all functional relationships defining $X$ and\n$Z$.\n", "date": "2024-03-12", "categories": []}
{"id": "2302.05451", "title": "Brain Effective Connectome based on fMRI and DTI Data: Bayesian Causal\n  Learning and Assessment", "abstract": "  Neuroscientific studies aim to find an accurate and reliable brain Effective\nConnectome (EC). Although current EC discovery methods have contributed to our\nunderstanding of brain organization, their performances are severely\nconstrained by the short sample size and poor temporal resolution of fMRI data,\nand high dimensionality of the brain connectome. By leveraging the DTI data as\nprior knowledge, we introduce two Bayesian causal discovery frameworks -- the\nBayesian GOLEM (BGOLEM) and Bayesian FGES (BFGES) methods -- that offer\nsignificantly more accurate and reliable ECs and address the shortcomings of\nthe existing causal discovery methods in discovering ECs based on only fMRI\ndata. Through a series of simulation studies on synthetic and hybrid (DTI of\nthe Human Connectome Project (HCP) subjects and synthetic fMRI) data, we\ndemonstrate the effectiveness of the proposed methods in discovering EC. To\nnumerically assess the improvement in the accuracy of ECs with our method on\nempirical data, we first introduce the Pseudo False Discovery Rate (PFDR) as a\nnew computational accuracy metric for causal discovery in the brain. We show\nthat our Bayesian methods achieve higher accuracy than traditional methods on\nHCP data. Additionally, we measure the reliability of discovered ECs using the\nRogers-Tanimoto index for test-retest data and show that our Bayesian methods\nprovide significantly more reproducible ECs than traditional methods. Overall,\nour study's numerical and graphical results highlight the potential for these\nframeworks to advance our understanding of brain function and organization\nsignificantly.\n", "date": "2023-02-10", "categories": []}
{"id": "2309.12819", "title": "Doubly Robust Proximal Causal Learning for Continuous Treatments", "abstract": "  Proximal causal learning is a promising framework for identifying the causal\neffect under the existence of unmeasured confounders. Within this framework,\nthe doubly robust (DR) estimator was derived and has shown its effectiveness in\nestimation, especially when the model assumption is violated. However, the\ncurrent form of the DR estimator is restricted to binary treatments, while the\ntreatment can be continuous in many real-world applications. The primary\nobstacle to continuous treatments resides in the delta function present in the\noriginal DR estimator, making it infeasible in causal effect estimation and\nintroducing a heavy computational burden in nuisance function estimation. To\naddress these challenges, we propose a kernel-based DR estimator that can well\nhandle continuous treatments. Equipped with its smoothness, we show that its\noracle form is a consistent approximation of the influence function. Further,\nwe propose a new approach to efficiently solve the nuisance functions. We then\nprovide a comprehensive convergence analysis in terms of the mean square error.\nWe demonstrate the utility of our estimator on synthetic datasets and\nreal-world applications.\n", "date": "2023-09-22", "categories": []}
{"id": "2312.06091", "title": "Learning Unknown Intervention Targets in Structural Causal Models from\n  Heterogeneous Data", "abstract": "  We study the problem of identifying the unknown intervention targets in\nstructural causal models where we have access to heterogeneous data collected\nfrom multiple environments. The unknown intervention targets are the set of\nendogenous variables whose corresponding exogenous noises change across the\nenvironments. We propose a two-phase approach which in the first phase recovers\nthe exogenous noises corresponding to unknown intervention targets whose\ndistributions have changed across environments. In the second phase, the\nrecovered noises are matched with the corresponding endogenous variables. For\nthe recovery phase, we provide sufficient conditions for learning these\nexogenous noises up to some component-wise invertible transformation. For the\nmatching phase, under the causal sufficiency assumption, we show that the\nproposed method uniquely identifies the intervention targets. In the presence\nof latent confounders, the intervention targets among the observed variables\ncannot be determined uniquely. We provide a candidate intervention target set\nwhich is a superset of the true intervention targets. Our approach improves\nupon the state of the art as the returned candidate set is always a subset of\nthe target set returned by previous work. Moreover, we do not require\nrestrictive assumptions such as linearity of the causal model or performing\ninvariance tests to learn whether a distribution is changing across\nenvironments which could be highly sample inefficient. Our experimental results\nshow the effectiveness of our proposed algorithm in practice.\n", "date": "2023-12-11", "categories": []}
{"id": "2403.05759", "title": "Membership Testing in Markov Equivalence Classes via Independence Query\n  Oracles", "abstract": "  Understanding causal relationships between variables is a fundamental problem\nwith broad impact in numerous scientific fields. While extensive research has\nbeen dedicated to learning causal graphs from data, its complementary concept\nof testing causal relationships has remained largely unexplored. While learning\ninvolves the task of recovering the Markov equivalence class (MEC) of the\nunderlying causal graph from observational data, the testing counterpart\naddresses the following critical question: Given a specific MEC and\nobservational data from some causal graph, can we determine if the\ndata-generating causal graph belongs to the given MEC?\n  We explore constraint-based testing methods by establishing bounds on the\nrequired number of conditional independence tests. Our bounds are in terms of\nthe size of the maximum undirected clique ($s$) of the given MEC. In the worst\ncase, we show a lower bound of $\\exp(\\Omega(s))$ independence tests. We then\ngive an algorithm that resolves the task with $\\exp(O(s))$ tests, matching our\nlower bound. Compared to the learning problem, where algorithms often use a\nnumber of independence tests that is exponential in the maximum in-degree, this\nshows that testing is relatively easier. In particular, it requires\nexponentially less independence tests in graphs featuring high in-degrees and\nsmall clique sizes. Additionally, using the DAG associahedron, we provide a\ngeometric interpretation of testing versus learning and discuss how our testing\nresult can aid learning.\n", "date": "2024-03-09", "categories": []}
{"id": "2403.06499", "title": "Detection of Unobserved Common Causes based on NML Code in Discrete,\n  Mixed, and Continuous Variables", "abstract": "  Causal discovery in the presence of unobserved common causes from\nobservational data only is a crucial but challenging problem. We categorize all\npossible causal relationships between two random variables into the following\nfour categories and aim to identify one from observed data: two cases in which\neither of the direct causality exists, a case that variables are independent,\nand a case that variables are confounded by latent confounders. Although\nexisting methods have been proposed to tackle this problem, they require\nunobserved variables to satisfy assumptions on the form of their equation\nmodels. In our previous study (Kobayashi et al., 2022), the first causal\ndiscovery method without such assumptions is proposed for discrete data and\nnamed CLOUD. Using Normalized Maximum Likelihood (NML) Code, CLOUD selects a\nmodel that yields the minimum codelength of the observed data from a set of\nmodel candidates. This paper extends CLOUD to apply for various data types\nacross discrete, mixed, and continuous. We not only performed theoretical\nanalysis to show the consistency of CLOUD in terms of the model selection, but\nalso demonstrated that CLOUD is more effective than existing methods in\ninferring causal relationships by extensive experiments on both synthetic and\nreal-world data.\n", "date": "2024-03-11", "categories": []}
{"id": "2311.04056", "title": "Multi-View Causal Representation Learning with Partial Observability", "abstract": "  We present a unified framework for studying the identifiability of\nrepresentations learned from simultaneously observed views, such as different\ndata modalities. We allow a partially observed setting in which each view\nconstitutes a nonlinear mixture of a subset of underlying latent variables,\nwhich can be causally related. We prove that the information shared across all\nsubsets of any number of views can be learned up to a smooth bijection using\ncontrastive learning and a single encoder per view. We also provide graphical\ncriteria indicating which latent variables can be identified through a simple\nset of rules, which we refer to as identifiability algebra. Our general\nframework and theoretical results unify and extend several previous works on\nmulti-view nonlinear ICA, disentanglement, and causal representation learning.\nWe experimentally validate our claims on numerical, image, and multi-modal data\nsets. Further, we demonstrate that the performance of prior methods is\nrecovered in different special cases of our setup. Overall, we find that access\nto multiple partial views enables us to identify a more fine-grained\nrepresentation, under the generally milder assumption of partial observability.\n", "date": "2023-11-07", "categories": []}
{"id": "2401.10632", "title": "Interventional Fairness on Partially Known Causal Graphs: A Constrained\n  Optimization Approach", "abstract": "  Fair machine learning aims to prevent discrimination against individuals or\nsub-populations based on sensitive attributes such as gender and race. In\nrecent years, causal inference methods have been increasingly used in fair\nmachine learning to measure unfairness by causal effects. However, current\nmethods assume that the true causal graph is given, which is often not true in\nreal-world applications. To address this limitation, this paper proposes a\nframework for achieving causal fairness based on the notion of interventions\nwhen the true causal graph is partially known. The proposed approach involves\nmodeling fair prediction using a Partially Directed Acyclic Graph (PDAG),\nspecifically, a class of causal DAGs that can be learned from observational\ndata combined with domain knowledge. The PDAG is used to measure causal\nfairness, and a constrained optimization problem is formulated to balance\nbetween fairness and accuracy. Results on both simulated and real-world\ndatasets demonstrate the effectiveness of this method.\n", "date": "2024-01-19", "categories": []}
{"id": "2403.04793", "title": "A Data-Driven Two-Phase Multi-Split Causal Ensemble Model for Time\n  Series", "abstract": "  Causal inference is a fundamental research topic for discovering the\ncause-effect relationships in many disciplines. However, not all algorithms are\nequally well-suited for a given dataset. For instance, some approaches may only\nbe able to identify linear relationships, while others are applicable for\nnon-linearities. Algorithms further vary in their sensitivity to noise and\ntheir ability to infer causal information from coupled vs. non-coupled time\nseries. Therefore, different algorithms often generate different causal\nrelationships for the same input. To achieve a more robust causal inference\nresult, this publication proposes a novel data-driven two-phase multi-split\ncausal ensemble model to combine the strengths of different causality base\nalgorithms. In comparison to existing approaches, the proposed ensemble method\nreduces the influence of noise through a data partitioning scheme in the first\nphase. To achieve this, the data are initially divided into several partitions\nand the base algorithms are applied to each partition. Subsequently, Gaussian\nmixture models are used to identify the causal relationships derived from the\ndifferent partitions that are likely to be valid. In the second phase, the\nidentified relationships from each base algorithm are then merged based on\nthree combination rules. The proposed ensemble approach is evaluated using\nmultiple metrics, among them a newly developed evaluation index for causal\nensemble approaches. We perform experiments using three synthetic datasets with\ndifferent volumes and complexity, which are specifically designed to test\ncausality detection methods under different circumstances while knowing the\nground truth causal relationships. In these experiments, our causality ensemble\noutperforms each of its base algorithms. In practical applications, the use of\nthe proposed method could hence lead to more robust and reliable causality\nresults.\n", "date": "2024-03-04", "categories": []}
{"id": "2212.01551", "title": "Quantify the Causes of Causal Emergence: Critical Conditions of\n  Uncertainty and Asymmetry in Causal Structure", "abstract": "  Beneficial to advanced computing devices, models with massive parameters are\nincreasingly employed to extract more information to enhance the precision in\ndescribing and predicting the patterns of objective systems. This phenomenon is\nparticularly pronounced in research domains associated with deep learning.\nHowever, investigations of causal relationships based on statistical and\ninformational theories have posed an interesting and valuable challenge to\nlarge-scale models in the recent decade. Macroscopic models with fewer\nparameters can outperform their microscopic counterparts with more parameters\nin effectively representing the system. This valuable situation is called\n\"Causal Emergence.\" This paper introduces a quantification framework, according\nto the Effective Information and Transition Probability Matrix, for assessing\nnumerical conditions of Causal Emergence as theoretical constraints of its\noccurrence. Specifically, our results quantitatively prove the cause of Causal\nEmergence. By a particular coarse-graining strategy, optimizing uncertainty and\nasymmetry within the model's causal structure is significantly more influential\nthan losing maximum information due to variations in model scales. Moreover, by\ndelving into the potential exhibited by Partial Information Decomposition and\nDeep Learning networks in the study of Causal Emergence, we discuss potential\napplication scenarios where our quantification framework could play a role in\nfuture investigations of Causal Emergence.\n", "date": "2022-12-03", "categories": []}
{"id": "2403.04039", "title": "Sample size planning for conditional counterfactual mean estimation with\n  a K-armed randomized experiment", "abstract": "  We cover how to determine a sufficiently large sample size for a $K$-armed\nrandomized experiment in order to estimate conditional counterfactual\nexpectations in data-driven subgroups. The sub-groups can be output by any\nfeature space partitioning algorithm, including as defined by binning users\nhaving similar predictive scores or as defined by a learned policy tree. After\ncarefully specifying the inference target, a minimum confidence level, and a\nmaximum margin of error, the key is to turn the original goal into a\nsimultaneous inference problem where the recommended sample size to offset an\nincreased possibility of estimation error is directly related to the number of\ninferences to be conducted. Given a fixed sample size budget, our result allows\nus to invert the question to one about the feasible number of treatment arms or\npartition complexity (e.g. number of decision tree leaves). Using policy trees\nto learn sub-groups, we evaluate our nominal guarantees on a large\npublicly-available randomized experiment test data set.\n", "date": "2024-03-06", "categories": []}
{"id": "2403.04236", "title": "Regularized DeepIV with Model Selection", "abstract": "  In this paper, we study nonparametric estimation of instrumental variable\n(IV) regressions. While recent advancements in machine learning have introduced\nflexible methods for IV estimation, they often encounter one or more of the\nfollowing limitations: (1) restricting the IV regression to be uniquely\nidentified; (2) requiring minimax computation oracle, which is highly unstable\nin practice; (3) absence of model selection procedure. In this paper, we\npresent the first method and analysis that can avoid all three limitations,\nwhile still enabling general function approximation. Specifically, we propose a\nminimax-oracle-free method called Regularized DeepIV (RDIV) regression that can\nconverge to the least-norm IV solution. Our method consists of two stages:\nfirst, we learn the conditional distribution of covariates, and by utilizing\nthe learned distribution, we learn the estimator by minimizing a\nTikhonov-regularized loss function. We further show that our method allows\nmodel selection procedures that can achieve the oracle rates in the\nmisspecified regime. When extended to an iterative estimator, our method\nmatches the current state-of-the-art convergence rate. Our method is a Tikhonov\nregularized variant of the popular DeepIV method with a non-parametric MLE\nfirst-stage estimator, and our results provide the first rigorous guarantees\nfor this empirically used method, showcasing the importance of regularization\nwhich was absent from the original work.\n", "date": "2024-03-07", "categories": []}
{"id": "2302.00064", "title": "Evaluating Temporal Observation-Based Causal Discovery Techniques\n  Applied to Road Driver Behaviour", "abstract": "  Autonomous robots are required to reason about the behaviour of dynamic\nagents in their environment. The creation of models to describe these\nrelationships is typically accomplished through the application of causal\ndiscovery techniques. However, as it stands observational causal discovery\ntechniques struggle to adequately cope with conditions such as causal sparsity\nand non-stationarity typically seen during online usage in autonomous agent\ndomains. Meanwhile, interventional techniques are not always feasible due to\ndomain restrictions. In order to better explore the issues facing observational\ntechniques and promote further discussion of these topics we carry out a\nbenchmark across 10 contemporary observational temporal causal discovery\nmethods in the domain of autonomous driving. By evaluating these methods upon\ncausal scenes drawn from real world datasets in addition to those generated\nsynthetically we highlight where improvements need to be made in order to\nfacilitate the application of causal discovery techniques to the aforementioned\nuse-cases. Finally, we discuss potential directions for future work that could\nhelp better tackle the difficulties currently experienced by state of the art\ntechniques.\n", "date": "2023-01-31", "categories": []}
{"id": "2306.03354", "title": "Simulation-Based Counterfactual Causal Discovery on Real World Driver\n  Behaviour", "abstract": "  Being able to reason about how one's behaviour can affect the behaviour of\nothers is a core skill required of intelligent driving agents. Despite this,\nthe state of the art struggles to meet the need of agents to discover causal\nlinks between themselves and others. Observational approaches struggle because\nof the non-stationarity of causal links in dynamic environments, and the\nsparsity of causal interactions while requiring the approaches to work in an\nonline fashion. Meanwhile interventional approaches are impractical as a\nvehicle cannot experiment with its actions on a public road. To counter the\nissue of non-stationarity we reformulate the problem in terms of extracted\nevents, while the previously mentioned restriction upon interventions can be\novercome with the use of counterfactual simulation. We present three variants\nof the proposed counterfactual causal discovery method and evaluate these\nagainst state of the art observational temporal causal discovery methods across\n3396 causal scenes extracted from a real world driving dataset. We find that\nthe proposed method significantly outperforms the state of the art on the\nproposed task quantitatively and can offer additional insights by comparing the\noutcome of an alternate series of decisions in a way that observational and\ninterventional approaches cannot.\n", "date": "2023-06-06", "categories": []}
{"id": "2403.03240", "title": "Triple/Debiased Lasso for Statistical Inference of Conditional Average\n  Treatment Effects", "abstract": "  This study investigates the estimation and the statistical inference about\nConditional Average Treatment Effects (CATEs), which have garnered attention as\na metric representing individualized causal effects. In our data-generating\nprocess, we assume linear models for the outcomes associated with binary\ntreatments and define the CATE as a difference between the expected outcomes of\nthese linear models. This study allows the linear models to be\nhigh-dimensional, and our interest lies in consistent estimation and\nstatistical inference for the CATE. In high-dimensional linear regression, one\ntypical approach is to assume sparsity. However, in our study, we do not assume\nsparsity directly. Instead, we consider sparsity only in the difference of the\nlinear models. We first use a doubly robust estimator to approximate this\ndifference and then regress the difference on covariates with Lasso\nregularization. Although this regression estimator is consistent for the CATE,\nwe further reduce the bias using the techniques in double/debiased machine\nlearning (DML) and debiased Lasso, leading to $\\sqrt{n}$-consistency and\nconfidence intervals. We refer to the debiased estimator as the triple/debiased\nLasso (TDL), applying both DML and debiased Lasso techniques. We confirm the\nsoundness of our proposed method through simulation studies.\n", "date": "2024-03-05", "categories": []}
{"id": "2403.03274", "title": "From Noise to Signal: Unveiling Treatment Effects from Digital Health\n  Data through Pharmacology-Informed Neural-SDE", "abstract": "  Digital health technologies (DHT), such as wearable devices, provide\npersonalized, continuous, and real-time monitoring of patient. These\ntechnologies are contributing to the development of novel therapies and\npersonalized medicine. Gaining insight from these technologies requires\nappropriate modeling techniques to capture clinically-relevant changes in\ndisease state. The data generated from these devices is characterized by being\nstochastic in nature, may have missing elements, and exhibits considerable\ninter-individual variability - thereby making it difficult to analyze using\ntraditional longitudinal modeling techniques. We present a novel\npharmacology-informed neural stochastic differential equation (SDE) model\ncapable of addressing these challenges. Using synthetic data, we demonstrate\nthat our approach is effective in identifying treatment effects and learning\ncausal relationships from stochastic data, thereby enabling counterfactual\nsimulation.\n", "date": "2024-03-05", "categories": []}
{"id": "2403.03772", "title": "AcceleratedLiNGAM: Learning Causal DAGs at the speed of GPUs", "abstract": "  Existing causal discovery methods based on combinatorial optimization or\nsearch are slow, prohibiting their application on large-scale datasets. In\nresponse, more recent methods attempt to address this limitation by formulating\ncausal discovery as structure learning with continuous optimization but such\napproaches thus far provide no statistical guarantees. In this paper, we show\nthat by efficiently parallelizing existing causal discovery methods, we can in\nfact scale them to thousands of dimensions, making them practical for\nsubstantially larger-scale problems. In particular, we parallelize the LiNGAM\nmethod, which is quadratic in the number of variables, obtaining up to a\n32-fold speed-up on benchmark datasets when compared with existing sequential\nimplementations. Specifically, we focus on the causal ordering subprocedure in\nDirectLiNGAM and implement GPU kernels to accelerate it. This allows us to\napply DirectLiNGAM to causal inference on large-scale gene expression data with\ngenetic interventions yielding competitive results compared with specialized\ncontinuous optimization methods, and Var-LiNGAM for causal discovery on U.S.\nstock data.\n", "date": "2024-03-06", "categories": []}
{"id": "2403.03791", "title": "KG-TREAT: Pre-training for Treatment Effect Estimation by Synergizing\n  Patient Data with Knowledge Graphs", "abstract": "  Treatment effect estimation (TEE) is the task of determining the impact of\nvarious treatments on patient outcomes. Current TEE methods fall short due to\nreliance on limited labeled data and challenges posed by sparse and\nhigh-dimensional observational patient data. To address the challenges, we\nintroduce a novel pre-training and fine-tuning framework, KG-TREAT, which\nsynergizes large-scale observational patient data with biomedical knowledge\ngraphs (KGs) to enhance TEE. Unlike previous approaches, KG-TREAT constructs\ndual-focus KGs and integrates a deep bi-level attention synergy method for\nin-depth information fusion, enabling distinct encoding of treatment-covariate\nand outcome-covariate relationships. KG-TREAT also incorporates two\npre-training tasks to ensure a thorough grounding and contextualization of\npatient data and KGs. Evaluation on four downstream TEE tasks shows KG-TREAT's\nsuperiority over existing methods, with an average improvement of 7% in Area\nunder the ROC Curve (AUC) and 9% in Influence Function-based Precision of\nEstimating Heterogeneous Effects (IF-PEHE). The effectiveness of our estimated\ntreatment effects is further affirmed by alignment with established randomized\nclinical trial findings.\n", "date": "2024-03-06", "categories": []}
{"id": "2211.17029", "title": "Directed Acyclic Graph Structure Learning from Dynamic Graphs", "abstract": "  Estimating the structure of directed acyclic graphs (DAGs) of features\n(variables) plays a vital role in revealing the latent data generation process\nand providing causal insights in various applications. Although there have been\nmany studies on structure learning with various types of data, the structure\nlearning on the dynamic graph has not been explored yet, and thus we study the\nlearning problem of node feature generation mechanism on such ubiquitous\ndynamic graph data. In a dynamic graph, we propose to simultaneously estimate\ncontemporaneous relationships and time-lagged interaction relationships between\nthe node features. These two kinds of relationships form a DAG, which could\neffectively characterize the feature generation process in a concise way. To\nlearn such a DAG, we cast the learning problem as a continuous score-based\noptimization problem, which consists of a differentiable score function to\nmeasure the validity of the learned DAGs and a smooth acyclicity constraint to\nensure the acyclicity of the learned DAGs. These two components are translated\ninto an unconstraint augmented Lagrangian objective which could be minimized by\nmature continuous optimization techniques. The resulting algorithm, named\nGraphNOTEARS, outperforms baselines on simulated data across a wide range of\nsettings that may encounter in real-world applications. We also apply the\nproposed approach on two dynamic graphs constructed from the real-world Yelp\ndataset, demonstrating our method could learn the connections between node\nfeatures, which conforms with the domain knowledge.\n", "date": "2022-11-30", "categories": []}
{"id": "2310.04295", "title": "Identifying Representations for Intervention Extrapolation", "abstract": "  The premise of identifiable and causal representation learning is to improve\nthe current representation learning paradigm in terms of generalizability or\nrobustness. Despite recent progress in questions of identifiability, more\ntheoretical results demonstrating concrete advantages of these methods for\ndownstream tasks are needed. In this paper, we consider the task of\nintervention extrapolation: predicting how interventions affect an outcome,\neven when those interventions are not observed at training time, and show that\nidentifiable representations can provide an effective solution to this task\neven if the interventions affect the outcome non-linearly. Our setup includes\nan outcome Y, observed features X, which are generated as a non-linear\ntransformation of latent features Z, and exogenous action variables A, which\ninfluence Z. The objective of intervention extrapolation is to predict how\ninterventions on A that lie outside the training support of A affect Y. Here,\nextrapolation becomes possible if the effect of A on Z is linear and the\nresidual when regressing Z on A has full support. As Z is latent, we combine\nthe task of intervention extrapolation with identifiable representation\nlearning, which we call Rep4Ex: we aim to map the observed features X into a\nsubspace that allows for non-linear extrapolation in A. We show that the hidden\nrepresentation is identifiable up to an affine transformation in Z-space, which\nis sufficient for intervention extrapolation. The identifiability is\ncharacterized by a novel constraint describing the linearity assumption of A on\nZ. Based on this insight, we propose a method that enforces the linear\ninvariance constraint and can be combined with any type of autoencoder. We\nvalidate our theoretical findings through synthetic experiments and show that\nour approach succeeds in predicting the effects of unseen interventions.\n", "date": "2023-10-06", "categories": []}
{"id": "2310.19794", "title": "Robust Causal Bandits for Linear Models", "abstract": "  Sequential design of experiments for optimizing a reward function in causal\nsystems can be effectively modeled by the sequential design of interventions in\ncausal bandits (CBs). In the existing literature on CBs, a critical assumption\nis that the causal models remain constant over time. However, this assumption\ndoes not necessarily hold in complex systems, which constantly undergo temporal\nmodel fluctuations. This paper addresses the robustness of CBs to such model\nfluctuations. The focus is on causal systems with linear structural equation\nmodels (SEMs). The SEMs and the time-varying pre- and post-interventional\nstatistical models are all unknown. Cumulative regret is adopted as the design\ncriteria, based on which the objective is to design a sequence of interventions\nthat incur the smallest cumulative regret with respect to an oracle aware of\nthe entire causal model and its fluctuations. First, it is established that the\nexisting approaches fail to maintain regret sub-linearity with even a few\ninstances of model deviation. Specifically, when the number of instances with\nmodel deviation is as few as $T^\\frac{1}{2L}$, where $T$ is the time horizon\nand $L$ is the longest causal path in the graph, the existing algorithms will\nhave linear regret in $T$. Next, a robust CB algorithm is designed, and its\nregret is analyzed, where upper and information-theoretic lower bounds on the\nregret are established. Specifically, in a graph with $N$ nodes and maximum\ndegree $d$, under a general measure of model deviation $C$, the cumulative\nregret is upper bounded by $\\tilde{\\mathcal{O}}(d^{L-\\frac{1}{2}}(\\sqrt{NT} +\nNC))$ and lower bounded by $\\Omega(d^{\\frac{L}{2}-2}\\max\\{\\sqrt{T},d^2C\\})$.\nComparing these bounds establishes that the proposed algorithm achieves nearly\noptimal $\\tilde{\\mathcal{O}}(\\sqrt{T})$ regret when $C$ is $o(\\sqrt{T})$ and\nmaintains sub-linear regret for a broader range of $C$.\n", "date": "2023-10-30", "categories": []}
{"id": "2311.03002", "title": "Estimating treatment effects from single-arm trials via latent-variable\n  modeling", "abstract": "  Randomized controlled trials (RCTs) are the accepted standard for treatment\neffect estimation but they can be infeasible due to ethical reasons and\nprohibitive costs. Single-arm trials, where all patients belong to the\ntreatment group, can be a viable alternative but require access to an external\ncontrol group. We propose an identifiable deep latent-variable model for this\nscenario that can also account for missing covariate observations by modeling\ntheir structured missingness patterns. Our method uses amortized variational\ninference to learn both group-specific and identifiable shared latent\nrepresentations, which can subsequently be used for {\\em (i)} patient matching\nif treatment outcomes are not available for the treatment group, or for {\\em\n(ii)} direct treatment effect estimation assuming outcomes are available for\nboth groups. We evaluate the model on a public benchmark as well as on a data\nset consisting of a published RCT study and real-world electronic health\nrecords. Compared to previous methods, our results show improved performance\nboth for direct treatment effect estimation as well as for effect estimation\nvia patient matching.\n", "date": "2023-11-06", "categories": []}
{"id": "2311.14058", "title": "Identification for Tree-shaped Structural Causal Models in Polynomial\n  Time", "abstract": "  Linear structural causal models (SCMs) are used to express and analyse the\nrelationships between random variables. Direct causal effects are represented\nas directed edges and confounding factors as bidirected edges. Identifying the\ncausal parameters from correlations between the nodes is an open problem in\nartificial intelligence. In this paper, we study SCMs whose directed component\nforms a tree. Van der Zander et al. (AISTATS'22, PLMR 151, pp. 6770--6792,\n2022) give a PSPACE-algorithm for the identification problem in this case,\nwhich is a significant improvement over the general Gr\\\"obner basis approach,\nwhich has doubly-exponential time complexity in the number of structural\nparameters. In this work, we present a randomized polynomial-time algorithm,\nwhich solves the identification problem for tree-shaped SCMs. For every\nstructural parameter, our algorithms decides whether it is generically\nidentifiable, generically 2-identifiable, or generically unidentifiable. (No\nother cases can occur.) In the first two cases, it provides one or two\nfractional affine square root terms of polynomials (FASTPs) for the\ncorresponding parameter, respectively.\n", "date": "2023-11-23", "categories": []}
{"id": "2403.02467", "title": "Applied Causal Inference Powered by ML and AI", "abstract": "  An introduction to the emerging fusion of machine learning and causal\ninference. The book presents ideas from classical structural equation models\n(SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and\nstructural causal models (SCMs), and covers Double/Debiased Machine Learning\nmethods to do inference in such models using modern predictive tools.\n", "date": "2024-03-04", "categories": []}
{"id": "2403.02698", "title": "Causal Walk: Debiasing Multi-Hop Fact Verification with Front-Door\n  Adjustment", "abstract": "  Conventional multi-hop fact verification models are prone to rely on spurious\ncorrelations from the annotation artifacts, leading to an obvious performance\ndecline on unbiased datasets. Among the various debiasing works, the causal\ninference-based methods become popular by performing theoretically guaranteed\ndebiasing such as casual intervention or counterfactual reasoning. However,\nexisting causal inference-based debiasing methods, which mainly formulate fact\nverification as a single-hop reasoning task to tackle shallow bias patterns,\ncannot deal with the complicated bias patterns hidden in multiple hops of\nevidence. To address the challenge, we propose Causal Walk, a novel method for\ndebiasing multi-hop fact verification from a causal perspective with front-door\nadjustment. Specifically, in the structural causal model, the reasoning path\nbetween the treatment (the input claim-evidence graph) and the outcome (the\nveracity label) is introduced as the mediator to block the confounder. With the\nfront-door adjustment, the causal effect between the treatment and the outcome\nis decomposed into the causal effect between the treatment and the mediator,\nwhich is estimated by applying the idea of random walk, and the causal effect\nbetween the mediator and the outcome, which is estimated with normalized\nweighted geometric mean approximation. To investigate the effectiveness of the\nproposed method, an adversarial multi-hop fact verification dataset and a\nsymmetric multi-hop fact verification dataset are proposed with the help of the\nlarge language model. Experimental results show that Causal Walk outperforms\nsome previous debiasing methods on both existing datasets and the newly\nconstructed datasets. Code and data will be released at\nhttps://github.com/zcccccz/CausalWalk.\n", "date": "2024-03-05", "categories": []}
{"id": "2307.16387", "title": "Relation-First Modeling Paradigm for Causal Representation Learning\n  toward the Development of AGI", "abstract": "  The traditional i.i.d.-based learning paradigm faces inherent challenges in\naddressing causal relationships, which has become increasingly evident with the\nrise of applications in causal representation learning.\n  Our understanding of causality naturally requires a perspective as the\ncreator rather than observer, as the ``what...if'' questions only hold within\nthe possible world we conceive. The traditional perspective limits capturing\ndynamic causal outcomes and leads to compensatory efforts such as the reliance\non hidden confounders. This paper lays the groundwork for the new perspective,\nwhich enables the \\emph{relation-first} modeling paradigm for causality. Also,\nit introduces the Relation-Indexed Representation Learning (RIRL) as a\npractical implementation, supported by experiments that validate its efficacy.\n", "date": "2023-07-31", "categories": []}
{"id": "2311.18274", "title": "Semiparametric Efficient Inference in Adaptive Experiments", "abstract": "  We consider the problem of efficient inference of the Average Treatment\nEffect in a sequential experiment where the policy governing the assignment of\nsubjects to treatment or control can change over time. We first provide a\ncentral limit theorem for the Adaptive Augmented Inverse-Probability Weighted\nestimator, which is semiparametric efficient, under weaker assumptions than\nthose previously made in the literature. This central limit theorem enables\nefficient inference at fixed sample sizes. We then consider a sequential\ninference setting, deriving both asymptotic and nonasymptotic confidence\nsequences that are considerably tighter than previous methods. These\nanytime-valid methods enable inference under data-dependent stopping times\n(sample sizes). Additionally, we use propensity score truncation techniques\nfrom the recent off-policy estimation literature to reduce the finite sample\nvariance of our estimator without affecting the asymptotic variance. Empirical\nresults demonstrate that our methods yield narrower confidence sequences than\nthose previously developed in the literature while maintaining time-uniform\nerror control.\n", "date": "2023-11-30", "categories": []}
{"id": "2403.00178", "title": "Causal Graph ODE: Continuous Treatment Effect Modeling in Multi-agent\n  Dynamical Systems", "abstract": "  Real-world multi-agent systems are often dynamic and continuous, where the\nagents co-evolve and undergo changes in their trajectories and interactions\nover time. For example, the COVID-19 transmission in the U.S. can be viewed as\na multi-agent system, where states act as agents and daily population movements\nbetween them are interactions. Estimating the counterfactual outcomes in such\nsystems enables accurate future predictions and effective decision-making, such\nas formulating COVID-19 policies. However, existing methods fail to model the\ncontinuous dynamic effects of treatments on the outcome, especially when\nmultiple treatments (e.g., \"stay-at-home\" and \"get-vaccine\" policies) are\napplied simultaneously. To tackle this challenge, we propose Causal Graph\nOrdinary Differential Equations (CAG-ODE), a novel model that captures the\ncontinuous interaction among agents using a Graph Neural Network (GNN) as the\nODE function. The key innovation of our model is to learn time-dependent\nrepresentations of treatments and incorporate them into the ODE function,\nenabling precise predictions of potential outcomes. To mitigate confounding\nbias, we further propose two domain adversarial learning-based objectives,\nwhich enable our model to learn balanced continuous representations that are\nnot affected by treatments or interference. Experiments on two datasets (i.e.,\nCOVID-19 and tumor growth) demonstrate the superior performance of our proposed\nmodel.\n", "date": "2024-02-29", "categories": []}
{"id": "2403.00233", "title": "Causal Bandits with General Causal Models and Interventions", "abstract": "  This paper considers causal bandits (CBs) for the sequential design of\ninterventions in a causal system. The objective is to optimize a reward\nfunction via minimizing a measure of cumulative regret with respect to the best\nsequence of interventions in hindsight. The paper advances the results on CBs\nin three directions. First, the structural causal models (SCMs) are assumed to\nbe unknown and drawn arbitrarily from a general class $\\mathcal{F}$ of\nLipschitz-continuous functions. Existing results are often focused on\n(generalized) linear SCMs. Second, the interventions are assumed to be\ngeneralized soft with any desired level of granularity, resulting in an\ninfinite number of possible interventions. The existing literature, in\ncontrast, generally adopts atomic and hard interventions. Third, we provide\ngeneral upper and lower bounds on regret. The upper bounds subsume (and\nimprove) known bounds for special cases. The lower bounds are generally\nhitherto unknown. These bounds are characterized as functions of the (i) graph\nparameters, (ii) eluder dimension of the space of SCMs, denoted by\n$\\operatorname{dim}(\\mathcal{F})$, and (iii) the covering number of the\nfunction space, denoted by ${\\rm cn}(\\mathcal{F})$. Specifically, the\ncumulative achievable regret over horizon $T$ is $\\mathcal{O}(K\nd^{L-1}\\sqrt{T\\operatorname{dim}(\\mathcal{F}) \\log({\\rm cn}(\\mathcal{F}))})$,\nwhere $K$ is related to the Lipschitz constants, $d$ is the graph's maximum\nin-degree, and $L$ is the length of the longest causal path. The upper bound is\nfurther refined for special classes of SCMs (neural network, polynomial, and\nlinear), and their corresponding lower bounds are provided.\n", "date": "2024-03-01", "categories": []}
{"id": "2403.00694", "title": "Defining Expertise: Applications to Treatment Effect Estimation", "abstract": "  Decision-makers are often experts of their domain and take actions based on\ntheir domain knowledge. Doctors, for instance, may prescribe treatments by\npredicting the likely outcome of each available treatment. Actions of an expert\nthus naturally encode part of their domain knowledge, and can help make\ninferences within the same domain: Knowing doctors try to prescribe the best\ntreatment for their patients, we can tell treatments prescribed more frequently\nare likely to be more effective. Yet in machine learning, the fact that most\ndecision-makers are experts is often overlooked, and \"expertise\" is seldom\nleveraged as an inductive bias. This is especially true for the literature on\ntreatment effect estimation, where often the only assumption made about actions\nis that of overlap. In this paper, we argue that expertise - particularly the\ntype of expertise the decision-makers of a domain are likely to have - can be\ninformative in designing and selecting methods for treatment effect estimation.\nWe formally define two types of expertise, predictive and prognostic, and\ndemonstrate empirically that: (i) the prominent type of expertise in a domain\nsignificantly influences the performance of different methods in treatment\neffect estimation, and (ii) it is possible to predict the type of expertise\npresent in a dataset, which can provide a quantitative basis for model\nselection.\n", "date": "2024-03-01", "categories": []}
{"id": "2402.03726", "title": "Learning Granger Causality from Instance-wise Self-attentive Hawkes\n  Processes", "abstract": "  We address the problem of learning Granger causality from asynchronous,\ninterdependent, multi-type event sequences. In particular, we are interested in\ndiscovering instance-level causal structures in an unsupervised manner.\nInstance-level causality identifies causal relationships among individual\nevents, providing more fine-grained information for decision-making. Existing\nwork in the literature either requires strong assumptions, such as linearity in\nthe intensity function, or heuristically defined model parameters that do not\nnecessarily meet the requirements of Granger causality. We propose\nInstance-wise Self-Attentive Hawkes Processes (ISAHP), a novel deep learning\nframework that can directly infer the Granger causality at the event instance\nlevel. ISAHP is the first neural point process model that meets the\nrequirements of Granger causality. It leverages the self-attention mechanism of\nthe transformer to align with the principles of Granger causality. We\nempirically demonstrate that ISAHP is capable of discovering complex\ninstance-level causal structures that cannot be handled by classical models. We\nalso show that ISAHP achieves state-of-the-art performance in proxy tasks\ninvolving type-level causal discovery and instance-level event type prediction.\n", "date": "2024-02-06", "categories": []}
{"id": "2306.05751", "title": "Advancing Counterfactual Inference through Nonlinear Quantile Regression", "abstract": "  The capacity to address counterfactual \"what if\" inquiries is crucial for\nunderstanding and making use of causal influences. Traditional counterfactual\ninference, under Pearls' counterfactual framework, typically depends on having\naccess to or estimating a structural causal model. Yet, in practice, this\ncausal model is often unknown and might be challenging to identify. Hence, this\npaper aims to perform reliable counterfactual inference based solely on\nobservational data and the (learned) qualitative causal structure, without\nnecessitating a predefined causal model or even direct estimations of\nconditional distributions. To this end, we establish a novel connection between\ncounterfactual inference and quantile regression and show that counterfactual\ninference can be reframed as an extended quantile regression problem. Building\non this insight, we propose a practical framework for efficient and effective\ncounterfactual inference implemented with neural networks under a bi-level\noptimization scheme. The proposed approach enhances the capacity to generalize\nestimated counterfactual outcomes to unseen data, thereby providing an upper\nbound on the generalization error. Furthermore, empirical evidence demonstrates\nits superior statistical efficiency in comparison to existing methods.\nEmpirical results conducted on multiple datasets offer compelling support for\nour theoretical assertions.\n", "date": "2023-06-09", "categories": []}
{"id": "2402.10240", "title": "A Dynamical View of the Question of Why", "abstract": "  We address causal reasoning in multivariate time series data generated by\nstochastic processes. Existing approaches are largely restricted to static\nsettings, ignoring the continuity and emission of variations across time. In\ncontrast, we propose a learning paradigm that directly establishes causation\nbetween events in the course of time. We present two key lemmas to compute\ncausal contributions and frame them as reinforcement learning problems. Our\napproach offers formal and computational tools for uncovering and quantifying\ncausal relationships in diffusion processes, subsuming various important\nsettings such as discrete-time Markov decision processes. Finally, in fairly\nintricate experiments and through sheer learning, our framework reveals and\nquantifies causal links, which otherwise seem inexplicable.\n", "date": "2024-02-14", "categories": []}
{"id": "2402.13241", "title": "Federated Causal Discovery from Heterogeneous Data", "abstract": "  Conventional causal discovery methods rely on centralized data, which is\ninconsistent with the decentralized nature of data in many real-world\nsituations. This discrepancy has motivated the development of federated causal\ndiscovery (FCD) approaches. However, existing FCD methods may be limited by\ntheir potentially restrictive assumptions of identifiable functional causal\nmodels or homogeneous data distributions, narrowing their applicability in\ndiverse scenarios. In this paper, we propose a novel FCD method attempting to\naccommodate arbitrary causal models and heterogeneous data. We first utilize a\nsurrogate variable corresponding to the client index to account for the data\nheterogeneity across different clients. We then develop a federated conditional\nindependence test (FCIT) for causal skeleton discovery and establish a\nfederated independent change principle (FICP) to determine causal directions.\nThese approaches involve constructing summary statistics as a proxy of the raw\ndata to protect data privacy. Owing to the nonparametric properties, FCIT and\nFICP make no assumption about particular functional forms, thereby facilitating\nthe handling of arbitrary causal models. We conduct extensive experiments on\nsynthetic and real datasets to show the efficacy of our method. The code is\navailable at https://github.com/lokali/FedCDH.git.\n", "date": "2024-02-20", "categories": []}
{"id": "2402.16909", "title": "Impact of Physical Activity on Quality of Life During Pregnancy: A\n  Causal ML Approach", "abstract": "  The concept of Quality of Life (QoL) refers to a holistic measurement of an\nindividual's well-being, incorporating psychological and social aspects.\nPregnant women, especially those with obesity and stress, often experience\nlower QoL. Physical activity (PA) has shown the potential to enhance the QoL.\nHowever, pregnant women who are overweight and obese rarely meet the\nrecommended level of PA. Studies have investigated the relationship between PA\nand QoL during pregnancy using correlation-based approaches. These methods aim\nto discover spurious correlations between variables rather than causal\nrelationships. Besides, the existing methods mainly rely on physical activity\nparameters and neglect the use of different factors such as maternal (medical)\nhistory and context data, leading to biased estimates. Furthermore, the\nestimations lack an understanding of mediators and counterfactual scenarios\nthat might affect them. In this paper, we investigate the causal relationship\nbetween being physically active (treatment variable) and the QoL (outcome)\nduring pregnancy and postpartum. To estimate the causal effect, we develop a\nCausal Machine Learning method, integrating causal discovery and causal\ninference components. The data for our investigation is derived from a\nlong-term wearable-based health monitoring study focusing on overweight and\nobese pregnant women. The machine learning (meta-learner) estimation technique\nis used to estimate the causal effect. Our result shows that performing\nadequate physical activity during pregnancy and postpartum improves the QoL by\nunits of 7.3 and 3.4 on average in physical health and psychological domains,\nrespectively. In the final step, four refutation analysis techniques are\nemployed to validate our estimation.\n", "date": "2024-02-25", "categories": []}
{"id": "2311.11463", "title": "Designing monitoring strategies for deployed machine learning\n  algorithms: navigating performativity through a causal lens", "abstract": "  After a machine learning (ML)-based system is deployed, monitoring its\nperformance is important to ensure the safety and effectiveness of the\nalgorithm over time. When an ML algorithm interacts with its environment, the\nalgorithm can affect the data-generating mechanism and be a major source of\nbias when evaluating its standalone performance, an issue known as\nperformativity. Although prior work has shown how to validate models in the\npresence of performativity using causal inference techniques, there has been\nlittle work on how to monitor models in the presence of performativity. Unlike\nthe setting of model validation, there is much less agreement on which\nperformance metrics to monitor. Different monitoring criteria impact how\ninterpretable the resulting test statistic is, what assumptions are needed for\nidentifiability, and the speed of detection. When this choice is further\ncoupled with the decision to use observational versus interventional data, ML\ndeployment teams are faced with a multitude of monitoring options. The aim of\nthis work is to highlight the relatively under-appreciated complexity of\ndesigning a monitoring strategy and how causal reasoning can provide a\nsystematic framework for choosing between these options. As a motivating\nexample, we consider an ML-based risk prediction algorithm for predicting\nunplanned readmissions. Bringing together tools from causal inference and\nstatistical process control, we consider six monitoring procedures (three\ncandidate monitoring criteria and two data sources) and investigate their\noperating characteristics in simulation studies. Results from this case study\nemphasize the seemingly simple (and obvious) fact that not all monitoring\nsystems are created equal, which has real-world impacts on the design and\ndocumentation of ML monitoring systems.\n", "date": "2023-11-20", "categories": []}
{"id": "2312.16815", "title": "Emergence and Causality in Complex Systems: A Survey on Causal Emergence\n  and Related Quantitative Studies", "abstract": "  Emergence and causality are two fundamental concepts for understanding\ncomplex systems. They are interconnected. On one hand, emergence refers to the\nphenomenon where macroscopic properties cannot be solely attributed to the\ncause of individual properties. On the other hand, causality can exhibit\nemergence, meaning that new causal laws may arise as we increase the level of\nabstraction. Causal emergence theory aims to bridge these two concepts and even\nemploys measures of causality to quantify emergence. This paper provides a\ncomprehensive review of recent advancements in quantitative theories and\napplications of causal emergence. Two key problems are addressed: quantifying\ncausal emergence and identifying it in data. Addressing the latter requires the\nuse of machine learning techniques, thus establishing a connection between\ncausal emergence and artificial intelligence. We highlighted that the\narchitectures used for identifying causal emergence are shared by causal\nrepresentation learning, causal model abstraction, and world model-based\nreinforcement learning. Consequently, progress in any of these areas can\nbenefit the others. Potential applications and future perspectives are also\ndiscussed in the final section of the review.\n", "date": "2023-12-28", "categories": []}
{"id": "2402.15625", "title": "Learning Cyclic Causal Models from Incomplete Data", "abstract": "  Causal learning is a fundamental problem in statistics and science, offering\ninsights into predicting the effects of unseen treatments on a system. Despite\nrecent advances in this topic, most existing causal discovery algorithms\noperate under two key assumptions: (i) the underlying graph is acyclic, and\n(ii) the available data is complete. These assumptions can be problematic as\nmany real-world systems contain feedback loops (e.g., biological systems), and\npractical scenarios frequently involve missing data. In this work, we propose a\nnovel framework, named MissNODAGS, for learning cyclic causal graphs from\npartially missing data. Under the additive noise model, MissNODAGS learns the\ncausal graph by alternating between imputing the missing data and maximizing\nthe expected log-likelihood of the visible part of the data in each training\nstep, following the principles of the expectation-maximization (EM) framework.\nThrough synthetic experiments and real-world single-cell perturbation data, we\ndemonstrate improved performance when compared to using state-of-the-art\nimputation techniques followed by causal learning on partially missing\ninterventional data.\n", "date": "2024-02-23", "categories": []}
{"id": "2402.16131", "title": "A VAE-based Framework for Learning Multi-Level Neural Granger-Causal\n  Connectivity", "abstract": "  Granger causality has been widely used in various application domains to\ncapture lead-lag relationships amongst the components of complex dynamical\nsystems, and the focus in extant literature has been on a single dynamical\nsystem. In certain applications in macroeconomics and neuroscience, one has\naccess to data from a collection of related such systems, wherein the modeling\ntask of interest is to extract the shared common structure that is embedded\nacross them, as well as to identify the idiosyncrasies within individual ones.\nThis paper introduces a Variational Autoencoder (VAE) based framework that\njointly learns Granger-causal relationships amongst components in a collection\nof related-yet-heterogeneous dynamical systems, and handles the aforementioned\ntask in a principled way. The performance of the proposed framework is\nevaluated on several synthetic data settings and benchmarked against existing\napproaches designed for individual system learning. The method is further\nillustrated on a real dataset involving time series data from a\nneurophysiological experiment and produces interpretable results.\n", "date": "2024-02-25", "categories": []}
{"id": "2209.11924", "title": "Interventional Causal Representation Learning", "abstract": "  Causal representation learning seeks to extract high-level latent factors\nfrom low-level sensory data. Most existing methods rely on observational data\nand structural assumptions (e.g., conditional independence) to identify the\nlatent factors. However, interventional data is prevalent across applications.\nCan interventional data facilitate causal representation learning? We explore\nthis question in this paper. The key observation is that interventional data\noften carries geometric signatures of the latent factors' support (i.e. what\nvalues each latent can possibly take). For example, when the latent factors are\ncausally connected, interventions can break the dependency between the\nintervened latents' support and their ancestors'. Leveraging this fact, we\nprove that the latent causal factors can be identified up to permutation and\nscaling given data from perfect $do$ interventions. Moreover, we can achieve\nblock affine identification, namely the estimated latent factors are only\nentangled with a few other latents if we have access to data from imperfect\ninterventions. These results highlight the unique power of interventional data\nin causal representation learning; they can enable provable identification of\nlatent factors without any assumptions about their distributions or dependency\nstructure.\n", "date": "2022-09-24", "categories": []}
{"id": "2301.12292", "title": "Zero-shot causal learning", "abstract": "  Predicting how different interventions will causally affect a specific\nindividual is important in a variety of domains such as personalized medicine,\npublic policy, and online marketing. There are a large number of methods to\npredict the effect of an existing intervention based on historical data from\nindividuals who received it. However, in many settings it is important to\npredict the effects of novel interventions (e.g., a newly invented drug), which\nthese methods do not address. Here, we consider zero-shot causal learning:\npredicting the personalized effects of a novel intervention. We propose CaML, a\ncausal meta-learning framework which formulates the personalized prediction of\neach intervention's effect as a task. CaML trains a single meta-model across\nthousands of tasks, each constructed by sampling an intervention, its\nrecipients, and its nonrecipients. By leveraging both intervention information\n(e.g., a drug's attributes) and individual features~(e.g., a patient's\nhistory), CaML is able to predict the personalized effects of novel\ninterventions that do not exist at the time of training. Experimental results\non real world datasets in large-scale medical claims and cell-line\nperturbations demonstrate the effectiveness of our approach. Most strikingly,\n\\method's zero-shot predictions outperform even strong baselines trained\ndirectly on data from the test interventions.\n", "date": "2023-01-28", "categories": []}
{"id": "2303.05279", "title": "Can large language models build causal graphs?", "abstract": "  Building causal graphs can be a laborious process. To ensure all relevant\ncausal pathways have been captured, researchers often have to discuss with\nclinicians and experts while also reviewing extensive relevant medical\nliterature. By encoding common and medical knowledge, large language models\n(LLMs) represent an opportunity to ease this process by automatically scoring\nedges (i.e., connections between two variables) in potential graphs. LLMs\nhowever have been shown to be brittle to the choice of probing words, context,\nand prompts that the user employs. In this work, we evaluate if LLMs can be a\nuseful tool in complementing causal graph development.\n", "date": "2023-03-07", "categories": []}
{"id": "2401.02602", "title": "Neural Causal Abstractions", "abstract": "  The abilities of humans to understand the world in terms of cause and effect\nrelationships, as well as to compress information into abstract concepts, are\ntwo hallmark features of human intelligence. These two topics have been studied\nin tandem in the literature under the rubric of causal abstractions theory. In\npractice, it remains an open problem how to best leverage abstraction theory in\nreal-world causal inference tasks, where the true mechanisms are unknown and\nonly limited data is available. In this paper, we develop a new family of\ncausal abstractions by clustering variables and their domains. This approach\nrefines and generalizes previous notions of abstractions to better accommodate\nindividual causal distributions that are spawned by Pearl's causal hierarchy.\nWe show that such abstractions are learnable in practical settings through\nNeural Causal Models (Xia et al., 2021), enabling the use of the deep learning\ntoolkit to solve various challenging causal inference tasks -- identification,\nestimation, sampling -- at different levels of granularity. Finally, we\nintegrate these results with representation learning to create more flexible\nabstractions, moving these results closer to practical applications. Our\nexperiments support the theory and illustrate how to scale causal inferences to\nhigh-dimensional settings involving image data.\n", "date": "2024-01-05", "categories": []}
{"id": "2402.15309", "title": "Counterfactual Generation with Identifiability Guarantees", "abstract": "  Counterfactual generation lies at the core of various machine learning tasks,\nincluding image translation and controllable text generation. This generation\nprocess usually requires the identification of the disentangled latent\nrepresentations, such as content and style, that underlie the observed data.\nHowever, it becomes more challenging when faced with a scarcity of paired data\nand labeling information. Existing disentangled methods crucially rely on\noversimplified assumptions, such as assuming independent content and style\nvariables, to identify the latent variables, even though such assumptions may\nnot hold for complex data distributions. For instance, food reviews tend to\ninvolve words like tasty, whereas movie reviews commonly contain words such as\nthrilling for the same positive sentiment. This problem is exacerbated when\ndata are sampled from multiple domains since the dependence between content and\nstyle may vary significantly over domains. In this work, we tackle the\ndomain-varying dependence between the content and the style variables inherent\nin the counterfactual generation task. We provide identification guarantees for\nsuch latent-variable models by leveraging the relative sparsity of the\ninfluences from different latent variables. Our theoretical insights enable the\ndevelopment of a doMain AdapTive counTerfactual gEneration model, called\n(MATTE). Our theoretically grounded framework achieves state-of-the-art\nperformance in unsupervised style transfer tasks, where neither paired data nor\nstyle labels are utilized, across four large-scale datasets. Code is available\nat https://github.com/hanqi-qi/Matte.git\n", "date": "2024-02-23", "categories": []}
{"id": "2303.02536", "title": "Finding Alignments Between Interpretable Causal Variables and\n  Distributed Neural Representations", "abstract": "  Causal abstraction is a promising theoretical framework for explainable\nartificial intelligence that defines when an interpretable high-level causal\nmodel is a faithful simplification of a low-level deep learning system.\nHowever, existing causal abstraction methods have two major limitations: they\nrequire a brute-force search over alignments between the high-level model and\nthe low-level one, and they presuppose that variables in the high-level model\nwill align with disjoint sets of neurons in the low-level one. In this paper,\nwe present distributed alignment search (DAS), which overcomes these\nlimitations. In DAS, we find the alignment between high-level and low-level\nmodels using gradient descent rather than conducting a brute-force search, and\nwe allow individual neurons to play multiple distinct roles by analyzing\nrepresentations in non-standard bases-distributed representations. Our\nexperiments show that DAS can discover internal structure that prior approaches\nmiss. Overall, DAS removes previous obstacles to conducting causal abstraction\nanalyses and allows us to find conceptual structure in trained neural nets.\n", "date": "2023-03-05", "categories": []}
{"id": "2304.11625", "title": "Meaningful Causal Aggregation and Paradoxical Confounding", "abstract": "  In aggregated variables the impact of interventions is typically ill-defined\nbecause different micro-realizations of the same macro-intervention can result\nin different changes of downstream macro-variables. We show that this\nill-definedness of causality on aggregated variables can turn unconfounded\ncausal relations into confounded ones and vice versa, depending on the\nrespective micro-realization. We argue that it is practically infeasible to\nonly use aggregated causal systems when we are free from this ill-definedness.\nInstead, we need to accept that macro causal relations are typically defined\nonly with reference to the micro states. On the positive side, we show that\ncause-effect relations can be aggregated when the macro interventions are such\nthat the distribution of micro states is the same as in the observational\ndistribution; we term this natural macro interventions. We also discuss\ngeneralizations of this observation.\n", "date": "2023-04-23", "categories": []}
{"id": "2402.14481", "title": "Towards Automated Causal Discovery: a case study on 5G telecommunication\n  data", "abstract": "  We introduce the concept of Automated Causal Discovery (AutoCD), defined as\nany system that aims to fully automate the application of causal discovery and\ncausal reasoning methods. AutoCD's goal is to deliver all causal information\nthat an expert human analyst would and answer a user's causal queries. We\ndescribe the architecture of such a platform, and illustrate its performance on\nsynthetic data sets. As a case study, we apply it on temporal telecommunication\ndata. The system is general and can be applied to a plethora of causal\ndiscovery problems.\n", "date": "2024-02-22", "categories": []}
{"id": "2305.19215", "title": "dotears: Scalable, consistent DAG estimation using observational and\n  interventional data", "abstract": "  New biological assays like Perturb-seq link highly parallel CRISPR\ninterventions to a high-dimensional transcriptomic readout, providing insight\ninto gene regulatory networks. Causal gene regulatory networks can be\nrepresented by directed acyclic graph (DAGs), but learning DAGs from\nobservational data is complicated by lack of identifiability and a\ncombinatorial solution space. Score-based structure learning improves practical\nscalability of inferring DAGs. Previous score-based methods are sensitive to\nerror variance structure; on the other hand, estimation of error variance is\ndifficult without prior knowledge of structure. Accordingly, we present\n$\\texttt{dotears}$ [doo-tairs], a continuous optimization framework which\nleverages observational and interventional data to infer a single causal\nstructure, assuming a linear Structural Equation Model (SEM).\n$\\texttt{dotears}$ exploits structural consequences of hard interventions to\ngive a marginal estimate of exogenous error structure, bypassing the circular\nestimation problem. We show that $\\texttt{dotears}$ is a provably consistent\nestimator of the true DAG under mild assumptions. $\\texttt{dotears}$\noutperforms other methods in varied simulations, and in real data infers edges\nthat validate with higher precision and recall than state-of-the-art methods\nthrough differential expression tests and high-confidence protein-protein\ninteractions.\n", "date": "2023-05-30", "categories": []}
{"id": "2402.13427", "title": "Quantitative causality, causality-guided scientific discovery, and\n  causal machine learning", "abstract": "  It has been said, arguably, that causality analysis should pave a promising\nway to interpretable deep learning and generalization. Incorporation of\ncausality into artificial intelligence (AI) algorithms, however, is challenged\nwith its vagueness, non-quantitiveness, computational inefficiency, etc. During\nthe past 18 years, these challenges have been essentially resolved, with the\nestablishment of a rigorous formalism of causality analysis initially motivated\nfrom atmospheric predictability. This not only opens a new field in the\natmosphere-ocean science, namely, information flow, but also has led to\nscientific discoveries in other disciplines, such as quantum mechanics,\nneuroscience, financial economics, etc., through various applications. This\nnote provides a brief review of the decade-long effort, including a list of\nmajor theoretical results, a sketch of the causal deep learning framework, and\nsome representative real-world applications in geoscience pertaining to this\njournal, such as those on the anthropogenic cause of global warming, the\ndecadal prediction of El Ni\\~no Modoki, the forecasting of an extreme drought\nin China, among others.\n", "date": "2024-02-20", "categories": []}
{"id": "2211.14221", "title": "Learning Large Causal Structures from Inverse Covariance Matrix via\n  Sparse Matrix Decomposition", "abstract": "  Learning causal structures from observational data is a fundamental problem\nfacing important computational challenges when the number of variables is\nlarge. In the context of linear structural equation models (SEMs), this paper\nfocuses on learning causal structures from the inverse covariance matrix. The\nproposed method, called ICID for Independence-preserving Decomposition from\nInverse Covariance matrix, is based on continuous optimization of a matrix\ndecomposition model that preserves the nonzero patterns of the inverse\ncovariance matrix. Through theoretical and empirical evidences, we show that\nICID efficiently identifies the sought directed acyclic graph (DAG) assuming\nthe knowledge of noise variances. Moreover, ICID is shown empirically to be\nrobust under bounded misspecification of noise variances in the case where the\nnoise variances are non-equal. The proposed method enjoys a low complexity, as\nreflected by its time efficiency in the experiments, and also enables a novel\nregularization scheme that yields highly accurate solutions on the Simulated\nfMRI data (Smith et al., 2011) in comparison with state-of-the-art algorithms.\n", "date": "2022-11-25", "categories": []}
{"id": "2306.13891", "title": "Estimating the Causal Effect of Early ArXiving on Paper Acceptance", "abstract": "  What is the effect of releasing a preprint of a paper before it is submitted\nfor peer review? No randomized controlled trial has been conducted, so we turn\nto observational data to answer this question. We use data from the ICLR\nconference (2018--2022) and apply methods from causal inference to estimate the\neffect of arXiving a paper before the reviewing period (early arXiving) on its\nacceptance to the conference. Adjusting for confounders such as topic, authors,\nand quality, we may estimate the causal effect. However, since quality is a\nchallenging construct to estimate, we use the negative outcome control method,\nusing paper citation count as a control variable to debias the quality\nconfounding effect. Our results suggest that early arXiving may have a small\neffect on a paper's chances of acceptance. However, this effect (when existing)\ndoes not differ significantly across different groups of authors, as grouped by\nauthor citation count and institute rank. This suggests that early arXiving\ndoes not provide an advantage to any particular group.\n", "date": "2023-06-24", "categories": []}
{"id": "2310.15627", "title": "Contextual Directed Acyclic Graphs", "abstract": "  Estimating the structure of directed acyclic graphs (DAGs) from observational\ndata remains a significant challenge in machine learning. Most research in this\narea concentrates on learning a single DAG for the entire population. This\npaper considers an alternative setting where the graph structure varies across\nindividuals based on available \"contextual\" features. We tackle this contextual\nDAG problem via a neural network that maps the contextual features to a DAG,\nrepresented as a weighted adjacency matrix. The neural network is equipped with\na novel projection layer that ensures the output matrices are sparse and\nsatisfy a recently developed characterization of acyclicity. We devise a\nscalable computational framework for learning contextual DAGs and provide a\nconvergence guarantee and an analytical gradient for backpropagating through\nthe projection layer. Our experiments suggest that the new approach can recover\nthe true context-specific graph where existing approaches fail.\n", "date": "2023-10-24", "categories": []}
{"id": "2310.18212", "title": "Robustness of Algorithms for Causal Structure Learning to Hyperparameter\n  Choice", "abstract": "  Hyperparameters play a critical role in machine learning. Hyperparameter\ntuning can make the difference between state-of-the-art and poor prediction\nperformance for any algorithm, but it is particularly challenging for structure\nlearning due to its unsupervised nature. As a result, hyperparameter tuning is\noften neglected in favour of using the default values provided by a particular\nimplementation of an algorithm. While there have been numerous studies on\nperformance evaluation of causal discovery algorithms, how hyperparameters\naffect individual algorithms, as well as the choice of the best algorithm for a\nspecific problem, has not been studied in depth before. This work addresses\nthis gap by investigating the influence of hyperparameters on causal structure\nlearning tasks. Specifically, we perform an empirical evaluation of\nhyperparameter selection for some seminal learning algorithms on datasets of\nvarying levels of complexity. We find that, while the choice of algorithm\nremains crucial to obtaining state-of-the-art performance, hyperparameter\nselection in ensemble settings strongly influences the choice of algorithm, in\nthat a poor choice of hyperparameters can lead to analysts using algorithms\nwhich do not give state-of-the-art performance for their data.\n", "date": "2023-10-27", "categories": []}
{"id": "2402.12400", "title": "Estimating the age-conditioned average treatment effects curves: An\n  application for assessing load-management strategies in the NBA", "abstract": "  In the realm of competitive sports, understanding the performance dynamics of\nathletes, represented by the age curve (showing progression, peak, and\ndecline), is vital. Our research introduces a novel framework for quantifying\nage-specific treatment effects, enhancing the granularity of performance\ntrajectory analysis. Firstly, we propose a methodology for estimating the age\ncurve using game-level data, diverging from traditional season-level data\napproaches, and tackling its inherent complexities with a meta-learner\nframework that leverages advanced machine learning models. This approach\nuncovers intricate non-linear patterns missed by existing methods. Secondly,\nour framework enables the identification of causal effects, allowing for a\ndetailed examination of age curves under various conditions. By defining the\nAge-Conditioned Treatment Effect (ACTE), we facilitate the exploration of\ncausal relationships regarding treatment impacts at specific ages. Finally,\napplying this methodology to study the effects of rest days on performance\nmetrics, particularly across different ages, offers valuable insights into load\nmanagement strategies' effectiveness. Our findings underscore the importance of\ntailored rest periods, highlighting their positive impact on athlete\nperformance and suggesting a reevaluation of current management practices for\noptimizing athlete performance.\n", "date": "2024-02-17", "categories": []}
{"id": "2402.12710", "title": "Integrating Active Learning in Causal Inference with Interference: A\n  Novel Approach in Online Experiments", "abstract": "  In the domain of causal inference research, the prevalent potential outcomes\nframework, notably the Rubin Causal Model (RCM), often overlooks individual\ninterference and assumes independent treatment effects. This assumption,\nhowever, is frequently misaligned with the intricate realities of real-world\nscenarios, where interference is not merely a possibility but a common\noccurrence. Our research endeavors to address this discrepancy by focusing on\nthe estimation of direct and spillover treatment effects under two assumptions:\n(1) network-based interference, where treatments on neighbors within connected\nnetworks affect one's outcomes, and (2) non-random treatment assignments\ninfluenced by confounders. To improve the efficiency of estimating potentially\ncomplex effects functions, we introduce an novel active learning approach:\nActive Learning in Causal Inference with Interference (ACI). This approach uses\nGaussian process to flexibly model the direct and spillover treatment effects\nas a function of a continuous measure of neighbors' treatment assignment. The\nACI framework sequentially identifies the experimental settings that demand\nfurther data. It further optimizes the treatment assignments under the network\ninterference structure using genetic algorithms to achieve efficient learning\noutcome. By applying our method to simulation data and a Tencent game dataset,\nwe demonstrate its feasibility in achieving accurate effects estimations with\nreduced data requirements. This ACI approach marks a significant advancement in\nthe realm of data efficiency for causal inference, offering a robust and\nefficient alternative to traditional methodologies, particularly in scenarios\ncharacterized by complex interference patterns.\n", "date": "2024-02-20", "categories": []}
{"id": "2206.01900", "title": "Estimating counterfactual treatment outcomes over time in complex\n  multiagent scenarios", "abstract": "  Evaluation of intervention in a multiagent system, e.g., when humans should\nintervene in autonomous driving systems and when a player should pass to\nteammates for a good shot, is challenging in various engineering and scientific\nfields. Estimating the individual treatment effect (ITE) using counterfactual\nlong-term prediction is practical to evaluate such interventions. However, most\nof the conventional frameworks did not consider the time-varying complex\nstructure of multiagent relationships and covariate counterfactual prediction.\nThis may lead to erroneous assessments of ITE and difficulty in interpretation.\nHere we propose an interpretable, counterfactual recurrent network in\nmultiagent systems to estimate the effect of the intervention. Our model\nleverages graph variational recurrent neural networks and theory-based\ncomputation with domain knowledge for the ITE estimation framework based on\nlong-term prediction of multiagent covariates and outcomes, which can confirm\nthe circumstances under which the intervention is effective. On simulated\nmodels of an automated vehicle and biological agents with time-varying\nconfounders, we show that our methods achieved lower estimation errors in\ncounterfactual covariates and the most effective treatment timing than the\nbaselines. Furthermore, using real basketball data, our methods performed\nrealistic counterfactual predictions and evaluated the counterfactual passes in\nshot scenarios.\n", "date": "2022-06-04", "categories": []}
{"id": "2308.07480", "title": "Order-based Structure Learning with Normalizing Flows", "abstract": "  Estimating the causal structure of observational data is a challenging\ncombinatorial search problem that scales super-exponentially with graph size.\nExisting methods use continuous relaxations to make this problem\ncomputationally tractable but often restrict the data-generating process to\nadditive noise models (ANMs) through explicit or implicit assumptions. We\npresent Order-based Structure Learning with Normalizing Flows (OSLow), a\nframework that relaxes these assumptions using autoregressive normalizing\nflows. We leverage the insight that searching over topological orderings is a\nnatural way to enforce acyclicity in structure discovery and propose a novel,\ndifferentiable permutation learning method to find such orderings. Through\nextensive experiments on synthetic and real-world data, we demonstrate that\nOSLow outperforms prior baselines and improves performance on the observational\nSachs and SynTReN datasets as measured by structural hamming distance and\nstructural intervention distance, highlighting the importance of relaxing the\nANM assumption made by existing methods.\n", "date": "2023-08-14", "categories": []}
{"id": "2402.00623", "title": "Bayesian Causal Inference with Gaussian Process Networks", "abstract": "  Causal discovery and inference from observational data is an essential\nproblem in statistics posing both modeling and computational challenges. These\nare typically addressed by imposing strict assumptions on the joint\ndistribution such as linearity. We consider the problem of the Bayesian\nestimation of the effects of hypothetical interventions in the Gaussian Process\nNetwork (GPN) model, a flexible causal framework which allows describing the\ncausal relationships nonparametrically. We detail how to perform causal\ninference on GPNs by simulating the effect of an intervention across the whole\nnetwork and propagating the effect of the intervention on downstream variables.\nWe further derive a simpler computational approximation by estimating the\nintervention distribution as a function of local variables only, modeling the\nconditional distributions via additive Gaussian processes. We extend both\nframeworks beyond the case of a known causal graph, incorporating uncertainty\nabout the causal structure via Markov chain Monte Carlo methods. Simulation\nstudies show that our approach is able to identify the effects of hypothetical\ninterventions with non-Gaussian, non-linear observational data and accurately\nreflect the posterior uncertainty of the causal estimates. Finally we compare\nthe results of our GPN-based causal inference approach to existing methods on a\ndataset of $A.~thaliana$ gene expressions.\n", "date": "2024-02-01", "categories": []}
{"id": "2402.01341", "title": "Fundamental Properties of Causal Entropy and Information Gain", "abstract": "  Recent developments enable the quantification of causal control given a\nstructural causal model (SCM). This has been accomplished by introducing\nquantities which encode changes in the entropy of one variable when intervening\non another. These measures, named causal entropy and causal information gain,\naim to address limitations in existing information theoretical approaches for\nmachine learning tasks where causality plays a crucial role. They have not yet\nbeen properly mathematically studied. Our research contributes to the formal\nunderstanding of the notions of causal entropy and causal information gain by\nestablishing and analyzing fundamental properties of these concepts, including\nbounds and chain rules. Furthermore, we elucidate the relationship between\ncausal entropy and stochastic interventions. We also propose definitions for\ncausal conditional entropy and causal conditional information gain. Overall,\nthis exploration paves the way for enhancing causal machine learning tasks\nthrough the study of recently-proposed information theoretic quantities\ngrounded in considerations about causality.\n", "date": "2024-02-02", "categories": []}
{"id": "2311.11229", "title": "Causal ATE Mitigates Unintended Bias in Controlled Text Generation", "abstract": "  We study attribute control in language models through the method of Causal\nAverage Treatment Effect (Causal ATE). Existing methods for the attribute\ncontrol task in Language Models (LMs) check for the co-occurrence of words in a\nsentence with the attribute of interest, and control for them. However,\nspurious correlation of the words with the attribute in the training dataset,\ncan cause models to hallucinate the presence of the attribute when presented\nwith the spurious correlate during inference. We show that the simple\nperturbation-based method of Causal ATE removes this unintended effect.\nSpecifically, we ground it in the problem of toxicity mitigation, where a\nsignificant challenge lies in the inadvertent bias that often emerges towards\nprotected groups post detoxification. We show that this unintended bias can be\nsolved by the use of the Causal ATE metric and rigorously prove our claim. We\nprovide experimental validations for our claims and release our code\n(anonymously) here:\nhttps://github.com/causalate-mitigates-bias/causal-ate-mitigates-bias.\n", "date": "2023-11-19", "categories": []}
{"id": "2306.16958", "title": "Identifiability of Direct Effects from Summary Causal Graphs", "abstract": "  Dynamic structural causal models (SCMs) are a powerful framework for\nreasoning in dynamic systems about direct effects which measure how a change in\none variable affects another variable while holding all other variables\nconstant. The causal relations in a dynamic structural causal model can be\nqualitatively represented with an acyclic full-time causal graph. Assuming\nlinearity and no hidden confounding and given the full-time causal graph, the\ndirect causal effect is always identifiable. However, in many application such\na graph is not available for various reasons but nevertheless experts have\naccess to the summary causal graph of the full-time causal graph which\nrepresents causal relations between time series while omitting temporal\ninformation and allowing cycles. This paper presents a complete identifiability\nresult which characterizes all cases for which the direct effect is graphically\nidentifiable from a summary causal graph and gives two sound finite adjustment\nsets that can be used to estimate the direct effect whenever it is\nidentifiable.\n", "date": "2023-06-29", "categories": []}
{"id": "2302.10809", "title": "Causal Explanations for Sequential Decision-Making in Multi-Agent\n  Systems", "abstract": "  We present CEMA: Causal Explanations in Multi-Agent systems; a framework for\ncreating causal natural language explanations of an agent's decisions in\ndynamic sequential multi-agent systems to build more trustworthy autonomous\nagents. Unlike prior work that assumes a fixed causal structure, CEMA only\nrequires a probabilistic model for forward-simulating the state of the system.\nUsing such a model, CEMA simulates counterfactual worlds that identify the\nsalient causes behind the agent's decisions. We evaluate CEMA on the task of\nmotion planning for autonomous driving and test it in diverse simulated\nscenarios. We show that CEMA correctly and robustly identifies the causes\nbehind the agent's decisions, even when a large number of other agents is\npresent, and show via a user study that CEMA's explanations have a positive\neffect on participants' trust in autonomous vehicles and are rated as high as\nhigh-quality baseline explanations elicited from other participants. We release\nthe collected explanations with annotations as the HEADD dataset.\n", "date": "2023-02-21", "categories": []}
{"id": "2303.02186", "title": "Causal Deep Learning", "abstract": "  Causality has the potential to truly transform the way we solve a large\nnumber of real-world problems. Yet, so far, its potential largely remains to be\nunlocked as causality often requires crucial assumptions which cannot be tested\nin practice. To address this challenge, we propose a new way of thinking about\ncausality -- we call this causal deep learning. Our causal deep learning\nframework spans three dimensions: (1) a structural dimension, which\nincorporates partial yet testable causal knowledge rather than assuming either\ncomplete or no causal knowledge among the variables of interest; (2) a\nparametric dimension, which encompasses parametric forms that capture the type\nof relationships among the variables of interest; and (3) a temporal dimension,\nwhich captures exposure times or how the variables of interest interact\n(possibly causally) over time. Causal deep learning enables us to make progress\non a variety of real-world problems by leveraging partial causal knowledge\n(including independencies among variables) and quantitatively characterising\ncausal relationships among variables of interest (possibly over time). Our\nframework clearly identifies which assumptions are testable and which ones are\nnot, such that the resulting solutions can be judiciously adopted in practice.\nUsing our formulation we can combine or chain together causal representations\nto solve specific problems without losing track of which assumptions are\nrequired to build these solutions, pushing real-world impact in healthcare,\neconomics and business, environmental sciences and education, through causal\ndeep learning.\n", "date": "2023-03-03", "categories": []}
{"id": "2306.10816", "title": "$\\texttt{causalAssembly}$: Generating Realistic Production Data for\n  Benchmarking Causal Discovery", "abstract": "  Algorithms for causal discovery have recently undergone rapid advances and\nincreasingly draw on flexible nonparametric methods to process complex data.\nWith these advances comes a need for adequate empirical validation of the\ncausal relationships learned by different algorithms. However, for most real\ndata sources true causal relations remain unknown. This issue is further\ncompounded by privacy concerns surrounding the release of suitable high-quality\ndata. To help address these challenges, we gather a complex dataset comprising\nmeasurements from an assembly line in a manufacturing context. This line\nconsists of numerous physical processes for which we are able to provide ground\ntruth causal relationships on the basis of a detailed study of the underlying\nphysics. We use the assembly line data and associated ground truth information\nto build a system for generation of semisynthetic manufacturing data that\nsupports benchmarking of causal discovery methods. To accomplish this, we\nemploy distributional random forests in order to flexibly estimate and\nrepresent conditional distributions that may be combined into joint\ndistributions that strictly adhere to a causal model over the observed\nvariables. The estimated conditionals and tools for data generation are made\navailable in our Python library $\\texttt{causalAssembly}$. Using the library,\nwe showcase how to benchmark several well-known causal discovery algorithms.\n", "date": "2023-06-19", "categories": []}
{"id": "2310.15450", "title": "General Identifiability and Achievability for Causal Representation\n  Learning", "abstract": "  This paper focuses on causal representation learning (CRL) under a general\nnonparametric latent causal model and a general transformation model that maps\nthe latent data to the observational data. It establishes identifiability and\nachievability results using two hard uncoupled interventions per node in the\nlatent causal graph. Notably, one does not know which pair of intervention\nenvironments have the same node intervened (hence, uncoupled). For\nidentifiability, the paper establishes that perfect recovery of the latent\ncausal model and variables is guaranteed under uncoupled interventions. For\nachievability, an algorithm is designed that uses observational and\ninterventional data and recovers the latent causal model and variables with\nprovable guarantees. This algorithm leverages score variations across different\nenvironments to estimate the inverse of the transformer and, subsequently, the\nlatent variables. The analysis, additionally, recovers the identifiability\nresult for two hard coupled interventions, that is when metadata about the pair\nof environments that have the same node intervened is known. This paper also\nshows that when observational data is available, additional faithfulness\nassumptions that are adopted by the existing literature are unnecessary.\n", "date": "2023-10-24", "categories": []}
{"id": "2402.09305", "title": "Embracing the black box: Heading towards foundation models for causal\n  discovery from time series data", "abstract": "  Causal discovery from time series data encompasses many existing solutions,\nincluding those based on deep learning techniques. However, these methods\ntypically do not endorse one of the most prevalent paradigms in deep learning:\nEnd-to-end learning. To address this gap, we explore what we call Causal\nPretraining. A methodology that aims to learn a direct mapping from\nmultivariate time series to the underlying causal graphs in a supervised\nmanner. Our empirical findings suggest that causal discovery in a supervised\nmanner is possible, assuming that the training and test time series samples\nshare most of their dynamics. More importantly, we found evidence that the\nperformance of Causal Pretraining can increase with data and model size, even\nif the additional data do not share the same dynamics. Further, we provide\nexamples where causal discovery for real-world data with causally pretrained\nneural networks is possible within limits. We argue that this hints at the\npossibility of a foundation model for causal discovery.\n", "date": "2024-02-14", "categories": []}
{"id": "2301.03180", "title": "Subset verification and search algorithms for causal DAGs", "abstract": "  Learning causal relationships between variables is a fundamental task in\ncausal inference and directed acyclic graphs (DAGs) are a popular choice to\nrepresent the causal relationships. As one can recover a causal graph only up\nto its Markov equivalence class from observations, interventions are often used\nfor the recovery task. Interventions are costly in general and it is important\nto design algorithms that minimize the number of interventions performed. In\nthis work, we study the problem of identifying the smallest set of\ninterventions required to learn the causal relationships between a subset of\nedges (target edges). Under the assumptions of faithfulness, causal\nsufficiency, and ideal interventions, we study this problem in two settings:\nwhen the underlying ground truth causal graph is known (subset verification)\nand when it is unknown (subset search). For the subset verification problem, we\nprovide an efficient algorithm to compute a minimum sized interventional set;\nwe further extend these results to bounded size non-atomic interventions and\nnode-dependent interventional costs. For the subset search problem, in the\nworst case, we show that no algorithm (even with adaptivity or randomization)\ncan achieve an approximation ratio that is asymptotically better than the\nvertex cover of the target edges when compared with the subset verification\nnumber. This result is surprising as there exists a logarithmic approximation\nalgorithm for the search problem when we wish to recover the whole causal\ngraph. To obtain our results, we prove several interesting structural\nproperties of interventional causal graphs that we believe have applications\nbeyond the subset verification/search problems studied here.\n", "date": "2023-01-09", "categories": []}
{"id": "2304.09981", "title": "Interpretable (not just posthoc-explainable) heterogeneous survivor\n  bias-corrected treatment effects for assignment of postdischarge\n  interventions to prevent readmissions", "abstract": "  We used survival analysis to quantify the impact of postdischarge evaluation\nand management (E/M) services in preventing hospital readmission or death. Our\napproach avoids a specific pitfall of applying machine learning to this\nproblem, which is an inflated estimate of the effect of interventions, due to\nsurvivors bias -- where the magnitude of inflation may be conditional on\nheterogeneous confounders in the population. This bias arises simply because in\norder to receive an intervention after discharge, a person must not have been\nreadmitted in the intervening period. After deriving an expression for this\nphantom effect, we controlled for this and other biases within an inherently\ninterpretable Bayesian survival framework. We identified case management\nservices as being the most impactful for reducing readmissions overall.\n", "date": "2023-04-19", "categories": []}
{"id": "2305.15984", "title": "Dynamic Inter-treatment Information Sharing for Individualized Treatment\n  Effects Estimation", "abstract": "  Estimation of individualized treatment effects (ITE) from observational\nstudies is a fundamental problem in causal inference and holds significant\nimportance across domains, including healthcare. However, limited observational\ndatasets pose challenges in reliable ITE estimation as data have to be split\namong treatment groups to train an ITE learner. While information sharing among\ntreatment groups can partially alleviate the problem, there is currently no\ngeneral framework for end-to-end information sharing in ITE estimation. To\ntackle this problem, we propose a deep learning framework based on\n`\\textit{soft weight sharing}' to train ITE learners, enabling \\textit{dynamic\nend-to-end} information sharing among treatment groups. The proposed framework\ncomplements existing ITE learners, and introduces a new class of ITE learners,\nreferred to as \\textit{HyperITE}. We extend state-of-the-art ITE learners with\n\\textit{HyperITE} versions and evaluate them on IHDP, ACIC-2016, and Twins\nbenchmarks. Our experimental results show that the proposed framework improves\nITE estimation error, with increasing effectiveness for smaller datasets.\n", "date": "2023-05-25", "categories": []}
{"id": "2310.13576", "title": "Tree Search in DAG Space with Model-based Reinforcement Learning for\n  Causal Discovery", "abstract": "  Identifying causal structure is central to many fields ranging from strategic\ndecision-making to biology and economics. In this work, we propose CD-UCT, a\nmodel-based reinforcement learning method for causal discovery based on tree\nsearch that builds directed acyclic graphs incrementally. We also formalize and\nprove the correctness of an efficient algorithm for excluding edges that would\nintroduce cycles, which enables deeper discrete search and sampling in DAG\nspace. The proposed method can be applied broadly to causal Bayesian networks\nwith both discrete and continuous random variables. We conduct a comprehensive\nevaluation on synthetic and real-world datasets, showing that CD-UCT\nsubstantially outperforms the state-of-the-art model-free reinforcement\nlearning technique and greedy search, constituting a promising advancement for\ncombinatorial methods.\n", "date": "2023-10-20", "categories": []}
{"id": "2402.08229", "title": "Causal Discovery under Off-Target Interventions", "abstract": "  Causal graph discovery is a significant problem with applications across\nvarious disciplines. However, with observational data alone, the underlying\ncausal graph can only be recovered up to its Markov equivalence class, and\nfurther assumptions or interventions are necessary to narrow down the true\ngraph. This work addresses the causal discovery problem under the setting of\nstochastic interventions with the natural goal of minimizing the number of\ninterventions performed. We propose the following stochastic intervention model\nwhich subsumes existing adaptive noiseless interventions in the literature\nwhile capturing scenarios such as fat-hand interventions and CRISPR gene\nknockouts: any intervention attempt results in an actual intervention on a\nrandom subset of vertices, drawn from a distribution dependent on attempted\naction. Under this model, we study the two fundamental problems in causal\ndiscovery of verification and search and provide approximation algorithms with\npolylogarithmic competitive ratios and provide some preliminary experimental\nresults.\n", "date": "2024-02-13", "categories": []}
{"id": "2211.03846", "title": "Federated Causal Discovery From Interventions", "abstract": "  Causal discovery serves a pivotal role in mitigating model uncertainty\nthrough recovering the underlying causal mechanisms among variables. In many\npractical domains, such as healthcare, access to the data gathered by\nindividual entities is limited, primarily for privacy and regulatory\nconstraints. However, the majority of existing causal discovery methods require\nthe data to be available in a centralized location. In response, researchers\nhave introduced federated causal discovery. While previous federated methods\nconsider distributed observational data, the integration of interventional data\nremains largely unexplored. We propose FedCDI, a federated framework for\ninferring causal structures from distributed data containing interventional\nsamples. In line with the federated learning framework, FedCDI improves privacy\nby exchanging belief updates rather than raw samples. Additionally, it\nintroduces a novel intervention-aware method for aggregating individual\nupdates. We analyze scenarios with shared or disjoint intervened covariates,\nand mitigate the adverse effects of interventional data heterogeneity. The\nperformance and scalability of FedCDI is rigorously tested across a variety of\nsynthetic and real-world graphs.\n", "date": "2022-11-07", "categories": []}
{"id": "2311.00886", "title": "COSTAR: Improved Temporal Counterfactual Estimation with Self-Supervised\n  Learning", "abstract": "  Estimation of temporal counterfactual outcomes from observed history is\ncrucial for decision-making in many domains such as healthcare and e-commerce,\nparticularly when randomized controlled trials (RCTs) suffer from high cost or\nimpracticality. For real-world datasets, modeling time-dependent confounders is\nchallenging due to complex dynamics, long-range dependencies and both past\ntreatments and covariates affecting the future outcomes. In this paper, we\nintroduce Counterfactual Self-Supervised Transformer (COSTAR), a novel approach\nthat integrates self-supervised learning for improved historical\nrepresentations. We propose a component-wise contrastive loss tailored for\ntemporal treatment outcome observations and explain its effectiveness from the\nview of unsupervised domain adaptation. COSTAR yields superior performance in\nestimation accuracy and generalization to out-of-distribution data compared to\nexisting models, as validated by empirical results on both synthetic and\nreal-world datasets.\n", "date": "2023-11-01", "categories": []}
{"id": "2312.05974", "title": "Learning the Causal Structure of Networked Dynamical Systems under\n  Latent Nodes and Structured Noise", "abstract": "  This paper considers learning the hidden causal network of a linear networked\ndynamical system (NDS) from the time series data at some of its nodes --\npartial observability. The dynamics of the NDS are driven by colored noise that\ngenerates spurious associations across pairs of nodes, rendering the problem\nmuch harder. To address the challenge of noise correlation and partial\nobservability, we assign to each pair of nodes a feature vector computed from\nthe time series data of observed nodes. The feature embedding is engineered to\nyield structural consistency: there exists an affine hyperplane that\nconsistently partitions the set of features, separating the feature vectors\ncorresponding to connected pairs of nodes from those corresponding to\ndisconnected pairs. The causal inference problem is thus addressed via\nclustering the designed features. We demonstrate with simple baseline\nsupervised methods the competitive performance of the proposed causal inference\nmechanism under broad connectivity regimes and noise correlation levels,\nincluding a real world network. Further, we devise novel technical guarantees\nof structural consistency for linear NDS under the considered regime.\n", "date": "2023-12-10", "categories": []}
{"id": "2402.05940", "title": "Causal Relationship Network of Risk Factors Impacting Workday Loss in\n  Underground Coal Mines", "abstract": "  This study aims to establish the causal relationship network between various\nfactors leading to workday loss in underground coal mines using a novel causal\nartificial intelligence (AI) method. The analysis utilizes data obtained from\nthe National Institute for Occupational Safety and Health (NIOSH). A total of\n101,010 injury records from 3,982 unique underground coal mines spanning the\nyears from 1990 to 2020 were extracted from the NIOSH database. Causal\nrelationships were analyzed and visualized using a novel causal AI method\ncalled Grouped Greedy Equivalence Search (GGES). The impact of each variable on\nworkday loss was assessed through intervention do-calculus adjustment (IDA)\nscores. Model training and validation were performed using the 10-fold\ncross-validation technique. Performance metrics, including adjacency precision\n(AP), adjacency recall (AR), arrowhead precision (AHP), and arrowhead recall\n(AHR), were utilized to evaluate the models. Findings revealed that after 2006,\nkey direct causes of workday loss among mining employees included total mining\nexperience, mean office employees, mean underground employees, county, and\ntotal mining experience (years). Total mining experience emerged as the most\ninfluential factor, whereas mean employees per mine exhibited the least\ninfluence. The analyses emphasized the significant role of total mining\nexperience in determining workday loss. The models achieved optimal\nperformance, with AP, AR, AHP, and AHR values measuring 0.694, 0.653, 0.386,\nand 0.345, respectively. This study demonstrates the feasibility of utilizing\nthe new GGES method to clarify the causal factors behind the workday loss by\nanalyzing employment demographics and injury records and establish their causal\nrelationship network.\n", "date": "2024-01-24", "categories": []}
{"id": "2402.06220", "title": "A Unified Causal View of Instruction Tuning", "abstract": "  Instruction tuning on a mixture of tasks has improved zero-shot capabilities\nin natural language processing (NLP). Nevertheless, existing methods often\nlearn features that exhibit correlations between instruction-formatted samples\nand target labels, rather than causal relationships. Termed as ``spurious\ncorrelation'' in statistics, such a correlation may change drastically in a new\ntask, making the effect from the learned features to be misleading. To this\nend, we develop a meta Structural Causal Model (meta-SCM) to integrate\ndifferent NLP tasks under a single causal structure of the data. Specifically,\nthe meta-SCM introduces multiple latent factors that represent properties of\nsource context, only some of which causally influence the target labels for a\nspecific task. The key idea is to learn task-required causal factors and only\nuse those to make predictions for a given task. Theoretically, we prove the\ncausal factor can be identified without mixing information from others. Guided\nby the identifiability, we propose a Structural Instruction Tuning (SIT) method\nto learn the task-required causal representations that can mimic the causal\nfactors for each task. The utility of our approach is verified by improvements\nof zero-shot ability on a range of unseen datasets and tasks.\n", "date": "2024-02-09", "categories": []}
{"id": "2304.07896", "title": "Out-of-Variable Generalization for Discriminative Models", "abstract": "  The ability of an agent to do well in new environments is a critical aspect\nof intelligence. In machine learning, this ability is known as\n$\\textit{strong}$ or $\\textit{out-of-distribution}$ generalization. However,\nmerely considering differences in data distributions is inadequate for fully\ncapturing differences between learning environments. In the present paper, we\ninvestigate $\\textit{out-of-variable}$ generalization, which pertains to an\nagent's generalization capabilities concerning environments with variables that\nwere never jointly observed before. This skill closely reflects the process of\nanimate learning: we, too, explore Nature by probing, observing, and measuring\n$\\textit{subsets}$ of variables at any given time. Mathematically,\n$\\textit{out-of-variable}$ generalization requires the efficient re-use of past\nmarginal information, i.e., information over subsets of previously observed\nvariables. We study this problem, focusing on prediction tasks across\nenvironments that contain overlapping, yet distinct, sets of causes. We show\nthat after fitting a classifier, the residual distribution in one environment\nreveals the partial derivative of the true generating function with respect to\nthe unobserved causal parent in that environment. We leverage this information\nand propose a method that exhibits non-trivial out-of-variable generalization\nperformance when facing an overlapping, yet distinct, set of causal predictors.\n", "date": "2023-04-16", "categories": []}
{"id": "2309.09323", "title": "Answering Causal Queries at Layer 3 with DiscoSCMs-Embracing\n  Heterogeneity", "abstract": "  In the realm of causal inference, Potential Outcomes (PO) and Structural\nCausal Models (SCM) are recognized as the principal frameworks.However, when it\ncomes to Layer 3 valuations -- counterfactual queries deeply entwined with\nindividual-level semantics -- both frameworks encounter limitations due to the\ndegenerative issues brought forth by the consistency rule. This paper advocates\nfor the Distribution-consistency Structural Causal Models (DiscoSCM) framework\nas a pioneering approach to counterfactual inference, skillfully integrating\nthe strengths of both PO and SCM. The DiscoSCM framework distinctively\nincorporates a unit selection variable $U$ and embraces the concept of\nuncontrollable exogenous noise realization. Through personalized incentive\nscenarios, we demonstrate the inadequacies of PO and SCM frameworks in\nrepresenting the probability of a user being a complier (a Layer 3 event)\nwithout degeneration, an issue adeptly resolved by adopting the assumption of\nindependent counterfactual noises within DiscoSCM. This innovative assumption\nbroadens the foundational counterfactual theory, facilitating the extension of\nnumerous theoretical results regarding the probability of causation to an\nindividual granularity level and leading to a comprehensive set of theories on\nheterogeneous counterfactual bounds. Ultimately, our paper posits that if one\nacknowledges and wishes to leverage the ubiquitous heterogeneity, understanding\ncausality as invariance across heterogeneous units, then DiscoSCM stands as a\nsignificant advancement in the methodology of counterfactual inference.\n", "date": "2023-09-17", "categories": []}
{"id": "2211.10805", "title": "On the Pointwise Behavior of Recursive Partitioning and Its Implications\n  for Heterogeneous Causal Effect Estimation", "abstract": "  Decision tree learning is increasingly being used for pointwise inference.\nImportant applications include causal heterogenous treatment effects and\ndynamic policy decisions, as well as conditional quantile regression and design\nof experiments, where tree estimation and inference is conducted at specific\nvalues of the covariates. In this paper, we call into question the use of\ndecision trees (trained by adaptive recursive partitioning) for such purposes\nby demonstrating that they can fail to achieve polynomial rates of convergence\nin uniform norm with non-vanishing probability, even with pruning. Instead, the\nconvergence may be arbitrarily slow or, in some important special cases, such\nas honest regression trees, fail completely. We show that random forests can\nremedy the situation, turning poor performing trees into nearly optimal\nprocedures, at the cost of losing interpretability and introducing two\nadditional tuning parameters. The two hallmarks of random forests, subsampling\nand the random feature selection mechanism, are seen to each distinctively\ncontribute to achieving nearly optimal performance for the model class\nconsidered.\n", "date": "2022-11-19", "categories": []}
{"id": "2402.04668", "title": "A Perspective on Individualized Treatment Effects Estimation from\n  Time-series Health Data", "abstract": "  The burden of diseases is rising worldwide, with unequal treatment efficacy\nfor patient populations that are underrepresented in clinical trials.\nHealthcare, however, is driven by the average population effect of medical\ntreatments and, therefore, operates in a \"one-size-fits-all\" approach, not\nnecessarily what best fits each patient. These facts suggest a pressing need\nfor methodologies to study individualized treatment effects (ITE) to drive\npersonalized treatment. Despite the increased interest in\nmachine-learning-driven ITE estimation models, the vast majority focus on\ntabular data with limited review and understanding of methodologies proposed\nfor time-series electronic health records (EHRs). To this end, this work\nprovides an overview of ITE works for time-series data and insights into future\nresearch. The work summarizes the latest work in the literature and reviews it\nin light of theoretical assumptions, types of treatment settings, and\ncomputational frameworks. Furthermore, this work discusses challenges and\nfuture research directions for ITEs in a time-series setting. We hope this work\nopens new directions and serves as a resource for understanding one of the\nexciting yet under-studied research areas.\n", "date": "2024-02-07", "categories": []}
{"id": "2402.04777", "title": "A fast score-based search algorithm for maximal ancestral graphs using\n  entropy", "abstract": "  \\emph{Maximal ancestral graph} (MAGs) is a class of graphical model that\nextend the famous \\emph{directed acyclic graph} in the presence of latent\nconfounders. Most score-based approaches to learn the unknown MAG from\nempirical data rely on BIC score which suffers from instability and heavy\ncomputations. We propose to use the framework of imsets\n\\citep{studeny2006probabilistic} to score MAGs using empirical entropy\nestimation and the newly proposed \\emph{refined Markov property}\n\\citep{hu2023towards}. Our graphical search procedure is similar to\n\\citet{claassen2022greedy} but improved from our theoretical results. We show\nthat our search algorithm is polynomial in number of nodes by restricting\ndegree, maximal head size and number of discriminating paths. In simulated\nexperiment, our algorithm shows superior performance compared to other state of\nart MAG learning algorithms.\n", "date": "2024-02-07", "categories": []}
{"id": "2310.19391", "title": "Causal Fair Metric: Bridging Causality, Individual Fairness, and\n  Adversarial Robustness", "abstract": "  Despite the essential need for comprehensive considerations in responsible\nAI, factors like robustness, fairness, and causality are often studied in\nisolation. Adversarial perturbation, used to identify vulnerabilities in\nmodels, and individual fairness, aiming for equitable treatment of similar\nindividuals, despite initial differences, both depend on metrics to generate\ncomparable input data instances. Previous attempts to define such joint metrics\noften lack general assumptions about data or structural causal models and were\nunable to reflect counterfactual proximity. To address this, our paper\nintroduces a causal fair metric formulated based on causal structures\nencompassing sensitive attributes and protected causal perturbation. To enhance\nthe practicality of our metric, we propose metric learning as a method for\nmetric estimation and deployment in real-world problems in the absence of\nstructural causal models. We also demonstrate the application of our novel\nmetric in classifiers. Empirical evaluation of real-world and synthetic\ndatasets illustrates the effectiveness of our proposed metric in achieving an\naccurate classifier with fairness, resilience to adversarial perturbations, and\na nuanced understanding of causal relationships.\n", "date": "2023-10-30", "categories": []}
{"id": "2402.02357", "title": "Multi-modal Causal Structure Learning and Root Cause Analysis", "abstract": "  Effective root cause analysis (RCA) is vital for swiftly restoring services,\nminimizing losses, and ensuring the smooth operation and management of complex\nsystems. Previous data-driven RCA methods, particularly those employing causal\ndiscovery techniques, have primarily focused on constructing dependency or\ncausal graphs for backtracking the root causes. However, these methods often\nfall short as they rely solely on data from a single modality, thereby\nresulting in suboptimal solutions. In this work, we propose Mulan, a unified\nmulti-modal causal structure learning method for root cause localization. We\nleverage a log-tailored language model to facilitate log representation\nlearning, converting log sequences into time-series data. To explore intricate\nrelationships across different modalities, we propose a contrastive\nlearning-based approach to extract modality-invariant and modality-specific\nrepresentations within a shared latent space. Additionally, we introduce a\nnovel key performance indicator-aware attention mechanism for assessing\nmodality reliability and co-learning a final causal graph. Finally, we employ\nrandom walk with restart to simulate system fault propagation and identify\npotential root causes. Extensive experiments on three real-world datasets\nvalidate the effectiveness of our proposed framework.\n", "date": "2024-02-04", "categories": []}
{"id": "2402.02663", "title": "Counterfactual Fairness Is Not Demographic Parity, and Other\n  Observations", "abstract": "  Blanket statements of equivalence between causal concepts and purely\nprobabilistic concepts should be approached with care. In this short note, I\nexamine a recent claim that counterfactual fairness is equivalent to\ndemographic parity. The claim fails to hold up upon closer examination. I will\ntake the opportunity to address some broader misunderstandings about\ncounterfactual fairness.\n", "date": "2024-02-05", "categories": []}
{"id": "2308.16262", "title": "Causal Strategic Learning with Competitive Selection", "abstract": "  We study the problem of agent selection in causal strategic learning under\nmultiple decision makers and address two key challenges that come with it.\nFirstly, while much of prior work focuses on studying a fixed pool of agents\nthat remains static regardless of their evaluations, we consider the impact of\nselection procedure by which agents are not only evaluated, but also selected.\nWhen each decision maker unilaterally selects agents by maximising their own\nutility, we show that the optimal selection rule is a trade-off between\nselecting the best agents and providing incentives to maximise the agents'\nimprovement. Furthermore, this optimal selection rule relies on incorrect\npredictions of agents' outcomes. Hence, we study the conditions under which a\ndecision maker's optimal selection rule will not lead to deterioration of\nagents' outcome nor cause unjust reduction in agents' selection chance. To that\nend, we provide an analytical form of the optimal selection rule and a\nmechanism to retrieve the causal parameters from observational data, under\ncertain assumptions on agents' behaviour. Secondly, when there are multiple\ndecision makers, the interference between selection rules introduces another\nsource of biases in estimating the underlying causal parameters. To address\nthis problem, we provide a cooperative protocol which all decision makers must\ncollectively adopt to recover the true causal parameters. Lastly, we complement\nour theoretical results with simulation studies. Our results highlight not only\nthe importance of causal modeling as a strategy to mitigate the effect of\ngaming, as suggested by previous work, but also the need of a benevolent\nregulator to enable it.\n", "date": "2023-08-30", "categories": []}
{"id": "2311.12267", "title": "Learning Causal Representations from General Environments:\n  Identifiability and Intrinsic Ambiguity", "abstract": "  We study causal representation learning, the task of recovering high-level\nlatent variables and their causal relationships in the form of a causal graph\nfrom low-level observed data (such as text and images), assuming access to\nobservations generated from multiple environments. Prior results on the\nidentifiability of causal representations typically assume access to\nsingle-node interventions which is rather unrealistic in practice, since the\nlatent variables are unknown in the first place. In this work, we provide the\nfirst identifiability results based on data that stem from general\nenvironments. We show that for linear causal models, while the causal graph can\nbe fully recovered, the latent variables are only identified up to the\nsurrounded-node ambiguity (SNA) \\citep{varici2023score}. We provide a\ncounterpart of our guarantee, showing that SNA is basically unavoidable in our\nsetting. We also propose an algorithm, \\texttt{LiNGCReL} which provably\nrecovers the ground-truth model up to SNA, and we demonstrate its effectiveness\nvia numerical experiments. Finally, we consider general non-parametric causal\nmodels and show that the same identification barrier holds when assuming access\nto groups of soft single-node interventions.\n", "date": "2023-11-21", "categories": []}
{"id": "2401.08875", "title": "DCRMTA: Unbiased Causal Representation for Multi-touch Attribution", "abstract": "  Multi-touch attribution (MTA) currently plays a pivotal role in achieving a\nfair estimation of the contributions of each advertising touchpoint to-wards\nconversion behavior, deeply influencing budget allocation and advertising\nrecommenda-tion. Previous works attempted to eliminate the bias caused by user\npreferences to achieve the unbiased assumption of the conversion model. The\nmulti-model collaboration method is not ef-ficient, and the complete\nelimination of user in-fluence also eliminates the causal effect of user\nfeatures on conversion, resulting in limited per-formance of the conversion\nmodel. This paper re-defines the causal effect of user features on con-versions\nand proposes a novel end-to-end ap-proach, Deep Causal Representation for MTA\n(DCRMTA). Our model focuses on extracting causa features between conversions\nand users while eliminating confounding variables. Fur-thermore, extensive\nexperiments demonstrate DCRMTA's superior performance in converting prediction\nacross varying data distributions, while also effectively attributing value\nacross dif-ferent advertising channels.\n", "date": "2024-01-16", "categories": []}
{"id": "2402.01785", "title": "DoubleMLDeep: Estimation of Causal Effects with Multimodal Data", "abstract": "  This paper explores the use of unstructured, multimodal data, namely text and\nimages, in causal inference and treatment effect estimation. We propose a\nneural network architecture that is adapted to the double machine learning\n(DML) framework, specifically the partially linear model. An additional\ncontribution of our paper is a new method to generate a semi-synthetic dataset\nwhich can be used to evaluate the performance of causal effect estimation in\nthe presence of text and images as confounders. The proposed methods and\narchitectures are evaluated on the semi-synthetic dataset and compared to\nstandard approaches, highlighting the potential benefit of using text and\nimages directly in causal studies. Our findings have implications for\nresearchers and practitioners in economics, marketing, finance, medicine and\ndata science in general who are interested in estimating causal quantities\nusing non-traditional data.\n", "date": "2024-02-01", "categories": []}
{"id": "2402.02696", "title": "Causal Feature Selection for Responsible Machine Learning", "abstract": "  Machine Learning (ML) has become an integral aspect of many real-world\napplications. As a result, the need for responsible machine learning has\nemerged, focusing on aligning ML models to ethical and social values, while\nenhancing their reliability and trustworthiness. Responsible ML involves many\nissues. This survey addresses four main issues: interpretability, fairness,\nadversarial robustness, and domain generalization. Feature selection plays a\npivotal role in the responsible ML tasks. However, building upon statistical\ncorrelations between variables can lead to spurious patterns with biases and\ncompromised performance. This survey focuses on the current study of causal\nfeature selection: what it is and how it can reinforce the four aspects of\nresponsible ML. By identifying features with causal impacts on outcomes and\ndistinguishing causality from correlation, causal feature selection is posited\nas a unique approach to ensuring ML models to be ethically and socially\nresponsible in high-stakes applications.\n", "date": "2024-02-05", "categories": []}
{"id": "2401.02930", "title": "Dagma-DCE: Interpretable, Non-Parametric Differentiable Causal Discovery", "abstract": "  We introduce Dagma-DCE, an interpretable and model-agnostic scheme for\ndifferentiable causal discovery. Current non- or over-parametric methods in\ndifferentiable causal discovery use opaque proxies of ``independence'' to\njustify the inclusion or exclusion of a causal relationship. We show\ntheoretically and empirically that these proxies may be arbitrarily different\nthan the actual causal strength. Juxtaposed to existing differentiable causal\ndiscovery algorithms, \\textsc{Dagma-DCE} uses an interpretable measure of\ncausal strength to define weighted adjacency matrices. In a number of simulated\ndatasets, we show our method achieves state-of-the-art level performance. We\nadditionally show that \\textsc{Dagma-DCE} allows for principled thresholding\nand sparsity penalties by domain-experts. The code for our method is available\nopen-source at https://github.com/DanWaxman/DAGMA-DCE, and can easily be\nadapted to arbitrary differentiable models.\n", "date": "2024-01-05", "categories": []}
{"id": "2402.01140", "title": "Root Cause Analysis In Microservice Using Neural Granger Causal\n  Discovery", "abstract": "  In recent years, microservices have gained widespread adoption in IT\noperations due to their scalability, maintenance, and flexibility. However, it\nbecomes challenging for site reliability engineers (SREs) to pinpoint the root\ncause due to the complex relationships in microservices when facing system\nmalfunctions. Previous research employed structured learning methods (e.g.,\nPC-algorithm) to establish causal relationships and derive root causes from\ncausal graphs. Nevertheless, they ignored the temporal order of time series\ndata and failed to leverage the rich information inherent in the temporal\nrelationships. For instance, in cases where there is a sudden spike in CPU\nutilization, it can lead to an increase in latency for other microservices.\nHowever, in this scenario, the anomaly in CPU utilization occurs before the\nlatency increase, rather than simultaneously. As a result, the PC-algorithm\nfails to capture such characteristics. To address these challenges, we propose\nRUN, a novel approach for root cause analysis using neural Granger causal\ndiscovery with contrastive learning. RUN enhances the backbone encoder by\nintegrating contextual information from time series, and leverages a time\nseries forecasting model to conduct neural Granger causal discovery. In\naddition, RUN incorporates Pagerank with a personalization vector to\nefficiently recommend the top-k root causes. Extensive experiments conducted on\nthe synthetic and real-world microservice-based datasets demonstrate that RUN\nnoticeably outperforms the state-of-the-art root cause analysis methods.\nMoreover, we provide an analysis scenario for the sock-shop case to showcase\nthe practicality and efficacy of RUN in microservice-based applications. Our\ncode is publicly available at https://github.com/zmlin1998/RUN.\n", "date": "2024-02-02", "categories": []}
{"id": "2311.18826", "title": "Geometry-Aware Normalizing Wasserstein Flows for Optimal Causal\n  Inference", "abstract": "  This paper presents a groundbreaking approach to causal inference by\nintegrating continuous normalizing flows (CNFs) with parametric submodels,\nenhancing their geometric sensitivity and improving upon traditional Targeted\nMaximum Likelihood Estimation (TMLE). Our method employs CNFs to refine TMLE,\noptimizing the Cram\\'er-Rao bound and transitioning from a predefined\ndistribution $p_0$ to a data-driven distribution $p_1$. We innovate further by\nembedding Wasserstein gradient flows within Fokker-Planck equations, thus\nimposing geometric structures that boost the robustness of CNFs, particularly\nin optimal transport theory.\n  Our approach addresses the disparity between sample and population\ndistributions, a critical factor in parameter estimation bias. We leverage\noptimal transport and Wasserstein gradient flows to develop causal inference\nmethodologies with minimal variance in finite-sample settings, outperforming\ntraditional methods like TMLE and AIPW. This novel framework, centered on\nWasserstein gradient flows, minimizes variance in efficient influence functions\nunder distribution $p_t$. Preliminary experiments showcase our method's\nsuperiority, yielding lower mean-squared errors compared to standard flows,\nthereby demonstrating the potential of geometry-aware normalizing Wasserstein\nflows in advancing statistical modeling and inference.\n", "date": "2023-11-30", "categories": []}
{"id": "2402.00043", "title": "Interactive and Intelligent Root Cause Analysis in Manufacturing with\n  Causal Bayesian Networks and Knowledge Graphs", "abstract": "  Root Cause Analysis (RCA) in the manufacturing of electric vehicles is the\nprocess of identifying fault causes. Traditionally, the RCA is conducted\nmanually, relying on process expert knowledge. Meanwhile, sensor networks\ncollect significant amounts of data in the manufacturing process. Using this\ndata for RCA makes it more efficient. However, purely data-driven methods like\nCausal Bayesian Networks have problems scaling to large-scale, real-world\nmanufacturing processes due to the vast amount of potential cause-effect\nrelationships (CERs). Furthermore, purely data-driven methods have the\npotential to leave out already known CERs or to learn spurious CERs. The paper\ncontributes by proposing an interactive and intelligent RCA tool that combines\nexpert knowledge of an electric vehicle manufacturing process and a data-driven\nmachine learning method. It uses reasoning over a large-scale Knowledge Graph\nof the manufacturing process while learning a Causal Bayesian Network. In\naddition, an Interactive User Interface enables a process expert to give\nfeedback to the root cause graph by adding and removing information to the\nKnowledge Graph. The interactive and intelligent RCA tool reduces the learning\ntime of the Causal Bayesian Network while decreasing the number of spurious\nCERs. Thus, the interactive and intelligent RCA tool closes the feedback loop\nbetween expert and machine learning method.\n", "date": "2024-01-20", "categories": []}
{"id": "2307.15176", "title": "RCT Rejection Sampling for Causal Estimation Evaluation", "abstract": "  Confounding is a significant obstacle to unbiased estimation of causal\neffects from observational data. For settings with high-dimensional covariates\n-- such as text data, genomics, or the behavioral social sciences --\nresearchers have proposed methods to adjust for confounding by adapting machine\nlearning methods to the goal of causal estimation. However, empirical\nevaluation of these adjustment methods has been challenging and limited. In\nthis work, we build on a promising empirical evaluation strategy that\nsimplifies evaluation design and uses real data: subsampling randomized\ncontrolled trials (RCTs) to create confounded observational datasets while\nusing the average causal effects from the RCTs as ground-truth. We contribute a\nnew sampling algorithm, which we call RCT rejection sampling, and provide\ntheoretical guarantees that causal identification holds in the observational\ndata to allow for valid comparisons to the ground-truth RCT. Using synthetic\ndata, we show our algorithm indeed results in low bias when oracle estimators\nare evaluated on the confounded samples, which is not always the case for a\npreviously proposed algorithm. In addition to this identification result, we\nhighlight several finite data considerations for evaluation designers who plan\nto use RCT rejection sampling on their own datasets. As a proof of concept, we\nimplement an example evaluation pipeline and walk through these finite data\nconsiderations with a novel, real-world RCT -- which we release publicly --\nconsisting of approximately 70k observations and text data as high-dimensional\ncovariates. Together, these contributions build towards a broader agenda of\nimproved empirical evaluation for causal estimation.\n", "date": "2023-07-27", "categories": []}
{"id": "2401.17737", "title": "Hierarchical Bias-Driven Stratification for Interpretable Causal Effect\n  Estimation", "abstract": "  Interpretability and transparency are essential for incorporating causal\neffect models from observational data into policy decision-making. They can\nprovide trust for the model in the absence of ground truth labels to evaluate\nthe accuracy of such models. To date, attempts at transparent causal effect\nestimation consist of applying post hoc explanation methods to black-box\nmodels, which are not interpretable. Here, we present BICauseTree: an\ninterpretable balancing method that identifies clusters where natural\nexperiments occur locally. Our approach builds on decision trees with a\ncustomized objective function to improve balancing and reduce treatment\nallocation bias. Consequently, it can additionally detect subgroups presenting\npositivity violations, exclude them, and provide a covariate-based definition\nof the target population we can infer from and generalize to. We evaluate the\nmethod's performance using synthetic and realistic datasets, explore its\nbias-interpretability tradeoff, and show that it is comparable with existing\napproaches.\n", "date": "2024-01-31", "categories": []}
{"id": "2401.18017", "title": "Causal Discovery by Kernel Deviance Measures with Heterogeneous\n  Transforms", "abstract": "  The discovery of causal relationships in a set of random variables is a\nfundamental objective of science and has also recently been argued as being an\nessential component towards real machine intelligence. One class of causal\ndiscovery techniques are founded based on the argument that there are inherent\nstructural asymmetries between the causal and anti-causal direction which could\nbe leveraged in determining the direction of causation. To go about capturing\nthese discrepancies between cause and effect remains to be a challenge and many\ncurrent state-of-the-art algorithms propose to compare the norms of the kernel\nmean embeddings of the conditional distributions. In this work, we argue that\nsuch approaches based on RKHS embeddings are insufficient in capturing\nprincipal markers of cause-effect asymmetry involving higher-order structural\nvariabilities of the conditional distributions. We propose Kernel Intrinsic\nInvariance Measure with Heterogeneous Transform (KIIM-HT) which introduces a\nnovel score measure based on heterogeneous transformation of RKHS embeddings to\nextract relevant higher-order moments of the conditional densities for causal\ndiscovery. Inference is made via comparing the score of each hypothetical\ncause-effect direction. Tests and comparisons on a synthetic dataset, a\ntwo-dimensional synthetic dataset and the real-world benchmark dataset\nT\\\"ubingen Cause-Effect Pairs verify our approach. In addition, we conduct a\nsensitivity analysis to the regularization parameter to faithfully compare\nprevious work to our method and an experiment with trials on varied\nhyperparameter values to showcase the robustness of our algorithm.\n", "date": "2024-01-31", "categories": []}
{"id": "2310.00489", "title": "Interpretable Imitation Learning with Dynamic Causal Relations", "abstract": "  Imitation learning, which learns agent policy by mimicking expert\ndemonstration, has shown promising results in many applications such as medical\ntreatment regimes and self-driving vehicles. However, it remains a difficult\ntask to interpret control policies learned by the agent. Difficulties mainly\ncome from two aspects: 1) agents in imitation learning are usually implemented\nas deep neural networks, which are black-box models and lack interpretability;\n2) the latent causal mechanism behind agents' decisions may vary along the\ntrajectory, rather than staying static throughout time steps. To increase\ntransparency and offer better interpretability of the neural agent, we propose\nto expose its captured knowledge in the form of a directed acyclic causal\ngraph, with nodes being action and state variables and edges denoting the\ncausal relations behind predictions. Furthermore, we design this causal\ndiscovery process to be state-dependent, enabling it to model the dynamics in\nlatent causal graphs. Concretely, we conduct causal discovery from the\nperspective of Granger causality and propose a self-explainable imitation\nlearning framework, {\\method}. The proposed framework is composed of three\nparts: a dynamic causal discovery module, a causality encoding module, and a\nprediction module, and is trained in an end-to-end manner. After the model is\nlearned, we can obtain causal relations among states and action variables\nbehind its decisions, exposing policies learned by it. Experimental results on\nboth synthetic and real-world datasets demonstrate the effectiveness of the\nproposed {\\method} in learning the dynamic causal graphs for understanding the\ndecision-making of imitation learning meanwhile maintaining high prediction\naccuracy.\n", "date": "2023-09-30", "categories": []}
{"id": "2312.15282", "title": "Causal Forecasting for Pricing", "abstract": "  This paper proposes a novel method for demand forecasting in a pricing\ncontext. Here, modeling the causal relationship between price as an input\nvariable to demand is crucial because retailers aim to set prices in a (profit)\noptimal manner in a downstream decision making problem. Our methods bring\ntogether the Double Machine Learning methodology for causal inference and\nstate-of-the-art transformer-based forecasting models. In extensive empirical\nexperiments, we show on the one hand that our method estimates the causal\neffect better in a fully controlled setting via synthetic, yet realistic data.\nOn the other hand, we demonstrate on real-world data that our method\noutperforms forecasting methods in off-policy settings (i.e., when there's a\nchange in the pricing policy) while only slightly trailing in the on-policy\nsetting.\n", "date": "2023-12-23", "categories": []}
{"id": "2401.16974", "title": "CORE: Towards Scalable and Efficient Causal Discovery with Reinforcement\n  Learning", "abstract": "  Causal discovery is the challenging task of inferring causal structure from\ndata. Motivated by Pearl's Causal Hierarchy (PCH), which tells us that passive\nobservations alone are not enough to distinguish correlation from causation,\nthere has been a recent push to incorporate interventions into machine learning\nresearch. Reinforcement learning provides a convenient framework for such an\nactive approach to learning. This paper presents CORE, a deep reinforcement\nlearning-based approach for causal discovery and intervention planning. CORE\nlearns to sequentially reconstruct causal graphs from data while learning to\nperform informative interventions. Our results demonstrate that CORE\ngeneralizes to unseen graphs and efficiently uncovers causal structures.\nFurthermore, CORE scales to larger graphs with up to 10 variables and\noutperforms existing approaches in structure estimation accuracy and sample\nefficiency. All relevant code and supplementary material can be found at\nhttps://github.com/sa-and/CORE\n", "date": "2024-01-30", "categories": []}
{"id": "2401.17027", "title": "Heterogeneous treatment effect estimation with subpopulation\n  identification for personalized medicine in opioid use disorder", "abstract": "  Deep learning models have demonstrated promising results in estimating\ntreatment effects (TEE). However, most of them overlook the variations in\ntreatment outcomes among subgroups with distinct characteristics. This\nlimitation hinders their ability to provide accurate estimations and treatment\nrecommendations for specific subgroups. In this study, we introduce a novel\nneural network-based framework, named SubgroupTE, which incorporates subgroup\nidentification and treatment effect estimation. SubgroupTE identifies diverse\nsubgroups and simultaneously estimates treatment effects for each subgroup,\nimproving the treatment effect estimation by considering the heterogeneity of\ntreatment responses. Comparative experiments on synthetic data show that\nSubgroupTE outperforms existing models in treatment effect estimation.\nFurthermore, experiments on a real-world dataset related to opioid use disorder\n(OUD) demonstrate the potential of our approach to enhance personalized\ntreatment recommendations for OUD patients.\n", "date": "2024-01-30", "categories": []}
{"id": "2309.13340", "title": "Towards LLM-guided Causal Explainability for Black-box Text Classifiers", "abstract": "  With the advent of larger and more complex deep learning models, such as in\nNatural Language Processing (NLP), model qualities like explainability and\ninterpretability, albeit highly desirable, are becoming harder challenges to\ntackle and solve. For example, state-of-the-art models in text classification\nare black-box by design. Although standard explanation methods provide some\ndegree of explainability, these are mostly correlation-based methods and do not\nprovide much insight into the model. The alternative of causal explainability\nis more desirable to achieve but extremely challenging in NLP due to a variety\nof reasons. Inspired by recent endeavors to utilize Large Language Models\n(LLMs) as experts, in this work, we aim to leverage the instruction-following\nand textual understanding capabilities of recent state-of-the-art LLMs to\nfacilitate causal explainability via counterfactual explanation generation for\nblack-box text classifiers. To do this, we propose a three-step pipeline via\nwhich, we use an off-the-shelf LLM to: (1) identify the latent or unobserved\nfeatures in the input text, (2) identify the input features associated with the\nlatent features, and finally (3) use the identified input features to generate\na counterfactual explanation. We experiment with our pipeline on multiple NLP\ntext classification datasets, with several recent LLMs, and present interesting\nand promising findings.\n", "date": "2023-09-23", "categories": []}
{"id": "2401.15447", "title": "Continuous Treatment Effect Estimation Using Gradient Interpolation and\n  Kernel Smoothing", "abstract": "  We address the Individualized continuous treatment effect (ICTE) estimation\nproblem where we predict the effect of any continuous-valued treatment on an\nindividual using observational data. The main challenge in this estimation task\nis the potential confounding of treatment assignment with an individual's\ncovariates in the training data, whereas during inference ICTE requires\nprediction on independently sampled treatments. In contrast to prior work that\nrelied on regularizers or unstable GAN training, we advocate the direct\napproach of augmenting training individuals with independently sampled\ntreatments and inferred counterfactual outcomes. We infer counterfactual\noutcomes using a two-pronged strategy: a Gradient Interpolation for\nclose-to-observed treatments, and a Gaussian Process based Kernel Smoothing\nwhich allows us to downweigh high variance inferences. We evaluate our method\non five benchmarks and show that our method outperforms six state-of-the-art\nmethods on the counterfactual estimation error. We analyze the superior\nperformance of our method by showing that (1) our inferred counterfactual\nresponses are more accurate, and (2) adding them to the training data reduces\nthe distributional distance between the confounded training distribution and\ntest distribution where treatment is independent of covariates. Our proposed\nmethod is model-agnostic and we show that it improves ICTE accuracy of several\nexisting models.\n", "date": "2024-01-27", "categories": []}
{"id": "2211.16468", "title": "Linear-Time Algorithms for Front-Door Adjustment in Causal Graphs", "abstract": "  Causal effect estimation from observational data is a fundamental task in\nempirical sciences. It becomes particularly challenging when unobserved\nconfounders are involved in a system. This paper focuses on front-door\nadjustment -- a classic technique which, using observed mediators allows to\nidentify causal effects even in the presence of unobserved confounding. While\nthe statistical properties of the front-door estimation are quite well\nunderstood, its algorithmic aspects remained unexplored for a long time. In\n2022, Jeong, Tian, and Bareinboim presented the first polynomial-time algorithm\nfor finding sets satisfying the front-door criterion in a given directed\nacyclic graph (DAG), with an $O(n^3(n+m))$ run time, where $n$ denotes the\nnumber of variables and $m$ the number of edges of the causal graph. In our\nwork, we give the first linear-time, i.e., $O(n+m)$, algorithm for this task,\nwhich thus reaches the asymptotically optimal time complexity. This result\nimplies an $O(n(n+m))$ delay enumeration algorithm of all front-door adjustment\nsets, again improving previous work by a factor of $n^3$. Moreover, we provide\nthe first linear-time algorithm for finding a minimal front-door adjustment\nset. We offer implementations of our algorithms in multiple programming\nlanguages to facilitate practical usage and empirically validate their\nfeasibility, even for large graphs.\n", "date": "2022-11-29", "categories": []}
{"id": "2309.07703", "title": "Causal Entropy and Information Gain for Measuring Causal Control", "abstract": "  Artificial intelligence models and methods commonly lack causal\ninterpretability. Despite the advancements in interpretable machine learning\n(IML) methods, they frequently assign importance to features which lack causal\ninfluence on the outcome variable. Selecting causally relevant features among\nthose identified as relevant by these methods, or even before model training,\nwould offer a solution. Feature selection methods utilizing information\ntheoretical quantities have been successful in identifying statistically\nrelevant features. However, the information theoretical quantities they are\nbased on do not incorporate causality, rendering them unsuitable for such\nscenarios. To address this challenge, this article proposes information\ntheoretical quantities that incorporate the causal structure of the system,\nwhich can be used to evaluate causal importance of features for some given\noutcome variable. Specifically, we introduce causal versions of entropy and\nmutual information, termed causal entropy and causal information gain, which\nare designed to assess how much control a feature provides over the outcome\nvariable. These newly defined quantities capture changes in the entropy of a\nvariable resulting from interventions on other variables. Fundamental results\nconnecting these quantities to the existence of causal effects are derived. The\nuse of causal information gain in feature selection is demonstrated,\nhighlighting its superiority over standard mutual information in revealing\nwhich features provide control over a chosen outcome variable. Our\ninvestigation paves the way for the development of methods with improved\ninterpretability in domains involving causation.\n", "date": "2023-09-14", "categories": []}
{"id": "2401.13758", "title": "Assumptions and Bounds in the Instrumental Variable Model", "abstract": "  In this note we give proofs for results relating to the Instrumental Variable\n(IV) model with binary response $Y$ and binary treatment $X$, but with an\ninstrument $Z$ with $K$ states. These results were originally stated in\nRichardson & Robins (2014), \"ACE Bounds; SEMS with Equilibrium Conditions,\"\narXiv:1410.0470.\n", "date": "2024-01-24", "categories": []}
{"id": "2401.14426", "title": "M$^3$TN: Multi-gate Mixture-of-Experts based Multi-valued Treatment\n  Network for Uplift Modeling", "abstract": "  Uplift modeling is a technique used to predict the effect of a treatment\n(e.g., discounts) on an individual's response. Although several methods have\nbeen proposed for multi-valued treatment, they are extended from binary\ntreatment methods. There are still some limitations. Firstly, existing methods\ncalculate uplift based on predicted responses, which may not guarantee a\nconsistent uplift distribution between treatment and control groups. Moreover,\nthis may cause cumulative errors for multi-valued treatment. Secondly, the\nmodel parameters become numerous with many prediction heads, leading to reduced\nefficiency. To address these issues, we propose a novel \\underline{M}ulti-gate\n\\underline{M}ixture-of-Experts based \\underline{M}ulti-valued\n\\underline{T}reatment \\underline{N}etwork (M$^3$TN). M$^3$TN consists of two\ncomponents: 1) a feature representation module with Multi-gate\nMixture-of-Experts to improve the efficiency; 2) a reparameterization module by\nmodeling uplift explicitly to improve the effectiveness. We also conduct\nextensive experiments to demonstrate the effectiveness and efficiency of our\nM$^3$TN.\n", "date": "2024-01-24", "categories": []}
{"id": "2210.14080", "title": "Learning Individual Treatment Effects under Heterogeneous Interference\n  in Networks", "abstract": "  Estimates of individual treatment effects from networked observational data\nare attracting increasing attention these days. One major challenge in network\nscenarios is the violation of the stable unit treatment value assumption\n(SUTVA), which assumes that the treatment assignment of a unit does not\ninfluence others' outcomes. In network data, due to interference, the outcome\nof a unit is influenced not only by its treatment (i.e., direct effects) but\nalso by others' treatments (i.e., spillover effects). Furthermore, the\ninfluences from other units are always heterogeneous (e.g., friends with\nsimilar interests affect a person differently than friends with different\ninterests). In this paper, we focus on the problem of estimating individual\ntreatment effects (both direct and spillover effects) under heterogeneous\ninterference. To address this issue, we propose a novel Dual Weighting\nRegression (DWR) algorithm by simultaneously learning attention weights that\ncapture the heterogeneous interference and sample weights to eliminate the\ncomplex confounding bias in networks. We formulate the entire learning process\nas a bi-level optimization problem. In theory, we present generalization error\nbounds for individual treatment effect estimation. Extensive experiments on\nfour benchmark datasets demonstrate that the proposed DWR algorithm outperforms\nstate-of-the-art methods for estimating individual treatment effects under\nheterogeneous interference.\n", "date": "2022-10-25", "categories": []}
{"id": "2401.13935", "title": "A New Paradigm for Counterfactual Reasoning in Fairness and Recourse", "abstract": "  Counterfactuals and counterfactual reasoning underpin numerous techniques for\nauditing and understanding artificial intelligence (AI) systems. The\ntraditional paradigm for counterfactual reasoning in this literature is the\ninterventional counterfactual, where hypothetical interventions are imagined\nand simulated. For this reason, the starting point for causal reasoning about\nlegal protections and demographic data in AI is an imagined intervention on a\nlegally-protected characteristic, such as ethnicity, race, gender, disability,\nage, etc. We ask, for example, what would have happened had your race been\ndifferent? An inherent limitation of this paradigm is that some demographic\ninterventions -- like interventions on race -- may not translate into the\nformalisms of interventional counterfactuals. In this work, we explore a new\nparadigm based instead on the backtracking counterfactual, where rather than\nimagine hypothetical interventions on legally-protected characteristics, we\nimagine alternate initial conditions while holding these characteristics fixed.\nWe ask instead, what would explain a counterfactual outcome for you as you\nactually are or could be? This alternate framework allows us to address many of\nthe same social concerns, but to do so while asking fundamentally different\nquestions that do not rely on demographic interventions.\n", "date": "2024-01-25", "categories": []}
{"id": "2302.11944", "title": "Counterfactual Situation Testing: Uncovering Discrimination under\n  Fairness given the Difference", "abstract": "  We present counterfactual situation testing (CST), a causal data mining\nframework for detecting discrimination in classifiers. CST aims to answer in an\nactionable and meaningful way the intuitive question \"what would have been the\nmodel outcome had the individual, or complainant, been of a different protected\nstatus?\" It extends the legally-grounded situation testing of Thanh et al.\n(2011) by operationalizing the notion of fairness given the difference using\ncounterfactual reasoning. For any complainant, we find and compare similar\nprotected and non-protected instances in the dataset used by the classifier to\nconstruct a control and test group, where a difference between the decision\noutcomes of the two groups implies potential individual discrimination. Unlike\nsituation testing, which builds both groups around the complainant, we build\nthe test group on the complainant's counterfactual generated using causal\nknowledge. The counterfactual is intended to reflect how the protected\nattribute when changed affects the seemingly neutral attributes used by the\nclassifier, which is taken for granted in many frameworks for discrimination.\nUnder CST, we compare similar individuals within each group but dissimilar\nindividuals across both groups due to the possible difference between the\ncomplainant and its counterfactual. Evaluating our framework on two\nclassification scenarios, we show that it uncovers a greater number of cases\nthan situation testing, even when the classifier satisfies the counterfactual\nfairness condition of Kusner et al. (2017).\n", "date": "2023-02-23", "categories": []}
{"id": "2310.03258", "title": "Assessing Electricity Service Unfairness with Transfer Counterfactual\n  Learning", "abstract": "  Energy justice is a growing area of interest in interdisciplinary energy\nresearch. However, identifying systematic biases in the energy sector remains\nchallenging due to confounding variables, intricate heterogeneity in\ncounterfactual effects, and limited data availability. First, this paper\ndemonstrates how one can evaluate counterfactual unfairness in a power system\nby analyzing the average causal effect of a specific protected attribute.\nSubsequently, we use subgroup analysis to handle model heterogeneity and\nintroduce a novel method for estimating counterfactual unfairness based on\ntransfer learning, which helps to alleviate the data scarcity in each subgroup.\nIn our numerical analysis, we apply our method to a unique large-scale\ncustomer-level power outage data set and investigate the counterfactual effect\nof demographic factors, such as income and age of the population, on power\noutage durations. Our results indicate that low-income and elderly-populated\nareas consistently experience longer power outages under both daily and\npost-disaster operations, and such discrimination is exacerbated under severe\nconditions. These findings suggest a widespread, systematic issue of injustice\nin the power service systems and emphasize the necessity for focused\ninterventions in disadvantaged communities.\n", "date": "2023-10-05", "categories": []}
{"id": "2401.12369", "title": "SubgroupTE: Advancing Treatment Effect Estimation with Subgroup\n  Identification", "abstract": "  Precise estimation of treatment effects is crucial for evaluating\nintervention effectiveness. While deep learning models have exhibited promising\nperformance in learning counterfactual representations for treatment effect\nestimation (TEE), a major limitation in most of these models is that they treat\nthe entire population as a homogeneous group, overlooking the diversity of\ntreatment effects across potential subgroups that have varying treatment\neffects. This limitation restricts the ability to precisely estimate treatment\neffects and provide subgroup-specific treatment recommendations. In this paper,\nwe propose a novel treatment effect estimation model, named SubgroupTE, which\nincorporates subgroup identification in TEE. SubgroupTE identifies\nheterogeneous subgroups with different treatment responses and more precisely\nestimates treatment effects by considering subgroup-specific causal effects. In\naddition, SubgroupTE iteratively optimizes subgrouping and treatment effect\nestimation networks to enhance both estimation and subgroup identification.\nComprehensive experiments on the synthetic and semi-synthetic datasets exhibit\nthe outstanding performance of SubgroupTE compared with the state-of-the-art\nmodels on treatment effect estimation. Additionally, a real-world study\ndemonstrates the capabilities of SubgroupTE in enhancing personalized treatment\nrecommendations for patients with opioid use disorder (OUD) by advancing\ntreatment effect estimation with subgroup identification.\n", "date": "2024-01-22", "categories": []}
{"id": "2312.16113", "title": "Task-Driven Causal Feature Distillation: Towards Trustworthy Risk\n  Prediction", "abstract": "  Since artificial intelligence has seen tremendous recent successes in many\nareas, it has sparked great interest in its potential for trustworthy and\ninterpretable risk prediction. However, most models lack causal reasoning and\nstruggle with class imbalance, leading to poor precision and recall. To address\nthis, we propose a Task-Driven Causal Feature Distillation model (TDCFD) to\ntransform original feature values into causal feature attributions for the\nspecific risk prediction task. The causal feature attribution helps describe\nhow much contribution the value of this feature can make to the risk prediction\nresult. After the causal feature distillation, a deep neural network is applied\nto produce trustworthy prediction results with causal interpretability and high\nprecision/recall. We evaluate the performance of our TDCFD method on several\nsynthetic and real datasets, and the results demonstrate its superiority over\nthe state-of-the-art methods regarding precision, recall, interpretability, and\ncausality.\n", "date": "2023-12-20", "categories": []}
{"id": "2401.10937", "title": "Subjective Causality", "abstract": "  We show that it is possible to understand and identify a decision maker's\nsubjective causal judgements by observing her preferences over interventions.\nFollowing Pearl [2000], we represent causality using causal models (also called\nstructural equations models), where the world is described by a collection of\nvariables, related by equations. We show that if a preference relation over\ninterventions satisfies certain axioms (related to standard axioms regarding\ncounterfactuals), then we can define (i) a causal model, (ii) a probability\ncapturing the decision-maker's uncertainty regarding the external factors in\nthe world and (iii) a utility on outcomes such that each intervention is\nassociated with an expected utility and such that intervention $A$ is preferred\nto $B$ iff the expected utility of $A$ is greater than that of $B$. In\naddition, we characterize when the causal model is unique. Thus, our results\nallow a modeler to test the hypothesis that a decision maker's preferences are\nconsistent with some causal model and to identify causal judgements from\nobserved behavior.\n", "date": "2024-01-17", "categories": []}
{"id": "2401.11394", "title": "Causal Generative Explainers using Counterfactual Inference: A Case\n  Study on the Morpho-MNIST Dataset", "abstract": "  In this paper, we propose leveraging causal generative learning as an\ninterpretable tool for explaining image classifiers. Specifically, we present a\ngenerative counterfactual inference approach to study the influence of visual\nfeatures (i.e., pixels) as well as causal factors through generative learning.\nTo this end, we first uncover the most influential pixels on a classifier's\ndecision by varying the value of a causal attribute via counterfactual\ninference and computing both Shapely and contrastive explanations for\ncounterfactual images with these different attribute values. We then establish\na Monte-Carlo mechanism using the generator of a causal generative model in\norder to adapt Shapley explainers to produce feature importances for the\nhuman-interpretable attributes of a causal dataset in the case where a\nclassifier has been trained exclusively on the images of the dataset. Finally,\nwe present optimization methods for creating counterfactual explanations of\nclassifiers by means of counterfactual inference, proposing straightforward\napproaches for both differentiable and arbitrary classifiers. We exploit the\nMorpho-MNIST causal dataset as a case study for exploring our proposed methods\nfor generating counterfacutl explantions. We employ visual explanation methods\nfrom OmnixAI open source toolkit to compare them with our proposed methods. By\nemploying quantitative metrics to measure the interpretability of\ncounterfactual explanations, we find that our proposed methods of\ncounterfactual explanation offer more interpretable explanations compared to\nthose generated from OmnixAI. This finding suggests that our methods are\nwell-suited for generating highly interpretable counterfactual explanations on\ncausal datasets.\n", "date": "2024-01-21", "categories": []}
{"id": "2401.10495", "title": "Causal Layering via Conditional Entropy", "abstract": "  Causal discovery aims to recover information about an unobserved causal graph\nfrom the observable data it generates. Layerings are orderings of the variables\nwhich place causes before effects. In this paper, we provide ways to recover\nlayerings of a graph by accessing the data via a conditional entropy oracle,\nwhen distributions are discrete. Our algorithms work by repeatedly removing\nsources or sinks from the graph. Under appropriate assumptions and\nconditioning, we can separate the sources or sinks from the remainder of the\nnodes by comparing their conditional entropy to the unconditional entropy of\ntheir noise. Our algorithms are provably correct and run in worst-case\nquadratic time. The main assumptions are faithfulness and injective noise, and\neither known noise entropies or weakly monotonically increasing noise entropies\nalong directed paths. In addition, we require one of either a very mild\nextension of faithfulness, or strictly monotonically increasing noise\nentropies, or expanding noise injectivity to include an additional single\nargument in the structural functions.\n", "date": "2024-01-19", "categories": []}
{"id": "2401.07231", "title": "Use of Prior Knowledge to Discover Causal Additive Models with\n  Unobserved Variables and its Application to Time Series Data", "abstract": "  This paper proposes two methods for causal additive models with unobserved\nvariables (CAM-UV). CAM-UV assumes that the causal functions take the form of\ngeneralized additive models and that latent confounders are present. First, we\npropose a method that leverages prior knowledge for efficient causal discovery.\nThen, we propose an extension of this method for inferring causality in time\nseries data. The original CAM-UV algorithm differs from other existing causal\nfunction models in that it does not seek the causal order between observed\nvariables, but rather aims to identify the causes for each observed variable.\nTherefore, the first proposed method in this paper utilizes prior knowledge,\nsuch as understanding that certain variables cannot be causes of specific\nothers. Moreover, by incorporating the prior knowledge that causes precedes\ntheir effects in time, we extend the first algorithm to the second method for\ncausal discovery in time series data. We validate the first proposed method by\nusing simulated data to demonstrate that the accuracy of causal discovery\nincreases as more prior knowledge is accumulated. Additionally, we test the\nsecond proposed method by comparing it with existing time series causal\ndiscovery methods, using both simulated data and real-world data.\n", "date": "2024-01-14", "categories": []}
{"id": "2208.07898", "title": "Collaborative causal inference on distributed data", "abstract": "  In recent years, the development of technologies for causal inference with\nprivacy preservation of distributed data has gained considerable attention.\nMany existing methods for distributed data focus on resolving the lack of\nsubjects (samples) and can only reduce random errors in estimating treatment\neffects. In this study, we propose a data collaboration quasi-experiment\n(DC-QE) that resolves the lack of both subjects and covariates, reducing random\nerrors and biases in the estimation. Our method involves constructing\ndimensionality-reduced intermediate representations from private data from\nlocal parties, sharing intermediate representations instead of private data for\nprivacy preservation, estimating propensity scores from the shared intermediate\nrepresentations, and finally, estimating the treatment effects from propensity\nscores. Through numerical experiments on both artificial and real-world data,\nwe confirm that our method leads to better estimation results than individual\nanalyses. While dimensionality reduction loses some information in the private\ndata and causes performance degradation, we observe that sharing intermediate\nrepresentations with many parties to resolve the lack of subjects and\ncovariates sufficiently improves performance to overcome the degradation caused\nby dimensionality reduction. Although external validity is not necessarily\nguaranteed, our results suggest that DC-QE is a promising method. With the\nwidespread use of our method, intermediate representations can be published as\nopen data to help researchers find causalities and accumulate a knowledge base.\n", "date": "2022-08-16", "categories": []}
{"id": "2305.17225", "title": "Causal Component Analysis", "abstract": "  Independent Component Analysis (ICA) aims to recover independent latent\nvariables from observed mixtures thereof. Causal Representation Learning (CRL)\naims instead to infer causally related (thus often statistically dependent)\nlatent variables, together with the unknown graph encoding their causal\nrelationships. We introduce an intermediate problem termed Causal Component\nAnalysis (CauCA). CauCA can be viewed as a generalization of ICA, modelling the\ncausal dependence among the latent components, and as a special case of CRL. In\ncontrast to CRL, it presupposes knowledge of the causal graph, focusing solely\non learning the unmixing function and the causal mechanisms. Any impossibility\nresults regarding the recovery of the ground truth in CauCA also apply for CRL,\nwhile possibility results may serve as a stepping stone for extensions to CRL.\nWe characterize CauCA identifiability from multiple datasets generated through\ndifferent types of interventions on the latent causal variables. As a\ncorollary, this interventional perspective also leads to new identifiability\nresults for nonlinear ICA -- a special case of CauCA with an empty graph --\nrequiring strictly fewer datasets than previous results. We introduce a\nlikelihood-based approach using normalizing flows to estimate both the unmixing\nfunction and the causal mechanisms, and demonstrate its effectiveness through\nextensive synthetic experiments in the CauCA and ICA setting.\n", "date": "2023-05-26", "categories": []}
{"id": "2312.04350", "title": "CLadder: Assessing Causal Reasoning in Language Models", "abstract": "  The ability to perform causal reasoning is widely considered a core feature\nof intelligence. In this work, we investigate whether large language models\n(LLMs) can coherently reason about causality. Much of the existing work in\nnatural language processing (NLP) focuses on evaluating commonsense causal\nreasoning in LLMs, thus failing to assess whether a model can perform causal\ninference in accordance with a set of well-defined formal rules. To address\nthis, we propose a new NLP task, causal inference in natural language, inspired\nby the \"causal inference engine\" postulated by Judea Pearl et al. We compose a\nlarge dataset, CLadder, with 10K samples: based on a collection of causal\ngraphs and queries (associational, interventional, and counterfactual), we\nobtain symbolic questions and ground-truth answers, through an oracle causal\ninference engine. These are then translated into natural language. We evaluate\nmultiple LLMs on our dataset, and we introduce and evaluate a bespoke\nchain-of-thought prompting strategy, CausalCoT. We show that our task is highly\nchallenging for LLMs, and we conduct an in-depth analysis to gain deeper\ninsights into the causal reasoning abilities of LLMs. Our data is open-sourced\nat https://huggingface.co/datasets/causalNLP/cladder, and our code can be found\nat https://github.com/causalNLP/cladder.\n", "date": "2023-12-07", "categories": []}
{"id": "2303.14226", "title": "Synthetic Combinations: A Causal Inference Framework for Combinatorial\n  Interventions", "abstract": "  Consider a setting where there are $N$ heterogeneous units and $p$\ninterventions. Our goal is to learn unit-specific potential outcomes for any\ncombination of these $p$ interventions, i.e., $N \\times 2^p$ causal parameters.\nChoosing a combination of interventions is a problem that naturally arises in a\nvariety of applications such as factorial design experiments, recommendation\nengines, combination therapies in medicine, conjoint analysis, etc. Running $N\n\\times 2^p$ experiments to estimate the various parameters is likely expensive\nand/or infeasible as $N$ and $p$ grow. Further, with observational data there\nis likely confounding, i.e., whether or not a unit is seen under a combination\nis correlated with its potential outcome under that combination. To address\nthese challenges, we propose a novel latent factor model that imposes structure\nacross units (i.e., the matrix of potential outcomes is approximately rank\n$r$), and combinations of interventions (i.e., the coefficients in the Fourier\nexpansion of the potential outcomes is approximately $s$ sparse). We establish\nidentification for all $N \\times 2^p$ parameters despite unobserved\nconfounding. We propose an estimation procedure, Synthetic Combinations, and\nestablish it is finite-sample consistent and asymptotically normal under\nprecise conditions on the observation pattern. Our results imply consistent\nestimation given $\\text{poly}(r) \\times \\left( N + s^2p\\right)$ observations,\nwhile previous methods have sample complexity scaling as $\\min(N \\times s^2p, \\\n\\ \\text{poly(r)} \\times (N + 2^p))$. We use Synthetic Combinations to propose a\ndata-efficient experimental design. Empirically, Synthetic Combinations\noutperforms competing approaches on a real-world dataset on movie\nrecommendations. Lastly, we extend our analysis to do causal inference where\nthe intervention is a permutation over $p$ items (e.g., rankings).\n", "date": "2023-03-24", "categories": []}
{"id": "2305.17727", "title": "Learning a Structural Causal Model for Intuition Reasoning in\n  Conversation", "abstract": "  Reasoning, a crucial aspect of NLP research, has not been adequately\naddressed by prevailing models including Large Language Model. Conversation\nreasoning, as a critical component of it, remains largely unexplored due to the\nabsence of a well-designed cognitive model. In this paper, inspired by\nintuition theory on conversation cognition, we develop a conversation cognitive\nmodel (CCM) that explains how each utterance receives and activates channels of\ninformation recursively. Besides, we algebraically transformed CCM into a\nstructural causal model (SCM) under some mild assumptions, rendering it\ncompatible with various causal discovery methods. We further propose a\nprobabilistic implementation of the SCM for utterance-level relation reasoning.\nBy leveraging variational inference, it explores substitutes for implicit\ncauses, addresses the issue of their unobservability, and reconstructs the\ncausal representations of utterances through the evidence lower bounds.\nMoreover, we constructed synthetic and simulated datasets incorporating\nimplicit causes and complete cause labels, alleviating the current situation\nwhere all available datasets are implicit-causes-agnostic. Extensive\nexperiments demonstrate that our proposed method significantly outperforms\nexisting methods on synthetic, simulated, and real-world datasets. Finally, we\nanalyze the performance of CCM under latent confounders and propose theoretical\nideas for addressing this currently unresolved issue.\n", "date": "2023-05-28", "categories": []}
{"id": "2306.17361", "title": "iSCAN: Identifying Causal Mechanism Shifts among Nonlinear Additive\n  Noise Models", "abstract": "  Structural causal models (SCMs) are widely used in various disciplines to\nrepresent causal relationships among variables in complex systems.\nUnfortunately, the underlying causal structure is often unknown, and estimating\nit from data remains a challenging task. In many situations, however, the end\ngoal is to localize the changes (shifts) in the causal mechanisms between\nrelated datasets instead of learning the full causal structure of the\nindividual datasets. Some applications include root cause analysis, analyzing\ngene regulatory network structure changes between healthy and cancerous\nindividuals, or explaining distribution shifts. This paper focuses on\nidentifying the causal mechanism shifts in two or more related datasets over\nthe same set of variables -- without estimating the entire DAG structure of\neach SCM. Prior work under this setting assumed linear models with Gaussian\nnoises; instead, in this work we assume that each SCM belongs to the more\ngeneral class of nonlinear additive noise models (ANMs). A key technical\ncontribution of this work is to show that the Jacobian of the score function\nfor the mixture distribution allows for the identification of shifts under\ngeneral non-parametric functional mechanisms. Once the shifted variables are\nidentified, we leverage recent work to estimate the structural differences, if\nany, for the shifted variables. Experiments on synthetic and real-world data\nare provided to showcase the applicability of this approach. Code implementing\nthe proposed method is open-source and publicly available at\nhttps://github.com/kevinsbello/iSCAN.\n", "date": "2023-06-30", "categories": []}
{"id": "2310.07958", "title": "Towards Causal Deep Learning for Vulnerability Detection", "abstract": "  Deep learning vulnerability detection has shown promising results in recent\nyears. However, an important challenge that still blocks it from being very\nuseful in practice is that the model is not robust under perturbation and it\ncannot generalize well over the out-of-distribution (OOD) data, e.g., applying\na trained model to unseen projects in real world. We hypothesize that this is\nbecause the model learned non-robust features, e.g., variable names, that have\nspurious correlations with labels. When the perturbed and OOD datasets no\nlonger have the same spurious features, the model prediction fails. To address\nthe challenge, in this paper, we introduced causality into deep learning\nvulnerability detection. Our approach CausalVul consists of two phases. First,\nwe designed novel perturbations to discover spurious features that the model\nmay use to make predictions. Second, we applied the causal learning algorithms,\nspecifically, do-calculus, on top of existing deep learning models to\nsystematically remove the use of spurious features and thus promote causal\nbased prediction. Our results show that CausalVul consistently improved the\nmodel accuracy, robustness and OOD performance for all the state-of-the-art\nmodels and datasets we experimented. To the best of our knowledge, this is the\nfirst work that introduces do calculus based causal learning to software\nengineering models and shows it's indeed useful for improving the model\naccuracy, robustness and generalization. Our replication package is located at\nhttps://figshare.com/s/0ffda320dcb96c249ef2.\n", "date": "2023-10-12", "categories": []}
{"id": "2401.06864", "title": "Deep Learning With DAGs", "abstract": "  Social science theories often postulate causal relationships among a set of\nvariables or events. Although directed acyclic graphs (DAGs) are increasingly\nused to represent these theories, their full potential has not yet been\nrealized in practice. As non-parametric causal models, DAGs require no\nassumptions about the functional form of the hypothesized relationships.\nNevertheless, to simplify the task of empirical evaluation, researchers tend to\ninvoke such assumptions anyway, even though they are typically arbitrary and do\nnot reflect any theoretical content or prior knowledge. Moreover, functional\nform assumptions can engender bias, whenever they fail to accurately capture\nthe complexity of the causal system under investigation. In this article, we\nintroduce causal-graphical normalizing flows (cGNFs), a novel approach to\ncausal inference that leverages deep neural networks to empirically evaluate\ntheories represented as DAGs. Unlike conventional approaches, cGNFs model the\nfull joint distribution of the data according to a DAG supplied by the analyst,\nwithout relying on stringent assumptions about functional form. In this way,\nthe method allows for flexible, semi-parametric estimation of any causal\nestimand that can be identified from the DAG, including total effects,\nconditional effects, direct and indirect effects, and path-specific effects. We\nillustrate the method with a reanalysis of Blau and Duncan's (1967) model of\nstatus attainment and Zhou's (2019) model of conditional versus controlled\nmobility. To facilitate adoption, we provide open-source software together with\na series of online tutorials for implementing cGNFs. The article concludes with\na discussion of current limitations and directions for future development.\n", "date": "2024-01-12", "categories": []}
{"id": "2401.07578", "title": "Confounded Budgeted Causal Bandits", "abstract": "  We study the problem of learning 'good' interventions in a stochastic\nenvironment modeled by its underlying causal graph. Good interventions refer to\ninterventions that maximize rewards. Specifically, we consider the setting of a\npre-specified budget constraint, where interventions can have non-uniform\ncosts. We show that this problem can be formulated as maximizing the expected\nreward for a stochastic multi-armed bandit with side information. We propose an\nalgorithm to minimize the cumulative regret in general causal graphs. This\nalgorithm trades off observations and interventions based on their costs to\nachieve the optimal reward. This algorithm generalizes the state-of-the-art\nmethods by allowing non-uniform costs and hidden confounders in the causal\ngraph. Furthermore, we develop an algorithm to minimize the simple regret in\nthe budgeted setting with non-uniform costs and also general causal graphs. We\nprovide theoretical guarantees, including both upper and lower bounds, as well\nas empirical evaluations of our algorithms. Our empirical results showcase that\nour algorithms outperform the state of the art.\n", "date": "2024-01-15", "categories": []}
{"id": "2401.08221", "title": "Towards Causal Relationship in Indefinite Data: Baseline Model and New\n  Datasets", "abstract": "  Integrating deep learning and causal discovery has encouraged us to spot that\nlearning causal structures and representations in dialogue and video is full of\nchallenges. We defined These data forms as \"Indefinite Data\", characterized by\nmulti-structure data and multi-value representations. Unlike existing adaptable\ndata forms, Indefinite Data still faces gaps in datasets and methods. To\naddress the dataset gap, we release two high-quality datasets - Causalogue and\nCausaction, containing text dialogue samples and video action samples with\ncausal annotations respectively. Moreover, the method gap arises from the\ncoexistence of multi-structure data and multi-value representations, breaking\nthe assumptions of all current methods and rendering them infeasible on\nIndefinite Data. To this end, we propose a probabilistic framework as a\nbaseline, incorporating three designed highlights for this gap: 1) establishing\nCausation Condition of representations using the independence of noise terms\nunder non-fixed causal structures, 2) treating causal strength as a latent\nvariable and measuring the reconstruction loss in the correlation space, and 3)\nestimating the effects of latent confounders. These highpoints make the\nprobabilistic model capable of overcoming challenges brought by the coexistence\nof multi-structure data and multi-value representations and pave the way for\nthe extension of latent confounders. Comprehensive experiments have evaluated\nbaseline results of causal structures, causal representations, and confounding\ndisentanglement.\n", "date": "2024-01-16", "categories": []}
{"id": "2401.08386", "title": "Deep Learning-based Group Causal Inference in Multivariate Time-series", "abstract": "  Causal inference in a nonlinear system of multivariate timeseries is\ninstrumental in disentangling the intricate web of relationships among\nvariables, enabling us to make more accurate predictions and gain deeper\ninsights into real-world complex systems. Causality methods typically identify\nthe causal structure of a multivariate system by considering the cause-effect\nrelationship of each pair of variables while ignoring the collective effect of\na group of variables or interactions involving more than two-time series\nvariables. In this work, we test model invariance by group-level interventions\non the trained deep networks to infer causal direction in groups of variables,\nsuch as climate and ecosystem, brain networks, etc. Extensive testing with\nsynthetic and real-world time series data shows a significant improvement of\nour method over other applied group causality methods and provides us insights\ninto real-world time series. The code for our method can be found\nat:https://github.com/wasimahmadpk/gCause.\n", "date": "2024-01-16", "categories": []}
{"id": "2401.06523", "title": "Boosting Causal Additive Models", "abstract": "  We present a boosting-based method to learn additive Structural Equation\nModels (SEMs) from observational data, with a focus on the theoretical aspects\nof determining the causal order among variables. We introduce a family of score\nfunctions based on arbitrary regression techniques, for which we establish\nnecessary conditions to consistently favor the true causal ordering. Our\nanalysis reveals that boosting with early stopping meets these criteria and\nthus offers a consistent score function for causal orderings. To address the\nchallenges posed by high-dimensional data sets, we adapt our approach through a\ncomponent-wise gradient descent in the space of additive SEMs. Our simulation\nstudy underlines our theoretical results for lower dimensions and demonstrates\nthat our high-dimensional adaptation is competitive with state-of-the-art\nmethods. In addition, it exhibits robustness with respect to the choice of the\nhyperparameters making the procedure easy to tune.\n", "date": "2024-01-12", "categories": []}
{"id": "2401.06557", "title": "Treatment-Aware Hyperbolic Representation Learning for Causal Effect\n  Estimation with Social Networks", "abstract": "  Estimating the individual treatment effect (ITE) from observational data is a\ncrucial research topic that holds significant value across multiple domains.\nHow to identify hidden confounders poses a key challenge in ITE estimation.\nRecent studies have incorporated the structural information of social networks\nto tackle this challenge, achieving notable advancements. However, these\nmethods utilize graph neural networks to learn the representation of hidden\nconfounders in Euclidean space, disregarding two critical issues: (1) the\nsocial networks often exhibit a scalefree structure, while Euclidean embeddings\nsuffer from high distortion when used to embed such graphs, and (2) each\nego-centric network within a social network manifests a treatment-related\ncharacteristic, implying significant patterns of hidden confounders. To address\nthese issues, we propose a novel method called Treatment-Aware Hyperbolic\nRepresentation Learning (TAHyper). Firstly, TAHyper employs the hyperbolic\nspace to encode the social networks, thereby effectively reducing the\ndistortion of confounder representation caused by Euclidean embeddings.\nSecondly, we design a treatment-aware relationship identification module that\nenhances the representation of hidden confounders by identifying whether an\nindividual and her neighbors receive the same treatment. Extensive experiments\non two benchmark datasets are conducted to demonstrate the superiority of our\nmethod.\n", "date": "2024-01-12", "categories": []}
{"id": "2306.01424", "title": "Partial Counterfactual Identification of Continuous Outcomes with a\n  Curvature Sensitivity Model", "abstract": "  Counterfactual inference aims to answer retrospective \"what if\" questions and\nthus belongs to the most fine-grained type of inference in Pearl's causality\nladder. Existing methods for counterfactual inference with continuous outcomes\naim at point identification and thus make strong and unnatural assumptions\nabout the underlying structural causal model. In this paper, we relax these\nassumptions and aim at partial counterfactual identification of continuous\noutcomes, i.e., when the counterfactual query resides in an ignorance interval\nwith informative bounds. We prove that, in general, the ignorance interval of\nthe counterfactual queries has non-informative bounds, already when functions\nof structural causal models are continuously differentiable. As a remedy, we\npropose a novel sensitivity model called Curvature Sensitivity Model. This\nallows us to obtain informative bounds by bounding the curvature of level sets\nof the functions. We further show that existing point counterfactual\nidentification methods are special cases of our Curvature Sensitivity Model\nwhen the bound of the curvature is set to zero. We then propose an\nimplementation of our Curvature Sensitivity Model in the form of a novel deep\ngenerative model, which we call Augmented Pseudo-Invertible Decoder. Our\nimplementation employs (i) residual normalizing flows with (ii) variational\naugmentations. We empirically demonstrate the effectiveness of our Augmented\nPseudo-Invertible Decoder. To the best of our knowledge, ours is the first\npartial identification model for Markovian structural causal models with\ncontinuous outcomes.\n", "date": "2023-06-02", "categories": []}
{"id": "2308.04365", "title": "SLEM: Machine Learning for Path Modeling and Causal Inference with Super\n  Learner Equation Modeling", "abstract": "  Causal inference is a crucial goal of science, enabling researchers to arrive\nat meaningful conclusions regarding the predictions of hypothetical\ninterventions using observational data. Path models, Structural Equation Models\n(SEMs), and, more generally, Directed Acyclic Graphs (DAGs), provide a means to\nunambiguously specify assumptions regarding the causal structure underlying a\nphenomenon. Unlike DAGs, which make very few assumptions about the functional\nand parametric form, SEM assumes linearity. This can result in functional\nmisspecification which prevents researchers from undertaking reliable effect\nsize estimation. In contrast, we propose Super Learner Equation Modeling, a\npath modeling technique integrating machine learning Super Learner ensembles.\nWe empirically demonstrate its ability to provide consistent and unbiased\nestimates of causal effects, its competitive performance for linear models when\ncompared with SEM, and highlight its superiority over SEM when dealing with\nnon-linear relationships. We provide open-source code, and a tutorial notebook\nwith example usage, accentuating the easy-to-use nature of the method.\n", "date": "2023-08-08", "categories": []}
{"id": "2308.07520", "title": "Nonlinearity, Feedback and Uniform Consistency in Causal Structural\n  Learning", "abstract": "  The goal of Causal Discovery is to find automated search methods for learning\ncausal structures from observational data. In some cases all variables of the\ninterested causal mechanism are measured, and the task is to predict the\neffects one measured variable has on another. In contrast, sometimes the\nvariables of primary interest are not directly observable but instead inferred\nfrom their manifestations in the data. These are referred to as latent\nvariables. One commonly known example is the psychological construct of\nintelligence, which cannot directly measured so researchers try to assess\nthrough various indicators such as IQ tests. In this case, casual discovery\nalgorithms can uncover underlying patterns and structures to reveal the causal\nconnections between the latent variables and between the latent and observed\nvariables. This thesis focuses on two questions in causal discovery: providing\nan alternative definition of k-Triangle Faithfulness that (i) is weaker than\nstrong faithfulness when applied to the Gaussian family of distributions, (ii)\ncan be applied to non-Gaussian families of distributions, and (iii) under the\nassumption that the modified version of Strong Faithfulness holds, can be used\nto show the uniform consistency of a modified causal discovery algorithm;\nrelaxing the sufficiency assumption to learn causal structures with latent\nvariables. Given the importance of inferring cause-and-effect relationships for\nunderstanding and forecasting complex systems, the work in this thesis of\nrelaxing various simplification assumptions is expected to extend the causal\ndiscovery method to be applicable in a wider range with diversified causal\nmechanism and statistical phenomena.\n", "date": "2023-08-15", "categories": []}
{"id": "2401.04890", "title": "Nonparametric Partial Disentanglement via Mechanism Sparsity: Sparse\n  Actions, Interventions and Sparse Temporal Dependencies", "abstract": "  This work introduces a novel principle for disentanglement we call mechanism\nsparsity regularization, which applies when the latent factors of interest\ndepend sparsely on observed auxiliary variables and/or past latent factors. We\npropose a representation learning method that induces disentanglement by\nsimultaneously learning the latent factors and the sparse causal graphical\nmodel that explains them. We develop a nonparametric identifiability theory\nthat formalizes this principle and shows that the latent factors can be\nrecovered by regularizing the learned causal graph to be sparse. More\nprecisely, we show identifiablity up to a novel equivalence relation we call\n\"consistency\", which allows some latent factors to remain entangled (hence the\nterm partial disentanglement). To describe the structure of this entanglement,\nwe introduce the notions of entanglement graphs and graph preserving functions.\nWe further provide a graphical criterion which guarantees complete\ndisentanglement, that is identifiability up to permutations and element-wise\ntransformations. We demonstrate the scope of the mechanism sparsity principle\nas well as the assumptions it relies on with several worked out examples. For\ninstance, the framework shows how one can leverage multi-node interventions\nwith unknown targets on the latent factors to disentangle them. We further draw\nconnections between our nonparametric results and the now popular exponential\nfamily assumption. Lastly, we propose an estimation procedure based on\nvariational autoencoders and a sparsity constraint and demonstrate it on\nvarious synthetic datasets. This work is meant to be a significantly extended\nversion of Lachapelle et al. (2022).\n", "date": "2024-01-10", "categories": []}
{"id": "2309.02281", "title": "s-ID: Causal Effect Identification in a Sub-Population", "abstract": "  Causal inference in a sub-population involves identifying the causal effect\nof an intervention on a specific subgroup, which is distinguished from the\nwhole population through the influence of systematic biases in the sampling\nprocess. However, ignoring the subtleties introduced by sub-populations can\neither lead to erroneous inference or limit the applicability of existing\nmethods. We introduce and advocate for a causal inference problem in\nsub-populations (henceforth called s-ID), in which we merely have access to\nobservational data of the targeted sub-population (as opposed to the entire\npopulation). Existing inference problems in sub-populations operate on the\npremise that the given data distributions originate from the entire population,\nthus, cannot tackle the s-ID problem. To address this gap, we provide necessary\nand sufficient conditions that must hold in the causal graph for a causal\neffect in a sub-population to be identifiable from the observational\ndistribution of that sub-population. Given these conditions, we present a sound\nand complete algorithm for the s-ID problem.\n", "date": "2023-09-05", "categories": []}
{"id": "2310.12803", "title": "Data Augmentations for Improved (Large) Language Model Generalization", "abstract": "  The reliance of text classifiers on spurious correlations can lead to poor\ngeneralization at deployment, raising concerns about their use in\nsafety-critical domains such as healthcare. In this work, we propose to use\ncounterfactual data augmentation, guided by knowledge of the causal structure\nof the data, to simulate interventions on spurious features and to learn more\nrobust text classifiers. We show that this strategy is appropriate in\nprediction problems where the label is spuriously correlated with an attribute.\nUnder the assumptions of such problems, we discuss the favorable sample\ncomplexity of counterfactual data augmentation, compared to importance\nre-weighting. Pragmatically, we match examples using auxiliary data, based on\ndiff-in-diff methodology, and use a large language model (LLM) to represent a\nconditional probability of text. Through extensive experimentation on learning\ncaregiver-invariant predictors of clinical diagnoses from medical narratives\nand on semi-synthetic data, we demonstrate that our method for simulating\ninterventions improves out-of-distribution (OOD) accuracy compared to baseline\ninvariant learning algorithms.\n", "date": "2023-10-19", "categories": []}
{"id": "2211.04370", "title": "NESTER: An Adaptive Neurosymbolic Method for Causal Effect Estimation", "abstract": "  Causal effect estimation from observational data is a central problem in\ncausal inference. Methods based on potential outcomes framework solve this\nproblem by exploiting inductive biases and heuristics from causal inference.\nEach of these methods addresses a specific aspect of causal effect estimation,\nsuch as controlling propensity score, enforcing randomization, etc., by\ndesigning neural network (NN) architectures and regularizers. In this paper, we\npropose an adaptive method called Neurosymbolic Causal Effect Estimator\n(NESTER), a generalized method for causal effect estimation. NESTER integrates\nthe ideas used in existing methods based on multi-head NNs for causal effect\nestimation into one framework. We design a Domain Specific Language (DSL)\ntailored for causal effect estimation based on causal inductive biases used in\nliterature. We conduct a theoretical analysis to investigate NESTER's efficacy\nin estimating causal effects. Our comprehensive empirical results show that\nNESTER performs better than state-of-the-art methods on benchmark datasets.\n", "date": "2022-11-08", "categories": []}
{"id": "2211.11183", "title": "Causal Fairness Assessment of Treatment Allocation with Electronic\n  Health Records", "abstract": "  Healthcare continues to grapple with the persistent issue of treatment\ndisparities, sparking concerns regarding the equitable allocation of treatments\nin clinical practice. While various fairness metrics have emerged to assess\nfairness in decision-making processes, a growing focus has been on\ncausality-based fairness concepts due to their capacity to mitigate confounding\neffects and reason about bias. However, the application of causal fairness\nnotions in evaluating the fairness of clinical decision-making with electronic\nhealth record (EHR) data remains an understudied domain. This study aims to\naddress the methodological gap in assessing causal fairness of treatment\nallocation with electronic health records data. We propose a causal fairness\nalgorithm to assess fairness in clinical decision-making. Our algorithm\naccounts for the heterogeneity of patient populations and identifies potential\nunfairness in treatment allocation by conditioning on patients who have the\nsame likelihood to benefit from the treatment. We apply this framework to a\npatient cohort with coronary artery disease derived from an EHR database to\nevaluate the fairness of treatment decisions. In addition, we investigate the\nimpact of social determinants of health on the assessment of causal fairness of\ntreatment allocation.\n", "date": "2022-11-21", "categories": []}
{"id": "2303.13850", "title": "Towards Learning and Explaining Indirect Causal Effects in Neural\n  Networks", "abstract": "  Recently, there has been a growing interest in learning and explaining causal\neffects within Neural Network (NN) models. By virtue of NN architectures,\nprevious approaches consider only direct and total causal effects assuming\nindependence among input variables. We view an NN as a structural causal model\n(SCM) and extend our focus to include indirect causal effects by introducing\nfeedforward connections among input neurons. We propose an ante-hoc method that\ncaptures and maintains direct, indirect, and total causal effects during NN\nmodel training. We also propose an algorithm for quantifying learned causal\neffects in an NN model and efficient approximation strategies for quantifying\ncausal effects in high-dimensional data. Extensive experiments conducted on\nsynthetic and real-world datasets demonstrate that the causal effects learned\nby our ante-hoc method better approximate the ground truth effects compared to\nexisting methods.\n", "date": "2023-03-24", "categories": []}
{"id": "2401.02154", "title": "Disentangle Estimation of Causal Effects from Cross-Silo Data", "abstract": "  Estimating causal effects among different events is of great importance to\ncritical fields such as drug development. Nevertheless, the data features\nassociated with events may be distributed across various silos and remain\nprivate within respective parties, impeding direct information exchange between\nthem. This, in turn, can result in biased estimations of local causal effects,\nwhich rely on the characteristics of only a subset of the covariates. To tackle\nthis challenge, we introduce an innovative disentangle architecture designed to\nfacilitate the seamless cross-silo transmission of model parameters, enriched\nwith causal mechanisms, through a combination of shared and private branches.\nBesides, we introduce global constraints into the equation to effectively\nmitigate bias within the various missing domains, thereby elevating the\naccuracy of our causal effect estimation. Extensive experiments conducted on\nnew semi-synthetic datasets show that our method outperforms state-of-the-art\nbaselines.\n", "date": "2024-01-04", "categories": []}
{"id": "2302.00293", "title": "A Survey of Methods, Challenges and Perspectives in Causality", "abstract": "  Deep Learning models have shown success in a large variety of tasks by\nextracting correlation patterns from high-dimensional data but still struggle\nwhen generalizing out of their initial distribution. As causal engines aim to\nlearn mechanisms independent from a data distribution, combining Deep Learning\nwith Causality can have a great impact on the two fields. In this paper, we\nfurther motivate this assumption. We perform an extensive overview of the\ntheories and methods for Causality from different perspectives, with an\nemphasis on Deep Learning and the challenges met by the two domains. We show\nearly attempts to bring the fields together and the possible perspectives for\nthe future. We finish by providing a large variety of applications for\ntechniques from Causality.\n", "date": "2023-02-01", "categories": []}
{"id": "2305.09126", "title": "Transfer Learning for Causal Effect Estimation", "abstract": "  We present a Transfer Causal Learning (TCL) framework when target and source\ndomains share the same covariate/feature spaces, aiming to improve causal\neffect estimation accuracy in limited data. Limited data is very common in\nmedical applications, where some rare medical conditions, such as sepsis, are\nof interest. Our proposed method, named \\texttt{$\\ell_1$-TCL}, incorporates\n$\\ell_1$ regularized TL for nuisance models (e.g., propensity score model); the\nTL estimator of the nuisance parameters is plugged into downstream average\ncausal/treatment effect estimators (e.g., inverse probability weighted\nestimator). We establish non-asymptotic recovery guarantees for the\n\\texttt{$\\ell_1$-TCL} with generalized linear model (GLM) under the sparsity\nassumption in the high-dimensional setting, and demonstrate the empirical\nbenefits of \\texttt{$\\ell_1$-TCL} through extensive numerical simulation for\nGLM and recent neural network nuisance models. Our method is subsequently\nextended to real data and generates meaningful insights consistent with medical\nliterature, a case where all baseline methods fail.\n", "date": "2023-05-16", "categories": []}
{"id": "2401.00781", "title": "Inferring Heterogeneous Treatment Effects of Crashes on Highway Traffic:\n  A Doubly Robust Causal Machine Learning Approach", "abstract": "  Highway traffic crashes exert a considerable impact on both transportation\nsystems and the economy. In this context, accurate and dependable emergency\nresponses are crucial for effective traffic management. However, the influence\nof crashes on traffic status varies across diverse factors and may be biased\ndue to selection bias. Therefore, there arises a necessity to accurately\nestimate the heterogeneous causal effects of crashes, thereby providing\nessential insights to facilitate individual-level emergency decision-making.\nThis paper proposes a novel causal machine learning framework to estimate the\ncausal effect of different types of crashes on highway speed. The Neyman-Rubin\nCausal Model (RCM) is employed to formulate this problem from a causal\nperspective. The Conditional Shapley Value Index (CSVI) is proposed based on\ncausal graph theory to filter adverse variables, and the Structural Causal\nModel (SCM) is then adopted to define the statistical estimand for causal\neffects. The treatment effects are estimated by Doubly Robust Learning (DRL)\nmethods, which combine doubly robust causal inference with classification and\nregression machine learning models. Experimental results from 4815 crashes on\nHighway Interstate 5 in Washington State reveal the heterogeneous treatment\neffects of crashes at varying distances and durations. The rear-end crashes\ncause more severe congestion and longer durations than other types of crashes,\nand the sideswipe crashes have the longest delayed impact. Additionally, the\nfindings show that rear-end crashes affect traffic greater at night, while\ncrash to objects has the most significant influence during peak hours.\nStatistical hypothesis tests, error metrics based on matched \"counterfactual\noutcomes\", and sensitive analyses are employed for assessment, and the results\nvalidate the accuracy and effectiveness of our method.\n", "date": "2024-01-01", "categories": []}
{"id": "2312.16211", "title": "An Explainable AI Approach to Large Language Model Assisted Causal Model\n  Auditing and Development", "abstract": "  Causal networks are widely used in many fields, including epidemiology,\nsocial science, medicine, and engineering, to model the complex relationships\nbetween variables. While it can be convenient to algorithmically infer these\nmodels directly from observational data, the resulting networks are often\nplagued with erroneous edges. Auditing and correcting these networks may\nrequire domain expertise frequently unavailable to the analyst. We propose the\nuse of large language models such as ChatGPT as an auditor for causal networks.\nOur method presents ChatGPT with a causal network, one edge at a time, to\nproduce insights about edge directionality, possible confounders, and mediating\nvariables. We ask ChatGPT to reflect on various aspects of each causal link and\nwe then produce visualizations that summarize these viewpoints for the human\nanalyst to direct the edge, gather more data, or test further hypotheses. We\nenvision a system where large language models, automated causal inference, and\nthe human analyst and domain expert work hand in hand as a team to derive\nholistic and comprehensive causal models for any given case scenario. This\npaper presents first results obtained with an emerging prototype.\n", "date": "2023-12-23", "categories": []}
{"id": "2312.16707", "title": "Modeling Systemic Risk: A Time-Varying Nonparametric Causal Inference\n  Framework", "abstract": "  We propose a nonparametric and time-varying directed information graph\n(TV-DIG) framework to estimate the evolving causal structure in time series\nnetworks, thereby addressing the limitations of traditional econometric models\nin capturing high-dimensional, nonlinear, and time-varying interconnections\namong series. This framework employs an information-theoretic measure rooted in\na generalized version of Granger-causality, which is applicable to both linear\nand nonlinear dynamics. Our framework offers advancements in measuring systemic\nrisk and establishes meaningful connections with established econometric\nmodels, including vector autoregression and switching models. We evaluate the\nefficacy of our proposed model through simulation experiments and empirical\nanalysis, reporting promising results in recovering simulated time-varying\nnetworks with nonlinear and multivariate structures. We apply this framework to\nidentify and monitor the evolution of interconnectedness and systemic risk\namong major assets and industrial sectors within the financial network. We\nfocus on cryptocurrencies' potential systemic risks to financial stability,\nincluding spillover effects on other sectors during crises like the COVID-19\npandemic and the Federal Reserve's 2020 emergency response. Our findings\nreveals significant, previously underrecognized pre-2020 influences of\ncryptocurrencies on certain financial sectors, highlighting their potential\nsystemic risks and offering a systematic approach in tracking evolving\ncross-sector interactions within financial networks.\n", "date": "2023-12-27", "categories": []}
{"id": "2305.05276", "title": "Causal Discovery from Subsampled Time Series with Proxy Variables", "abstract": "  Inferring causal structures from time series data is the central interest of\nmany scientific inquiries. A major barrier to such inference is the problem of\nsubsampling, i.e., the frequency of measurement is much lower than that of\ncausal influence. To overcome this problem, numerous methods have been\nproposed, yet either was limited to the linear case or failed to achieve\nidentifiability. In this paper, we propose a constraint-based algorithm that\ncan identify the entire causal structure from subsampled time series, without\nany parametric constraint. Our observation is that the challenge of subsampling\narises mainly from hidden variables at the unobserved time steps. Meanwhile,\nevery hidden variable has an observed proxy, which is essentially itself at\nsome observable time in the future, benefiting from the temporal structure.\nBased on these, we can leverage the proxies to remove the bias induced by the\nhidden variables and hence achieve identifiability. Following this intuition,\nwe propose a proxy-based causal discovery algorithm. Our algorithm is\nnonparametric and can achieve full causal identification. Theoretical\nadvantages are reflected in synthetic and real-world experiments.\n", "date": "2023-05-09", "categories": []}
{"id": "2312.15919", "title": "Review on Causality Detection Based on Empirical Dynamic Modeling", "abstract": "  In contemporary scientific research, understanding the distinction between\ncorrelation and causation is crucial. While correlation is a widely used\nanalytical standard, it does not inherently imply causation. This paper\naddresses the potential for misinterpretation in relying solely on correlation,\nespecially in the context of nonlinear dynamics. Despite the rapid development\nof various correlation research methodologies, including machine learning, the\nexploration into mining causal correlations between variables remains ongoing.\nEmpirical Dynamic Modeling (EDM) emerges as a data-driven framework for\nmodeling dynamic systems, distinguishing itself by eschewing traditional\nformulaic methods in data analysis. Instead, it reconstructs dynamic system\nbehavior directly from time series data. The fundamental premise of EDM is that\ndynamic systems can be conceptualized as processes where a set of states,\ngoverned by specific rules, evolve over time in a high-dimensional space. By\nreconstructing these evolving states, dynamic systems can be effectively\nmodeled. Using EDM, this paper explores the detection of causal relationships\nbetween variables within dynamic systems through their time series data. It\nposits that if variable X causes variable Y, then the information about X is\ninherent in Y and can be extracted from Y's data. This study begins by\nexamining the dialectical relationship between correlation and causation,\nemphasizing that correlation does not equate to causation, and the absence of\ncorrelation does not necessarily indicate a lack of causation.\n", "date": "2023-12-26", "categories": []}
{"id": "2312.14333", "title": "Behaviour Modelling of Social Animals via Causal Structure Discovery and\n  Graph Neural Networks", "abstract": "  Better understanding the natural world is a crucial task with a wide range of\napplications. In environments with close proximity between humans and animals,\nsuch as zoos, it is essential to better understand the causes behind animal\nbehaviour and what interventions are responsible for changes in their\nbehaviours. This can help to predict unusual behaviours, mitigate detrimental\neffects and increase the well-being of animals. There has been work on\nmodelling the dynamics behind swarms of birds and insects but the complex\nsocial behaviours of mammalian groups remain less explored. In this work, we\npropose a method to build behavioural models using causal structure discovery\nand graph neural networks for time series. We apply this method to a mob of\nmeerkats in a zoo environment and study its ability to predict future actions\nand model the behaviour distribution at an individual-level and at a group\nlevel. We show that our method can match and outperform standard deep learning\narchitectures and generate more realistic data, while using fewer parameters\nand providing increased interpretability.\n", "date": "2023-12-21", "categories": []}
{"id": "2312.14670", "title": "Zero-shot Causal Graph Extrapolation from Text via LLMs", "abstract": "  We evaluate the ability of large language models (LLMs) to infer causal\nrelations from natural language. Compared to traditional natural language\nprocessing and deep learning techniques, LLMs show competitive performance in a\nbenchmark of pairwise relations without needing (explicit) training samples.\nThis motivates us to extend our approach to extrapolating causal graphs through\niterated pairwise queries. We perform a preliminary analysis on a benchmark of\nbiomedical abstracts with ground-truth causal graphs validated by experts. The\nresults are promising and support the adoption of LLMs for such a crucial step\nin causal inference, especially in medical domains, where the amount of\nscientific text to analyse might be huge, and the causal statements are often\nimplicit.\n", "date": "2023-12-22", "categories": []}
{"id": "2306.03625", "title": "Fair and Robust Estimation of Heterogeneous Treatment Effects for Policy\n  Learning", "abstract": "  We propose a simple and general framework for nonparametric estimation of\nheterogeneous treatment effects under fairness constraints. Under standard\nregularity conditions, we show that the resulting estimators possess the double\nrobustness property. We use this framework to characterize the trade-off\nbetween fairness and the maximum welfare achievable by the optimal policy. We\nevaluate the methods in a simulation study and illustrate them in a real-world\ncase study.\n", "date": "2023-06-06", "categories": []}
{"id": "2312.12678", "title": "Causal Discovery for fMRI data: Challenges, Solutions, and a Case Study", "abstract": "  Designing studies that apply causal discovery requires navigating many\nresearcher degrees of freedom. This complexity is exacerbated when the study\ninvolves fMRI data. In this paper we (i) describe nine challenges that occur\nwhen applying causal discovery to fMRI data, (ii) discuss the space of\ndecisions that need to be made, (iii) review how a recent case study made those\ndecisions, (iv) and identify existing gaps that could potentially be solved by\nthe development of new methods. Overall, causal discovery is a promising\napproach for analyzing fMRI data, and multiple successful applications have\nindicated that it is superior to traditional fMRI functional connectivity\nmethods, but current causal discovery methods for fMRI leave room for\nimprovement.\n", "date": "2023-12-20", "categories": []}
{"id": "2312.12731", "title": "Robustly Improving Bandit Algorithms with Confounded and Selection\n  Biased Offline Data: A Causal Approach", "abstract": "  This paper studies bandit problems where an agent has access to offline data\nthat might be utilized to potentially improve the estimation of each arm's\nreward distribution. A major obstacle in this setting is the existence of\ncompound biases from the observational data. Ignoring these biases and blindly\nfitting a model with the biased data could even negatively affect the online\nlearning phase. In this work, we formulate this problem from a causal\nperspective. First, we categorize the biases into confounding bias and\nselection bias based on the causal structure they imply. Next, we extract the\ncausal bound for each arm that is robust towards compound biases from biased\nobservational data. The derived bounds contain the ground truth mean reward and\ncan effectively guide the bandit agent to learn a nearly-optimal decision\npolicy. We also conduct regret analysis in both contextual and non-contextual\nbandit settings and show that prior causal bounds could help consistently\nreduce the asymptotic regret.\n", "date": "2023-12-20", "categories": []}
{"id": "2212.09897", "title": "Inducing Character-level Structure in Subword-based Language Models with\n  Type-level Interchange Intervention Training", "abstract": "  Language tasks involving character-level manipulations (e.g., spelling\ncorrections, arithmetic operations, word games) are challenging for models\noperating on subword units. To address this, we develop a causal intervention\nframework to learn robust and interpretable character representations inside\nsubword-based language models. Our method treats each character as a typed\nvariable in a causal model and learns such causal structures by adapting the\ninterchange intervention training method of Geiger et al. (2021). We\nadditionally introduce a suite of character-level tasks that systematically\nvary in their dependence on meaning and sequence-level context. While\ncharacter-level models still perform best on purely form-based tasks like\nstring reversal, our method outperforms character-level models on more complex\ntasks that blend form, meaning, and context, such as spelling correction in\ncontext and word search games. Compared with standard subword-based models, our\napproach also significantly improves robustness on unseen token sequences and\nleads to human-interpretable internal representations of characters.\n", "date": "2022-12-19", "categories": []}
{"id": "2312.08670", "title": "Temporal-Spatial Entropy Balancing for Causal Continuous\n  Treatment-Effect Estimation", "abstract": "  In the field of intracity freight transportation, changes in order volume are\nsignificantly influenced by temporal and spatial factors. When building subsidy\nand pricing strategies, predicting the causal effects of these strategies on\norder volume is crucial. In the process of calculating causal effects,\nconfounding variables can have an impact. Traditional methods to control\nconfounding variables handle data from a holistic perspective, which cannot\nensure the precision of causal effects in specific temporal and spatial\ndimensions. However, temporal and spatial dimensions are extremely critical in\nthe logistics field, and this limitation may directly affect the precision of\nsubsidy and pricing strategies. To address these issues, this study proposes a\ntechnique based on flexible temporal-spatial grid partitioning. Furthermore,\nbased on the flexible grid partitioning technique, we further propose a\ncontinuous entropy balancing method in the temporal-spatial domain, which named\nTS-EBCT (Temporal-Spatial Entropy Balancing for Causal Continue Treatments).\nThe method proposed in this paper has been tested on two simulation datasets\nand two real datasets, all of which have achieved excellent performance. In\nfact, after applying the TS-EBCT method to the intracity freight transportation\nfield, the prediction accuracy of the causal effect has been significantly\nimproved. It brings good business benefits to the company's subsidy and pricing\nstrategies.\n", "date": "2023-12-14", "categories": []}
{"id": "2312.11573", "title": "Estimation of individual causal effects in network setup for multiple\n  treatments", "abstract": "  We study the problem of estimation of Individual Treatment Effects (ITE) in\nthe context of multiple treatments and networked observational data. Leveraging\nthe network information, we aim to utilize hidden confounders that may not be\ndirectly accessible in the observed data, thereby enhancing the practical\napplicability of the strong ignorability assumption. To achieve this, we first\nemploy Graph Convolutional Networks (GCN) to learn a shared representation of\nthe confounders. Then, our approach utilizes separate neural networks to infer\npotential outcomes for each treatment. We design a loss function as a weighted\ncombination of two components: representation loss and Mean Squared Error (MSE)\nloss on the factual outcomes. To measure the representation loss, we extend\nexisting metrics such as Wasserstein and Maximum Mean Discrepancy (MMD) from\nthe binary treatment setting to the multiple treatments scenario. To validate\nthe effectiveness of our proposed methodology, we conduct a series of\nexperiments on the benchmark datasets such as BlogCatalog and Flickr. The\nexperimental results consistently demonstrate the superior performance of our\nmodels when compared to baseline methods.\n", "date": "2023-12-18", "categories": []}
{"id": "2312.11818", "title": "Root Cause Explanation of Outliers under Noisy Mechanisms", "abstract": "  Identifying root causes of anomalies in causal processes is vital across\ndisciplines. Once identified, one can isolate the root causes and implement\nnecessary measures to restore the normal operation. Causal processes are often\nmodelled as graphs with entities being nodes and their paths/interconnections\nas edge. Existing work only consider the contribution of nodes in the\ngenerative process, thus can not attribute the outlier score to the edges of\nthe mechanism if the anomaly occurs in the connections. In this paper, we\nconsider both individual edge and node of each mechanism when identifying the\nroot causes. We introduce a noisy functional causal model to account for this\npurpose. Then, we employ Bayesian learning and inference methods to infer the\nnoises of the nodes and edges. We then represent the functional form of a\ntarget outlier leaf as a function of the node and edge noises. Finally, we\npropose an efficient gradient-based attribution method to compute the anomaly\nattribution scores which scales linearly with the number of nodes and edges.\nExperiments on simulated datasets and two real-world scenario datasets show\nbetter anomaly attribution performance of the proposed method compared to the\nbaselines. Our method scales to larger graphs with more nodes and edges.\n", "date": "2023-12-19", "categories": []}
{"id": "2312.11934", "title": "Identification of Causal Structure with Latent Variables Based on Higher\n  Order Cumulants", "abstract": "  Causal discovery with latent variables is a crucial but challenging task.\nDespite the emergence of numerous methods aimed at addressing this challenge,\nthey are not fully identified to the structure that two observed variables are\ninfluenced by one latent variable and there might be a directed edge in\nbetween. Interestingly, we notice that this structure can be identified through\nthe utilization of higher-order cumulants. By leveraging the higher-order\ncumulants of non-Gaussian data, we provide an analytical solution for\nestimating the causal coefficients or their ratios. With the estimated (ratios\nof) causal coefficients, we propose a novel approach to identify the existence\nof a causal edge between two observed variables subject to latent variable\ninfluence. In case when such a causal edge exits, we introduce an asymmetry\ncriterion to determine the causal direction. The experimental results\ndemonstrate the effectiveness of our proposed method.\n", "date": "2023-12-19", "categories": []}
{"id": "2312.12206", "title": "Identification of Causal Structure in the Presence of Missing Data with\n  Additive Noise Model", "abstract": "  Missing data are an unavoidable complication frequently encountered in many\ncausal discovery tasks. When a missing process depends on the missing values\nthemselves (known as self-masking missingness), the recovery of the joint\ndistribution becomes unattainable, and detecting the presence of such\nself-masking missingness remains a perplexing challenge. Consequently, due to\nthe inability to reconstruct the original distribution and to discern the\nunderlying missingness mechanism, simply applying existing causal discovery\nmethods would lead to wrong conclusions. In this work, we found that the recent\nadvances additive noise model has the potential for learning causal structure\nunder the existence of the self-masking missingness. With this observation, we\naim to investigate the identification problem of learning causal structure from\nmissing data under an additive noise model with different missingness\nmechanisms, where the `no self-masking missingness' assumption can be\neliminated appropriately. Specifically, we first elegantly extend the scope of\nidentifiability of causal skeleton to the case with weak self-masking\nmissingness (i.e., no other variable could be the cause of self-masking\nindicators except itself). We further provide the sufficient and necessary\nidentification conditions of the causal direction under additive noise model\nand show that the causal structure can be identified up to an IN-equivalent\npattern. We finally propose a practical algorithm based on the above\ntheoretical results on learning the causal skeleton and causal direction.\nExtensive experiments on synthetic and real data demonstrate the efficiency and\neffectiveness of the proposed algorithms.\n", "date": "2023-12-19", "categories": []}
{"id": "2301.12212", "title": "Efficient Enumeration of Markov Equivalent DAGs", "abstract": "  Enumerating the directed acyclic graphs (DAGs) of a Markov equivalence class\n(MEC) is an important primitive in causal analysis. The central resource from\nthe perspective of computational complexity is the delay, that is, the time an\nalgorithm that lists all members of the class requires between two consecutive\noutputs. Commonly used algorithms for this task utilize the rules proposed by\nMeek (1995) or the transformational characterization by Chickering (1995), both\nresulting in superlinear delay. In this paper, we present the first linear-time\ndelay algorithm. On the theoretical side, we show that our algorithm can be\ngeneralized to enumerate DAGs represented by models that incorporate background\nknowledge, such as MPDAGs; on the practical side, we provide an efficient\nimplementation and evaluate it in a series of experiments. Complementary to the\nlinear-time delay algorithm, we also provide intriguing insights into Markov\nequivalence itself: All members of an MEC can be enumerated such that two\nsuccessive DAGs have structural Hamming distance at most three.\n", "date": "2023-01-28", "categories": []}
{"id": "2302.09694", "title": "Disentangled Representation for Causal Mediation Analysis", "abstract": "  Estimating direct and indirect causal effects from observational data is\ncrucial to understanding the causal mechanisms and predicting the behaviour\nunder different interventions. Causal mediation analysis is a method that is\noften used to reveal direct and indirect effects. Deep learning shows promise\nin mediation analysis, but the current methods only assume latent confounders\nthat affect treatment, mediator and outcome simultaneously, and fail to\nidentify different types of latent confounders (e.g., confounders that only\naffect the mediator or outcome). Furthermore, current methods are based on the\nsequential ignorability assumption, which is not feasible for dealing with\nmultiple types of latent confounders. This work aims to circumvent the\nsequential ignorability assumption and applies the piecemeal deconfounding\nassumption as an alternative. We propose the Disentangled Mediation Analysis\nVariational AutoEncoder (DMAVAE), which disentangles the representations of\nlatent confounders into three types to accurately estimate the natural direct\neffect, natural indirect effect and total effect. Experimental results show\nthat the proposed method outperforms existing methods and has strong\ngeneralisation ability. We further apply the method to a real-world dataset to\nshow its potential application.\n", "date": "2023-02-19", "categories": []}
{"id": "2306.02235", "title": "Learning Linear Causal Representations from Interventions under General\n  Nonlinear Mixing", "abstract": "  We study the problem of learning causal representations from unknown, latent\ninterventions in a general setting, where the latent distribution is Gaussian\nbut the mixing function is completely general. We prove strong identifiability\nresults given unknown single-node interventions, i.e., without having access to\nthe intervention targets. This generalizes prior works which have focused on\nweaker classes, such as linear maps or paired counterfactual data. This is also\nthe first instance of causal identifiability from non-paired interventions for\ndeep neural network embeddings. Our proof relies on carefully uncovering the\nhigh-dimensional geometric structure present in the data distribution after a\nnon-linear density transformation, which we capture by analyzing quadratic\nforms of precision matrices of the latent distributions. Finally, we propose a\ncontrastive algorithm to identify the latent variables in practice and evaluate\nits performance on various tasks.\n", "date": "2023-06-04", "categories": []}
{"id": "2312.10102", "title": "Robust Estimation of Causal Heteroscedastic Noise Models", "abstract": "  Distinguishing the cause and effect from bivariate observational data is the\nfoundational problem that finds applications in many scientific disciplines.\nOne solution to this problem is assuming that cause and effect are generated\nfrom a structural causal model, enabling identification of the causal direction\nafter estimating the model in each direction. The heteroscedastic noise model\nis a type of structural causal model where the cause can contribute to both the\nmean and variance of the noise. Current methods for estimating heteroscedastic\nnoise models choose the Gaussian likelihood as the optimization objective which\ncan be suboptimal and unstable when the data has a non-Gaussian distribution.\nTo address this limitation, we propose a novel approach to estimating this\nmodel with Student's $t$-distribution, which is known for its robustness in\naccounting for sampling variability with smaller sample sizes and extreme\nvalues without significantly altering the overall distribution shape. This\nadaptability is beneficial for capturing the parameters of the noise\ndistribution in heteroscedastic noise models. Our empirical evaluations\ndemonstrate that our estimators are more robust and achieve better overall\nperformance across synthetic and real benchmarks.\n", "date": "2023-12-15", "categories": []}
{"id": "2312.10388", "title": "The Causal Impact of Credit Lines on Spending Distributions", "abstract": "  Consumer credit services offered by e-commerce platforms provide customers\nwith convenient loan access during shopping and have the potential to stimulate\nsales. To understand the causal impact of credit lines on spending, previous\nstudies have employed causal estimators, based on direct regression (DR),\ninverse propensity weighting (IPW), and double machine learning (DML) to\nestimate the treatment effect. However, these estimators do not consider the\nnotion that an individual's spending can be understood and represented as a\ndistribution, which captures the range and pattern of amounts spent across\ndifferent orders. By disregarding the outcome as a distribution, valuable\ninsights embedded within the outcome distribution might be overlooked. This\npaper develops a distribution-valued estimator framework that extends existing\nreal-valued DR-, IPW-, and DML-based estimators to distribution-valued\nestimators within Rubin's causal framework. We establish their consistency and\napply them to a real dataset from a large e-commerce platform. Our findings\nreveal that credit lines positively influence spending across all quantiles;\nhowever, as credit lines increase, consumers allocate more to luxuries (higher\nquantiles) than necessities (lower quantiles).\n", "date": "2023-12-16", "categories": []}
{"id": "2312.10435", "title": "Uncertainty Quantification in Heterogeneous Treatment Effect Estimation\n  with Gaussian-Process-Based Partially Linear Model", "abstract": "  Estimating heterogeneous treatment effects across individuals has attracted\ngrowing attention as a statistical tool for performing critical\ndecision-making. We propose a Bayesian inference framework that quantifies the\nuncertainty in treatment effect estimation to support decision-making in a\nrelatively small sample size setting. Our proposed model places Gaussian\nprocess priors on the nonparametric components of a semiparametric model called\na partially linear model. This model formulation has three advantages. First,\nwe can analytically compute the posterior distribution of a treatment effect\nwithout relying on the computationally demanding posterior approximation.\nSecond, we can guarantee that the posterior distribution concentrates around\nthe true one as the sample size goes to infinity. Third, we can incorporate\nprior knowledge about a treatment effect into the prior distribution, improving\nthe estimation efficiency. Our experimental results show that even in the small\nsample size setting, our method can accurately estimate the heterogeneous\ntreatment effects and effectively quantify its estimation uncertainty.\n", "date": "2023-12-16", "categories": []}
{"id": "2312.10570", "title": "Adversarially Balanced Representation for Continuous Treatment Effect\n  Estimation", "abstract": "  Individual treatment effect (ITE) estimation requires adjusting for the\ncovariate shift between populations with different treatments, and deep\nrepresentation learning has shown great promise in learning a balanced\nrepresentation of covariates. However the existing methods mostly consider the\nscenario of binary treatments. In this paper, we consider the more practical\nand challenging scenario in which the treatment is a continuous variable (e.g.\ndosage of a medication), and we address the two main challenges of this setup.\nWe propose the adversarial counterfactual regression network (ACFR) that\nadversarially minimizes the representation imbalance in terms of KL divergence,\nand also maintains the impact of the treatment value on the outcome prediction\nby leveraging an attention mechanism. Theoretically we demonstrate that ACFR\nobjective function is grounded in an upper bound on counterfactual outcome\nprediction error. Our experimental evaluation on semi-synthetic datasets\ndemonstrates the empirical superiority of ACFR over a range of state-of-the-art\nmethods.\n", "date": "2023-12-17", "categories": []}
{"id": "2312.11001", "title": "A Versatile Causal Discovery Framework to Allow Causally-Related Hidden\n  Variables", "abstract": "  Most existing causal discovery methods rely on the assumption of no latent\nconfounders, limiting their applicability in solving real-life problems. In\nthis paper, we introduce a novel, versatile framework for causal discovery that\naccommodates the presence of causally-related hidden variables almost\neverywhere in the causal network (for instance, they can be effects of observed\nvariables), based on rank information of covariance matrix over observed\nvariables. We start by investigating the efficacy of rank in comparison to\nconditional independence and, theoretically, establish necessary and sufficient\nconditions for the identifiability of certain latent structural patterns.\nFurthermore, we develop a Rank-based Latent Causal Discovery algorithm, RLCD,\nthat can efficiently locate hidden variables, determine their cardinalities,\nand discover the entire causal structure over both measured and hidden ones. We\nalso show that, under certain graphical conditions, RLCD correctly identifies\nthe Markov Equivalence Class of the whole latent causal graph asymptotically.\nExperimental results on both synthetic and real-world personality data sets\ndemonstrate the efficacy of the proposed approach in finite-sample cases.\n", "date": "2023-12-18", "categories": []}
{"id": "2312.11158", "title": "Interventionally Consistent Surrogates for Agent-based Simulators", "abstract": "  Agent-based simulators provide granular representations of complex\nintelligent systems by directly modelling the interactions of the system's\nconstituent agents. Their high-fidelity nature enables hyper-local policy\nevaluation and testing of what-if scenarios, but is associated with large\ncomputational costs that inhibits their widespread use. Surrogate models can\naddress these computational limitations, but they must behave consistently with\nthe agent-based model under policy interventions of interest. In this paper, we\ncapitalise on recent developments on causal abstractions to develop a framework\nfor learning interventionally consistent surrogate models for agent-based\nsimulators. Our proposed approach facilitates rapid experimentation with policy\ninterventions in complex systems, while inducing surrogates to behave\nconsistently with high probability with respect to the agent-based simulator\nacross interventions of interest. We demonstrate with empirical studies that\nobservationally trained surrogates can misjudge the effect of interventions and\nmisguide policymakers towards suboptimal policies, while surrogates trained for\ninterventional consistency with our proposed method closely mimic the behaviour\nof an agent-based model under interventions of interest.\n", "date": "2023-12-18", "categories": []}
{"id": "2308.16599", "title": "Using machine learning to understand causal relationships between urban\n  form and travel CO2 emissions across continents", "abstract": "  Climate change mitigation in urban mobility requires policies reconfiguring\nurban form to increase accessibility and facilitate low-carbon modes of\ntransport. However, current policy research has insufficiently assessed urban\nform effects on car travel at three levels: (1) Causality -- Can causality be\nestablished beyond theoretical and correlation-based analyses? (2)\nGeneralizability -- Do relationships hold across different cities and world\nregions? (3) Context specificity -- How do relationships vary across\nneighborhoods of a city? Here, we address all three gaps via causal graph\ndiscovery and explainable machine learning to detect urban form effects on\nintra-city car travel, based on mobility data of six cities across three\ncontinents. We find significant causal effects of urban form on trip emissions\nand inter-feature effects, which had been neglected in previous work. Our\nresults demonstrate that destination accessibility matters most overall, while\nlow density and low connectivity also sharply increase CO$_2$ emissions. These\ngeneral trends are similar across cities but we find idiosyncratic effects that\ncan lead to substantially different recommendations. In more monocentric\ncities, we identify spatial corridors -- about 10--50 km from the city center\n-- where subcenter-oriented development is more relevant than increased access\nto the main center. Our work demonstrates a novel application of machine\nlearning that enables new research addressing the needs of causality,\ngeneralizability, and contextual specificity for scaling evidence-based urban\nclimate solutions.\n", "date": "2023-08-31", "categories": []}
{"id": "2312.09639", "title": "Multiple Instance Learning for Uplift Modeling", "abstract": "  Uplift modeling is widely used in performance marketing to estimate effects\nof promotion campaigns (e.g., increase of customer retention rate). Since it is\nimpossible to observe outcomes of a recipient in treatment (e.g., receiving a\ncertain promotion) and control (e.g., without promotion) groups simultaneously\n(i.e., counter-factual), uplift models are mainly trained on instances of\ntreatment and control groups separately to form two models respectively, and\nuplifts are predicted by the difference of predictions from these two models\n(i.e., two-model method). When responses are noisy and the treatment effect is\nfractional, induced individual uplift predictions will be inaccurate, resulting\nin targeting undesirable customers. Though it is impossible to obtain the ideal\nground-truth individual uplifts, known as Individual Treatment Effects (ITEs),\nalternatively, an average uplift of a group of users, called Average Treatment\nEffect (ATE), can be observed from experimental deliveries. Upon this, similar\nto Multiple Instance Learning (MIL) in which each training sample is a bag of\ninstances, our framework sums up individual user uplift predictions for each\nbag of users as its bag-wise ATE prediction, and regularizes it to its ATE\nlabel, thus learning more accurate individual uplifts. Additionally, to amplify\nthe fractional treatment effect, bags are composed of instances with adjacent\nindividual uplift predictions, instead of random instances. Experiments\nconducted on two datasets show the effectiveness and universality of the\nproposed framework.\n", "date": "2023-12-15", "categories": []}
{"id": "2303.04416", "title": "Inference on Optimal Dynamic Policies via Softmax Approximation", "abstract": "  Estimating optimal dynamic policies from offline data is a fundamental\nproblem in dynamic decision making. In the context of causal inference, the\nproblem is known as estimating the optimal dynamic treatment regime. Even\nthough there exists a plethora of methods for estimation, constructing\nconfidence intervals for the value of the optimal regime and structural\nparameters associated with it is inherently harder, as it involves non-linear\nand non-differentiable functionals of unknown quantities that need to be\nestimated. Prior work resorted to sub-sample approaches that can deteriorate\nthe quality of the estimate. We show that a simple soft-max approximation to\nthe optimal treatment regime, for an appropriately fast growing temperature\nparameter, can achieve valid inference on the truly optimal regime. We\nillustrate our result for a two-period optimal dynamic regime, though our\napproach should directly extend to the finite horizon case. Our work combines\ntechniques from semi-parametric inference and $g$-estimation, together with an\nappropriate triangular array central limit theorem, as well as a novel analysis\nof the asymptotic influence and asymptotic bias of softmax approximations.\n", "date": "2023-03-08", "categories": []}
{"id": "2308.03669", "title": "Diffusion Model in Causal Inference with Unmeasured Confounders", "abstract": "  We study how to extend the use of the diffusion model to answer the causal\nquestion from the observational data under the existence of unmeasured\nconfounders. In Pearl's framework of using a Directed Acyclic Graph (DAG) to\ncapture the causal intervention, a Diffusion-based Causal Model (DCM) was\nproposed incorporating the diffusion model to answer the causal questions more\naccurately, assuming that all of the confounders are observed. However,\nunmeasured confounders in practice exist, which hinders DCM from being\napplicable. To alleviate this limitation of DCM, we propose an extended model\ncalled Backdoor Criterion based DCM (BDCM), whose idea is rooted in the\nBackdoor criterion to find the variables in DAG to be included in the decoding\nprocess of the diffusion model so that we can extend DCM to the case with\nunmeasured confounders. Synthetic data experiment demonstrates that our\nproposed model captures the counterfactual distribution more precisely than DCM\nunder the unmeasured confounders.\n", "date": "2023-08-07", "categories": []}
{"id": "2312.08107", "title": "Causal Optimal Transport of Abstractions", "abstract": "  Causal abstraction (CA) theory establishes formal criteria for relating\nmultiple structural causal models (SCMs) at different levels of granularity by\ndefining maps between them. These maps have significant relevance for\nreal-world challenges such as synthesizing causal evidence from multiple\nexperimental environments, learning causally consistent representations at\ndifferent resolutions, and linking interventions across multiple SCMs. In this\nwork, we propose COTA, the first method to learn abstraction maps from\nobservational and interventional data without assuming complete knowledge of\nthe underlying SCMs. In particular, we introduce a multi-marginal Optimal\nTransport (OT) formulation that enforces do-calculus causal constraints,\ntogether with a cost function that relies on interventional information. We\nextensively evaluate COTA on synthetic and real world problems, and showcase\nits advantages over non-causal, independent and aggregated COTA formulations.\nFinally, we demonstrate the efficiency of our method as a data augmentation\ntool by comparing it against the state-of-the-art CA learning framework, which\nassumes fully specified SCMs, on a real-world downstream task.\n", "date": "2023-12-13", "categories": []}
{"id": "2312.05549", "title": "Multi-granularity Causal Structure Learning", "abstract": "  Unveil, model, and comprehend the causal mechanisms underpinning natural\nphenomena stand as fundamental endeavors across myriad scientific disciplines.\nMeanwhile, new knowledge emerges when discovering causal relationships from\ndata. Existing causal learning algorithms predominantly focus on the isolated\neffects of variables, overlook the intricate interplay of multiple variables\nand their collective behavioral patterns. Furthermore, the ubiquity of\nhigh-dimensional data exacts a substantial temporal cost for causal algorithms.\nIn this paper, we develop a novel method called MgCSL (Multi-granularity Causal\nStructure Learning), which first leverages sparse auto-encoder to explore\ncoarse-graining strategies and causal abstractions from micro-variables to\nmacro-ones. MgCSL then takes multi-granularity variables as inputs to train\nmultilayer perceptrons and to delve the causality between variables. To enhance\nthe efficacy on high-dimensional data, MgCSL introduces a simplified acyclicity\nconstraint to adeptly search the directed acyclic graph among variables.\nExperimental results show that MgCSL outperforms competitive baselines, and\nfinds out explainable causal connections on fMRI datasets.\n", "date": "2023-12-09", "categories": []}
{"id": "2312.06820", "title": "Extracting Self-Consistent Causal Insights from Users Feedback with LLMs\n  and In-context Learning", "abstract": "  Microsoft Windows Feedback Hub is designed to receive customer feedback on a\nwide variety of subjects including critical topics such as power and battery.\nFeedback is one of the most effective ways to have a grasp of users' experience\nwith Windows and its ecosystem. However, the sheer volume of feedback received\nby Feedback Hub makes it immensely challenging to diagnose the actual cause of\nreported issues. To better understand and triage issues, we leverage Double\nMachine Learning (DML) to associate users' feedback with telemetry signals. One\nof the main challenges we face in the DML pipeline is the necessity of domain\nknowledge for model design (e.g., causal graph), which sometimes is either not\navailable or hard to obtain. In this work, we take advantage of reasoning\ncapabilities in Large Language Models (LLMs) to generate a prior model that\nwhich to some extent compensates for the lack of domain knowledge and could be\nused as a heuristic for measuring feedback informativeness. Our LLM-based\napproach is able to extract previously known issues, uncover new bugs, and\nidentify sequences of events that lead to a bug, while minimizing out-of-domain\noutputs.\n", "date": "2023-12-11", "categories": []}
{"id": "2312.07175", "title": "Instrumental Variable Estimation for Causal Inference in Longitudinal\n  Data with Time-Dependent Latent Confounders", "abstract": "  Causal inference from longitudinal observational data is a challenging\nproblem due to the difficulty in correctly identifying the time-dependent\nconfounders, especially in the presence of latent time-dependent confounders.\nInstrumental variable (IV) is a powerful tool for addressing the latent\nconfounders issue, but the traditional IV technique cannot deal with latent\ntime-dependent confounders in longitudinal studies. In this work, we propose a\nnovel Time-dependent Instrumental Factor Model (TIFM) for time-varying causal\neffect estimation from data with latent time-dependent confounders. At each\ntime-step, the proposed TIFM method employs the Recurrent Neural Network (RNN)\narchitecture to infer latent IV, and then uses the inferred latent IV factor\nfor addressing the confounding bias caused by the latent time-dependent\nconfounders. We provide a theoretical analysis for the proposed TIFM method\nregarding causal effect estimation in longitudinal data. Extensive evaluation\nwith synthetic datasets demonstrates the effectiveness of TIFM in addressing\ncausal effect estimation over time. We further apply TIFM to a climate dataset\nto showcase the potential of the proposed method in tackling real-world\nproblems.\n", "date": "2023-12-12", "categories": []}
{"id": "2211.16393", "title": "Bayesian Semiparametric Model for Sequential Treatment Decisions with\n  Informative Timing", "abstract": "  We develop a Bayesian semi-parametric model for the estimating the impact of\ndynamic treatment rules on survival among patients diagnosed with pediatric\nacute myeloid leukemia (AML). The data consist of a subset of patients enrolled\nin the phase III AAML1031 clinical trial in which patients move through a\nsequence of four treatment courses. At each course, they undergo treatment that\nmay or may not include anthracyclines (ACT). While ACT is known to be effective\nat treating AML, it is also cardiotoxic and can lead to early death for some\npatients. Our task is to estimate the potential survival probability under\nhypothetical dynamic ACT treatment strategies, but there are several\nimpediments. First, since ACT was not randomized in the trial, its effect on\nsurvival is confounded over time. Second, subjects initiate the next course\ndepending on when they recover from the previous course, making timing\npotentially informative of subsequent treatment and survival. Third, patients\nmay die or drop out before ever completing the full treatment sequence. We\ndevelop a generative Bayesian semi-parametric model based on Gamma Process\npriors to address these complexities. At each treatment course, the model\ncaptures subjects' transition to subsequent treatment or death in continuous\ntime under a given rule. A g-computation procedure is used to compute a\nposterior over potential survival probability that is adjusted for time-varying\nconfounding. Using this approach, we conduct posterior inference for the\nefficacy of hypothetical treatment rules that dynamically modify ACT based on\nevolving cardiac function.\n", "date": "2022-11-29", "categories": []}
{"id": "2306.02822", "title": "Discovering Dynamic Causal Space for DAG Structure Learning", "abstract": "  Discovering causal structure from purely observational data (i.e., causal\ndiscovery), aiming to identify causal relationships among variables, is a\nfundamental task in machine learning. The recent invention of differentiable\nscore-based DAG learners is a crucial enabler, which reframes the combinatorial\noptimization problem into a differentiable optimization with a DAG constraint\nover directed graph space. Despite their great success, these cutting-edge DAG\nlearners incorporate DAG-ness independent score functions to evaluate the\ndirected graph candidates, lacking in considering graph structure. As a result,\nmeasuring the data fitness alone regardless of DAG-ness inevitably leads to\ndiscovering suboptimal DAGs and model vulnerabilities. Towards this end, we\npropose a dynamic causal space for DAG structure learning, coined CASPER, that\nintegrates the graph structure into the score function as a new measure in the\ncausal space to faithfully reflect the causal distance between estimated and\nground truth DAG. CASPER revises the learning process as well as enhances the\nDAG structure learning via adaptive attention to DAG-ness. Grounded by\nempirical visualization, CASPER, as a space, satisfies a series of desired\nproperties, such as structure awareness and noise robustness. Extensive\nexperiments on both synthetic and real-world datasets clearly validate the\nsuperiority of our CASPER over the state-of-the-art causal discovery methods in\nterms of accuracy and robustness.\n", "date": "2023-06-05", "categories": []}
{"id": "2310.02854", "title": "Multi-Domain Causal Representation Learning via Weak Distributional\n  Invariances", "abstract": "  Causal representation learning has emerged as the center of action in causal\nmachine learning research. In particular, multi-domain datasets present a\nnatural opportunity for showcasing the advantages of causal representation\nlearning over standard unsupervised representation learning. While recent works\nhave taken crucial steps towards learning causal representations, they often\nlack applicability to multi-domain datasets due to over-simplifying assumptions\nabout the data; e.g. each domain comes from a different single-node perfect\nintervention. In this work, we relax these assumptions and capitalize on the\nfollowing observation: there often exists a subset of latents whose certain\ndistributional properties (e.g., support, variance) remain stable across\ndomains; this property holds when, for example, each domain comes from a\nmulti-node imperfect intervention. Leveraging this observation, we show that\nautoencoders that incorporate such invariances can provably identify the stable\nset of latents from the rest across different settings.\n", "date": "2023-10-04", "categories": []}
{"id": "2312.05404", "title": "Disentangled Latent Representation Learning for Tackling the Confounding\n  M-Bias Problem in Causal Inference", "abstract": "  In causal inference, it is a fundamental task to estimate the causal effect\nfrom observational data. However, latent confounders pose major challenges in\ncausal inference in observational data, for example, confounding bias and\nM-bias. Recent data-driven causal effect estimators tackle the confounding bias\nproblem via balanced representation learning, but assume no M-bias in the\nsystem, thus they fail to handle the M-bias. In this paper, we identify a\nchallenging and unsolved problem caused by a variable that leads to confounding\nbias and M-bias simultaneously. To address this problem with co-occurring\nM-bias and confounding bias, we propose a novel Disentangled Latent\nRepresentation learning framework for learning latent representations from\nproxy variables for unbiased Causal effect Estimation (DLRCE) from\nobservational data. Specifically, DLRCE learns three sets of latent\nrepresentations from the measured proxy variables to adjust for the confounding\nbias and M-bias. Extensive experiments on both synthetic and three real-world\ndatasets demonstrate that DLRCE significantly outperforms the state-of-the-art\nestimators in the case of the presence of both confounding bias and M-bias.\n", "date": "2023-12-08", "categories": []}
{"id": "2312.05540", "title": "Federated Causality Learning with Explainable Adaptive Optimization", "abstract": "  Discovering the causality from observational data is a crucial task in\nvarious scientific domains. With increasing awareness of privacy, data are not\nallowed to be exposed, and it is very hard to learn causal graphs from\ndispersed data, since these data may have different distributions. In this\npaper, we propose a federated causal discovery strategy (FedCausal) to learn\nthe unified global causal graph from decentralized heterogeneous data. We\ndesign a global optimization formula to naturally aggregate the causal graphs\nfrom client data and constrain the acyclicity of the global graph without\nexposing local data. Unlike other federated causal learning algorithms,\nFedCausal unifies the local and global optimizations into a complete directed\nacyclic graph (DAG) learning process with a flexible optimization objective. We\nprove that this optimization objective has a high interpretability and can\nadaptively handle homogeneous and heterogeneous data. Experimental results on\nsynthetic and real datasets show that FedCausal can effectively deal with\nnon-independently and identically distributed (non-iid) data and has a superior\nperformance.\n", "date": "2023-12-09", "categories": []}
{"id": "2306.05415", "title": "Causal normalizing flows: from theory to practice", "abstract": "  In this work, we deepen on the use of normalizing flows for causal reasoning.\nSpecifically, we first leverage recent results on non-linear ICA to show that\ncausal models are identifiable from observational data given a causal ordering,\nand thus can be recovered using autoregressive normalizing flows (NFs). Second,\nwe analyze different design and learning choices for causal normalizing flows\nto capture the underlying causal data-generating process. Third, we describe\nhow to implement the do-operator in causal NFs, and thus, how to answer\ninterventional and counterfactual questions. Finally, in our experiments, we\nvalidate our design and training choices through a comprehensive ablation\nstudy; compare causal NFs to other approaches for approximating causal models;\nand empirically demonstrate that causal NFs can be used to address real-world\nproblems, where the presence of mixed discrete-continuous data and partial\nknowledge on the causal graph is the norm. The code for this work can be found\nat https://github.com/psanch21/causal-flows.\n", "date": "2023-06-08", "categories": []}
{"id": "2306.11072", "title": "Causal Effect Regularization: Automated Detection and Removal of\n  Spurious Attributes", "abstract": "  In many classification datasets, the task labels are spuriously correlated\nwith some input attributes. Classifiers trained on such datasets often rely on\nthese attributes for prediction, especially when the spurious correlation is\nhigh, and thus fail to generalize whenever there is a shift in the attributes'\ncorrelation at deployment. If we assume that the spurious attributes are known\na priori, several methods have been proposed to learn a classifier that is\ninvariant to the specified attributes. However, in real-world data, information\nabout spurious attributes is typically unavailable. Therefore, we propose a\nmethod to automatically identify spurious attributes by estimating their causal\neffect on the label and then use a regularization objective to mitigate the\nclassifier's reliance on them. Compared to a recent method for identifying\nspurious attributes, we find that our method is more accurate in removing the\nattribute from the learned model, especially when spurious correlation is high.\nSpecifically, across synthetic, semi-synthetic, and real-world datasets, our\nmethod shows significant improvement in a metric used to quantify the\ndependence of a classifier on spurious attributes ($\\Delta$Prob), while\nobtaining better or similar accuracy. In addition, our method mitigates the\nreliance on spurious attributes even under noisy estimation of causal effects.\nTo explain the empirical robustness of our method, we create a simple linear\nclassification task with two sets of attributes: causal and spurious. We prove\nthat our method only requires that the ranking of estimated causal effects is\ncorrect across attributes to select the correct classifier.\n", "date": "2023-06-19", "categories": []}
{"id": "2307.13917", "title": "BayesDAG: Gradient-Based Posterior Inference for Causal Discovery", "abstract": "  Bayesian causal discovery aims to infer the posterior distribution over\ncausal models from observed data, quantifying epistemic uncertainty and\nbenefiting downstream tasks. However, computational challenges arise due to\njoint inference over combinatorial space of Directed Acyclic Graphs (DAGs) and\nnonlinear functions. Despite recent progress towards efficient posterior\ninference over DAGs, existing methods are either limited to variational\ninference on node permutation matrices for linear causal models, leading to\ncompromised inference accuracy, or continuous relaxation of adjacency matrices\nconstrained by a DAG regularizer, which cannot ensure resulting graphs are\nDAGs. In this work, we introduce a scalable Bayesian causal discovery framework\nbased on a combination of stochastic gradient Markov Chain Monte Carlo\n(SG-MCMC) and Variational Inference (VI) that overcomes these limitations. Our\napproach directly samples DAGs from the posterior without requiring any DAG\nregularization, simultaneously draws function parameter samples and is\napplicable to both linear and nonlinear causal models. To enable our approach,\nwe derive a novel equivalence to the permutation-based DAG learning, which\nopens up possibilities of using any relaxed gradient estimator defined over\npermutations. To our knowledge, this is the first framework applying\ngradient-based MCMC sampling for causal discovery. Empirical evaluation on\nsynthetic and real-world datasets demonstrate our approach's effectiveness\ncompared to state-of-the-art baselines.\n", "date": "2023-07-26", "categories": []}
{"id": "2312.05234", "title": "The impact of heteroskedasticity on uplift modeling", "abstract": "  There are various applications, where companies need to decide to which\nindividuals they should best allocate treatment. To support such decisions,\nuplift models are applied to predict treatment effects on an individual level.\nBased on the predicted treatment effects, individuals can be ranked and\ntreatment allocation can be prioritized according to this ranking. An implicit\nassumption, which has not been doubted in the previous uplift modeling\nliterature, is that this treatment prioritization approach tends to bring\nindividuals with high treatment effects to the top and individuals with low\ntreatment effects to the bottom of the ranking. In our research, we show that\nheteroskedastictity in the training data can cause a bias of the uplift model\nranking: individuals with the highest treatment effects can get accumulated in\nlarge numbers at the bottom of the ranking. We explain theoretically how\nheteroskedasticity can bias the ranking of uplift models and show this process\nin a simulation and on real-world data. We argue that this problem of ranking\nbias due to heteroskedasticity might occur in many real-world applications and\nrequires modification of the treatment prioritization to achieve an efficient\ntreatment allocation.\n", "date": "2023-12-08", "categories": []}
{"id": "2311.09137", "title": "Causal prediction models for medication safety monitoring: The diagnosis\n  of vancomycin-induced acute kidney injury", "abstract": "  The current best practice approach for the retrospective diagnosis of adverse\ndrug events (ADEs) in hospitalized patients relies on a full patient chart\nreview and a formal causality assessment by multiple medical experts. This\nevaluation serves to qualitatively estimate the probability of causation (PC);\nthe probability that a drug was a necessary cause of an adverse event. This\npractice is manual, resource intensive and prone to human biases, and may thus\nbenefit from data-driven decision support. Here, we pioneer a causal modeling\napproach using observational data to estimate a lower bound of the PC\n(PC$_{low}$). This method includes two key causal inference components: (1) the\ntarget trial emulation framework and (2) estimation of individualized treatment\neffects using machine learning. We apply our method to the clinically relevant\nuse-case of vancomycin-induced acute kidney injury in intensive care patients,\nand compare our causal model-based PC$_{low}$ estimates to qualitative\nestimates of the PC provided by a medical expert. Important limitations and\npotential improvements are discussed, and we conclude that future improved\ncausal models could provide essential data-driven support for medication safety\nmonitoring in hospitalized patients.\n", "date": "2023-11-15", "categories": []}
{"id": "2312.03843", "title": "Exposing Disparities in Flood Adaptation for Equitable Future\n  Interventions", "abstract": "  As governments race to implement new climate adaptation policies that prepare\nfor more frequent flooding, they must seek policies that are effective for all\ncommunities and uphold climate justice. This requires evaluating policies not\nonly on their overall effectiveness but also on whether their benefits are felt\nacross all communities. We illustrate the importance of considering such\ndisparities for flood adaptation using the FEMA National Flood Insurance\nProgram Community Rating System and its dataset of $\\sim$2.5 million flood\ninsurance claims. We use ${\\rm C{\\scriptsize AUSAL}F{\\scriptsize LOW}}$, a\ncausal inference method based on deep generative models, to estimate the\ntreatment effect of flood adaptation interventions based on a community's\nincome, diversity, population, flood risk, educational attainment, and\nprecipitation. We find that the program saves communities \\$5,000--15,000 per\nhousehold. However, these savings are not evenly spread across communities. For\nexample, for low-income communities savings sharply decline as flood-risk\nincreases in contrast to their high-income counterparts with all else equal.\nEven among low-income communities, there is a gap in savings between\npredominantly white and non-white communities: savings of predominantly white\ncommunities can be higher by more than \\$6000 per household. As communities\nworldwide ramp up efforts to reduce losses inflicted by floods, simply\nprescribing a series flood adaptation measures is not enough. Programs must\nprovide communities with the necessary technical and economic support to\ncompensate for historical patterns of disenfranchisement, racism, and\ninequality. Future flood adaptation efforts should go beyond reducing losses\noverall and aim to close existing gaps to equitably support communities in the\nrace for climate adaptation.\n", "date": "2023-12-06", "categories": []}
{"id": "2312.04343", "title": "Causality and Explainability for Trustworthy Integrated Pest Management", "abstract": "  Pesticides serve as a common tool in agricultural pest control but\nsignificantly contribute to the climate crisis. To combat this, Integrated Pest\nManagement (IPM) stands as a climate-smart alternative. Despite its potential,\nIPM faces low adoption rates due to farmers' skepticism about its\neffectiveness. To address this challenge, we introduce an advanced data\nanalysis framework tailored to enhance IPM adoption. Our framework provides i)\nrobust pest population predictions across diverse environments with invariant\nand causal learning, ii) interpretable pest presence predictions using\ntransparent models, iii) actionable advice through counterfactual explanations\nfor in-season IPM interventions, iv) field-specific treatment effect\nestimations, and v) assessments of the effectiveness of our advice using causal\ninference. By incorporating these features, our framework aims to alleviate\nskepticism and encourage wider adoption of IPM practices among farmers.\n", "date": "2023-12-07", "categories": []}
{"id": "2312.02858", "title": "Towards Causal Representations of Climate Model Data", "abstract": "  Climate models, such as Earth system models (ESMs), are crucial for\nsimulating future climate change based on projected Shared Socioeconomic\nPathways (SSP) greenhouse gas emissions scenarios. While ESMs are sophisticated\nand invaluable, machine learning-based emulators trained on existing simulation\ndata can project additional climate scenarios much faster and are\ncomputationally efficient. However, they often lack generalizability and\ninterpretability. This work delves into the potential of causal representation\nlearning, specifically the \\emph{Causal Discovery with Single-parent Decoding}\n(CDSD) method, which could render climate model emulation efficient\n\\textit{and} interpretable. We evaluate CDSD on multiple climate datasets,\nfocusing on emissions, temperature, and precipitation. Our findings shed light\non the challenges, limitations, and promise of using CDSD as a stepping stone\ntowards more interpretable and robust climate model emulation.\n", "date": "2023-12-05", "categories": []}
{"id": "2312.03580", "title": "Invariance & Causal Representation Learning: Prospects and Limitations", "abstract": "  In causal models, a given mechanism is assumed to be invariant to changes of\nother mechanisms. While this principle has been utilized for inference in\nsettings where the causal variables are observed, theoretical insights when the\nvariables of interest are latent are largely missing. We assay the connection\nbetween invariance and causal representation learning by establishing\nimpossibility results which show that invariance alone is insufficient to\nidentify latent causal variables. Together with practical considerations, we\nuse these theoretical findings to highlight the need for additional constraints\nin order to identify representations by exploiting invariance.\n", "date": "2023-12-06", "categories": []}
{"id": "2312.02438", "title": "Adaptive Instrument Design for Indirect Experiments", "abstract": "  Indirect experiments provide a valuable framework for estimating treatment\neffects in situations where conducting randomized control trials (RCTs) is\nimpractical or unethical. Unlike RCTs, indirect experiments estimate treatment\neffects by leveraging (conditional) instrumental variables, enabling estimation\nthrough encouragement and recommendation rather than strict treatment\nassignment. However, the sample efficiency of such estimators depends not only\non the inherent variability in outcomes but also on the varying compliance\nlevels of users with the instrumental variables and the choice of estimator\nbeing used, especially when dealing with numerous instrumental variables. While\nadaptive experiment design has a rich literature for direct experiments, in\nthis paper we take the initial steps towards enhancing sample efficiency for\nindirect experiments by adaptively designing a data collection policy over\ninstrumental variables. Our main contribution is a practical computational\nprocedure that utilizes influence functions to search for an optimal data\ncollection policy, minimizing the mean-squared error of the desired\n(non-linear) estimator. Through experiments conducted in various domains\ninspired by real-world applications, we showcase how our method can\nsignificantly improve the sample efficiency of indirect experiments.\n", "date": "2023-12-05", "categories": []}
{"id": "2208.09590", "title": "Data-Driven Causal Effect Estimation Based on Graphical Causal\n  Modelling: A Survey", "abstract": "  In many fields of scientific research and real-world applications, unbiased\nestimation of causal effects from non-experimental data is crucial for\nunderstanding the mechanism underlying the data and for decision-making on\neffective responses or interventions. A great deal of research has been\nconducted to address this challenging problem from different angles. For\nestimating causal effect in observational data, assumptions such as Markov\ncondition, faithfulness and causal sufficiency are always made. Under the\nassumptions, full knowledge such as, a set of covariates or an underlying\ncausal graph, is typically required. A practical challenge is that in many\napplications, no such full knowledge or only some partial knowledge is\navailable. In recent years, research has emerged to use search strategies based\non graphical causal modelling to discover useful knowledge from data for causal\neffect estimation, with some mild assumptions, and has shown promise in\ntackling the practical challenge. In this survey, we review these data-driven\nmethods on causal effect estimation for a single treatment with a single\noutcome of interest and focus on the challenges faced by data-driven causal\neffect estimation. We concisely summarise the basic concepts and theories that\nare essential for data-driven causal effect estimation using graphical causal\nmodelling but are scattered around the literature. We identify and discuss the\nchallenges faced by data-driven causal effect estimation and characterise the\nexisting methods by their assumptions and the approaches to tackling the\nchallenges. We analyse the strengths and limitations of the different types of\nmethods and present an empirical evaluation to support the discussions. We hope\nthis review will motivate more researchers to design better data-driven methods\nbased on graphical causal modelling for the challenging problem of causal\neffect estimation.\n", "date": "2022-08-20", "categories": []}
{"id": "2306.09433", "title": "Towards Practical Federated Causal Structure Learning", "abstract": "  Understanding causal relations is vital in scientific discovery. The process\nof causal structure learning involves identifying causal graphs from\nobservational data to understand such relations. Usually, a central server\nperforms this task, but sharing data with the server poses privacy risks.\nFederated learning can solve this problem, but existing solutions for federated\ncausal structure learning make unrealistic assumptions about data and lack\nconvergence guarantees. FedC2SL is a federated constraint-based causal\nstructure learning scheme that learns causal graphs using a federated\nconditional independence test, which examines conditional independence between\ntwo variables under a condition set without collecting raw data from clients.\nFedC2SL requires weaker and more realistic assumptions about data and offers\nstronger resistance to data variability among clients. FedPC and FedFCI are the\ntwo variants of FedC2SL for causal structure learning in causal sufficiency and\ncausal insufficiency, respectively. The study evaluates FedC2SL using both\nsynthetic datasets and real-world data against existing solutions and finds it\ndemonstrates encouraging performance and strong resilience to data\nheterogeneity among clients.\n", "date": "2023-06-15", "categories": []}
{"id": "2307.08304", "title": "Efficient Computation of Counterfactual Bounds", "abstract": "  We assume to be given structural equations over discrete variables inducing a\ndirected acyclic graph, namely, a structural causal model, together with data\nabout its internal nodes. The question we want to answer is how we can compute\nbounds for partially identifiable counterfactual queries from such an input. We\nstart by giving a map from structural casual models to credal networks. This\nallows us to compute exact counterfactual bounds via algorithms for credal nets\non a subclass of structural causal models. Exact computation is going to be\ninefficient in general given that, as we show, causal inference is NP-hard even\non polytrees. We target then approximate bounds via a causal EM scheme. We\nevaluate their accuracy by providing credible intervals on the quality of the\napproximation; we show through a synthetic benchmark that the EM scheme\ndelivers accurate results in a fair number of runs. In the course of the\ndiscussion, we also point out what seems to be a neglected limitation to the\ntrending idea that counterfactual bounds can be computed without knowledge of\nthe structural equations. We also present a real case study on palliative care\nto show how our algorithms can readily be used for practical purposes.\n", "date": "2023-07-17", "categories": []}
{"id": "2312.00509", "title": "Bayesian causal discovery from unknown general interventions", "abstract": "  We consider the problem of learning causal Directed Acyclic Graphs (DAGs)\nusing combinations of observational and interventional experimental data.\nCurrent methods tailored to this setting assume that interventions either\ndestroy parent-child relations of the intervened (target) nodes or only alter\nsuch relations without modifying the parent sets, even when the intervention\ntargets are unknown. We relax this assumption by proposing a Bayesian method\nfor causal discovery from general interventions, which allow for modifications\nof the parent sets of the unknown targets. Even in this framework, DAGs and\ngeneral interventions may be identifiable only up to some equivalence classes.\nWe provide graphical characterizations of such interventional Markov\nequivalence and devise compatible priors for Bayesian inference that guarantee\nscore equivalence of indistinguishable structures. We then develop a Markov\nChain Monte Carlo (MCMC) scheme to approximate the posterior distribution over\nDAGs, intervention targets and induced parent sets. Finally, we evaluate the\nproposed methodology on both simulated and real protein expression data.\n", "date": "2023-12-01", "categories": []}
{"id": "2307.13868", "title": "Learning sources of variability from high-dimensional observational\n  studies", "abstract": "  Causal inference studies whether the presence of a variable influences an\nobserved outcome. As measured by quantities such as the \"average treatment\neffect,\" this paradigm is employed across numerous biological fields, from\nvaccine and drug development to policy interventions. Unfortunately, the\nmajority of these methods are often limited to univariate outcomes. Our work\ngeneralizes causal estimands to outcomes with any number of dimensions or any\nmeasurable space, and formulates traditional causal estimands for nominal\nvariables as causal discrepancy tests. We propose a simple technique for\nadjusting universally consistent conditional independence tests and prove that\nthese tests are universally consistent causal discrepancy tests. Numerical\nexperiments illustrate that our method, Causal CDcorr, leads to improvements in\nboth finite sample validity and power when compared to existing strategies. Our\nmethods are all open source and available at github.com/ebridge2/cdcorr.\n", "date": "2023-07-26", "categories": []}
{"id": "2206.10479", "title": "Policy Learning with Asymmetric Counterfactual Utilities", "abstract": "  Data-driven decision making plays an important role even in high stakes\nsettings like medicine and public policy. Learning optimal policies from\nobserved data requires a careful formulation of the utility function whose\nexpected value is maximized across a population. Although researchers typically\nuse utilities that depend on observed outcomes alone, in many settings the\ndecision maker's utility function is more properly characterized by the joint\nset of potential outcomes under all actions. For example, the Hippocratic\nprinciple to \"do no harm\" implies that the cost of causing death to a patient\nwho would otherwise survive without treatment is greater than the cost of\nforgoing life-saving treatment. We consider optimal policy learning with\nasymmetric counterfactual utility functions of this form that consider the\njoint set of potential outcomes. We show that asymmetric counterfactual\nutilities lead to an unidentifiable expected utility function, and so we first\npartially identify it. Drawing on statistical decision theory, we then derive\nminimax decision rules by minimizing the maximum expected utility loss relative\nto different alternative policies. We show that one can learn minimax loss\ndecision rules from observed data by solving intermediate classification\nproblems, and establish that the finite sample excess expected utility loss of\nthis procedure is bounded by the regret of these intermediate classifiers. We\napply this conceptual framework and methodology to the decision about whether\nor not to use right heart catheterization for patients with possible pulmonary\nhypertension.\n", "date": "2022-06-21", "categories": []}
{"id": "2311.16616", "title": "Adversarial Distribution Balancing for Counterfactual Reasoning", "abstract": "  The development of causal prediction models is challenged by the fact that\nthe outcome is only observable for the applied (factual) intervention and not\nfor its alternatives (the so-called counterfactuals); in medicine we only know\npatients' survival for the administered drug and not for other therapeutic\noptions. Machine learning approaches for counterfactual reasoning have to deal\nwith both unobserved outcomes and distributional differences due to non-random\ntreatment administration. Unsupervised domain adaptation (UDA) addresses\nsimilar issues; one has to deal with unobserved outcomes -- the labels of the\ntarget domain -- and distributional differences between source and target\ndomain. We propose Adversarial Distribution Balancing for Counterfactual\nReasoning (ADBCR), which directly uses potential outcome estimates of the\ncounterfactuals to remove spurious causal relations. We show that ADBCR\noutcompetes state-of-the-art methods on three benchmark datasets, and\ndemonstrate that ADBCR's performance can be further improved if unlabeled\nvalidation data are included in the training procedure to better adapt the\nmodel to the validation domain.\n", "date": "2023-11-28", "categories": []}
{"id": "2311.14994", "title": "Exploring Causal Learning through Graph Neural Networks: An In-depth\n  Review", "abstract": "  In machine learning, exploring data correlations to predict outcomes is a\nfundamental task. Recognizing causal relationships embedded within data is\npivotal for a comprehensive understanding of system dynamics, the significance\nof which is paramount in data-driven decision-making processes. Beyond\ntraditional methods, there has been a surge in the use of graph neural networks\n(GNNs) for causal learning, given their capabilities as universal data\napproximators. Thus, a thorough review of the advancements in causal learning\nusing GNNs is both relevant and timely. To structure this review, we introduce\na novel taxonomy that encompasses various state-of-the-art GNN methods employed\nin studying causality. GNNs are further categorized based on their applications\nin the causality domain. We further provide an exhaustive compilation of\ndatasets integral to causal learning with GNNs to serve as a resource for\npractical study. This review also touches upon the application of causal\nlearning across diverse sectors. We conclude the review with insights into\npotential challenges and promising avenues for future exploration in this\nrapidly evolving field of machine learning.\n", "date": "2023-11-25", "categories": []}
{"id": "2311.05286", "title": "Causal Inference from Text: Unveiling Interactions between Variables", "abstract": "  Adjusting for latent covariates is crucial for estimating causal effects from\nobservational textual data. Most existing methods only account for confounding\ncovariates that affect both treatment and outcome, potentially leading to\nbiased causal effects. This bias arises from insufficient consideration of\nnon-confounding covariates, which are relevant only to either the treatment or\nthe outcome. In this work, we aim to mitigate the bias by unveiling\ninteractions between different variables to disentangle the non-confounding\ncovariates when estimating causal effects from text. The disentangling process\nensures covariates only contribute to their respective objectives, enabling\nindependence between variables. Additionally, we impose a constraint to balance\nrepresentations from the treatment group and control group to alleviate\nselection bias. We conduct experiments on two different treatment factors under\nvarious scenarios, and the proposed model significantly outperforms recent\nstrong baselines. Furthermore, our thorough analysis on earnings call\ntranscripts demonstrates that our model can effectively disentangle the\nvariables, and further investigations into real-world scenarios provide\nguidance for investors to make informed decisions.\n", "date": "2023-11-09", "categories": []}
{"id": "2208.04589", "title": "Long-term Causal Effects Estimation via Latent Surrogates Representation\n  Learning", "abstract": "  Estimating long-term causal effects based on short-term surrogates is a\nsignificant but challenging problem in many real-world applications, e.g.,\nmarketing and medicine. Despite its success in certain domains, most existing\nmethods estimate causal effects in an idealistic and simplistic way - ignoring\nthe causal structure among short-term outcomes and treating all of them as\nsurrogates. However, such methods cannot be well applied to real-world\nscenarios, in which the partially observed surrogates are mixed with their\nproxies among short-term outcomes. To this end, we develop our flexible method,\nLaser, to estimate long-term causal effects in the more realistic situation\nthat the surrogates are observed or have observed proxies.Given the\nindistinguishability between the surrogates and proxies, we utilize\nidentifiable variational auto-encoder (iVAE) to recover the whole valid\nsurrogates on all the surrogates candidates without the need of distinguishing\nthe observed surrogates or the proxies of latent surrogates. With the help of\nthe recovered surrogates, we further devise an unbiased estimation of long-term\ncausal effects. Extensive experimental results on the real-world and\nsemi-synthetic datasets demonstrate the effectiveness of our proposed method.\n", "date": "2022-08-09", "categories": []}
{"id": "2305.18183", "title": "On Counterfactual Data Augmentation Under Confounding", "abstract": "  Counterfactual data augmentation has recently emerged as a method to mitigate\nconfounding biases in the training data. These biases, such as spurious\ncorrelations, arise due to various observed and unobserved confounding\nvariables in the data generation process. In this paper, we formally analyze\nhow confounding biases impact downstream classifiers and present a causal\nviewpoint to the solutions based on counterfactual data augmentation. We\nexplore how removing confounding biases serves as a means to learn invariant\nfeatures, ultimately aiding in generalization beyond the observed data\ndistribution. Additionally, we present a straightforward yet powerful algorithm\nfor generating counterfactual images, which effectively mitigates the influence\nof confounding effects on downstream classifiers. Through experiments on MNIST\nvariants and the CelebA datasets, we demonstrate how our simple augmentation\nmethod helps existing state-of-the-art methods achieve good results.\n", "date": "2023-05-29", "categories": []}
{"id": "2307.01452", "title": "Causal Reinforcement Learning: A Survey", "abstract": "  Reinforcement learning is an essential paradigm for solving sequential\ndecision problems under uncertainty. Despite many remarkable achievements in\nrecent decades, applying reinforcement learning methods in the real world\nremains challenging. One of the main obstacles is that reinforcement learning\nagents lack a fundamental understanding of the world and must therefore learn\nfrom scratch through numerous trial-and-error interactions. They may also face\nchallenges in providing explanations for their decisions and generalizing the\nacquired knowledge. Causality, however, offers a notable advantage as it can\nformalize knowledge in a systematic manner and leverage invariance for\neffective knowledge transfer. This has led to the emergence of causal\nreinforcement learning, a subfield of reinforcement learning that seeks to\nenhance existing algorithms by incorporating causal relationships into the\nlearning process. In this survey, we comprehensively review the literature on\ncausal reinforcement learning. We first introduce the basic concepts of\ncausality and reinforcement learning, and then explain how causality can\naddress core challenges in non-causal reinforcement learning. We categorize and\nsystematically review existing causal reinforcement learning approaches based\non their target problems and methodologies. Finally, we outline open issues and\nfuture directions in this emerging field.\n", "date": "2023-07-04", "categories": []}
{"id": "2311.12307", "title": "Causality is all you need", "abstract": "  In the fundamental statistics course, students are taught to remember the\nwell-known saying: \"Correlation is not Causation\". Till now, statistics (i.e.,\ncorrelation) have developed various successful frameworks, such as Transformer\nand Pre-training large-scale models, which have stacked multiple parallel\nself-attention blocks to imitate a wide range of tasks. However, in the\ncausation community, how to build an integrated causal framework still remains\nan untouched domain despite its excellent intervention capabilities. In this\npaper, we propose the Causal Graph Routing (CGR) framework, an integrated\ncausal scheme relying entirely on the intervention mechanisms to reveal the\ncause-effect forces hidden in data. Specifically, CGR is composed of a stack of\ncausal layers. Each layer includes a set of parallel deconfounding blocks from\ndifferent causal graphs. We combine these blocks via the concept of the\nproposed sufficient cause, which allows the model to dynamically select the\nsuitable deconfounding methods in each layer. CGR is implemented as the stacked\nnetworks, integrating no confounder, back-door adjustment, front-door\nadjustment, and probability of sufficient cause. We evaluate this framework on\ntwo classical tasks of CV and NLP. Experiments show CGR can surpass the current\nstate-of-the-art methods on both Visual Question Answer and Long Document\nClassification tasks. In particular, CGR has great potential in building the\n\"causal\" pre-training large-scale model that effectively generalizes to diverse\ntasks. It will improve the machines' comprehension of causal relationships\nwithin a broader semantic space.\n", "date": "2023-11-21", "categories": []}
{"id": "2208.14989", "title": "Learning Multiscale Non-stationary Causal Structures", "abstract": "  This paper addresses a gap in the current state of the art by providing a\nsolution for modeling causal relationships that evolve over time and occur at\ndifferent time scales. Specifically, we introduce the multiscale non-stationary\ndirected acyclic graph (MN-DAG), a framework for modeling multivariate time\nseries data. Our contribution is twofold. Firstly, we expose a probabilistic\ngenerative model by leveraging results from spectral and causality theories.\nOur model allows sampling an MN-DAG according to user-specified priors on the\ntime-dependence and multiscale properties of the causal graph. Secondly, we\ndevise a Bayesian method named Multiscale Non-stationary Causal Structure\nLearner (MN-CASTLE) that uses stochastic variational inference to estimate\nMN-DAGs. The method also exploits information from the local partial\ncorrelation between time series over different time resolutions. The data\ngenerated from an MN-DAG reproduces well-known features of time series in\ndifferent domains, such as volatility clustering and serial correlation.\nAdditionally, we show the superior performance of MN-CASTLE on synthetic data\nwith different multiscale and non-stationary properties compared to baseline\nmodels. Finally, we apply MN-CASTLE to identify the drivers of the natural gas\nprices in the US market. Causal relationships have strengthened during the\nCOVID-19 outbreak and the Russian invasion of Ukraine, a fact that baseline\nmethods fail to capture. MN-CASTLE identifies the causal impact of critical\neconomic drivers on natural gas prices, such as seasonal factors, economic\nuncertainty, oil prices, and gas storage deviations.\n", "date": "2022-08-31", "categories": []}
{"id": "2311.08427", "title": "Towards a Transportable Causal Network Model Based on Observational\n  Healthcare Data", "abstract": "  Over the last decades, many prognostic models based on artificial\nintelligence techniques have been used to provide detailed predictions in\nhealthcare. Unfortunately, the real-world observational data used to train and\nvalidate these models are almost always affected by biases that can strongly\nimpact the outcomes validity: two examples are values missing not-at-random and\nselection bias. Addressing them is a key element in achieving transportability\nand in studying the causal relationships that are critical in clinical decision\nmaking, going beyond simpler statistical approaches based on probabilistic\nassociation.\n  In this context, we propose a novel approach that combines selection\ndiagrams, missingness graphs, causal discovery and prior knowledge into a\nsingle graphical model to estimate the cardiovascular risk of adolescent and\nyoung females who survived breast cancer. We learn this model from data\ncomprising two different cohorts of patients. The resulting causal network\nmodel is validated by expert clinicians in terms of risk assessment, accuracy\nand explainability, and provides a prognostic model that outperforms competing\nmachine learning methods.\n", "date": "2023-11-13", "categories": []}
{"id": "2311.11212", "title": "Can We Utilize Pre-trained Language Models within Causal Discovery\n  Algorithms?", "abstract": "  Scaling laws have allowed Pre-trained Language Models (PLMs) into the field\nof causal reasoning. Causal reasoning of PLM relies solely on text-based\ndescriptions, in contrast to causal discovery which aims to determine the\ncausal relationships between variables utilizing data. Recently, there has been\ncurrent research regarding a method that mimics causal discovery by aggregating\nthe outcomes of repetitive causal reasoning, achieved through specifically\ndesigned prompts. It highlights the usefulness of PLMs in discovering cause and\neffect, which is often limited by a lack of data, especially when dealing with\nmultiple variables. Conversely, the characteristics of PLMs which are that PLMs\ndo not analyze data and they are highly dependent on prompt design leads to a\ncrucial limitation for directly using PLMs in causal discovery. Accordingly,\nPLM-based causal reasoning deeply depends on the prompt design and carries out\nthe risk of overconfidence and false predictions in determining causal\nrelationships. In this paper, we empirically demonstrate the aforementioned\nlimitations of PLM-based causal reasoning through experiments on\nphysics-inspired synthetic data. Then, we propose a new framework that\nintegrates prior knowledge obtained from PLM with a causal discovery algorithm.\nThis is accomplished by initializing an adjacency matrix for causal discovery\nand incorporating regularization using prior knowledge. Our proposed framework\nnot only demonstrates improved performance through the integration of PLM and\ncausal discovery but also suggests how to leverage PLM-extracted prior\nknowledge with existing causal discovery algorithms.\n", "date": "2023-11-19", "categories": []}
{"id": "2311.11689", "title": "Causal Structure Learning Supervised by Large Language Model", "abstract": "  Causal discovery from observational data is pivotal for deciphering complex\nrelationships. Causal Structure Learning (CSL), which focuses on deriving\ncausal Directed Acyclic Graphs (DAGs) from data, faces challenges due to vast\nDAG spaces and data sparsity. The integration of Large Language Models (LLMs),\nrecognized for their causal reasoning capabilities, offers a promising\ndirection to enhance CSL by infusing it with knowledge-based causal inferences.\nHowever, existing approaches utilizing LLMs for CSL have encountered issues,\nincluding unreliable constraints from imperfect LLM inferences and the\ncomputational intensity of full pairwise variable analyses. In response, we\nintroduce the Iterative LLM Supervised CSL (ILS-CSL) framework. ILS-CSL\ninnovatively integrates LLM-based causal inference with CSL in an iterative\nprocess, refining the causal DAG using feedback from LLMs. This method not only\nutilizes LLM resources more efficiently but also generates more robust and\nhigh-quality structural constraints compared to previous methodologies. Our\ncomprehensive evaluation across eight real-world datasets demonstrates\nILS-CSL's superior performance, setting a new standard in CSL efficacy and\nshowcasing its potential to significantly advance the field of causal\ndiscovery. The codes are available at\n\\url{https://github.com/tyMadara/ILS-CSL}.\n", "date": "2023-11-20", "categories": []}
{"id": "2311.10512", "title": "Causal Fairness-Guided Dataset Reweighting using Neural Networks", "abstract": "  The importance of achieving fairness in machine learning models cannot be\noverstated. Recent research has pointed out that fairness should be examined\nfrom a causal perspective, and several fairness notions based on the on Pearl's\ncausal framework have been proposed. In this paper, we construct a reweighting\nscheme of datasets to address causal fairness. Our approach aims at mitigating\nbias by considering the causal relationships among variables and incorporating\nthem into the reweighting process. The proposed method adopts two neural\nnetworks, whose structures are intentionally used to reflect the structures of\na causal graph and of an interventional graph. The two neural networks can\napproximate the causal model of the data, and the causal model of\ninterventions. Furthermore, reweighting guided by a discriminator is applied to\nachieve various fairness notions. Experiments on real-world datasets show that\nour method can achieve causal fairness on the data while remaining close to the\noriginal data for downstream tasks.\n", "date": "2023-11-17", "categories": []}
{"id": "2311.08434", "title": "Uplift Modeling based on Graph Neural Network Combined with Causal\n  Knowledge", "abstract": "  Uplift modeling is a fundamental component of marketing effect modeling,\nwhich is commonly employed to evaluate the effects of treatments on outcomes.\nThrough uplift modeling, we can identify the treatment with the greatest\nbenefit. On the other side, we can identify clients who are likely to make\nfavorable decisions in response to a certain treatment. In the past, uplift\nmodeling approaches relied heavily on the difference-in-difference (DID)\narchitecture, paired with a machine learning model as the estimation learner,\nwhile neglecting the link and confidential information between features. We\nproposed a framework based on graph neural networks that combine causal\nknowledge with an estimate of uplift value. Firstly, we presented a causal\nrepresentation technique based on CATE (conditional average treatment effect)\nestimation and adjacency matrix structure learning. Secondly, we suggested a\nmore scalable uplift modeling framework based on graph convolution networks for\ncombining causal knowledge. Our findings demonstrate that this method works\neffectively for predicting uplift values, with small errors in typical\nsimulated data, and its effectiveness has been verified in actual industry\nmarketing data.\n", "date": "2023-11-14", "categories": []}
{"id": "2305.18427", "title": "Interpretable Reward Redistribution in Reinforcement Learning: A Causal\n  Approach", "abstract": "  A major challenge in reinforcement learning is to determine which\nstate-action pairs are responsible for future rewards that are delayed. Reward\nredistribution serves as a solution to re-assign credits for each time step\nfrom observed sequences. While the majority of current approaches construct the\nreward redistribution in an uninterpretable manner, we propose to explicitly\nmodel the contributions of state and action from a causal perspective,\nresulting in an interpretable reward redistribution and preserving policy\ninvariance. In this paper, we start by studying the role of causal generative\nmodels in reward redistribution by characterizing the generation of Markovian\nrewards and trajectory-wise long-term return and further propose a framework,\ncalled Generative Return Decomposition (GRD), for policy optimization in\ndelayed reward scenarios. Specifically, GRD first identifies the unobservable\nMarkovian rewards and causal relations in the generative process. Then, GRD\nmakes use of the identified causal generative model to form a compact\nrepresentation to train policy over the most favorable subspace of the state\nspace of the agent. Theoretically, we show that the unobservable Markovian\nreward function is identifiable, as well as the underlying causal structure and\ncausal models. Experimental results show that our method outperforms\nstate-of-the-art methods and the provided visualization further demonstrates\nthe interpretability of our method. The project page is located at\nhttps://reedzyd.github.io/GenerativeReturnDecomposition/.\n", "date": "2023-05-28", "categories": []}
{"id": "2311.06567", "title": "SCADI: Self-supervised Causal Disentanglement in Latent Variable Models", "abstract": "  Causal disentanglement has great potential for capturing complex situations.\nHowever, there is a lack of practical and efficient approaches. It is already\nknown that most unsupervised disentangling methods are unable to produce\nidentifiable results without additional information, often leading to randomly\ndisentangled output. Therefore, most existing models for disentangling are\nweakly supervised, providing information about intrinsic factors, which incurs\nexcessive costs. Therefore, we propose a novel model, SCADI(SElf-supervised\nCAusal DIsentanglement), that enables the model to discover semantic factors\nand learn their causal relationships without any supervision. This model\ncombines a masked structural causal model (SCM) with a pseudo-label generator\nfor causal disentanglement, aiming to provide a new direction for\nself-supervised causal disentanglement models.\n", "date": "2023-11-11", "categories": []}
{"id": "2311.07191", "title": "Applying Large Language Models for Causal Structure Learning in Non\n  Small Cell Lung Cancer", "abstract": "  Causal discovery is becoming a key part in medical AI research. These methods\ncan enhance healthcare by identifying causal links between biomarkers,\ndemographics, treatments and outcomes. They can aid medical professionals in\nchoosing more impactful treatments and strategies. In parallel, Large Language\nModels (LLMs) have shown great potential in identifying patterns and generating\ninsights from text data. In this paper we investigate applying LLMs to the\nproblem of determining the directionality of edges in causal discovery.\nSpecifically, we test our approach on a deidentified set of Non Small Cell Lung\nCancer(NSCLC) patients that have both electronic health record and genomic\npanel data. Graphs are validated using Bayesian Dirichlet estimators using\ntabular data. Our result shows that LLMs can accurately predict the\ndirectionality of edges in causal graphs, outperforming existing\nstate-of-the-art methods. These findings suggests that LLMs can play a\nsignificant role in advancing causal discovery and help us better understand\ncomplex systems.\n", "date": "2023-11-13", "categories": []}
{"id": "2310.00233", "title": "CausalImages: An R Package for Causal Inference with Earth Observation,\n  Bio-medical, and Social Science Images", "abstract": "  The causalimages R package enables causal inference with image and image\nsequence data, providing new tools for integrating novel data sources like\nsatellite and bio-medical imagery into the study of cause and effect. One set\nof functions enables image-based causal inference analyses. For example, one\nkey function decomposes treatment effect heterogeneity by images using an\ninterpretable Bayesian framework. This allows for determining which types of\nimages or image sequences are most responsive to interventions. A second\nmodeling function allows researchers to control for confounding using images.\nThe package also allows investigators to produce embeddings that serve as\nvector summaries of the image or video content. Finally, infrastructural\nfunctions are also provided, such as tools for writing large-scale image and\nimage sequence data as sequentialized byte strings for more rapid image\nanalysis. causalimages therefore opens new capabilities for causal inference in\nR, letting researchers use informative imagery in substantive analyses in a\nfast and accessible manner.\n", "date": "2023-09-30", "categories": []}
{"id": "2306.04027", "title": "Intervention Generalization: A View from Factor Graph Models", "abstract": "  One of the goals of causal inference is to generalize from past experiments\nand observational data to novel conditions. While it is in principle possible\nto eventually learn a mapping from a novel experimental condition to an outcome\nof interest, provided a sufficient variety of experiments is available in the\ntraining data, coping with a large combinatorial space of possible\ninterventions is hard. Under a typical sparse experimental design, this mapping\nis ill-posed without relying on heavy regularization or prior distributions.\nSuch assumptions may or may not be reliable, and can be hard to defend or test.\nIn this paper, we take a close look at how to warrant a leap from past\nexperiments to novel conditions based on minimal assumptions about the\nfactorization of the distribution of the manipulated system, communicated in\nthe well-understood language of factor graph models. A postulated\n$\\textit{interventional factor model}$ (IFM) may not always be informative, but\nit conveniently abstracts away a need for explicitly modeling unmeasured\nconfounding and feedback mechanisms, leading to directly testable claims. Given\nan IFM and datasets from a collection of experimental regimes, we derive\nconditions for identifiability of the expected outcomes of new regimes never\nobserved in these training data. We implement our framework using several\nefficient algorithms, and apply them on a range of semi-synthetic experiments.\n", "date": "2023-06-06", "categories": []}
{"id": "2307.06250", "title": "Identifiability Guarantees for Causal Disentanglement from Soft\n  Interventions", "abstract": "  Causal disentanglement aims to uncover a representation of data using latent\nvariables that are interrelated through a causal model. Such a representation\nis identifiable if the latent model that explains the data is unique. In this\npaper, we focus on the scenario where unpaired observational and interventional\ndata are available, with each intervention changing the mechanism of a latent\nvariable. When the causal variables are fully observed, statistically\nconsistent algorithms have been developed to identify the causal model under\nfaithfulness assumptions. We here show that identifiability can still be\nachieved with unobserved causal variables, given a generalized notion of\nfaithfulness. Our results guarantee that we can recover the latent causal model\nup to an equivalence class and predict the effect of unseen combinations of\ninterventions, in the limit of infinite data. We implement our causal\ndisentanglement framework by developing an autoencoding variational Bayes\nalgorithm and apply it to the problem of predicting combinatorial perturbation\neffects in genomics.\n", "date": "2023-07-12", "categories": []}
{"id": "2309.04069", "title": "Inferring physical laws by artificial intelligence based causal models", "abstract": "  The advances in Artificial Intelligence (AI) and Machine Learning (ML) have\nopened up many avenues for scientific research, and are adding new dimensions\nto the process of knowledge creation. However, even the most powerful and\nversatile of ML applications till date are primarily in the domain of analysis\nof associations and boil down to complex data fitting. Judea Pearl has pointed\nout that Artificial General Intelligence must involve interventions involving\nthe acts of doing and imagining. Any machine assisted scientific discovery thus\nmust include casual analysis and interventions. In this context, we propose a\ncausal learning model of physical principles, which not only recognizes\ncorrelations but also brings out casual relationships. We use the principles of\ncausal inference and interventions to study the cause-and-effect relationships\nin the context of some well-known physical phenomena. We show that this\ntechnique can not only figure out associations among data, but is also able to\ncorrectly ascertain the cause-and-effect relations amongst the variables,\nthereby strengthening (or weakening) our confidence in the proposed model of\nthe underlying physical process.\n", "date": "2023-09-08", "categories": []}
{"id": "2311.05420", "title": "Counterfactually Fair Representation", "abstract": "  The use of machine learning models in high-stake applications (e.g.,\nhealthcare, lending, college admission) has raised growing concerns due to\npotential biases against protected social groups. Various fairness notions and\nmethods have been proposed to mitigate such biases. In this work, we focus on\nCounterfactual Fairness (CF), a fairness notion that is dependent on an\nunderlying causal graph and first proposed by Kusner \\textit{et\nal.}~\\cite{kusner2017counterfactual}; it requires that the outcome an\nindividual perceives is the same in the real world as it would be in a\n\"counterfactual\" world, in which the individual belongs to another social\ngroup. Learning fair models satisfying CF can be challenging. It was shown in\n\\cite{kusner2017counterfactual} that a sufficient condition for satisfying CF\nis to \\textbf{not} use features that are descendants of sensitive attributes in\nthe causal graph. This implies a simple method that learns CF models only using\nnon-descendants of sensitive attributes while eliminating all descendants.\nAlthough several subsequent works proposed methods that use all features for\ntraining CF models, there is no theoretical guarantee that they can satisfy CF.\nIn contrast, this work proposes a new algorithm that trains models using all\nthe available features. We theoretically and empirically show that models\ntrained with this method can satisfy CF\\footnote{The code repository for this\nwork can be found in\n\\url{https://github.com/osu-srml/CF_Representation_Learning}}.\n", "date": "2023-11-09", "categories": []}
{"id": "2311.05421", "title": "Diffusion Based Causal Representation Learning", "abstract": "  Causal reasoning can be considered a cornerstone of intelligent systems.\nHaving access to an underlying causal graph comes with the promise of\ncause-effect estimation and the identification of efficient and safe\ninterventions. However, learning causal representations remains a major\nchallenge, due to the complexity of many real-world systems. Previous works on\ncausal representation learning have mostly focused on Variational Auto-Encoders\n(VAE). These methods only provide representations from a point estimate, and\nthey are unsuitable to handle high dimensions. To overcome these problems, we\nproposed a new Diffusion-based Causal Representation Learning (DCRL) algorithm.\nThis algorithm uses diffusion-based representations for causal discovery. DCRL\noffers access to infinite dimensional latent codes, which encode different\nlevels of information in the latent code. In a first proof of principle, we\ninvestigate the use of DCRL for causal representation learning. We further\ndemonstrate experimentally that this approach performs comparably well in\nidentifying the causal structure and causal variables.\n", "date": "2023-11-09", "categories": []}
{"id": "2304.10577", "title": "B-Learner: Quasi-Oracle Bounds on Heterogeneous Causal Effects Under\n  Hidden Confounding", "abstract": "  Estimating heterogeneous treatment effects from observational data is a\ncrucial task across many fields, helping policy and decision-makers take better\nactions. There has been recent progress on robust and efficient methods for\nestimating the conditional average treatment effect (CATE) function, but these\nmethods often do not take into account the risk of hidden confounding, which\ncould arbitrarily and unknowingly bias any causal estimate based on\nobservational data. We propose a meta-learner called the B-Learner, which can\nefficiently learn sharp bounds on the CATE function under limits on the level\nof hidden confounding. We derive the B-Learner by adapting recent results for\nsharp and valid bounds of the average treatment effect (Dorn et al., 2021) into\nthe framework given by Kallus & Oprescu (2023) for robust and model-agnostic\nlearning of conditional distributional treatment effects. The B-Learner can use\nany function estimator such as random forests and deep neural networks, and we\nprove its estimates are valid, sharp, efficient, and have a quasi-oracle\nproperty with respect to the constituent estimators under more general\nconditions than existing methods. Semi-synthetic experimental comparisons\nvalidate the theoretical findings, and we use real-world data to demonstrate\nhow the method might be used in practice.\n", "date": "2023-04-20", "categories": []}
{"id": "2310.18471", "title": "Causal disentanglement of multimodal data", "abstract": "  Causal representation learning algorithms discover lower-dimensional\nrepresentations of data that admit a decipherable interpretation of cause and\neffect; as achieving such interpretable representations is challenging, many\ncausal learning algorithms utilize elements indicating prior information, such\nas (linear) structural causal models, interventional data, or weak supervision.\nUnfortunately, in exploratory causal representation learning, such elements and\nprior information may not be available or warranted. Alternatively, scientific\ndatasets often have multiple modalities or physics-based constraints, and the\nuse of such scientific, multimodal data has been shown to improve\ndisentanglement in fully unsupervised settings. Consequently, we introduce a\ncausal representation learning algorithm (causalPIMA) that can use multimodal\ndata and known physics to discover important features with causal\nrelationships. Our innovative algorithm utilizes a new differentiable\nparametrization to learn a directed acyclic graph (DAG) together with a latent\nspace of a variational autoencoder in an end-to-end differentiable framework\nvia a single, tractable evidence lower bound loss function. We place a Gaussian\nmixture prior on the latent space and identify each of the mixtures with an\noutcome of the DAG nodes; this novel identification enables feature discovery\nwith causal relationships. Tested against a synthetic and a scientific dataset,\nour results demonstrate the capability of learning an interpretable causal\nstructure while simultaneously discovering key features in a fully unsupervised\nsetting.\n", "date": "2023-10-27", "categories": []}
{"id": "2311.03989", "title": "Learned Causal Method Prediction", "abstract": "  For a given causal question, it is important to efficiently decide which\ncausal inference method to use for a given dataset. This is challenging because\ncausal methods typically rely on complex and difficult-to-verify assumptions,\nand cross-validation is not applicable since ground truth causal quantities are\nunobserved. In this work, we propose CAusal Method Predictor (CAMP), a\nframework for predicting the best method for a given dataset. To this end, we\ngenerate datasets from a diverse set of synthetic causal models, score the\ncandidate methods, and train a model to directly predict the highest-scoring\nmethod for that dataset. Next, by formulating a self-supervised pre-training\nobjective centered on dataset assumptions relevant for causal inference, we\nsignificantly reduce the need for costly labeled data and enhance training\nefficiency. Our strategy learns to map implicit dataset properties to the best\nmethod in a data-driven manner. In our experiments, we focus on method\nprediction for causal discovery. CAMP outperforms selecting any individual\ncandidate method and demonstrates promising generalization to unseen\nsemi-synthetic and real-world benchmarks.\n", "date": "2023-11-07", "categories": []}
{"id": "2311.04636", "title": "Learning Linear Gaussian Polytree Models with Interventions", "abstract": "  We present a consistent and highly scalable local approach to learn the\ncausal structure of a linear Gaussian polytree using data from interventional\nexperiments with known intervention targets. Our methods first learn the\nskeleton of the polytree and then orient its edges. The output is a CPDAG\nrepresenting the interventional equivalence class of the polytree of the true\nunderlying distribution. The skeleton and orientation recovery procedures we\nuse rely on second order statistics and low-dimensional marginal distributions.\nWe assess the performance of our methods under different scenarios in synthetic\ndata sets and apply our algorithm to learn a polytree in a gene expression\ninterventional data set. Our simulation studies demonstrate that our approach\nis fast, has good accuracy in terms of structural Hamming distance, and handles\nproblems with thousands of nodes.\n", "date": "2023-11-08", "categories": []}
{"id": "2212.04922", "title": "Doubly Robust Kernel Statistics for Testing Distributional Treatment\n  Effects", "abstract": "  With the widespread application of causal inference, it is increasingly\nimportant to have tools which can test for the presence of causal effects in a\ndiverse array of circumstances. In this vein we focus on the problem of testing\nfor \\emph{distributional} causal effects, where the treatment affects not just\nthe mean, but also higher order moments of the distribution, as well as\nmultidimensional or structured outcomes. We build upon a previously introduced\nframework, Counterfactual Mean Embeddings, for representing causal\ndistributions within Reproducing Kernel Hilbert Spaces (RKHS) by proposing new,\nimproved, estimators for the distributional embeddings. These improved\nestimators are inspired by doubly robust estimators of the causal mean, using a\nsimilar form within the kernel space. We analyse these estimators, proving they\nretain the doubly robust property and have improved convergence rates compared\nto the original estimators. This leads to new permutation based tests for\ndistributional causal effects, using the estimators we propose as tests\nstatistics. We experimentally and theoretically demonstrate the validity of our\ntests.\n", "date": "2022-12-09", "categories": []}
{"id": "2306.03929", "title": "Finding Counterfactually Optimal Action Sequences in Continuous State\n  Spaces", "abstract": "  Whenever a clinician reflects on the efficacy of a sequence of treatment\ndecisions for a patient, they may try to identify critical time steps where,\nhad they made different decisions, the patient's health would have improved.\nWhile recent methods at the intersection of causal inference and reinforcement\nlearning promise to aid human experts, as the clinician above, to\nretrospectively analyze sequential decision making processes, they have focused\non environments with finitely many discrete states. However, in many practical\napplications, the state of the environment is inherently continuous in nature.\nIn this paper, we aim to fill this gap. We start by formally characterizing a\nsequence of discrete actions and continuous states using finite horizon Markov\ndecision processes and a broad class of bijective structural causal models.\nBuilding upon this characterization, we formalize the problem of finding\ncounterfactually optimal action sequences and show that, in general, we cannot\nexpect to solve it in polynomial time. Then, we develop a search method based\non the $A^*$ algorithm that, under a natural form of Lipschitz continuity of\nthe environment's dynamics, is guaranteed to return the optimal solution to the\nproblem. Experiments on real clinical data show that our method is very\nefficient in practice, and it has the potential to offer interesting insights\nfor sequential decision making tasks.\n", "date": "2023-06-06", "categories": []}
{"id": "2311.02221", "title": "Structured Neural Networks for Density Estimation and Causal Inference", "abstract": "  Injecting structure into neural networks enables learning functions that\nsatisfy invariances with respect to subsets of inputs. For instance, when\nlearning generative models using neural networks, it is advantageous to encode\nthe conditional independence structure of observed variables, often in the form\nof Bayesian networks. We propose the Structured Neural Network (StrNN), which\ninjects structure through masking pathways in a neural network. The masks are\ndesigned via a novel relationship we explore between neural network\narchitectures and binary matrix factorization, to ensure that the desired\nindependencies are respected. We devise and study practical algorithms for this\notherwise NP-hard design problem based on novel objectives that control the\nmodel architecture. We demonstrate the utility of StrNN in three applications:\n(1) binary and Gaussian density estimation with StrNN, (2) real-valued density\nestimation with Structured Autoregressive Flows (StrAFs) and Structured\nContinuous Normalizing Flows (StrCNF), and (3) interventional and\ncounterfactual analysis with StrAFs for causal inference. Our work opens up new\navenues for learning neural networks that enable data-efficient generative\nmodeling and the use of normalizing flows for causal effect estimation.\n", "date": "2023-11-03", "categories": []}
{"id": "2305.10050", "title": "The Impact of Missing Data on Causal Discovery: A Multicentric Clinical\n  Study", "abstract": "  Causal inference for testing clinical hypotheses from observational data\npresents many difficulties because the underlying data-generating model and the\nassociated causal graph are not usually available. Furthermore, observational\ndata may contain missing values, which impact the recovery of the causal graph\nby causal discovery algorithms: a crucial issue often ignored in clinical\nstudies. In this work, we use data from a multi-centric study on endometrial\ncancer to analyze the impact of different missingness mechanisms on the\nrecovered causal graph. This is achieved by extending state-of-the-art causal\ndiscovery algorithms to exploit expert knowledge without sacrificing\ntheoretical soundness. We validate the recovered graph with expert physicians,\nshowing that our approach finds clinically-relevant solutions. Finally, we\ndiscuss the goodness of fit of our graph and its consistency from a clinical\ndecision-making perspective using graphical separation to validate causal\npathways.\n", "date": "2023-05-17", "categories": []}
{"id": "2306.02899", "title": "Learning nonparametric latent causal graphs with unknown interventions", "abstract": "  We establish conditions under which latent causal graphs are\nnonparametrically identifiable and can be reconstructed from unknown\ninterventions in the latent space. Our primary focus is the identification of\nthe latent structure in measurement models without parametric assumptions such\nas linearity or Gaussianity. Moreover, we do not assume the number of hidden\nvariables is known, and we show that at most one unknown intervention per\nhidden variable is needed. This extends a recent line of work on learning\ncausal representations from observations and interventions. The proofs are\nconstructive and introduce two new graphical concepts -- imaginary subsets and\nisolated edges -- that may be useful in their own right. As a matter of\nindependent interest, the proofs also involve a novel characterization of the\nlimits of edge orientations within the equivalence class of DAGs induced by\nunknown interventions. These are the first results to characterize the\nconditions under which causal representations are identifiable without making\nany parametric assumptions in a general setting with unknown interventions and\nwithout faithfulness.\n", "date": "2023-06-05", "categories": []}
{"id": "2311.01902", "title": "High Precision Causal Model Evaluation with Conditional Randomization", "abstract": "  The gold standard for causal model evaluation involves comparing model\npredictions with true effects estimated from randomized controlled trials\n(RCT). However, RCTs are not always feasible or ethical to perform. In\ncontrast, conditionally randomized experiments based on inverse probability\nweighting (IPW) offer a more realistic approach but may suffer from high\nestimation variance. To tackle this challenge and enhance causal model\nevaluation in real-world conditional randomization settings, we introduce a\nnovel low-variance estimator for causal error, dubbed as the pairs estimator.\nBy applying the same IPW estimator to both the model and true experimental\neffects, our estimator effectively cancels out the variance due to IPW and\nachieves a smaller asymptotic variance. Empirical studies demonstrate the\nimproved of our estimator, highlighting its potential on achieving near-RCT\nperformance. Our method offers a simple yet powerful solution to evaluate\ncausal inference models in conditional randomization settings without\ncomplicated modification of the IPW estimator itself, paving the way for more\nrobust and reliable model assessments.\n", "date": "2023-11-03", "categories": []}
{"id": "2301.07210", "title": "Causal Falsification of Digital Twins", "abstract": "  Digital twins are virtual systems designed to predict how a real-world\nprocess will evolve in response to interventions. This modelling paradigm holds\nsubstantial promise in many applications, but rigorous procedures for assessing\ntheir accuracy are essential for safety-critical settings. We consider how to\nassess the accuracy of a digital twin using real-world data. We formulate this\nas causal inference problem, which leads to a precise definition of what it\nmeans for a twin to be \"correct\" appropriate for many applications.\nUnfortunately, fundamental results from causal inference mean observational\ndata cannot be used to certify that a twin is correct in this sense unless\npotentially tenuous assumptions are made, such as that the data are\nunconfounded. To avoid these assumptions, we propose instead to find situations\nin which the twin is not correct, and present a general-purpose statistical\nprocedure for doing so. Our approach yields reliable and actionable information\nabout the twin under only the assumption of an i.i.d. dataset of observational\ntrajectories, and remains sound even if the data are confounded. We apply our\nmethodology to a large-scale, real-world case study involving sepsis modelling\nwithin the Pulse Physiology Engine, which we assess using the MIMIC-III dataset\nof ICU patients.\n", "date": "2023-01-17", "categories": []}
{"id": "2311.00923", "title": "A Review and Roadmap of Deep Causal Model from Different Causal\n  Structures and Representations", "abstract": "  The fusion of causal models with deep learning introducing increasingly\nintricate data sets, such as the causal associations within images or between\ntextual components, has surfaced as a focal research area. Nonetheless, the\nbroadening of original causal concepts and theories to such complex,\nnon-statistical data has been met with serious challenges. In response, our\nstudy proposes redefinitions of causal data into three distinct categories from\nthe standpoint of causal structure and representation: definite data,\nsemi-definite data, and indefinite data. Definite data chiefly pertains to\nstatistical data used in conventional causal scenarios, while semi-definite\ndata refers to a spectrum of data formats germane to deep learning, including\ntime-series, images, text, and others. Indefinite data is an emergent research\nsphere inferred from the progression of data forms by us. To comprehensively\npresent these three data paradigms, we elaborate on their formal definitions,\ndifferences manifested in datasets, resolution pathways, and development of\nresearch. We summarize key tasks and achievements pertaining to definite and\nsemi-definite data from myriad research undertakings, present a roadmap for\nindefinite data, beginning with its current research conundrums. Lastly, we\nclassify and scrutinize the key datasets presently utilized within these three\nparadigms.\n", "date": "2023-11-02", "categories": []}
{"id": "2311.00927", "title": "Scalable Counterfactual Distribution Estimation in Multivariate Causal\n  Models", "abstract": "  We consider the problem of estimating the counterfactual joint distribution\nof multiple quantities of interests (e.g., outcomes) in a multivariate causal\nmodel extended from the classical difference-in-difference design. Existing\nmethods for this task either ignore the correlation structures among dimensions\nof the multivariate outcome by considering univariate causal models on each\ndimension separately and hence produce incorrect counterfactual distributions,\nor poorly scale even for moderate-size datasets when directly dealing with such\nmultivariate causal model. We propose a method that alleviates both issues\nsimultaneously by leveraging a robust latent one-dimensional subspace of the\noriginal high-dimension space and exploiting the efficient estimation from the\nunivariate causal model on such space. Since the construction of the\none-dimensional subspace uses information from all the dimensions, our method\ncan capture the correlation structures and produce good estimates of the\ncounterfactual distribution. We demonstrate the advantages of our approach over\nexisting methods on both synthetic and real-world data.\n", "date": "2023-11-02", "categories": []}
{"id": "2301.12389", "title": "On Learning Necessary and Sufficient Causal Graphs", "abstract": "  The causal revolution has stimulated interest in understanding complex\nrelationships in various fields. Most of the existing methods aim to discover\ncausal relationships among all variables within a complex large-scale graph.\nHowever, in practice, only a small subset of variables in the graph are\nrelevant to the outcomes of interest. Consequently, causal estimation with the\nfull causal graph -- particularly given limited data -- could lead to numerous\nfalsely discovered, spurious variables that exhibit high correlation with, but\nexert no causal impact on, the target outcome. In this paper, we propose\nlearning a class of necessary and sufficient causal graphs (NSCG) that\nexclusively comprises causally relevant variables for an outcome of interest,\nwhich we term causal features. The key idea is to employ probabilities of\ncausation to systematically evaluate the importance of features in the causal\ngraph, allowing us to identify a subgraph relevant to the outcome of interest.\nTo learn NSCG from data, we develop a necessary and sufficient causal\nstructural learning (NSCSL) algorithm, by establishing theoretical properties\nand relationships between probabilities of causation and natural causal effects\nof features. Across empirical studies of simulated and real data, we\ndemonstrate that NSCSL outperforms existing algorithms and can reveal crucial\nyeast genes for target heritable traits of interest.\n", "date": "2023-01-29", "categories": []}
{"id": "2303.18211", "title": "A Scale-Invariant Sorting Criterion to Find a Causal Order in Additive\n  Noise Models", "abstract": "  Additive Noise Models (ANMs) are a common model class for causal discovery\nfrom observational data and are often used to generate synthetic data for\ncausal discovery benchmarking. Specifying an ANM requires choosing all\nparameters, including those not fixed by explicit assumptions. Reisach et al.\n(2021) show that sorting variables by increasing variance often yields an\nordering close to a causal order and introduce var-sortability to quantify this\nalignment. Since increasing variances may be unrealistic and are\nscale-dependent, ANM data are often standardized in benchmarks.\n  We show that synthetic ANM data are characterized by another pattern that is\nscale-invariant: the explainable fraction of a variable's variance, as captured\nby the coefficient of determination $R^2$, tends to increase along the causal\norder. The result is high $R^2$-sortability, meaning that sorting the variables\nby increasing $R^2$ yields an ordering close to a causal order. We propose an\nefficient baseline algorithm termed $R^2$-SortnRegress that exploits high\n$R^2$-sortability and that can match and exceed the performance of established\ncausal discovery algorithms. We show analytically that sufficiently high edge\nweights lead to a relative decrease of the noise contributions along causal\nchains, resulting in increasingly deterministic relationships and high $R^2$.\nWe characterize $R^2$-sortability for different simulation parameters and find\nhigh values in common settings. Our findings reveal high $R^2$-sortability as\nan assumption about the data generating process relevant to causal discovery\nand implicit in many ANM sampling schemes. It should be made explicit, as its\nprevalence in real-world data is unknown. For causal discovery benchmarking, we\nimplement $R^2$-sortability, the $R^2$-SortnRegress algorithm, and ANM\nsimulation procedures in our library CausalDisco at\nhttps://causaldisco.github.io/CausalDisco/.\n", "date": "2023-03-31", "categories": []}
{"id": "2306.09520", "title": "Ensembled Prediction Intervals for Causal Outcomes Under Hidden\n  Confounding", "abstract": "  Causal inference of exact individual treatment outcomes in the presence of\nhidden confounders is rarely possible. Recent work has extended prediction\nintervals with finite-sample guarantees to partially identifiable causal\noutcomes, by means of a sensitivity model for hidden confounding. In deep\nlearning, predictors can exploit their inductive biases for better\ngeneralization out of sample. We argue that the structure inherent to a deep\nensemble should inform a tighter partial identification of the causal outcomes\nthat they predict. We therefore introduce an approach termed Caus-Modens, for\ncharacterizing causal outcome intervals by modulated ensembles. We present a\nsimple approach to partial identification using existing causal sensitivity\nmodels and show empirically that Caus-Modens gives tighter outcome intervals,\nas measured by the necessary interval size to achieve sufficient coverage. The\nlast of our three diverse benchmarks is a novel usage of GPT-4 for\nobservational experiments with unknown but probeable ground truth.\n", "date": "2023-06-15", "categories": []}
{"id": "2311.00577", "title": "Personalized Assignment to One of Many Treatment Arms via Regularized\n  and Clustered Joint Assignment Forests", "abstract": "  We consider learning personalized assignments to one of many treatment arms\nfrom a randomized controlled trial. Standard methods that estimate\nheterogeneous treatment effects separately for each arm may perform poorly in\nthis case due to excess variance. We instead propose methods that pool\ninformation across treatment arms: First, we consider a regularized\nforest-based assignment algorithm based on greedy recursive partitioning that\nshrinks effect estimates across arms. Second, we augment our algorithm by a\nclustering scheme that combines treatment arms with consistently similar\noutcomes. In a simulation study, we compare the performance of these approaches\nto predicting arm-wise outcomes separately, and document gains of directly\noptimizing the treatment assignment with regularization and clustering. In a\ntheoretical model, we illustrate how a high number of treatment arms makes\nfinding the best arm hard, while we can achieve sizable utility gains from\npersonalization by regularized optimization.\n", "date": "2023-11-01", "categories": []}
{"id": "2311.00599", "title": "Structure Learning with Adaptive Random Neighborhood Informed MCMC", "abstract": "  In this paper, we introduce a novel MCMC sampler, PARNI-DAG, for a\nfully-Bayesian approach to the problem of structure learning under\nobservational data. Under the assumption of causal sufficiency, the algorithm\nallows for approximate sampling directly from the posterior distribution on\nDirected Acyclic Graphs (DAGs). PARNI-DAG performs efficient sampling of DAGs\nvia locally informed, adaptive random neighborhood proposal that results in\nbetter mixing properties. In addition, to ensure better scalability with the\nnumber of nodes, we couple PARNI-DAG with a pre-tuning procedure of the\nsampler's parameters that exploits a skeleton graph derived through some\nconstraint-based or scoring-based algorithms. Thanks to these novel features,\nPARNI-DAG quickly converges to high-probability regions and is less likely to\nget stuck in local modes in the presence of high correlation between nodes in\nhigh-dimensional settings. After introducing the technical novelties in\nPARNI-DAG, we empirically demonstrate its mixing efficiency and accuracy in\nlearning DAG structures on a variety of experiments.\n", "date": "2023-11-01", "categories": []}
{"id": "2311.00674", "title": "Recovering Linear Causal Models with Latent Variables via Cholesky\n  Factorization of Covariance Matrix", "abstract": "  Discovering the causal relationship via recovering the directed acyclic graph\n(DAG) structure from the observed data is a well-known challenging\ncombinatorial problem. When there are latent variables, the problem becomes\neven more difficult. In this paper, we first propose a DAG structure recovering\nalgorithm, which is based on the Cholesky factorization of the covariance\nmatrix of the observed data. The algorithm is fast and easy to implement and\nhas theoretical grantees for exact recovery. On synthetic and real-world\ndatasets, the algorithm is significantly faster than previous methods and\nachieves the state-of-the-art performance. Furthermore, under the equal error\nvariances assumption, we incorporate an optimization procedure into the\nCholesky factorization based algorithm to handle the DAG recovering problem\nwith latent variables. Numerical simulations show that the modified \"Cholesky +\noptimization\" algorithm is able to recover the ground truth graph in most cases\nand outperforms existing algorithms.\n", "date": "2023-11-01", "categories": []}
{"id": "2305.14083", "title": "Counterfactual Augmentation for Multimodal Learning Under Presentation\n  Bias", "abstract": "  In real-world machine learning systems, labels are often derived from user\nbehaviors that the system wishes to encourage. Over time, new models must be\ntrained as new training examples and features become available. However,\nfeedback loops between users and models can bias future user behavior, inducing\na presentation bias in the labels that compromises the ability to train new\nmodels. In this paper, we propose counterfactual augmentation, a novel causal\nmethod for correcting presentation bias using generated counterfactual labels.\nOur empirical evaluations demonstrate that counterfactual augmentation yields\nbetter downstream performance compared to both uncorrected models and existing\nbias-correction methods. Model analyses further indicate that the generated\ncounterfactuals align closely with true counterfactuals in an oracle setting.\n", "date": "2023-05-23", "categories": []}
{"id": "2306.07916", "title": "Identification of Nonlinear Latent Hierarchical Models", "abstract": "  Identifying latent variables and causal structures from observational data is\nessential to many real-world applications involving biological data, medical\ndata, and unstructured data such as images and languages. However, this task\ncan be highly challenging, especially when observed variables are generated by\ncausally related latent variables and the relationships are nonlinear. In this\nwork, we investigate the identification problem for nonlinear latent\nhierarchical causal models in which observed variables are generated by a set\nof causally related latent variables, and some latent variables may not have\nobserved children.\n  We show that the identifiability of causal structures and latent variables\n(up to invertible transformations) can be achieved under mild assumptions: on\ncausal structures, we allow for multiple paths between any pair of variables in\nthe graph, which relaxes latent tree assumptions in prior work; on structural\nfunctions, we permit general nonlinearity and multi-dimensional continuous\nvariables, alleviating existing work's parametric assumptions. Specifically, we\nfirst develop an identification criterion in the form of novel identifiability\nguarantees for an elementary latent variable model. Leveraging this criterion,\nwe show that both causal structures and latent variables of the hierarchical\nmodel can be identified asymptotically by explicitly constructing an estimation\nprocedure. To the best of our knowledge, our work is the first to establish\nidentifiability guarantees for both causal structures and latent variables in\nnonlinear latent hierarchical models.\n", "date": "2023-06-13", "categories": []}
{"id": "2310.20075", "title": "Meek Separators and Their Applications in Targeted Causal Discovery", "abstract": "  Learning causal structures from interventional data is a fundamental problem\nwith broad applications across various fields. While many previous works have\nfocused on recovering the entire causal graph, in practice, there are scenarios\nwhere learning only part of the causal graph suffices. This is called\n$targeted$ causal discovery. In our work, we focus on two such well-motivated\nproblems: subset search and causal matching. We aim to minimize the number of\ninterventions in both cases.\n  Towards this, we introduce the $Meek~separator$, which is a subset of\nvertices that, when intervened, decomposes the remaining unoriented edges into\nsmaller connected components. We then present an efficient algorithm to find\nMeek separators that are of small sizes. Such a procedure is helpful in\ndesigning various divide-and-conquer-based approaches. In particular, we\npropose two randomized algorithms that achieve logarithmic approximation for\nsubset search and causal matching, respectively. Our results provide the first\nknown average-case provable guarantees for both problems. We believe that this\nopens up possibilities to design near-optimal methods for many other targeted\ncausal structure learning problems arising from various applications.\n", "date": "2023-10-30", "categories": []}
{"id": "2310.20307", "title": "Causal Interpretation of Self-Attention in Pre-Trained Transformers", "abstract": "  We propose a causal interpretation of self-attention in the Transformer\nneural network architecture. We interpret self-attention as a mechanism that\nestimates a structural equation model for a given input sequence of symbols\n(tokens). The structural equation model can be interpreted, in turn, as a\ncausal structure over the input symbols under the specific context of the input\nsequence. Importantly, this interpretation remains valid in the presence of\nlatent confounders. Following this interpretation, we estimate conditional\nindependence relations between input symbols by calculating partial\ncorrelations between their corresponding representations in the deepest\nattention layer. This enables learning the causal structure over an input\nsequence using existing constraint-based algorithms. In this sense, existing\npre-trained Transformers can be utilized for zero-shot causal-discovery. We\ndemonstrate this method by providing causal explanations for the outcomes of\nTransformers in two tasks: sentiment classification (NLP) and recommendation.\n", "date": "2023-10-31", "categories": []}
{"id": "2310.20697", "title": "Text-Transport: Toward Learning Causal Effects of Natural Language", "abstract": "  As language technologies gain prominence in real-world settings, it is\nimportant to understand how changes to language affect reader perceptions. This\ncan be formalized as the causal effect of varying a linguistic attribute (e.g.,\nsentiment) on a reader's response to the text. In this paper, we introduce\nText-Transport, a method for estimation of causal effects from natural language\nunder any text distribution. Current approaches for valid causal effect\nestimation require strong assumptions about the data, meaning the data from\nwhich one can estimate valid causal effects often is not representative of the\nactual target domain of interest. To address this issue, we leverage the notion\nof distribution shift to describe an estimator that transports causal effects\nbetween domains, bypassing the need for strong assumptions in the target\ndomain. We derive statistical guarantees on the uncertainty of this estimator,\nand we report empirical results and analyses that support the validity of\nText-Transport across data settings. Finally, we use Text-Transport to study a\nrealistic setting--hate speech on social media--in which causal effects do\nshift significantly between text domains, demonstrating the necessity of\ntransport when conducting causal inference on natural language.\n", "date": "2023-10-31", "categories": []}
{"id": "2208.10027", "title": "Learning Invariant Representations under General Interventions on the\n  Response", "abstract": "  It has become increasingly common nowadays to collect observations of feature\nand response pairs from different environments. As a consequence, one has to\napply learned predictors to data with a different distribution due to\ndistribution shifts. One principled approach is to adopt the structural causal\nmodels to describe training and test models, following the invariance principle\nwhich says that the conditional distribution of the response given its\npredictors remains the same across environments. However, this principle might\nbe violated in practical settings when the response is intervened. A natural\nquestion is whether it is still possible to identify other forms of invariance\nto facilitate prediction in unseen environments. To shed light on this\nchallenging scenario, we focus on linear structural causal models (SCMs) and\nintroduce invariant matching property (IMP), an explicit relation to capture\ninterventions through an additional feature, leading to an alternative form of\ninvariance that enables a unified treatment of general interventions on the\nresponse as well as the predictors. We analyze the asymptotic generalization\nerrors of our method under both the discrete and continuous environment\nsettings, where the continuous case is handled by relating it to the\nsemiparametric varying coefficient models. We present algorithms that show\ncompetitive performance compared to existing methods over various experimental\nsettings including a COVID dataset.\n", "date": "2022-08-22", "categories": []}
{"id": "2301.09028", "title": "Characterization and Learning of Causal Graphs with Small Conditioning\n  Sets", "abstract": "  Constraint-based causal discovery algorithms learn part of the causal graph\nstructure by systematically testing conditional independences observed in the\ndata. These algorithms, such as the PC algorithm and its variants, rely on\ngraphical characterizations of the so-called equivalence class of causal graphs\nproposed by Pearl. However, constraint-based causal discovery algorithms\nstruggle when data is limited since conditional independence tests quickly lose\ntheir statistical power, especially when the conditioning set is large. To\naddress this, we propose using conditional independence tests where the size of\nthe conditioning set is upper bounded by some integer $k$ for robust causal\ndiscovery. The existing graphical characterizations of the equivalence classes\nof causal graphs are not applicable when we cannot leverage all the conditional\nindependence statements. We first define the notion of $k$-Markov equivalence:\nTwo causal graphs are $k$-Markov equivalent if they entail the same conditional\nindependence constraints where the conditioning set size is upper bounded by\n$k$. We propose a novel representation that allows us to graphically\ncharacterize $k$-Markov equivalence between two causal graphs. We propose a\nsound constraint-based algorithm called the $k$-PC algorithm for learning this\nequivalence class. Finally, we conduct synthetic, and semi-synthetic\nexperiments to demonstrate that the $k$-PC algorithm enables more robust causal\ndiscovery in the small sample regime compared to the baseline algorithms.\n", "date": "2023-01-22", "categories": []}
{"id": "2306.00542", "title": "Nonparametric Identifiability of Causal Representations from Unknown\n  Interventions", "abstract": "  We study causal representation learning, the task of inferring latent causal\nvariables and their causal relations from high-dimensional mixtures of the\nvariables. Prior work relies on weak supervision, in the form of counterfactual\npre- and post-intervention views or temporal structure; places restrictive\nassumptions, such as linearity, on the mixing function or latent causal model;\nor requires partial knowledge of the generative process, such as the causal\ngraph or intervention targets. We instead consider the general setting in which\nboth the causal model and the mixing function are nonparametric. The learning\nsignal takes the form of multiple datasets, or environments, arising from\nunknown interventions in the underlying causal model. Our goal is to identify\nboth the ground truth latents and their causal graph up to a set of ambiguities\nwhich we show to be irresolvable from interventional data. We study the\nfundamental setting of two causal variables and prove that the observational\ndistribution and one perfect intervention per node suffice for identifiability,\nsubject to a genericity condition. This condition rules out spurious solutions\nthat involve fine-tuning of the intervened and observational distributions,\nmirroring similar conditions for nonlinear cause-effect inference. For an\narbitrary number of variables, we show that at least one pair of distinct\nperfect interventional domains per node guarantees identifiability. Further, we\ndemonstrate that the strengths of causal influences among the latent variables\nare preserved by all equivalent solutions, rendering the inferred\nrepresentation appropriate for drawing causal conclusions from new data. Our\nstudy provides the first identifiability results for the general nonparametric\nsetting with unknown interventions, and elucidates what is possible and\nimpossible for causal representation learning without more direct supervision.\n", "date": "2023-06-01", "categories": []}
{"id": "2306.11839", "title": "Should I Stop or Should I Go: Early Stopping with Heterogeneous\n  Populations", "abstract": "  Randomized experiments often need to be stopped prematurely due to the\ntreatment having an unintended harmful effect. Existing methods that determine\nwhen to stop an experiment early are typically applied to the data in aggregate\nand do not account for treatment effect heterogeneity. In this paper, we study\nthe early stopping of experiments for harm on heterogeneous populations. We\nfirst establish that current methods often fail to stop experiments when the\ntreatment harms a minority group of participants. We then use causal machine\nlearning to develop CLASH, the first broadly-applicable method for\nheterogeneous early stopping. We demonstrate CLASH's performance on simulated\nand real data and show that it yields effective early stopping for both\nclinical trials and A/B tests.\n", "date": "2023-06-20", "categories": []}
{"id": "2310.18654", "title": "Causal discovery in a complex industrial system: A time series benchmark", "abstract": "  Causal discovery outputs a causal structure, represented by a graph, from\nobserved data. For time series data, there is a variety of methods, however, it\nis difficult to evaluate these on real data as realistic use cases very rarely\ncome with a known causal graph to which output can be compared. In this paper,\nwe present a dataset from an industrial subsystem at the European Spallation\nSource along with its causal graph which has been constructed from expert\nknowledge. This provides a testbed for causal discovery from time series\nobservations of complex systems, and we believe this can help inform the\ndevelopment of causal discovery methodology.\n", "date": "2023-10-28", "categories": []}
{"id": "2310.19054", "title": "Object-centric architectures enable efficient causal representation\n  learning", "abstract": "  Causal representation learning has showed a variety of settings in which we\ncan disentangle latent variables with identifiability guarantees (up to some\nreasonable equivalence class). Common to all of these approaches is the\nassumption that (1) the latent variables are represented as $d$-dimensional\nvectors, and (2) that the observations are the output of some injective\ngenerative function of these latent variables. While these assumptions appear\nbenign, we show that when the observations are of multiple objects, the\ngenerative function is no longer injective and disentanglement fails in\npractice. We can address this failure by combining recent developments in\nobject-centric learning and causal representation learning. By modifying the\nSlot Attention architecture arXiv:2006.15055, we develop an object-centric\narchitecture that leverages weak supervision from sparse perturbations to\ndisentangle each object's properties. This approach is more data-efficient in\nthe sense that it requires significantly fewer perturbations than a comparable\napproach that encodes to a Euclidean space and we show that this approach\nsuccessfully disentangles the properties of a set of objects in a series of\nsimple image-based disentanglement experiments.\n", "date": "2023-10-29", "categories": []}
{"id": "2310.19519", "title": "A General Neural Causal Model for Interactive Recommendation", "abstract": "  Survivor bias in observational data leads the optimization of recommender\nsystems towards local optima. Currently most solutions re-mines existing\nhuman-system collaboration patterns to maximize longer-term satisfaction by\nreinforcement learning. However, from the causal perspective, mitigating\nsurvivor effects requires answering a counterfactual problem, which is\ngenerally unidentifiable and inestimable. In this work, we propose a neural\ncausal model to achieve counterfactual inference. Specifically, we first build\na learnable structural causal model based on its available graphical\nrepresentations which qualitatively characterizes the preference transitions.\nMitigation of the survivor bias is achieved though counterfactual consistency.\nTo identify the consistency, we use the Gumbel-max function as structural\nconstrains. To estimate the consistency, we apply reinforcement optimizations,\nand use Gumbel-Softmax as a trade-off to get a differentiable function. Both\ntheoretical and empirical studies demonstrate the effectiveness of our\nsolution.\n", "date": "2023-10-30", "categories": []}
{"id": "2310.19691", "title": "Causal Context Connects Counterfactual Fairness to Robust Prediction and\n  Group Fairness", "abstract": "  Counterfactual fairness requires that a person would have been classified in\nthe same way by an AI or other algorithmic system if they had a different\nprotected class, such as a different race or gender. This is an intuitive\nstandard, as reflected in the U.S. legal system, but its use is limited because\ncounterfactuals cannot be directly observed in real-world data. On the other\nhand, group fairness metrics (e.g., demographic parity or equalized odds) are\nless intuitive but more readily observed. In this paper, we use $\\textit{causal\ncontext}$ to bridge the gaps between counterfactual fairness, robust\nprediction, and group fairness. First, we motivate counterfactual fairness by\nshowing that there is not necessarily a fundamental trade-off between fairness\nand accuracy because, under plausible conditions, the counterfactually fair\npredictor is in fact accuracy-optimal in an unbiased target distribution.\nSecond, we develop a correspondence between the causal graph of the\ndata-generating process and which, if any, group fairness metrics are\nequivalent to counterfactual fairness. Third, we show that in three common\nfairness contexts$\\unicode{x2013}$measurement error, selection on label, and\nselection on predictors$\\unicode{x2013}$counterfactual fairness is equivalent\nto demographic parity, equalized odds, and calibration, respectively.\nCounterfactual fairness can sometimes be tested by measuring relatively simple\ngroup fairness metrics.\n", "date": "2023-10-30", "categories": []}
{"id": "2208.04627", "title": "Causal Effect Identification in Uncertain Causal Networks", "abstract": "  Causal identification is at the core of the causal inference literature,\nwhere complete algorithms have been proposed to identify causal queries of\ninterest. The validity of these algorithms hinges on the restrictive assumption\nof having access to a correctly specified causal structure. In this work, we\nstudy the setting where a probabilistic model of the causal structure is\navailable. Specifically, the edges in a causal graph exist with uncertainties\nwhich may, for example, represent degree of belief from domain experts.\nAlternatively, the uncertainty about an edge may reflect the confidence of a\nparticular statistical test. The question that naturally arises in this setting\nis: Given such a probabilistic graph and a specific causal effect of interest,\nwhat is the subgraph which has the highest plausibility and for which the\ncausal effect is identifiable? We show that answering this question reduces to\nsolving an NP-complete combinatorial optimization problem which we call the\nedge ID problem. We propose efficient algorithms to approximate this problem\nand evaluate them against both real-world networks and randomly generated\ngraphs.\n", "date": "2022-08-09", "categories": []}
{"id": "2302.00993", "title": "Unpaired Multi-Domain Causal Representation Learning", "abstract": "  The goal of causal representation learning is to find a representation of\ndata that consists of causally related latent variables. We consider a setup\nwhere one has access to data from multiple domains that potentially share a\ncausal representation. Crucially, observations in different domains are assumed\nto be unpaired, that is, we only observe the marginal distribution in each\ndomain but not their joint distribution. In this paper, we give sufficient\nconditions for identifiability of the joint distribution and the shared causal\ngraph in a linear setup. Identifiability holds if we can uniquely recover the\njoint distribution and the shared causal representation from the marginal\ndistributions in each domain. We transform our identifiability results into a\npractical method to recover the shared latent causal graph.\n", "date": "2023-02-02", "categories": []}
{"id": "2305.14695", "title": "A Causal View of Entity Bias in (Large) Language Models", "abstract": "  Entity bias widely affects pretrained (large) language models, causing them\nto rely on (biased) parametric knowledge to make unfaithful predictions.\nAlthough causality-inspired methods have shown great potential to mitigate\nentity bias, it is hard to precisely estimate the parameters of underlying\ncausal models in practice. The rise of black-box LLMs also makes the situation\neven worse, because of their inaccessible parameters and uncalibrated logits.\nTo address these problems, we propose a specific structured causal model (SCM)\nwhose parameters are comparatively easier to estimate. Building upon this SCM,\nwe propose causal intervention techniques to mitigate entity bias for both\nwhite-box and black-box settings. The proposed causal intervention perturbs the\noriginal entity with neighboring entities. This intervention reduces specific\nbiasing information pertaining to the original entity while still preserving\nsufficient semantic information from similar entities. Under the white-box\nsetting, our training-time intervention improves OOD performance of PLMs on\nrelation extraction (RE) and machine reading comprehension (MRC) by 5.7 points\nand by 9.1 points, respectively. Under the black-box setting, our in-context\nintervention effectively reduces the entity-based knowledge conflicts of\nGPT-3.5, achieving up to 20.5 points of improvement of exact match accuracy on\nMRC and up to 17.6 points of reduction in memorization ratio on RE. Our code is\navailable at https://github.com/luka-group/Causal-View-of-Entity-Bias.\n", "date": "2023-05-24", "categories": []}
{"id": "2310.17679", "title": "Fast Scalable and Accurate Discovery of DAGs Using the Best Order Score\n  Search and Grow-Shrink Trees", "abstract": "  Learning graphical conditional independence structures is an important\nmachine learning problem and a cornerstone of causal discovery. However, the\naccuracy and execution time of learning algorithms generally struggle to scale\nto problems with hundreds of highly connected variables -- for instance,\nrecovering brain networks from fMRI data. We introduce the best order score\nsearch (BOSS) and grow-shrink trees (GSTs) for learning directed acyclic graphs\n(DAGs) in this paradigm. BOSS greedily searches over permutations of variables,\nusing GSTs to construct and score DAGs from permutations. GSTs efficiently\ncache scores to eliminate redundant calculations. BOSS achieves\nstate-of-the-art performance in accuracy and execution time, comparing\nfavorably to a variety of combinatorial and gradient-based learning algorithms\nunder a broad range of conditions. To demonstrate its practicality, we apply\nBOSS to two sets of resting-state fMRI data: simulated data with\npseudo-empirical noise distributions derived from randomized empirical fMRI\ncortical signals and clinical data from 3T fMRI scans processed into cortical\nparcels. BOSS is available for use within the TETRAD project which includes\nPython and R wrappers.\n", "date": "2023-10-26", "categories": []}
{"id": "2310.18123", "title": "Sample Complexity Bounds for Score-Matching: Causal Discovery and\n  Generative Modeling", "abstract": "  This paper provides statistical sample complexity bounds for score-matching\nand its applications in causal discovery. We demonstrate that accurate\nestimation of the score function is achievable by training a standard deep ReLU\nneural network using stochastic gradient descent. We establish bounds on the\nerror rate of recovering causal relationships using the score-matching-based\ncausal discovery method of Rolland et al. [2022], assuming a sufficiently good\nestimation of the score function. Finally, we analyze the upper bound of\nscore-matching estimation within the score-based generative modeling, which has\nbeen applied for causal discovery but is also of independent interest within\nthe domain of generative models.\n", "date": "2023-10-27", "categories": []}
{"id": "2301.12930", "title": "Cause-Effect Inference in Location-Scale Noise Models: Maximum\n  Likelihood vs. Independence Testing", "abstract": "  A fundamental problem of causal discovery is cause-effect inference, learning\nthe correct causal direction between two random variables. Significant progress\nhas been made through modelling the effect as a function of its cause and a\nnoise term, which allows us to leverage assumptions about the generating\nfunction class. The recently introduced heteroscedastic location-scale noise\nfunctional models (LSNMs) combine expressive power with identifiability\nguarantees. LSNM model selection based on maximizing likelihood achieves\nstate-of-the-art accuracy, when the noise distributions are correctly\nspecified. However, through an extensive empirical evaluation, we demonstrate\nthat the accuracy deteriorates sharply when the form of the noise distribution\nis misspecified by the user. Our analysis shows that the failure occurs mainly\nwhen the conditional variance in the anti-causal direction is smaller than that\nin the causal direction. As an alternative, we find that causal model selection\nthrough residual independence testing is much more robust to noise\nmisspecification and misleading conditional variance.\n", "date": "2023-01-26", "categories": []}
{"id": "2310.17325", "title": "C-Disentanglement: Discovering Causally-Independent Generative Factors\n  under an Inductive Bias of Confounder", "abstract": "  Representation learning assumes that real-world data is generated by a few\nsemantically meaningful generative factors (i.e., sources of variation) and\naims to discover them in the latent space. These factors are expected to be\ncausally disentangled, meaning that distinct factors are encoded into separate\nlatent variables, and changes in one factor will not affect the values of the\nothers. Compared to statistical independence, causal disentanglement allows\nmore controllable data generation, improved robustness, and better\ngeneralization. However, most existing work assumes unconfoundedness in the\ndiscovery process, that there are no common causes to the generative factors\nand thus obtain only statistical independence. In this paper, we recognize the\nimportance of modeling confounders in discovering causal generative factors.\nUnfortunately, such factors are not identifiable without proper inductive bias.\nWe fill the gap by introducing a framework entitled Confounded-Disentanglement\n(C-Disentanglement), the first framework that explicitly introduces the\ninductive bias of confounder via labels from domain expertise. In addition, we\naccordingly propose an approach to sufficiently identify the causally\ndisentangled factors under any inductive bias of the confounder. We conduct\nextensive experiments on both synthetic and real-world datasets. Our method\ndemonstrates competitive results compared to various SOTA baselines in\nobtaining causally disentangled features and downstream tasks under domain\nshifts.\n", "date": "2023-10-26", "categories": []}
{"id": "2310.16819", "title": "CATE Lasso: Conditional Average Treatment Effect Estimation with\n  High-Dimensional Linear Regression", "abstract": "  In causal inference about two treatments, Conditional Average Treatment\nEffects (CATEs) play an important role as a quantity representing an\nindividualized causal effect, defined as a difference between the expected\noutcomes of the two treatments conditioned on covariates. This study assumes\ntwo linear regression models between a potential outcome and covariates of the\ntwo treatments and defines CATEs as a difference between the linear regression\nmodels. Then, we propose a method for consistently estimating CATEs even under\nhigh-dimensional and non-sparse parameters. In our study, we demonstrate that\ndesirable theoretical properties, such as consistency, remain attainable even\nwithout assuming sparsity explicitly if we assume a weaker assumption called\nimplicit sparsity originating from the definition of CATEs. In this assumption,\nwe suppose that parameters of linear models in potential outcomes can be\ndivided into treatment-specific and common parameters, where the\ntreatment-specific parameters take difference values between each linear\nregression model, while the common parameters remain identical. Thus, in a\ndifference between two linear regression models, the common parameters\ndisappear, leaving only differences in the treatment-specific parameters.\nConsequently, the non-zero parameters in CATEs correspond to the differences in\nthe treatment-specific parameters. Leveraging this assumption, we develop a\nLasso regression method specialized for CATE estimation and present that the\nestimator is consistent. Finally, we confirm the soundness of the proposed\nmethod by simulation studies.\n", "date": "2023-10-25", "categories": []}
{"id": "2310.14925", "title": "Efficient Causal Discovery for Robotics Applications", "abstract": "  Using robots for automating tasks in environments shared with humans, such as\nwarehouses, shopping centres, or hospitals, requires these robots to comprehend\nthe fundamental physical interactions among nearby agents and objects.\nSpecifically, creating models to represent cause-and-effect relationships among\nthese elements can aid in predicting unforeseen human behaviours and anticipate\nthe outcome of particular robot actions. To be suitable for robots, causal\nanalysis must be both fast and accurate, meeting real-time demands and the\nlimited computational resources typical in most robotics applications. In this\npaper, we present a practical demonstration of our approach for fast and\naccurate causal analysis, known as Filtered PCMCI (F-PCMCI), along with a\nreal-world robotics application. The provided application illustrates how our\nF-PCMCI can accurately and promptly reconstruct the causal model of a\nhuman-robot interaction scenario, which can then be leveraged to enhance the\nquality of the interaction.\n", "date": "2023-10-23", "categories": []}
{"id": "2310.13892", "title": "Specify Robust Causal Representation from Mixed Observations", "abstract": "  Learning representations purely from observations concerns the problem of\nlearning a low-dimensional, compact representation which is beneficial to\nprediction models. Under the hypothesis that the intrinsic latent factors\nfollow some casual generative models, we argue that by learning a causal\nrepresentation, which is the minimal sufficient causes of the whole system, we\ncan improve the robustness and generalization performance of machine learning\nmodels. In this paper, we develop a learning method to learn such\nrepresentation from observational data by regularizing the learning procedure\nwith mutual information measures, according to the hypothetical factored causal\ngraph. We theoretically and empirically show that the models trained with the\nlearned causal representations are more robust under adversarial attacks and\ndistribution shifts compared with baselines. The supplementary materials are\navailable at https://github.com/ymy $4323460 / \\mathrm{CaRI} /$.\n", "date": "2023-10-21", "categories": []}
{"id": "2310.13977", "title": "Continual Invariant Risk Minimization", "abstract": "  Empirical risk minimization can lead to poor generalization behavior on\nunseen environments if the learned model does not capture invariant feature\nrepresentations. Invariant risk minimization (IRM) is a recent proposal for\ndiscovering environment-invariant representations. IRM was introduced by\nArjovsky et al. (2019) and extended by Ahuja et al. (2020). IRM assumes that\nall environments are available to the learning system at the same time. With\nthis work, we generalize the concept of IRM to scenarios where environments are\nobserved sequentially. We show that existing approaches, including those\ndesigned for continual learning, fail to identify the invariant features and\nmodels across sequentially presented environments. We extend IRM under a\nvariational Bayesian and bilevel framework, creating a general approach to\ncontinual invariant risk minimization. We also describe a strategy to solve the\noptimization problems using a variant of the alternating direction method of\nmultiplier (ADMM). We show empirically using multiple datasets and with\nmultiple sequential environments that the proposed methods outperform or is\ncompetitive with prior approaches.\n", "date": "2023-10-21", "categories": []}
{"id": "2310.14064", "title": "Counterfactual Prediction Under Selective Confounding", "abstract": "  This research addresses the challenge of conducting interpretable causal\ninference between a binary treatment and its resulting outcome when not all\nconfounders are known. Confounders are factors that have an influence on both\nthe treatment and the outcome. We relax the requirement of knowing all\nconfounders under desired treatment, which we refer to as Selective\nConfounding, to enable causal inference in diverse real-world scenarios. Our\nproposed scheme is designed to work in situations where multiple\ndecision-makers with different policies are involved and where there is a\nre-evaluation mechanism after the initial decision to ensure consistency. These\nassumptions are more practical to fulfill compared to the availability of all\nconfounders under all treatments. To tackle the issue of Selective Confounding,\nwe propose the use of dual-treatment samples. These samples allow us to employ\ntwo-step procedures, such as Regression Adjustment or Doubly-Robust, to learn\ncounterfactual predictors. We provide both theoretical error bounds and\nempirical evidence of the effectiveness of our proposed scheme using synthetic\nand real-world child placement data. Furthermore, we introduce three evaluation\nmethods specifically tailored to assess the performance in child placement\nscenarios. By emphasizing transparency and interpretability, our approach aims\nto provide decision-makers with a valuable tool. The source code repository of\nthis work is located at https://github.com/sohaib730/CausalML.\n", "date": "2023-10-21", "categories": []}
{"id": "2310.14246", "title": "Shortcuts for causal discovery of nonlinear models by score matching", "abstract": "  The use of simulated data in the field of causal discovery is ubiquitous due\nto the scarcity of annotated real data. Recently, Reisach et al., 2021\nhighlighted the emergence of patterns in simulated linear data, which displays\nincreasing marginal variance in the casual direction. As an ablation in their\nexperiments, Montagna et al., 2023 found that similar patterns may emerge in\nnonlinear models for the variance of the score vector $\\nabla \\log\np_{\\mathbf{X}}$, and introduced the ScoreSort algorithm. In this work, we\nformally define and characterize this score-sortability pattern of nonlinear\nadditive noise models. We find that it defines a class of identifiable\n(bivariate) causal models overlapping with nonlinear additive noise models. We\ntheoretically demonstrate the advantages of ScoreSort in terms of statistical\nefficiency compared to prior state-of-the-art score matching-based methods and\nempirically show the score-sortability of the most common synthetic benchmarks\nin the literature. Our findings remark (1) the lack of diversity in the data as\nan important limitation in the evaluation of nonlinear causal discovery\napproaches, (2) the importance of thoroughly testing different settings within\na problem class, and (3) the importance of analyzing statistical properties in\ncausal discovery, where research is often limited to defining identifiability\nconditions of the model.\n", "date": "2023-10-22", "categories": []}
{"id": "2310.14935", "title": "Causal machine learning for single-cell genomics", "abstract": "  Advances in single-cell omics allow for unprecedented insights into the\ntranscription profiles of individual cells. When combined with large-scale\nperturbation screens, through which specific biological mechanisms can be\ntargeted, these technologies allow for measuring the effect of targeted\nperturbations on the whole transcriptome. These advances provide an opportunity\nto better understand the causative role of genes in complex biological\nprocesses such as gene regulation, disease progression or cellular development.\nHowever, the high-dimensional nature of the data, coupled with the intricate\ncomplexity of biological systems renders this task nontrivial. Within the\nmachine learning community, there has been a recent increase of interest in\ncausality, with a focus on adapting established causal techniques and\nalgorithms to handle high-dimensional data. In this perspective, we delineate\nthe application of these methodologies within the realm of single-cell genomics\nand their challenges. We first present the model that underlies most of current\ncausal approaches to single-cell biology and discuss and challenge the\nassumptions it entails from the biological point of view. We then identify open\nproblems in the application of causal approaches to single-cell data:\ngeneralising to unseen environments, learning interpretable models, and\nlearning causal models of dynamics. For each problem, we discuss how various\nresearch directions - including the development of computational approaches and\nthe adaptation of experimental protocols - may offer ways forward, or on the\ncontrary pose some difficulties. With the advent of single cell atlases and\nincreasing perturbation data, we expect causal models to become a crucial tool\nfor informed experimental design.\n", "date": "2023-10-23", "categories": []}
{"id": "2207.09920", "title": "DESCN: Deep Entire Space Cross Networks for Individual Treatment Effect\n  Estimation", "abstract": "  Causal Inference has wide applications in various areas such as E-commerce\nand precision medicine, and its performance heavily relies on the accurate\nestimation of the Individual Treatment Effect (ITE). Conventionally, ITE is\npredicted by modeling the treated and control response functions separately in\ntheir individual sample spaces. However, such an approach usually encounters\ntwo issues in practice, i.e. divergent distribution between treated and control\ngroups due to treatment bias, and significant sample imbalance of their\npopulation sizes. This paper proposes Deep Entire Space Cross Networks (DESCN)\nto model treatment effects from an end-to-end perspective. DESCN captures the\nintegrated information of the treatment propensity, the response, and the\nhidden treatment effect through a cross network in a multi-task learning\nmanner. Our method jointly learns the treatment and response functions in the\nentire sample space to avoid treatment bias and employs an intermediate pseudo\ntreatment effect prediction network to relieve sample imbalance. Extensive\nexperiments are conducted on a synthetic dataset and a large-scaled production\ndataset from the E-commerce voucher distribution business. The results indicate\nthat DESCN can successfully enhance the accuracy of ITE estimation and improve\nthe uplift ranking performance. A sample of the production dataset and the\nsource code are released to facilitate future research in the community, which\nis, to the best of our knowledge, the first large-scale public biased treatment\ndataset for causal inference.\n", "date": "2022-07-19", "categories": []}
{"id": "2310.13364", "title": "Dissecting Causal Biases", "abstract": "  Accurately measuring discrimination in machine learning-based automated\ndecision systems is required to address the vital issue of fairness between\nsubpopulations and/or individuals. Any bias in measuring discrimination can\nlead to either amplification or underestimation of the true value of\ndiscrimination. This paper focuses on a class of bias originating in the way\ntraining data is generated and/or collected. We call such class causal biases\nand use tools from the field of causality to formally define and analyze such\nbiases. Four sources of bias are considered, namely, confounding, selection,\nmeasurement, and interaction. The main contribution of this paper is to\nprovide, for each source of bias, a closed-form expression in terms of the\nmodel parameters. This makes it possible to analyze the behavior of each source\nof bias, in particular, in which cases they are absent and in which other cases\nthey are maximized. We hope that the provided characterizations help the\ncommunity better understand the sources of bias in machine learning\napplications.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.13553", "title": "On sample complexity of conditional independence testing with Von Mises\n  estimator with application to causal discovery", "abstract": "  Motivated by conditional independence testing, an essential step in\nconstraint-based causal discovery algorithms, we study the nonparametric Von\nMises estimator for the entropy of multivariate distributions built on a kernel\ndensity estimator. We establish an exponential concentration inequality for\nthis estimator. We design a test for conditional independence (CI) based on our\nestimator, called VM-CI, which achieves optimal parametric rates under\nsmoothness assumptions. Leveraging the exponential concentration, we prove a\ntight upper bound for the overall error of VM-CI. This, in turn, allows us to\ncharacterize the sample complexity of any constraint-based causal discovery\nalgorithm that uses VM-CI for CI tests. To the best of our knowledge, this is\nthe first sample complexity guarantee for causal discovery for continuous\nvariables. Furthermore, we empirically show that VM-CI outperforms other\npopular CI tests in terms of either time or sample complexity (or both), which\ntranslates to a better performance in structure learning as well.\n", "date": "2023-10-20", "categories": []}
{"id": "2306.02479", "title": "Contagion Effect Estimation Using Proximal Embeddings", "abstract": "  Contagion effect refers to the causal effect of peers' behavior on the\noutcome of an individual in social networks. Contagion can be confounded due to\nlatent homophily which makes contagion effect estimation very hard: nodes in a\nhomophilic network tend to have ties to peers with similar attributes and can\nbehave similarly without influencing one another. One way to account for latent\nhomophily is by considering proxies for the unobserved confounders. However, as\nwe demonstrate in this paper, existing proxy-based methods for contagion effect\nestimation have a very high variance when the proxies are high-dimensional. To\naddress this issue, we introduce a novel framework, Proximal Embeddings\n(ProEmb), that integrates variational autoencoders with adversarial networks to\ncreate low-dimensional representations of high-dimensional proxies and help\nwith identifying contagion effects. While VAEs have been used previously for\nrepresentation learning in causal inference, a novel aspect of our approach is\nthe additional component of adversarial networks to balance the representations\nof different treatment groups, which is essential in causal inference from\nobservational data where these groups typically come from different\ndistributions. We empirically show that our method significantly increases the\naccuracy and reduces the variance of contagion effect estimation in\nobservational network data compared to state-of-the-art methods.\n", "date": "2023-06-04", "categories": []}
{"id": "2212.06312", "title": "Policy learning for many outcomes of interest: Combining optimal policy\n  trees with multi-objective Bayesian optimisation", "abstract": "  Methods for learning optimal policies use causal machine learning models to\ncreate human-interpretable rules for making choices around the allocation of\ndifferent policy interventions. However, in realistic policy-making contexts,\ndecision-makers often care about trade-offs between outcomes, not just\nsingle-mindedly maximising utility for one outcome. This paper proposes an\napproach termed Multi-Objective Policy Learning (MOPoL) which combines optimal\ndecision trees for policy learning with a multi-objective Bayesian optimisation\napproach to explore the trade-off between multiple outcomes. It does this by\nbuilding a Pareto frontier of non-dominated models for different hyperparameter\nsettings which govern outcome weighting. The key here is that a low-cost greedy\ntree can be an accurate proxy for the very computationally costly optimal tree\nfor the purposes of making decisions which means models can be repeatedly fit\nto learn a Pareto frontier. The method is applied to a real-world case-study of\nnon-price rationing of anti-malarial medication in Kenya.\n", "date": "2022-12-13", "categories": []}
{"id": "2302.03246", "title": "CDANs: Temporal Causal Discovery from Autocorrelated and Non-Stationary\n  Time Series Data", "abstract": "  Time series data are found in many areas of healthcare such as medical time\nseries, electronic health records (EHR), measurements of vitals, and wearable\ndevices. Causal discovery, which involves estimating causal relationships from\nobservational data, holds the potential to play a significant role in\nextracting actionable insights about human health. In this study, we present a\nnovel constraint-based causal discovery approach for autocorrelated and\nnon-stationary time series data (CDANs). Our proposed method addresses several\nlimitations of existing causal discovery methods for autocorrelated and\nnon-stationary time series data, such as high dimensionality, the inability to\nidentify lagged causal relationships, and overlooking changing modules. Our\napproach identifies lagged and instantaneous/contemporaneous causal\nrelationships along with changing modules that vary over time. The method\noptimizes the conditioning sets in a constraint-based search by considering\nlagged parents instead of conditioning on the entire past that addresses high\ndimensionality. The changing modules are detected by considering both\ncontemporaneous and lagged parents. The approach first detects the lagged\nadjacencies, then identifies the changing modules and contemporaneous\nadjacencies, and finally determines the causal direction. We extensively\nevaluated our proposed method on synthetic and real-world clinical datasets,\nand compared its performance with several baseline approaches. The experimental\nresults demonstrate the effectiveness of the proposed method in detecting\ncausal relationships and changing modules for autocorrelated and non-stationary\ntime series data.\n", "date": "2023-02-07", "categories": []}
{"id": "2310.10666", "title": "Extracting Physical Causality from Measurements to Detect and Localize\n  False Data Injection Attacks", "abstract": "  False Data Injection Attack (FDIA) has become a growing concern in modern\ncyber-physical power systems. Most existing FDIA detection techniques project\nthe raw measurement data into a high-dimensional latent space to separate\nnormal and attacked samples. These approaches focus more on the statistical\ncorrelations of data values and are therefore susceptible to data distribution\ndrifts induced by changes in system operating points or changes in FDIA types\nand strengths, especially for FDIA localization tasks. Causal inference, on the\nother hand, extracts the causality behind the coordinated fluctuations of\ndifferent measurements. The causality patterns are determined by fundamental\nphysical laws such as Ohm's Law and Kirchhoff's Law. They are sensitive to the\nviolation of physical laws caused by FDIA, but tend to remain stable with the\ndrift of system operating points. Leveraging this advantage, this paper\nproposes a joint FDIA detection and localization framework based on causal\ninference and the Graph Attention Network (GAT) to identify the attacked system\nnodes. The proposed framework consists of two levels. The lower level uses the\nX-learner algorithm to estimate the causality strength between measurements and\ngenerate Measurement Causality Graphs (MCGs). The upper level then applies a\nGAT to identify the anomaly patterns in the MCGs. Since the extracted causality\npatterns are intrinsically related to the measurements, it is easier for the\nupper level to figure out the attacked nodes than the existing FDIA\nlocalization approaches. The performance of the proposed framework is evaluated\non the IEEE 39-bus system. Experimental results show that the causality-based\nFDIA detection and localization mechanism is highly interpretable and robust.\n", "date": "2023-09-21", "categories": []}
{"id": "2310.11059", "title": "Causal Feature Selection via Transfer Entropy", "abstract": "  Machine learning algorithms are designed to capture complex relationships\nbetween features. In this context, the high dimensionality of data often\nresults in poor model performance, with the risk of overfitting. Feature\nselection, the process of selecting a subset of relevant and non-redundant\nfeatures, is, therefore, an essential step to mitigate these issues. However,\nclassical feature selection approaches do not inspect the causal relationship\nbetween selected features and target, which can lead to misleading results in\nreal-world applications. Causal discovery, instead, aims to identify causal\nrelationships between features with observational data. In this paper, we\npropose a novel methodology at the intersection between feature selection and\ncausal discovery, focusing on time series. We introduce a new causal feature\nselection approach that relies on the forward and backward feature selection\nprocedures and leverages transfer entropy to estimate the causal flow of\ninformation from the features to the target in time series. Our approach\nenables the selection of features not only in terms of mere model performance\nbut also captures the causal information flow. In this context, we provide\ntheoretical guarantees on the regression and classification errors for both the\nexact and the finite-sample cases. Finally, we present numerical validations on\nsynthetic and real-world regression problems, showing results competitive\nw.r.t. the considered baselines.\n", "date": "2023-10-17", "categories": []}
{"id": "2310.11154", "title": "Causal discovery using dynamically requested knowledge", "abstract": "  Causal Bayesian Networks (CBNs) are an important tool for reasoning under\nuncertainty in complex real-world systems. Determining the graphical structure\nof a CBN remains a key challenge and is undertaken either by eliciting it from\nhumans, using machine learning to learn it from data, or using a combination of\nthese two approaches. In the latter case, human knowledge is generally provided\nto the algorithm before it starts, but here we investigate a novel approach\nwhere the structure learning algorithm itself dynamically identifies and\nrequests knowledge for relationships that the algorithm identifies as uncertain\nduring structure learning. We integrate this approach into the Tabu structure\nlearning algorithm and show that it offers considerable gains in structural\naccuracy, which are generally larger than those offered by existing approaches\nfor integrating knowledge. We suggest that a variant which requests only arc\norientation information may be particularly useful where the practitioner has\nlittle preexisting knowledge of the causal relationships. As well as offering\nimproved accuracy, the approach can use human expertise more effectively and\ncontributes to making the structure learning process more transparent.\n", "date": "2023-10-17", "categories": []}
{"id": "2310.11340", "title": "Contextualized Machine Learning", "abstract": "  We examine Contextualized Machine Learning (ML), a paradigm for learning\nheterogeneous and context-dependent effects. Contextualized ML estimates\nheterogeneous functions by applying deep learning to the meta-relationship\nbetween contextual information and context-specific parametric models. This is\na form of varying-coefficient modeling that unifies existing frameworks\nincluding cluster analysis and cohort modeling by introducing two reusable\nconcepts: a context encoder which translates sample context into model\nparameters, and sample-specific model which operates on sample predictors. We\nreview the process of developing contextualized models, nonparametric inference\nfrom contextualized models, and identifiability conditions of contextualized\nmodels. Finally, we present the open-source PyTorch package ContextualizedML.\n", "date": "2023-10-17", "categories": []}
{"id": "2303.04038", "title": "Root Cause Identification for Collective Anomalies in Time Series given\n  an Acyclic Summary Causal Graph with Loops", "abstract": "  This paper presents an approach for identifying the root causes of collective\nanomalies given observational time series and an acyclic summary causal graph\nwhich depicts an abstraction of causal relations present in a dynamic system at\nits normal regime. The paper first shows how the problem of root cause\nidentification can be divided into many independent subproblems by grouping\nrelated anomalies using d-separation. Further, it shows how, under this\nsetting, some root causes can be found directly from the graph and from the\ntime of appearance of anomalies. Finally, it shows, how the rest of the root\ncauses can be found by comparing direct effects in the normal and in the\nanomalous regime. To this end, an adjustment set for identifying direct effects\nis introduced. Extensive experiments conducted on both simulated and real-world\ndatasets demonstrate the effectiveness of the proposed method.\n", "date": "2023-03-07", "categories": []}
{"id": "2305.16988", "title": "Sharp Bounds for Generalized Causal Sensitivity Analysis", "abstract": "  Causal inference from observational data is crucial for many disciplines such\nas medicine and economics. However, sharp bounds for causal effects under\nrelaxations of the unconfoundedness assumption (causal sensitivity analysis)\nare subject to ongoing research. So far, works with sharp bounds are restricted\nto fairly simple settings (e.g., a single binary treatment). In this paper, we\npropose a unified framework for causal sensitivity analysis under unobserved\nconfounding in various settings. For this, we propose a flexible generalization\nof the marginal sensitivity model (MSM) and then derive sharp bounds for a\nlarge class of causal effects. This includes (conditional) average treatment\neffects, effects for mediation analysis and path analysis, and distributional\neffects. Furthermore, our sensitivity model is applicable to discrete,\ncontinuous, and time-varying treatments. It allows us to interpret the partial\nidentification problem under unobserved confounding as a distribution shift in\nthe latent confounders while evaluating the causal effect of interest. In the\nspecial case of a single binary treatment, our bounds for (conditional) average\ntreatment effects coincide with recent optimality results for causal\nsensitivity analysis. Finally, we propose a scalable algorithm to estimate our\nsharp bounds from observational data.\n", "date": "2023-05-26", "categories": []}
{"id": "2310.10239", "title": "Structural transfer learning of non-Gaussian DAG", "abstract": "  Directed acyclic graph (DAG) has been widely employed to represent\ndirectional relationships among a set of collected nodes. Yet, the available\ndata in one single study is often limited for accurate DAG reconstruction,\nwhereas heterogeneous data may be collected from multiple relevant studies. It\nremains an open question how to pool the heterogeneous data together for better\nDAG structure reconstruction in the target study. In this paper, we first\nintroduce a novel set of structural similarity measures for DAG and then\npresent a transfer DAG learning framework by effectively leveraging information\nfrom auxiliary DAGs of different levels of similarities. Our theoretical\nanalysis shows substantial improvement in terms of DAG reconstruction in the\ntarget study, even when no auxiliary DAG is overall similar to the target DAG,\nwhich is in sharp contrast to most existing transfer learning methods. The\nadvantage of the proposed transfer DAG learning is also supported by extensive\nnumerical experiments on both synthetic data and multi-site brain functional\nconnectivity network data.\n", "date": "2023-10-16", "categories": []}
{"id": "2302.11715", "title": "Variable Importance Matching for Causal Inference", "abstract": "  Our goal is to produce methods for observational causal inference that are\nauditable, easy to troubleshoot, accurate for treatment effect estimation, and\nscalable to high-dimensional data. We describe a general framework called\nModel-to-Match that achieves these goals by (i) learning a distance metric via\noutcome modeling, (ii) creating matched groups using the distance metric, and\n(iii) using the matched groups to estimate treatment effects. Model-to-Match\nuses variable importance measurements to construct a distance metric, making it\na flexible framework that can be adapted to various applications. Concentrating\non the scalability of the problem in the number of potential confounders, we\noperationalize the Model-to-Match framework with LASSO. We derive performance\nguarantees for settings where LASSO outcome modeling consistently identifies\nall confounders (importantly without requiring the linear model to be correctly\nspecified). We also provide experimental results demonstrating the method's\nauditability, accuracy, and scalability as well as extensions to more general\nnonparametric outcome modeling.\n", "date": "2023-02-23", "categories": []}
{"id": "2305.02615", "title": "How to Enhance Causal Discrimination of Utterances: A Case on Affective\n  Reasoning", "abstract": "  Our investigation into the Affective Reasoning in Conversation (ARC) task\nhighlights the challenge of causal discrimination. Almost all existing models,\nincluding large language models (LLMs), excel at capturing semantic\ncorrelations within utterance embeddings but fall short in determining the\nspecific causal relationships. To overcome this limitation, we propose the\nincorporation of \\textit{i.i.d.} noise terms into the conversation process,\nthereby constructing a structural causal model (SCM). It explores how distinct\ncausal relationships of fitted embeddings can be discerned through independent\nconditions. To facilitate the implementation of deep learning, we introduce the\ncogn frameworks to handle unstructured conversation data, and employ an\nautoencoder architecture to regard the unobservable noise as learnable\n\"implicit causes.\" Moreover, we curate a synthetic dataset that includes i.i.d.\nnoise. Through comprehensive experiments, we validate the effectiveness and\ninterpretability of our approach. Our code is available in\nhttps://github.com/Zodiark-ch/mater-of-our-EMNLP2023-paper.\n", "date": "2023-05-04", "categories": []}
{"id": "2210.11355", "title": "Network Synthetic Interventions: A Causal Framework for Panel Data Under\n  Network Interference", "abstract": "  We propose a generalization of the synthetic controls and synthetic\ninterventions methodology to incorporate network interference. We consider the\nestimation of unit-specific potential outcomes from panel data in the presence\nof spillover across units and unobserved confounding. Key to our approach is a\nnovel latent factor model that takes into account network interference and\ngeneralizes the factor models typically used in panel data settings. We propose\nan estimator, Network Synthetic Interventions (NSI), and show that it\nconsistently estimates the mean outcomes for a unit under an arbitrary set of\ncounterfactual treatments for the network. We further establish that the\nestimator is asymptotically normal. We furnish two validity tests for whether\nthe NSI estimator reliably generalizes to produce accurate counterfactual\nestimates. We provide a novel graph-based experiment design that guarantees the\nNSI estimator produces accurate counterfactual estimates, and also analyze the\nsample complexity of the proposed design. We conclude with simulations that\ncorroborate our theoretical findings.\n", "date": "2022-10-20", "categories": []}
{"id": "2310.08377", "title": "Do Not Marginalize Mechanisms, Rather Consolidate!", "abstract": "  Structural causal models (SCMs) are a powerful tool for understanding the\ncomplex causal relationships that underlie many real-world systems. As these\nsystems grow in size, the number of variables and complexity of interactions\nbetween them does, too. Thus, becoming convoluted and difficult to analyze.\nThis is particularly true in the context of machine learning and artificial\nintelligence, where an ever increasing amount of data demands for new methods\nto simplify and compress large scale SCM. While methods for marginalizing and\nabstracting SCM already exist today, they may destroy the causality of the\nmarginalized model. To alleviate this, we introduce the concept of\nconsolidating causal mechanisms to transform large-scale SCM while preserving\nconsistent interventional behaviour. We show consolidation is a powerful method\nfor simplifying SCM, discuss reduction of computational complexity and give a\nperspective on generalizing abilities of consolidated SCM.\n", "date": "2023-10-12", "categories": []}
{"id": "2310.08406", "title": "Tightening Bounds on Probabilities of Causation By Merging Datasets", "abstract": "  Probabilities of Causation (PoC) play a fundamental role in decision-making\nin law, health care and public policy. Nevertheless, their point identification\nis challenging, requiring strong assumptions, in the absence of which only\nbounds can be derived. Existing work to further tighten these bounds by\nleveraging extra information either provides numerical bounds, symbolic bounds\nfor fixed dimensionality, or requires access to multiple datasets that contain\nthe same treatment and outcome variables. However, in many clinical,\nepidemiological and public policy applications, there exist external datasets\nthat examine the effect of different treatments on the same outcome variable,\nor study the association between covariates and the outcome variable. These\nexternal datasets cannot be used in conjunction with the aforementioned bounds,\nsince the former may entail different treatment assignment mechanisms, or even\nobey different causal structures. Here, we provide symbolic bounds on the PoC\nfor this challenging scenario. We focus on combining either two randomized\nexperiments studying different treatments, or a randomized experiment and an\nobservational study, assuming causal sufficiency. Our symbolic bounds work for\narbitrary dimensionality of covariates and treatment, and we discuss the\nconditions under which these bounds are tighter than existing bounds in\nliterature. Finally, our bounds parameterize the difference in treatment\nassignment mechanism across datasets, allowing the mechanisms to vary across\ndatasets while still allowing causal information to be transferred from the\nexternal dataset to the target dataset.\n", "date": "2023-10-12", "categories": []}
{"id": "2309.11942", "title": "On the Probability of Immunity", "abstract": "  This work is devoted to the study of the probability of immunity, i.e. the\neffect occurs whether exposed or not. We derive necessary and sufficient\nconditions for non-immunity and $\\epsilon$-bounded immunity, i.e. the\nprobability of immunity is zero and $\\epsilon$-bounded, respectively. The\nformer allows us to estimate the probability of benefit (i.e., the effect\noccurs if and only if exposed) from a randomized controlled trial, and the\nlatter allows us to produce bounds of the probability of benefit that are\ntighter than the existing ones. We also introduce the concept of indirect\nimmunity (i.e., through a mediator) and repeat our previous analysis for it.\nFinally, we propose a method for sensitivity analysis of the probability of\nimmunity under unmeasured confounding.\n", "date": "2023-09-21", "categories": []}
{"id": "2310.06969", "title": "Positivity-free Policy Learning with Observational Data", "abstract": "  Policy learning utilizing observational data is pivotal across various\ndomains, with the objective of learning the optimal treatment assignment policy\nwhile adhering to specific constraints such as fairness, budget, and\nsimplicity. This study introduces a novel positivity-free (stochastic) policy\nlearning framework designed to address the challenges posed by the\nimpracticality of the positivity assumption in real-world scenarios. This\nframework leverages incremental propensity score policies to adjust propensity\nscore values instead of assigning fixed values to treatments. We characterize\nthese incremental propensity score policies and establish identification\nconditions, employing semiparametric efficiency theory to propose efficient\nestimators capable of achieving rapid convergence rates, even when integrated\nwith advanced machine learning algorithms. This paper provides a thorough\nexploration of the theoretical guarantees associated with policy learning and\nvalidates the proposed framework's finite-sample performance through\ncomprehensive numerical experiments, ensuring the identification of causal\neffects from observational data is both robust and reliable.\n", "date": "2023-10-10", "categories": []}
{"id": "2209.05598", "title": "Learning domain-specific causal discovery from time series", "abstract": "  Causal discovery (CD) from time-varying data is important in neuroscience,\nmedicine, and machine learning. Techniques for CD encompass randomized\nexperiments, which are generally unbiased but expensive, and algorithms such as\nGranger causality, conditional-independence-based, structural-equation-based,\nand score-based methods that are only accurate under strong assumptions made by\nhuman designers. However, as demonstrated in other areas of machine learning,\nhuman expertise is often not entirely accurate and tends to be outperformed in\ndomains with abundant data. In this study, we examine whether we can enhance\ndomain-specific causal discovery for time series using a data-driven approach.\nOur findings indicate that this procedure significantly outperforms\nhuman-designed, domain-agnostic causal discovery methods, such as Mutual\nInformation, VAR-LiNGAM, and Granger Causality on the MOS 6502 microprocessor,\nthe NetSim fMRI dataset, and the Dream3 gene dataset. We argue that, when\nfeasible, the causality field should consider a supervised approach in which\ndomain-specific CD procedures are learned from extensive datasets with known\ncausal relationships, rather than being designed by human specialists. Our\nfindings promise a new approach toward improving CD in neural and medical data\nand for the broader machine learning community.\n", "date": "2022-09-12", "categories": []}
{"id": "2310.06100", "title": "High Dimensional Causal Inference with Variational Backdoor Adjustment", "abstract": "  Backdoor adjustment is a technique in causal inference for estimating\ninterventional quantities from purely observational data. For example, in\nmedical settings, backdoor adjustment can be used to control for confounding\nand estimate the effectiveness of a treatment. However, high dimensional\ntreatments and confounders pose a series of potential pitfalls: tractability,\nidentifiability, optimization. In this work, we take a generative modeling\napproach to backdoor adjustment for high dimensional treatments and\nconfounders. We cast backdoor adjustment as an optimization problem in\nvariational inference without reliance on proxy variables and hidden\nconfounders. Empirically, our method is able to estimate interventional\nlikelihood in a variety of high dimensional settings, including semi-synthetic\nX-ray medical data. To the best of our knowledge, this is the first application\nof backdoor adjustment in which all the relevant variables are high\ndimensional.\n", "date": "2023-10-09", "categories": []}
{"id": "2310.06237", "title": "Differentially Private Multi-Site Treatment Effect Estimation", "abstract": "  Patient privacy is a major barrier to healthcare AI. For confidentiality\nreasons, most patient data remains in silo in separate hospitals, preventing\nthe design of data-driven healthcare AI systems that need large volumes of\npatient data to make effective decisions. A solution to this is collective\nlearning across multiple sites through federated learning with differential\nprivacy. However, literature in this space typically focuses on differentially\nprivate statistical estimation and machine learning, which is different from\nthe causal inference-related problems that arise in healthcare. In this work,\nwe take a fresh look at federated learning with a focus on causal inference;\nspecifically, we look at estimating the average treatment effect (ATE), an\nimportant task in causal inference for healthcare applications, and provide a\nfederated analytics approach to enable ATE estimation across multiple sites\nalong with differential privacy (DP) guarantees at each site. The main\nchallenge comes from site heterogeneity -- different sites have different\nsample sizes and privacy budgets. We address this through a class of per-site\nestimation algorithms that reports the ATE estimate and its variance as a\nquality measure, and an aggregation algorithm on the server side that minimizes\nthe overall variance of the final ATE estimate. Our experiments on real and\nsynthetic data show that our method reliably aggregates private statistics\nacross sites and provides better privacy-utility tradeoff under site\nheterogeneity than baselines.\n", "date": "2023-10-10", "categories": []}
{"id": "2302.11737", "title": "Causally Disentangled Generative Variational AutoEncoder", "abstract": "  We present a new supervised learning technique for the Variational\nAutoEncoder (VAE) that allows it to learn a causally disentangled\nrepresentation and generate causally disentangled outcomes simultaneously. We\ncall this approach Causally Disentangled Generation (CDG). CDG is a generative\nmodel that accurately decodes an output based on a causally disentangled\nrepresentation. Our research demonstrates that adding supervised regularization\nto the encoder alone is insufficient for achieving a generative model with CDG,\neven for a simple task. Therefore, we explore the necessary and sufficient\nconditions for achieving CDG within a specific model. Additionally, we\nintroduce a universal metric for evaluating the causal disentanglement of a\ngenerative model. Empirical results from both image and tabular datasets\nsupport our findings.\n", "date": "2023-02-23", "categories": []}
{"id": "2310.05526", "title": "Projecting infinite time series graphs to finite marginal graphs using\n  number theory", "abstract": "  In recent years, a growing number of method and application works have\nadapted and applied the causal-graphical-model framework to time series data.\nMany of these works employ time-resolved causal graphs that extend infinitely\ninto the past and future and whose edges are repetitive in time, thereby\nreflecting the assumption of stationary causal relationships. However, most\nresults and algorithms from the causal-graphical-model framework are not\ndesigned for infinite graphs. In this work, we develop a method for projecting\ninfinite time series graphs with repetitive edges to marginal graphical models\non a finite time window. These finite marginal graphs provide the answers to\n$m$-separation queries with respect to the infinite graph, a task that was\npreviously unresolved. Moreover, we argue that these marginal graphs are useful\nfor causal discovery and causal effect estimation in time series, effectively\nenabling to apply results developed for finite graphs to the infinite graphs.\nThe projection procedure relies on finding common ancestors in the\nto-be-projected graph and is, by itself, not new. However, the projection\nprocedure has not yet been algorithmically implemented for time series graphs\nsince in these infinite graphs there can be infinite sets of paths that might\ngive rise to common ancestors. We solve the search over these possibly infinite\nsets of paths by an intriguing combination of path-finding techniques for\nfinite directed graphs and solution theory for linear Diophantine equations. By\nproviding an algorithm that carries out the projection, our paper makes an\nimportant step towards a theoretically-grounded and method-agnostic\ngeneralization of a range of causal inference methods and results to time\nseries.\n", "date": "2023-10-09", "categories": []}
{"id": "2309.10514", "title": "Partially Specified Causal Simulations", "abstract": "  Simulation studies play a key role in the validation of causal inference\nmethods. The simulation results are reliable only if the study is designed\naccording to the promised operational conditions of the method-in-test. Still,\nmany causal inference literature tend to design over-restricted or misspecified\nstudies. In this paper, we elaborate on the problem of improper simulation\ndesign for causal methods and compile a list of desiderata for an effective\nsimulation framework. We then introduce partially randomized causal simulation\n(PARCS), a simulation framework that meets those desiderata. PARCS synthesizes\ndata based on graphical causal models and a wide range of adjustable\nparameters. There is a legible mapping from usual causal assumptions to the\nparameters, thus, users can identify and specify the subset of related\nparameters and randomize the remaining ones to generate a range of complying\ndata-generating processes for their causal method. The result is a more\ncomprehensive and inclusive empirical investigation for causal claims. Using\nPARCS, we reproduce and extend the simulation studies of two well-known causal\ndiscovery and missing data analysis papers to emphasize the necessity of a\nproper simulation design. Our results show that those papers would have\nimproved and extended the findings, had they used PARCS for simulation. The\nframework is implemented as a Python package, too. By discussing the\ncomprehensiveness and transparency of PARCS, we encourage causal inference\nresearchers to utilize it as a standard tool for future works.\n", "date": "2023-09-19", "categories": []}
{"id": "2310.03352", "title": "Tractable Bounding of Counterfactual Queries by Knowledge Compilation", "abstract": "  We discuss the problem of bounding partially identifiable queries, such as\ncounterfactuals, in Pearlian structural causal models. A recently proposed\niterated EM scheme yields an inner approximation of those bounds by sampling\nthe initialisation parameters. Such a method requires multiple (Bayesian\nnetwork) queries over models sharing the same structural equations and\ntopology, but different exogenous probabilities. This setup makes a compilation\nof the underlying model to an arithmetic circuit advantageous, thus inducing a\nsizeable inferential speed-up. We show how a single symbolic knowledge\ncompilation allows us to obtain the circuit structure with symbolic parameters\nto be replaced by their actual values when computing the different queries. We\nalso discuss parallelisation techniques to further speed up the bound\ncomputation. Experiments against standard Bayesian network inference show clear\ncomputational advantages with up to an order of magnitude of speed-up.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.03579", "title": "Causal Inference in Gene Regulatory Networks with GFlowNet: Towards\n  Scalability in Large Systems", "abstract": "  Understanding causal relationships within Gene Regulatory Networks (GRNs) is\nessential for unraveling the gene interactions in cellular processes. However,\ncausal discovery in GRNs is a challenging problem for multiple reasons\nincluding the existence of cyclic feedback loops and uncertainty that yields\ndiverse possible causal structures. Previous works in this area either ignore\ncyclic dynamics (assume acyclic structure) or struggle with scalability. We\nintroduce Swift-DynGFN as a novel framework that enhances causal structure\nlearning in GRNs while addressing scalability concerns. Specifically,\nSwift-DynGFN exploits gene-wise independence to boost parallelization and to\nlower computational cost. Experiments on real single-cell RNA velocity and\nsynthetic GRN datasets showcase the advancement in learning causal structure in\nGRNs and scalability in larger systems.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.01011", "title": "Towards Fixing Clever-Hans Predictors with Counterfactual Knowledge\n  Distillation", "abstract": "  This paper introduces a novel technique called counterfactual knowledge\ndistillation (CFKD) to detect and remove reliance on confounders in deep\nlearning models with the help of human expert feedback. Confounders are\nspurious features that models tend to rely on, which can result in unexpected\nerrors in regulated or safety-critical domains. The paper highlights the\nbenefit of CFKD in such domains and shows some advantages of counterfactual\nexplanations over other types of explanations. We propose an experiment scheme\nto quantitatively evaluate the success of CFKD and different teachers that can\ngive feedback to the model. We also introduce a new metric that is better\ncorrelated with true test performance than validation accuracy. The paper\ndemonstrates the effectiveness of CFKD on synthetically augmented datasets and\non real-world histopathological datasets.\n", "date": "2023-10-02", "categories": []}
{"id": "2210.16751", "title": "Formalizing Statistical Causality via Modal Logic", "abstract": "  We propose a formal language for describing and explaining statistical\ncausality. Concretely, we define Statistical Causality Language (StaCL) for\nexpressing causal effects and specifying the requirements for causal inference.\nStaCL incorporates modal operators for interventions to express causal\nproperties between probability distributions in different possible worlds in a\nKripke model. We formalize axioms for probability distributions, interventions,\nand causal predicates using StaCL formulas. These axioms are expressive enough\nto derive the rules of Pearl's do-calculus. Finally, we demonstrate by examples\nthat StaCL can be used to specify and explain the correctness of statistical\ncausal inference.\n", "date": "2022-10-30", "categories": []}
{"id": "2303.05628", "title": "On the Unlikelihood of D-Separation", "abstract": "  Causal discovery aims to recover a causal graph from data generated by it;\nconstraint based methods do so by searching for a d-separating conditioning set\nof nodes in the graph via an oracle. In this paper, we provide analytic\nevidence that on large graphs, d-separation is a rare phenomenon, even when\nguaranteed to exist, unless the graph is extremely sparse. We then provide an\nanalytic average case analysis of the PC Algorithm for causal discovery, as\nwell as a variant of the SGS Algorithm we call UniformSGS. We consider a set\n$V=\\{v_1,\\ldots,v_n\\}$ of nodes, and generate a random DAG $G=(V,E)$ where\n$(v_a, v_b) \\in E$ with i.i.d. probability $p_1$ if $a<b$ and $0$ if $a > b$.\nWe provide upper bounds on the probability that a subset of $V-\\{x,y\\}$\nd-separates $x$ and $y$, conditional on $x$ and $y$ being d-separable; our\nupper bounds decay exponentially fast to $0$ as $|V| \\rightarrow \\infty$. For\nthe PC Algorithm, while it is known that its worst-case guarantees fail on\nnon-sparse graphs, we show that the same is true for the average case, and that\nthe sparsity requirement is quite demanding: for good performance, the density\nmust go to $0$ as $|V| \\rightarrow \\infty$ even in the average case. For\nUniformSGS, while it is known that the running time is exponential for existing\nedges, we show that in the average case, that is the expected running time for\nmost non-existing edges as well.\n", "date": "2023-03-10", "categories": []}
{"id": "2305.13057", "title": "Causality-Aided Trade-off Analysis for Machine Learning Fairness", "abstract": "  There has been an increasing interest in enhancing the fairness of machine\nlearning (ML). Despite the growing number of fairness-improving methods, we\nlack a systematic understanding of the trade-offs among factors considered in\nthe ML pipeline when fairness-improving methods are applied. This understanding\nis essential for developers to make informed decisions regarding the provision\nof fair ML services. Nonetheless, it is extremely difficult to analyze the\ntrade-offs when there are multiple fairness parameters and other crucial\nmetrics involved, coupled, and even in conflict with one another.\n  This paper uses causality analysis as a principled method for analyzing\ntrade-offs between fairness parameters and other crucial metrics in ML\npipelines. To ractically and effectively conduct causality analysis, we propose\na set of domain-specific optimizations to facilitate accurate causal discovery\nand a unified, novel interface for trade-off analysis based on well-established\ncausal inference methods. We conduct a comprehensive empirical study using\nthree real-world datasets on a collection of widelyused fairness-improving\ntechniques. Our study obtains actionable suggestions for users and developers\nof fair ML. We further demonstrate the versatile usage of our approach in\nselecting the optimal fairness-improving method, paving the way for more\nethical and socially responsible AI technologies.\n", "date": "2023-05-22", "categories": []}
{"id": "2306.00374", "title": "CFL: Causally Fair Language Models Through Token-level Attribute\n  Controlled Generation", "abstract": "  We propose a method to control the attributes of Language Models (LMs) for\nthe text generation task using Causal Average Treatment Effect (ATE) scores and\ncounterfactual augmentation. We explore this method, in the context of LM\ndetoxification, and propose the Causally Fair Language (CFL) architecture for\ndetoxifying pre-trained LMs in a plug-and-play manner. Our architecture is\nbased on a Structural Causal Model (SCM) that is mathematically transparent and\ncomputationally efficient as compared with many existing detoxification\ntechniques. We also propose several new metrics that aim to better understand\nthe behaviour of LMs in the context of toxic text generation. Further, we\nachieve state of the art performance for toxic degeneration, which are computed\nusing \\RTP (RTP) benchmark. Our experiments show that CFL achieves such a\ndetoxification without much impact on the model perplexity. We also show that\nCFL mitigates the unintended bias problem through experiments on the BOLD\ndataset.\n", "date": "2023-06-01", "categories": []}
{"id": "2310.01753", "title": "CausalTime: Realistically Generated Time-series for Benchmarking of\n  Causal Discovery", "abstract": "  Time-series causal discovery (TSCD) is a fundamental problem of machine\nlearning. However, existing synthetic datasets cannot properly evaluate or\npredict the algorithms' performance on real data. This study introduces the\nCausalTime pipeline to generate time-series that highly resemble the real data\nand with ground truth causal graphs for quantitative performance evaluation.\nThe pipeline starts from real observations in a specific scenario and produces\na matching benchmark dataset. Firstly, we harness deep neural networks along\nwith normalizing flow to accurately capture realistic dynamics. Secondly, we\nextract hypothesized causal graphs by performing importance analysis on the\nneural network or leveraging prior knowledge. Thirdly, we derive the ground\ntruth causal graphs by splitting the causal model into causal term, residual\nterm, and noise term. Lastly, using the fitted network and the derived causal\ngraph, we generate corresponding versatile time-series proper for algorithm\nassessment. In the experiments, we validate the fidelity of the generated data\nthrough qualitative and quantitative experiments, followed by a benchmarking of\nexisting TSCD algorithms using these generated datasets. CausalTime offers a\nfeasible solution to evaluating TSCD algorithms in real applications and can be\ngeneralized to a wide range of fields. For easy use of the proposed approach,\nwe also provide a user-friendly website, hosted on www.causaltime.cc.\n", "date": "2023-10-03", "categories": []}
{"id": "2310.01865", "title": "Conditional Instrumental Variable Regression with Representation\n  Learning for Causal Inference", "abstract": "  This paper studies the challenging problem of estimating causal effects from\nobservational data, in the presence of unobserved confounders. The two-stage\nleast square (TSLS) method and its variants with a standard instrumental\nvariable (IV) are commonly used to eliminate confounding bias, including the\nbias caused by unobserved confounders, but they rely on the linearity\nassumption. Besides, the strict condition of unconfounded instruments posed on\na standard IV is too strong to be practical. To address these challenging and\npractical problems of the standard IV method (linearity assumption and the\nstrict condition), in this paper, we use a conditional IV (CIV) to relax the\nunconfounded instrument condition of standard IV and propose a non-linear CIV\nregression with Confounding Balancing Representation Learning, CBRL.CIV, for\njointly eliminating the confounding bias from unobserved confounders and\nbalancing the observed confounders, without the linearity assumption. We\ntheoretically demonstrate the soundness of CBRL.CIV. Extensive experiments on\nsynthetic and two real-world datasets show the competitive performance of\nCBRL.CIV against state-of-the-art IV-based estimators and superiority in\ndealing with the non-linear situation.\n", "date": "2023-10-03", "categories": []}
{"id": "2310.01937", "title": "Causal Inference with Conditional Front-Door Adjustment and Identifiable\n  Variational Autoencoder", "abstract": "  An essential and challenging problem in causal inference is causal effect\nestimation from observational data. The problem becomes more difficult with the\npresence of unobserved confounding variables. The front-door adjustment is a\npractical approach for dealing with unobserved confounding variables. However,\nthe restriction for the standard front-door adjustment is difficult to satisfy\nin practice. In this paper, we relax some of the restrictions by proposing the\nconcept of conditional front-door (CFD) adjustment and develop the theorem that\nguarantees the causal effect identifiability of CFD adjustment. Furthermore, as\nit is often impossible for a CFD variable to be given in practice, it is\ndesirable to learn it from data. By leveraging the ability of deep generative\nmodels, we propose CFDiVAE to learn the representation of the CFD adjustment\nvariable directly from data with the identifiable Variational AutoEncoder and\nformally prove the model identifiability. Extensive experiments on synthetic\ndatasets validate the effectiveness of CFDiVAE and its superiority over\nexisting methods. The experiments also show that the performance of CFDiVAE is\nless sensitive to the causal strength of unobserved confounding variables. We\nfurther apply CFDiVAE to a real-world dataset to demonstrate its potential\napplication.\n", "date": "2023-10-03", "categories": []}
{"id": "2305.16183", "title": "Passive learning of active causal strategies in agents and language\n  models", "abstract": "  What can be learned about causality and experimentation from passive data?\nThis question is salient given recent successes of passively-trained language\nmodels in interactive domains such as tool use. Passive learning is inherently\nlimited. However, we show that purely passive learning can in fact allow an\nagent to learn generalizable strategies for determining and using causal\nstructures, as long as the agent can intervene at test time. We formally\nillustrate that learning a strategy of first experimenting, then seeking goals,\ncan allow generalization from passive learning in principle. We then show\nempirically that agents trained via imitation on expert data can indeed\ngeneralize at test time to infer and use causal links which are never present\nin the training data; these agents can also generalize experimentation\nstrategies to novel variable sets never observed in training. We then show that\nstrategies for causal intervention and exploitation can be generalized from\npassive data even in a more complex environment with high-dimensional\nobservations, with the support of natural language explanations. Explanations\ncan even allow passive learners to generalize out-of-distribution from\nperfectly-confounded training data. Finally, we show that language models,\ntrained only on passive next-word prediction, can generalize causal\nintervention strategies from a few-shot prompt containing examples of\nexperimentation, together with explanations and reasoning. These results\nhighlight the surprising power of passive learning of active causal strategies,\nand may help to understand the behaviors and capabilities of language models.\n", "date": "2023-05-25", "categories": []}
{"id": "2308.10047", "title": "Towards Probabilistic Causal Discovery, Inference & Explanations for\n  Autonomous Drones in Mine Surveying Tasks", "abstract": "  Causal modelling offers great potential to provide autonomous agents the\nability to understand the data-generation process that governs their\ninteractions with the world. Such models capture formal knowledge as well as\nprobabilistic representations of noise and uncertainty typically encountered by\nautonomous robots in real-world environments. Thus, causality can aid\nautonomous agents in making decisions and explaining outcomes, but deploying\ncausality in such a manner introduces new challenges. Here we identify\nchallenges relating to causality in the context of a drone system operating in\na salt mine. Such environments are challenging for autonomous agents because of\nthe presence of confounders, non-stationarity, and a difficulty in building\ncomplete causal models ahead of time. To address these issues, we propose a\nprobabilistic causal framework consisting of: causally-informed POMDP planning,\nonline SCM adaptation, and post-hoc counterfactual explanations. Further, we\noutline planned experimentation to evaluate the framework integrated with a\ndrone system in simulated mine environments and on a real-world mine dataset.\n", "date": "2023-08-19", "categories": []}
{"id": "2310.00570", "title": "LaPLACE: Probabilistic Local Model-Agnostic Causal Explanations", "abstract": "  Machine learning models have undeniably achieved impressive performance\nacross a range of applications. However, their often perceived black-box\nnature, and lack of transparency in decision-making, have raised concerns about\nunderstanding their predictions. To tackle this challenge, researchers have\ndeveloped methods to provide explanations for machine learning models. In this\npaper, we introduce LaPLACE-explainer, designed to provide probabilistic\ncause-and-effect explanations for any classifier operating on tabular data, in\na human-understandable manner. The LaPLACE-Explainer component leverages the\nconcept of a Markov blanket to establish statistical boundaries between\nrelevant and non-relevant features automatically. This approach results in the\nautomatic generation of optimal feature subsets, serving as explanations for\npredictions. Importantly, this eliminates the need to predetermine a fixed\nnumber N of top features as explanations, enhancing the flexibility and\nadaptability of our methodology. Through the incorporation of conditional\nprobabilities, our approach offers probabilistic causal explanations and\noutperforms LIME and SHAP (well-known model-agnostic explainers) in terms of\nlocal accuracy and consistency of explained features. LaPLACE's soundness,\nconsistency, local accuracy, and adaptability are rigorously validated across\nvarious classification models. Furthermore, we demonstrate the practical\nutility of these explanations via experiments with both simulated and\nreal-world datasets. This encompasses addressing trust-related issues, such as\nevaluating prediction reliability, facilitating model selection, enhancing\ntrustworthiness, and identifying fairness-related concerns within classifiers.\n", "date": "2023-10-01", "categories": []}
{"id": "2211.07533", "title": "Generalized Balancing Weights via Deep Neural Networks", "abstract": "  Estimating causal effects from observational data is a central problem in\nmany domains. A general approach is to balance covariates with weights such\nthat the distribution of the data mimics randomization. We present generalized\nbalancing weights, Neural Balancing Weights (NBW), to estimate the causal\neffects of an arbitrary mixture of discrete and continuous interventions. The\nweights were obtained through direct estimation of the density ratio between\nthe source and balanced distributions by optimizing the variational\nrepresentation of $f$-divergence. For this, we selected $\\alpha$-divergence as\nit presents efficient optimization because it has an estimator whose sample\ncomplexity is independent of its ground truth value and unbiased mini-batch\ngradients; moreover, it is advantageous for the vanishing-gradient problem. In\naddition, we provide the following two methods for estimating the balancing\nweights: improving the generalization performance of the balancing weights and\nchecking the balance of the distribution changed by the weights. Finally, we\ndiscuss the sample size requirements for the weights as a general problem of a\ncurse of dimensionality when balancing multidimensional data. Our study\nprovides a basic approach for estimating the balancing weights of\nmultidimensional data using variational $f$-divergences.\n", "date": "2022-11-14", "categories": []}
{"id": "2309.16409", "title": "Constructing Synthetic Treatment Groups without the Mean Exchangeability\n  Assumption", "abstract": "  The purpose of this work is to transport the information from multiple\nrandomized controlled trials to the target population where we only have the\ncontrol group data. Previous works rely critically on the mean exchangeability\nassumption. However, as pointed out by many current studies, the mean\nexchangeability assumption might be violated. Motivated by the synthetic\ncontrol method, we construct a synthetic treatment group for the target\npopulation by a weighted mixture of treatment groups of source populations. We\nestimate the weights by minimizing the conditional maximum mean discrepancy\nbetween the weighted control groups of source populations and the target\npopulation. We establish the asymptotic normality of the synthetic treatment\ngroup estimator based on the sieve semiparametric theory. Our method can serve\nas a novel complementary approach when the mean exchangeability assumption is\nviolated. Experiments are conducted on synthetic and real-world datasets to\ndemonstrate the effectiveness of our methods.\n", "date": "2023-09-28", "categories": []}
{"id": "2303.07122", "title": "Quantifying Causes of Arctic Amplification via Deep Learning based\n  Time-series Causal Inference", "abstract": "  The warming of the Arctic, also known as Arctic amplification, is led by\nseveral atmospheric and oceanic drivers. However, the details of its underlying\nthermodynamic causes are still unknown. Inferring the causal effects of\natmospheric processes on sea ice melt using fixed treatment effect strategies\nleads to unrealistic counterfactual estimations. Such models are also prone to\nbias due to time-varying confoundedness. Further, the complex non-linearity in\nEarth science data makes it infeasible to perform causal inference using\nexisting marginal structural techniques. In order to tackle these challenges,\nwe propose TCINet - time-series causal inference model to infer causation under\ncontinuous treatment using recurrent neural networks and a novel probabilistic\nbalancing technique. Through experiments on synthetic and observational data,\nwe show how our research can substantially improve the ability to quantify\nleading causes of Arctic sea ice melt, further paving paths for causal\ninference in observational Earth science.\n", "date": "2023-02-22", "categories": []}
{"id": "2211.08573", "title": "Realization of Causal Representation Learning to Adjust Confounding Bias\n  in Latent Space", "abstract": "  Causal DAGs(Directed Acyclic Graphs) are usually considered in a 2D plane.\nEdges indicate causal effects' directions and imply their corresponding\ntime-passings. Due to the natural restriction of statistical models, effect\nestimation is usually approximated by averaging the individuals' correlations,\ni.e., observational changes over a specific time. However, in the context of\nMachine Learning on large-scale questions with complex DAGs, such slight biases\ncan snowball to distort global models - More importantly, it has practically\nimpeded the development of AI, for instance, the weak generalizability of\ncausal models. In this paper, we redefine causal DAG as \\emph{do-DAG}, in which\nvariables' values are no longer time-stamp-dependent, and timelines can be seen\nas axes. By geometric explanation of multi-dimensional do-DAG, we identify the\n\\emph{Causal Representation Bias} and its necessary factors, differentiated\nfrom common confounding biases. Accordingly, a DL(Deep Learning)-based\nframework will be proposed as the general solution, along with a realization\nmethod and experiments to verify its feasibility.\n", "date": "2022-11-15", "categories": []}
{"id": "2301.10859", "title": "Salesforce CausalAI Library: A Fast and Scalable Framework for Causal\n  Analysis of Time Series and Tabular Data", "abstract": "  We introduce the Salesforce CausalAI Library, an open-source library for\ncausal analysis using observational data. It supports causal discovery and\ncausal inference for tabular and time series data, of discrete, continuous and\nheterogeneous types. This library includes algorithms that handle linear and\nnon-linear causal relationships between variables, and uses multi-processing\nfor speed-up. We also include a data generator capable of generating synthetic\ndata with specified structural equation model for the aforementioned data\nformats and types, that helps users control the ground-truth causal process\nwhile investigating various algorithms. Finally, we provide a user interface\n(UI) that allows users to perform causal analysis on data without coding. The\ngoal of this library is to provide a fast and flexible solution for a variety\nof problems in the domain of causality. This technical report describes the\nSalesforce CausalAI API along with its capabilities, the implementations of the\nsupported algorithms, and experiments demonstrating their performance and\nspeed. Our library is available at\n\\url{https://github.com/salesforce/causalai}.\n", "date": "2023-01-25", "categories": []}
{"id": "2302.00869", "title": "Disentanglement of Latent Representations via Causal Interventions", "abstract": "  The process of generating data such as images is controlled by independent\nand unknown factors of variation. The retrieval of these variables has been\nstudied extensively in the disentanglement, causal representation learning, and\nindependent component analysis fields. Recently, approaches merging these\ndomains together have shown great success. Instead of directly representing the\nfactors of variation, the problem of disentanglement can be seen as finding the\ninterventions on one image that yield a change to a single factor. Following\nthis assumption, we introduce a new method for disentanglement inspired by\ncausal dynamics that combines causality theory with vector-quantized\nvariational autoencoders. Our model considers the quantized vectors as causal\nvariables and links them in a causal graph. It performs causal interventions on\nthe graph and generates atomic transitions affecting a unique factor of\nvariation in the image. We also introduce a new task of action retrieval that\nconsists of finding the action responsible for the transition between two\nimages. We test our method on standard synthetic and real-world disentanglement\ndatasets. We show that it can effectively disentangle the factors of variation\nand perform precise interventions on high-level semantic attributes of an image\nwithout affecting its quality, even with imbalanced data distributions.\n", "date": "2023-02-02", "categories": []}
{"id": "2309.13066", "title": "Causal Discovery and Counterfactual Explanations for Personalized\n  Student Learning", "abstract": "  The paper focuses on identifying the causes of student performance to provide\npersonalized recommendations for improving pass rates. We introduce the need to\nmove beyond predictive models and instead identify causal relationships. We\npropose using causal discovery techniques to achieve this. The study's main\ncontributions include using causal discovery to identify causal predictors of\nstudent performance and applying counterfactual analysis to provide\npersonalized recommendations. The paper describes the application of causal\ndiscovery methods, specifically the PC algorithm, to real-life student\nperformance data. It addresses challenges such as sample size limitations and\nemphasizes the role of domain knowledge in causal discovery. The results reveal\nthe identified causal relationships, such as the influence of earlier test\ngrades and mathematical ability on final student performance. Limitations of\nthis study include the reliance on domain expertise for accurate causal\ndiscovery, and the necessity of larger sample sizes for reliable results. The\npotential for incorrect causal structure estimations is acknowledged. A major\nchallenge remains, which is the real-time implementation and validation of\ncounterfactual recommendations. In conclusion, the paper demonstrates the value\nof causal discovery for understanding student performance and providing\npersonalized recommendations. It highlights the challenges, benefits, and\nlimitations of using causal inference in an educational context, setting the\nstage for future studies to further explore and refine these methods.\n", "date": "2023-09-18", "categories": []}
{"id": "2309.13103", "title": "OpportunityFinder: A Framework for Automated Causal Inference", "abstract": "  We introduce OpportunityFinder, a code-less framework for performing a\nvariety of causal inference studies with panel data for non-expert users. In\nits current state, OpportunityFinder only requires users to provide raw\nobservational data and a configuration file. A pipeline is then triggered that\ninspects/processes data, chooses the suitable algorithm(s) to execute the\ncausal study. It returns the causal impact of the treatment on the configured\noutcome, together with sensitivity and robustness results. Causal inference is\nwidely studied and used to estimate the downstream impact of individual's\ninteractions with products and features. It is common that these causal studies\nare performed by scientists and/or economists periodically. Business\nstakeholders are often bottle-necked on scientist or economist bandwidth to\nconduct causal studies. We offer OpportunityFinder as a solution for commonly\nperformed causal studies with four key features: (1) easy to use for both\nBusiness Analysts and Scientists, (2) abstraction of multiple algorithms under\na single I/O interface, (3) support for causal impact analysis under binary\ntreatment with panel data and (4) dynamic selection of algorithm based on scale\nof data.\n", "date": "2023-09-22", "categories": []}
{"id": "2309.13884", "title": "Estimating Treatment Effects Under Heterogeneous Interference", "abstract": "  Treatment effect estimation can assist in effective decision-making in\ne-commerce, medicine, and education. One popular application of this estimation\nlies in the prediction of the impact of a treatment (e.g., a promotion) on an\noutcome (e.g., sales) of a particular unit (e.g., an item), known as the\nindividual treatment effect (ITE). In many online applications, the outcome of\na unit can be affected by the treatments of other units, as units are often\nassociated, which is referred to as interference. For example, on an online\nshopping website, sales of an item will be influenced by an advertisement of\nits co-purchased item. Prior studies have attempted to model interference to\nestimate the ITE accurately, but they often assume a homogeneous interference,\ni.e., relationships between units only have a single view. However, in\nreal-world applications, interference may be heterogeneous, with multi-view\nrelationships. For instance, the sale of an item is usually affected by the\ntreatment of its co-purchased and co-viewed items. We hypothesize that ITE\nestimation will be inaccurate if this heterogeneous interference is not\nproperly modeled. Therefore, we propose a novel approach to model heterogeneous\ninterference by developing a new architecture to aggregate information from\ndiverse neighbors. Our proposed method contains graph neural networks that\naggregate same-view information, a mechanism that aggregates information from\ndifferent views, and attention mechanisms. In our experiments on multiple\ndatasets with heterogeneous interference, the proposed method significantly\noutperforms existing methods for ITE estimation, confirming the importance of\nmodeling heterogeneous interference.\n", "date": "2023-09-25", "categories": []}
{"id": "2309.13924", "title": "Recursive Counterfactual Deconfounding for Object Recognition", "abstract": "  Image recognition is a classic and common task in the computer vision field,\nwhich has been widely applied in the past decade. Most existing methods in\nliterature aim to learn discriminative features from labeled images for\nclassification, however, they generally neglect confounders that infiltrate\ninto the learned features, resulting in low performances for discriminating\ntest images. To address this problem, we propose a Recursive Counterfactual\nDeconfounding model for object recognition in both closed-set and open-set\nscenarios based on counterfactual analysis, called RCD. The proposed model\nconsists of a factual graph and a counterfactual graph, where the relationships\namong image features, model predictions, and confounders are built and updated\nrecursively for learning more discriminative features. It performs in a\nrecursive manner so that subtler counterfactual features could be learned and\neliminated progressively, and both the discriminability and generalization of\nthe proposed model could be improved accordingly. In addition, a negative\ncorrelation constraint is designed for alleviating the negative effects of the\ncounterfactual features further at the model training stage. Extensive\nexperimental results on both closed-set recognition task and open-set\nrecognition task demonstrate that the proposed RCD model performs better than\n11 state-of-the-art baselines significantly in most cases.\n", "date": "2023-09-25", "categories": []}
{"id": "2302.13348", "title": "Kernel Conditional Moment Constraints for Confounding Robust Inference", "abstract": "  We study policy evaluation of offline contextual bandits subject to\nunobserved confounders. Sensitivity analysis methods are commonly used to\nestimate the policy value under the worst-case confounding over a given\nuncertainty set. However, existing work often resorts to some coarse relaxation\nof the uncertainty set for the sake of tractability, leading to overly\nconservative estimation of the policy value. In this paper, we propose a\ngeneral estimator that provides a sharp lower bound of the policy value. It can\nbe shown that our estimator contains the recently proposed sharp estimator by\nDorn and Guo (2022) as a special case, and our method enables a novel extension\nof the classical marginal sensitivity model using f-divergence. To construct\nour estimator, we leverage the kernel method to obtain a tractable\napproximation to the conditional moment constraints, which traditional\nnon-sharp estimators failed to take into account. In the theoretical analysis,\nwe provide a condition for the choice of the kernel which guarantees no\nspecification error that biases the lower bound estimation. Furthermore, we\nprovide consistency guarantees of policy evaluation and learning. In the\nexperiments with synthetic and real-world data, we demonstrate the\neffectiveness of the proposed method.\n", "date": "2023-02-26", "categories": []}
{"id": "2308.09735", "title": "CTP:A Causal Interpretable Model for Non-Communicable Disease\n  Progression Prediction", "abstract": "  Non-communicable disease is the leading cause of death, emphasizing the need\nfor accurate prediction of disease progression and informed clinical\ndecision-making. Machine learning (ML) models have shown promise in this domain\nby capturing non-linear patterns within patient features. However, existing\nML-based models cannot provide causal interpretable predictions and estimate\ntreatment effects, limiting their decision-making perspective. In this study,\nwe propose a novel model called causal trajectory prediction (CTP) to tackle\nthe limitation. The CTP model combines trajectory prediction and causal\ndiscovery to enable accurate prediction of disease progression trajectories and\nuncover causal relationships between features. By incorporating a causal graph\ninto the prediction process, CTP ensures that ancestor features are not\ninfluenced by the treatment of descendant features, thereby enhancing the\ninterpretability of the model. By estimating the bounds of treatment effects,\neven in the presence of unmeasured confounders, the CTP provides valuable\ninsights for clinical decision-making. We evaluate the performance of the CTP\nusing simulated and real medical datasets. Experimental results demonstrate\nthat our model achieves satisfactory performance, highlighting its potential to\nassist clinical decisions. Source code is in\n\\href{https://github.com/DanielSun94/CFPA}{here}.\n", "date": "2023-08-18", "categories": []}
{"id": "2309.12600", "title": "Multiply Robust Federated Estimation of Targeted Average Treatment\n  Effects", "abstract": "  Federated or multi-site studies have distinct advantages over single-site\nstudies, including increased generalizability, the ability to study\nunderrepresented populations, and the opportunity to study rare exposures and\noutcomes. However, these studies are challenging due to the need to preserve\nthe privacy of each individual's data and the heterogeneity in their covariate\ndistributions. We propose a novel federated approach to derive valid causal\ninferences for a target population using multi-site data. We adjust for\ncovariate shift and covariate mismatch between sites by developing\nmultiply-robust and privacy-preserving nuisance function estimation. Our\nmethodology incorporates transfer learning to estimate ensemble weights to\ncombine information from source sites. We show that these learned weights are\nefficient and optimal under different scenarios. We showcase the finite sample\nadvantages of our approach in terms of efficiency and robustness compared to\nexisting approaches.\n", "date": "2023-09-22", "categories": []}
{"id": "2309.11509", "title": "Using causal inference to avoid fallouts in data-driven parametric\n  analysis: a case study in the architecture, engineering, and construction\n  industry", "abstract": "  The decision-making process in real-world implementations has been affected\nby a growing reliance on data-driven models. We investigated the synergetic\npattern between the data-driven methods, empirical domain knowledge, and\nfirst-principles simulations. We showed the potential risk of biased results\nwhen using data-driven models without causal analysis. Using a case study\nassessing the implication of several design solutions on the energy consumption\nof a building, we proved the necessity of causal analysis during the\ndata-driven modeling process. We concluded that: (a) Data-driven models'\naccuracy assessment or domain knowledge screening may not rule out biased and\nspurious results; (b) Data-driven models' feature selection should involve\ncareful consideration of causal relationships, especially colliders; (c) Causal\nanalysis results can be used as an aid to first-principles simulation design\nand parameter checking to avoid cognitive biases. We proved the benefits of\ncausal analysis when applied to data-driven models in building engineering.\n", "date": "2023-09-11", "categories": []}
{"id": "2309.12036", "title": "Uplift vs. predictive modeling: a theoretical analysis", "abstract": "  Despite the growing popularity of machine-learning techniques in\ndecision-making, the added value of causal-oriented strategies with respect to\npure machine-learning approaches has rarely been quantified in the literature.\nThese strategies are crucial for practitioners in various domains, such as\nmarketing, telecommunications, health care and finance. This paper presents a\ncomprehensive treatment of the subject, starting from firm theoretical\nfoundations and highlighting the parameters that influence the performance of\nthe uplift and predictive approaches. The focus of the paper is on a binary\noutcome case and a binary action, and the paper presents a theoretical analysis\nof uplift modeling, comparing it with the classical predictive approach. The\nmain research contributions of the paper include a new formulation of the\nmeasure of profit, a formal proof of the convergence of the uplift curve to the\nmeasure of profit ,and an illustration, through simulations, of the conditions\nunder which predictive approaches still outperform uplift modeling. We show\nthat the mutual information between the features and the outcome plays a\nsignificant role, along with the variance of the estimators, the distribution\nof the potential outcomes and the underlying costs and benefits of the\ntreatment and the outcome.\n", "date": "2023-09-21", "categories": []}
{"id": "2309.09416", "title": "Causal Discovery and Prediction: Methods and Algorithms", "abstract": "  We are not only observers but also actors of reality. Our capability to\nintervene and alter the course of some events in the space and time surrounding\nus is an essential component of how we build our model of the world. In this\ndoctoral thesis we introduce a generic a-priori assessment of each possible\nintervention, in order to select the most cost-effective interventions only,\nand avoid unnecessary systematic experimentation on the real world. Based on\nthis a-priori assessment, we propose an active learning algorithm that\nidentifies the causal relations in any given causal model, using a least cost\nsequence of interventions. There are several novel aspects introduced by our\nalgorithm. It is, in most case scenarios, able to discard many causal model\ncandidates using relatively inexpensive interventions that only test one value\nof the intervened variables. Also, the number of interventions performed by the\nalgorithm can be bounded by the number of causal model candidates. Hence, fewer\ninitial candidates (or equivalently, more prior knowledge) lead to fewer\ninterventions for causal discovery.\n  Causality is intimately related to time, as causes appear to precede their\neffects. Cyclical causal processes are a very interesting case of causality in\nrelation to time. In this doctoral thesis we introduce a formal analysis of\ntime cyclical causal settings by defining a causal analog to the purely\nobservational Dynamic Bayesian Networks, and provide a sound and complete\nalgorithm for the identification of causal effects in the cyclic setting. We\nintroduce the existence of two types of hidden confounder variables in this\nframework, which affect in substantially different ways the identification\nprocedures, a distinction with no analog in either Dynamic Bayesian Networks or\nstandard causal graphs.\n", "date": "2023-09-18", "categories": []}
{"id": "2309.09901", "title": "The role of causality in explainable artificial intelligence", "abstract": "  Causality and eXplainable Artificial Intelligence (XAI) have developed as\nseparate fields in computer science, even though the underlying concepts of\ncausation and explanation share common ancient roots. This is further enforced\nby the lack of review works jointly covering these two fields. In this paper,\nwe investigate the literature to try to understand how and to what extent\ncausality and XAI are intertwined. More precisely, we seek to uncover what\nkinds of relationships exist between the two concepts and how one can benefit\nfrom them, for instance, in building trust in AI systems. As a result, three\nmain perspectives are identified. In the first one, the lack of causality is\nseen as one of the major limitations of current AI and XAI approaches, and the\n\"optimal\" form of explanations is investigated. The second is a pragmatic\nperspective and considers XAI as a tool to foster scientific exploration for\ncausal inquiry, via the identification of pursue-worthy experimental\nmanipulations. Finally, the third perspective supports the idea that causality\nis propaedeutic to XAI in three possible manners: exploiting concepts borrowed\nfrom causality to support or improve XAI, utilizing counterfactuals for\nexplainability, and considering accessing a causal model as explaining itself.\nTo complement our analysis, we also provide relevant software solutions used to\nautomate causal tasks. We believe our work provides a unified view of the two\nfields of causality and XAI by highlighting potential domain bridges and\nuncovering possible limitations.\n", "date": "2023-09-18", "categories": []}
{"id": "2211.08209", "title": "On counterfactual inference with unobserved confounding", "abstract": "  Given an observational study with $n$ independent but heterogeneous units,\nour goal is to learn the counterfactual distribution for each unit using only\none $p$-dimensional sample per unit containing covariates, interventions, and\noutcomes. Specifically, we allow for unobserved confounding that introduces\nstatistical biases between interventions and outcomes as well as exacerbates\nthe heterogeneity across units. Modeling the conditional distribution of the\noutcomes as an exponential family, we reduce learning the unit-level\ncounterfactual distributions to learning $n$ exponential family distributions\nwith heterogeneous parameters and only one sample per distribution. We\nintroduce a convex objective that pools all $n$ samples to jointly learn all\n$n$ parameter vectors, and provide a unit-wise mean squared error bound that\nscales linearly with the metric entropy of the parameter space. For example,\nwhen the parameters are $s$-sparse linear combination of $k$ known vectors, the\nerror is $O(s\\log k/p)$. En route, we derive sufficient conditions for\ncompactly supported distributions to satisfy the logarithmic Sobolev\ninequality. As an application of the framework, our results enable consistent\nimputation of sparsely missing covariates.\n", "date": "2022-11-14", "categories": []}
{"id": "2309.08165", "title": "To Predict or to Reject: Causal Effect Estimation with Uncertainty on\n  Networked Data", "abstract": "  Due to the imbalanced nature of networked observational data, the causal\neffect predictions for some individuals can severely violate the\npositivity/overlap assumption, rendering unreliable estimations. Nevertheless,\nthis potential risk of individual-level treatment effect estimation on\nnetworked data has been largely under-explored. To create a more trustworthy\ncausal effect estimator, we propose the uncertainty-aware graph deep kernel\nlearning (GraphDKL) framework with Lipschitz constraint to model the prediction\nuncertainty with Gaussian process and identify unreliable estimations. To the\nbest of our knowledge, GraphDKL is the first framework to tackle the violation\nof positivity assumption when performing causal effect estimation with graphs.\nWith extensive experiments, we demonstrate the superiority of our proposed\nmethod in uncertainty-aware causal effect estimation on networked data.\n", "date": "2023-09-15", "categories": []}
{"id": "2309.08332", "title": "Estimation of Counterfactual Interventions under Uncertainties", "abstract": "  Counterfactual analysis is intuitively performed by humans on a daily basis\neg. \"What should I have done differently to get the loan approved?\". Such\ncounterfactual questions also steer the formulation of scientific hypotheses.\nMore formally it provides insights about potential improvements of a system by\ninferring the effects of hypothetical interventions into a past observation of\nthe system's behaviour which plays a prominent role in a variety of industrial\napplications. Due to the hypothetical nature of such analysis, counterfactual\ndistributions are inherently ambiguous. This ambiguity is particularly\nchallenging in continuous settings in which a continuum of explanations exist\nfor the same observation. In this paper, we address this problem by following a\nhierarchical Bayesian approach which explicitly models such uncertainty. In\nparticular, we derive counterfactual distributions for a Bayesian Warped\nGaussian Process thereby allowing for non-Gaussian distributions and\nnon-additive noise. We illustrate the properties our approach on a synthetic\nand on a semi-synthetic example and show its performance when used within an\nalgorithmic recourse downstream task.\n", "date": "2023-09-15", "categories": []}
{"id": "2309.08406", "title": "Constraint-Free Structure Learning with Smooth Acyclic Orientations", "abstract": "  The structure learning problem consists of fitting data generated by a\nDirected Acyclic Graph (DAG) to correctly reconstruct its arcs. In this\ncontext, differentiable approaches constrain or regularize the optimization\nproblem using a continuous relaxation of the acyclicity property. The\ncomputational cost of evaluating graph acyclicity is cubic on the number of\nnodes and significantly affects scalability. In this paper we introduce COSMO,\na constraint-free continuous optimization scheme for acyclic structure\nlearning. At the core of our method, we define a differentiable approximation\nof an orientation matrix parameterized by a single priority vector. Differently\nfrom previous work, our parameterization fits a smooth orientation matrix and\nthe resulting acyclic adjacency matrix without evaluating acyclicity at any\nstep. Despite the absence of explicit constraints, we prove that COSMO always\nconverges to an acyclic solution. In addition to being asymptotically faster,\nour empirical analysis highlights how COSMO performance on graph reconstruction\ncompares favorably with competing structure learning methods.\n", "date": "2023-09-15", "categories": []}
{"id": "2308.05969", "title": "Learning nonparametric DAGs with incremental information via high-order\n  HSIC", "abstract": "  Score-based methods for learning Bayesain networks(BN) aim to maximizing the\nglobal score functions. However, if local variables have direct and indirect\ndependence simultaneously, the global optimization on score functions misses\nedges between variables with indirect dependent relationship, of which scores\nare smaller than those with direct dependent relationship. In this paper, we\npresent an identifiability condition based on a determined subset of parents to\nidentify the underlying DAG. By the identifiability condition, we develop a\ntwo-phase algorithm namely optimal-tuning (OT) algorithm to locally amend the\nglobal optimization. In the optimal phase, an optimization problem based on\nfirst-order Hilbert-Schmidt independence criterion (HSIC) gives an estimated\nskeleton as the initial determined parents subset. In the tuning phase, the\nskeleton is locally tuned by deletion, addition and DAG-formalization\nstrategies using the theoretically proved incremental properties of high-order\nHSIC. Numerical experiments for different synthetic datasets and real-world\ndatasets show that the OT algorithm outperforms existing methods. Especially in\nSigmoid Mix model with the size of the graph being ${\\rm\\bf d=40}$, the\nstructure intervention distance (SID) of the OT algorithm is 329.7 smaller than\nthe one obtained by CAM, which indicates that the graph estimated by the OT\nalgorithm misses fewer edges compared with CAM.Source code of the OT algorithm\nis available at https://github.com/YafeiannWang/optimal-tune-algorithm.\n", "date": "2023-08-11", "categories": []}
{"id": "2309.06739", "title": "MCNS: Mining Causal Natural Structures Inside Time Series via A Novel\n  Internal Causality Scheme", "abstract": "  Causal inference permits us to discover covert relationships of various\nvariables in time series. However, in most existing works, the variables\nmentioned above are the dimensions. The causality between dimensions could be\ncursory, which hinders the comprehension of the internal relationship and the\nbenefit of the causal graph to the neural networks (NNs). In this paper, we\nfind that causality exists not only outside but also inside the time series\nbecause it reflects a succession of events in the real world. It inspires us to\nseek the relationship between internal subsequences. However, the challenges\nare the hardship of discovering causality from subsequences and utilizing the\ncausal natural structures to improve NNs. To address these challenges, we\npropose a novel framework called Mining Causal Natural Structure (MCNS), which\nis automatic and domain-agnostic and helps to find the causal natural\nstructures inside time series via the internal causality scheme. We evaluate\nthe MCNS framework and impregnation NN with MCNS on time series classification\ntasks. Experimental results illustrate that our impregnation, by refining\nattention, shape selection classification, and pruning datasets, drives NN,\neven the data itself preferable accuracy and interpretability. Besides, MCNS\nprovides an in-depth, solid summary of the time series and datasets.\n", "date": "2023-09-13", "categories": []}
{"id": "2209.09423", "title": "Fairness and robustness in anti-causal prediction", "abstract": "  Robustness to distribution shift and fairness have independently emerged as\ntwo important desiderata required of modern machine learning models. While\nthese two desiderata seem related, the connection between them is often unclear\nin practice. Here, we discuss these connections through a causal lens, focusing\non anti-causal prediction tasks, where the input to a classifier (e.g., an\nimage) is assumed to be generated as a function of the target label and the\nprotected attribute. By taking this perspective, we draw explicit connections\nbetween a common fairness criterion - separation - and a common notion of\nrobustness - risk invariance. These connections provide new motivation for\napplying the separation criterion in anticausal settings, and inform old\ndiscussions regarding fairness-performance tradeoffs. In addition, our findings\nsuggest that robustness-motivated approaches can be used to enforce separation,\nand that they often work better in practice than methods designed to directly\nenforce separation. Using a medical dataset, we empirically validate our\nfindings on the task of detecting pneumonia from X-rays, in a setting where\ndifferences in prevalence across sex groups motivates a fairness mitigation.\nOur findings highlight the importance of considering causal structure when\nchoosing and enforcing fairness criteria.\n", "date": "2022-09-20", "categories": []}
{"id": "2309.05107", "title": "Nonlinear Granger Causality using Kernel Ridge Regression", "abstract": "  I introduce a novel algorithm and accompanying Python library, named\nmlcausality, designed for the identification of nonlinear Granger causal\nrelationships. This novel algorithm uses a flexible plug-in architecture that\nenables researchers to employ any nonlinear regressor as the base prediction\nmodel. Subsequently, I conduct a comprehensive performance analysis of\nmlcausality when the prediction regressor is the kernel ridge regressor with\nthe radial basis function kernel. The results demonstrate that mlcausality\nemploying kernel ridge regression achieves competitive AUC scores across a\ndiverse set of simulated data. Furthermore, mlcausality with kernel ridge\nregression yields more finely calibrated $p$-values in comparison to rival\nalgorithms. This enhancement enables mlcausality to attain superior accuracy\nscores when using intuitive $p$-value-based thresholding criteria. Finally,\nmlcausality with the kernel ridge regression exhibits significantly reduced\ncomputation times compared to existing nonlinear Granger causality algorithms.\nIn fact, in numerous instances, this innovative approach achieves superior\nsolutions within computational timeframes that are an order of magnitude\nshorter than those required by competing algorithms.\n", "date": "2023-09-10", "categories": []}
{"id": "2306.14115", "title": "Towards Trustworthy Explanation: On Causal Rationalization", "abstract": "  With recent advances in natural language processing, rationalization becomes\nan essential self-explaining diagram to disentangle the black box by selecting\na subset of input texts to account for the major variation in prediction. Yet,\nexisting association-based approaches on rationalization cannot identify true\nrationales when two or more snippets are highly inter-correlated and thus\nprovide a similar contribution to prediction accuracy, so-called spuriousness.\nTo address this limitation, we novelly leverage two causal desiderata,\nnon-spuriousness and efficiency, into rationalization from the causal inference\nperspective. We formally define a series of probabilities of causation based on\na newly proposed structural causal model of rationalization, with its\ntheoretical identification established as the main component of learning\nnecessary and sufficient rationales. The superior performance of the proposed\ncausal rationalization is demonstrated on real-world review and medical\ndatasets with extensive experiments compared to state-of-the-art methods.\n", "date": "2023-06-25", "categories": []}
{"id": "2309.04676", "title": "Flexible and Robust Counterfactual Explanations with Minimal Satisfiable\n  Perturbations", "abstract": "  Counterfactual explanations (CFEs) exemplify how to minimally modify a\nfeature vector to achieve a different prediction for an instance. CFEs can\nenhance informational fairness and trustworthiness, and provide suggestions for\nusers who receive adverse predictions. However, recent research has shown that\nmultiple CFEs can be offered for the same instance or instances with slight\ndifferences. Multiple CFEs provide flexible choices and cover diverse\ndesiderata for user selection. However, individual fairness and model\nreliability will be damaged if unstable CFEs with different costs are returned.\nExisting methods fail to exploit flexibility and address the concerns of\nnon-robustness simultaneously. To address these issues, we propose a\nconceptually simple yet effective solution named Counterfactual Explanations\nwith Minimal Satisfiable Perturbations (CEMSP). Specifically, CEMSP constrains\nchanging values of abnormal features with the help of their semantically\nmeaningful normal ranges. For efficiency, we model the problem as a Boolean\nsatisfiability problem to modify as few features as possible. Additionally,\nCEMSP is a general framework and can easily accommodate more practical\nrequirements, e.g., casualty and actionability. Compared to existing methods,\nwe conduct comprehensive experiments on both synthetic and real-world datasets\nto demonstrate that our method provides more robust explanations while\npreserving flexibility.\n", "date": "2023-09-09", "categories": []}
{"id": "2309.05264", "title": "Enabling Runtime Verification of Causal Discovery Algorithms with\n  Automated Conditional Independence Reasoning (Extended Version)", "abstract": "  Causal discovery is a powerful technique for identifying causal relationships\namong variables in data. It has been widely used in various applications in\nsoftware engineering. Causal discovery extensively involves conditional\nindependence (CI) tests. Hence, its output quality highly depends on the\nperformance of CI tests, which can often be unreliable in practice. Moreover,\nprivacy concerns arise when excessive CI tests are performed.\n  Despite the distinct nature between unreliable and excessive CI tests, this\npaper identifies a unified and principled approach to addressing both of them.\nGenerally, CI statements, the outputs of CI tests, adhere to Pearl's axioms,\nwhich are a set of well-established integrity constraints on conditional\nindependence. Hence, we can either detect erroneous CI statements if they\nviolate Pearl's axioms or prune excessive CI statements if they are logically\nentailed by Pearl's axioms. Holistically, both problems boil down to reasoning\nabout the consistency of CI statements under Pearl's axioms (referred to as CIR\nproblem).\n  We propose a runtime verification tool called CICheck, designed to harden\ncausal discovery algorithms from reliability and privacy perspectives. CICheck\nemploys a sound and decidable encoding scheme that translates CIR into SMT\nproblems. To solve the CIR problem efficiently, CICheck introduces a four-stage\ndecision procedure with three lightweight optimizations that actively prove or\nrefute consistency, and only resort to costly SMT-based reasoning when\nnecessary. Based on the decision procedure to CIR, CICheck includes two\nvariants: ED-CICheck and ED-CICheck, which detect erroneous CI tests (to\nenhance reliability) and prune excessive CI tests (to enhance privacy),\nrespectively. [abridged due to length limit]\n", "date": "2023-09-11", "categories": []}
{"id": "2302.12341", "title": "Rank-Based Causal Discovery for Post-Nonlinear Models", "abstract": "  Learning causal relationships from empirical observations is a central task\nin scientific research. A common method is to employ structural causal models\nthat postulate noisy functional relations among a set of interacting variables.\nTo ensure unique identifiability of causal directions, researchers consider\nrestricted subclasses of structural causal models. Post-nonlinear (PNL) causal\nmodels constitute one of the most flexible options for such restricted\nsubclasses, containing in particular the popular additive noise models as a\nfurther subclass. However, learning PNL models is not well studied beyond the\nbivariate case. The existing methods learn non-linear functional relations by\nminimizing residual dependencies and subsequently test independence from\nresiduals to determine causal orientations. However, these methods can be prone\nto overfitting and, thus, difficult to tune appropriately in practice. As an\nalternative, we propose a new approach for PNL causal discovery that uses\nrank-based methods to estimate the functional parameters. This new approach\nexploits natural invariances of PNL models and disentangles the estimation of\nthe non-linear functions from the independence tests used to find causal\norientations. We prove consistency of our method and validate our results in\nnumerical experiments.\n", "date": "2023-02-23", "categories": []}
{"id": "2309.02571", "title": "Causal Structure Recovery of Linear Dynamical Systems: An FFT based\n  Approach", "abstract": "  Learning causal effects from data is a fundamental and well-studied problem\nacross science, especially when the cause-effect relationship is static in\nnature. However, causal effect is less explored when there are dynamical\ndependencies, i.e., when dependencies exist between entities across time.\nIdentifying dynamic causal effects from time-series observations is\ncomputationally expensive when compared to the static scenario. We demonstrate\nthat the computational complexity of recovering the causation structure for the\nvector auto-regressive (VAR) model is $O(Tn^3N^2)$, where $n$ is the number of\nnodes, $T$ is the number of samples, and $N$ is the largest time-lag in the\ndependency between entities. We report a method, with a reduced complexity of\n$O(Tn^3 \\log N)$, to recover the causation structure to obtain frequency-domain\n(FD) representations of time-series. Since FFT accumulates all the time\ndependencies on every frequency, causal inference can be performed efficiently\nby considering the state variables as random variables at any given frequency.\nWe additionally show that, for systems with interactions that are LTI,\ndo-calculus machinery can be realized in the FD resulting in versions of the\nclassical single-door (with cycles), front and backdoor criteria. We\ndemonstrate, for a large class of problems, graph reconstruction using\nmultivariate Wiener projections results in a significant computational\nadvantage with $O(n)$ complexity over reconstruction algorithms such as the PC\nalgorithm which has $O(n^q)$ complexity, where $q$ is the maximum neighborhood\nsize. This advantage accrues due to some remarkable properties of the phase\nresponse of the frequency-dependent Wiener coefficients which is not present in\nany time-domain approach.\n", "date": "2023-09-05", "categories": []}
{"id": "2309.04470", "title": "On the Actionability of Outcome Prediction", "abstract": "  Predicting future outcomes is a prevalent application of machine learning in\nsocial impact domains. Examples range from predicting student success in\neducation to predicting disease risk in healthcare. Practitioners recognize\nthat the ultimate goal is not just to predict but to act effectively.\nIncreasing evidence suggests that relying on outcome predictions for downstream\ninterventions may not have desired results.\n  In most domains there exists a multitude of possible interventions for each\nindividual, making the challenge of taking effective action more acute. Even\nwhen causal mechanisms connecting the individual's latent states to outcomes is\nwell understood, in any given instance (a specific student or patient),\npractitioners still need to infer -- from budgeted measurements of latent\nstates -- which of many possible interventions will be most effective for this\nindividual. With this in mind, we ask: when are accurate predictors of outcomes\nhelpful for identifying the most suitable intervention?\n  Through a simple model encompassing actions, latent states, and measurements,\nwe demonstrate that pure outcome prediction rarely results in the most\neffective policy for taking actions, even when combined with other\nmeasurements. We find that except in cases where there is a single decisive\naction for improving the outcome, outcome prediction never maximizes \"action\nvalue\", the utility of taking actions. Making measurements of actionable latent\nstates, where specific actions lead to desired outcomes, considerably enhances\nthe action value compared to outcome prediction, and the degree of improvement\ndepends on action costs and the outcome model. This analysis emphasizes the\nneed to go beyond generic outcome prediction in interventional settings by\nincorporating knowledge of plausible actions and latent states.\n", "date": "2023-09-08", "categories": []}
{"id": "2212.04031", "title": "On Root Cause Localization and Anomaly Mitigation through Causal\n  Inference", "abstract": "  Due to a wide spectrum of applications in the real world, such as security,\nfinancial surveillance, and health risk, various deep anomaly detection models\nhave been proposed and achieved state-of-the-art performance. However, besides\nbeing effective, in practice, the practitioners would further like to know what\ncauses the abnormal outcome and how to further fix it. In this work, we propose\nRootCLAM, which aims to achieve Root Cause Localization and Anomaly Mitigation\nfrom a causal perspective. Especially, we formulate anomalies caused by\nexternal interventions on the normal causal mechanism and aim to locate the\nabnormal features with external interventions as root causes. After that, we\nfurther propose an anomaly mitigation approach that aims to recommend\nmitigation actions on abnormal features to revert the abnormal outcomes such\nthat the counterfactuals guided by the causal mechanism are normal. Experiments\non three datasets show that our approach can locate the root causes and further\nflip the abnormal labels.\n", "date": "2022-12-08", "categories": []}
{"id": "2308.00904", "title": "VLUCI: Variational Learning of Unobserved Confounders for Counterfactual\n  Inference", "abstract": "  Causal inference plays a vital role in diverse domains like epidemiology,\nhealthcare, and economics. De-confounding and counterfactual prediction in\nobservational data has emerged as a prominent concern in causal inference\nresearch. While existing models tackle observed confounders, the presence of\nunobserved confounders remains a significant challenge, distorting causal\ninference and impacting counterfactual outcome accuracy. To address this, we\npropose a novel variational learning model of unobserved confounders for\ncounterfactual inference (VLUCI), which generates the posterior distribution of\nunobserved confounders. VLUCI relaxes the unconfoundedness assumption often\noverlooked by most causal inference methods. By disentangling observed and\nunobserved confounders, VLUCI constructs a doubly variational inference model\nto approximate the distribution of unobserved confounders, which are used for\ninferring more accurate counterfactual outcomes. Extensive experiments on\nsynthetic and semi-synthetic datasets demonstrate VLUCI's superior performance\nin inferring unobserved confounders. It is compatible with state-of-the-art\ncounterfactual inference models, significantly improving inference accuracy at\nboth group and individual levels. Additionally, VLUCI provides confidence\nintervals for counterfactual outcomes, aiding decision-making in risk-sensitive\ndomains. We further clarify the considerations when applying VLUCI to cases\nwhere unobserved confounders don't strictly conform to our model assumptions\nusing the public IHDP dataset as an example, highlighting the practical\nadvantages of VLUCI.\n", "date": "2023-08-02", "categories": []}
{"id": "2309.03730", "title": "A Causal Perspective on Loan Pricing: Investigating the Impacts of\n  Selection Bias on Identifying Bid-Response Functions", "abstract": "  In lending, where prices are specific to both customers and products, having\na well-functioning personalized pricing policy in place is essential to\neffective business making. Typically, such a policy must be derived from\nobservational data, which introduces several challenges. While the problem of\n``endogeneity'' is prominently studied in the established pricing literature,\nthe problem of selection bias (or, more precisely, bid selection bias) is not.\nWe take a step towards understanding the effects of selection bias by posing\npricing as a problem of causal inference. Specifically, we consider the\nreaction of a customer to price a treatment effect. In our experiments, we\nsimulate varying levels of selection bias on a semi-synthetic dataset on\nmortgage loan applications in Belgium. We investigate the potential of\nparametric and nonparametric methods for the identification of individual\nbid-response functions. Our results illustrate how conventional methods such as\nlogistic regression and neural networks suffer adversely from selection bias.\nIn contrast, we implement state-of-the-art methods from causal machine learning\nand show their capability to overcome selection bias in pricing data.\n", "date": "2023-09-07", "categories": []}
{"id": "2211.06919", "title": "Towards Privacy-Aware Causal Structure Learning in Federated Setting", "abstract": "  Causal structure learning has been extensively studied and widely used in\nmachine learning and various applications. To achieve an ideal performance,\nexisting causal structure learning algorithms often need to centralize a large\namount of data from multiple data sources. However, in the privacy-preserving\nsetting, it is impossible to centralize data from all sources and put them\ntogether as a single dataset. To preserve data privacy, federated learning as a\nnew learning paradigm has attracted much attention in machine learning in\nrecent years. In this paper, we study a privacy-aware causal structure learning\nproblem in the federated setting and propose a novel Federated PC (FedPC)\nalgorithm with two new strategies for preserving data privacy without\ncentralizing data. Specifically, we first propose a novel layer-wise\naggregation strategy for a seamless adaptation of the PC algorithm into the\nfederated learning paradigm for federated skeleton learning, then we design an\neffective strategy for learning consistent separation sets for federated edge\norientation. The extensive experiments validate that FedPC is effective for\ncausal structure learning in a federated learning setting.\n", "date": "2022-11-13", "categories": []}
{"id": "2305.03859", "title": "Open problems in causal structure learning: A case study of COVID-19 in\n  the UK", "abstract": "  Causal machine learning (ML) algorithms recover graphical structures that\ntell us something about cause-and-effect relationships. The causal\nrepresentation praovided by these algorithms enables transparency and\nexplainability, which is necessary for decision making in critical real-world\nproblems. Yet, causal ML has had limited impact in practice compared to\nassociational ML. This paper investigates the challenges of causal ML with\napplication to COVID-19 UK pandemic data. We collate data from various public\nsources and investigate what the various structure learning algorithms learn\nfrom these data. We explore the impact of different data formats on algorithms\nspanning different classes of learning, and assess the results produced by each\nalgorithm, and groups of algorithms, in terms of graphical structure, model\ndimensionality, sensitivity analysis, confounding variables, predictive and\ninterventional inference. We use these results to highlight open problems in\ncausal structure learning and directions for future research. To facilitate\nfuture work, we make all graphs, models, data sets, and source code publicly\navailable online.\n", "date": "2023-05-05", "categories": []}
{"id": "2309.01392", "title": "Differentiable Bayesian Structure Learning with Acyclicity Assurance", "abstract": "  Score-based approaches in the structure learning task are thriving because of\ntheir scalability. Continuous relaxation has been the key reason for this\nadvancement. Despite achieving promising outcomes, most of these methods are\nstill struggling to ensure that the graphs generated from the latent space are\nacyclic by minimizing a defined score. There has also been another trend of\npermutation-based approaches, which concern the search for the topological\nordering of the variables in the directed acyclic graph in order to limit the\nsearch space of the graph. In this study, we propose an alternative approach\nfor strictly constraining the acyclicty of the graphs with an integration of\nthe knowledge from the topological orderings. Our approach can reduce inference\ncomplexity while ensuring the structures of the generated graphs to be acyclic.\nOur empirical experiments with simulated and real-world data show that our\napproach can outperform related Bayesian score-based approaches.\n", "date": "2023-09-04", "categories": []}
{"id": "2309.03092", "title": "Establishing Markov Equivalence in Cyclic Directed Graphs", "abstract": "  We present a new, efficient procedure to establish Markov equivalence between\ndirected graphs that may or may not contain cycles under the\n\\textit{d}-separation criterion. It is based on the Cyclic Equivalence Theorem\n(CET) in the seminal works on cyclic models by Thomas Richardson in the mid\n'90s, but now rephrased from an ancestral perspective. The resulting\ncharacterization leads to a procedure for establishing Markov equivalence\nbetween graphs that no longer requires tests for d-separation, leading to a\nsignificantly reduced algorithmic complexity. The conceptually simplified\ncharacterization may help to reinvigorate theoretical research towards sound\nand complete cyclic discovery in the presence of latent confounders. This\nversion includes a correction to rule (iv) in Theorem 1, and the subsequent\nadjustment in part 2 of Algorithm 2.\n", "date": "2023-09-01", "categories": []}
{"id": "2301.13348", "title": "A Reinforcement Learning Framework for Dynamic Mediation Analysis", "abstract": "  Mediation analysis learns the causal effect transmitted via mediator\nvariables between treatments and outcomes and receives increasing attention in\nvarious scientific domains to elucidate causal relations. Most existing works\nfocus on point-exposure studies where each subject only receives one treatment\nat a single time point. However, there are a number of applications (e.g.,\nmobile health) where the treatments are sequentially assigned over time and the\ndynamic mediation effects are of primary interest. Proposing a reinforcement\nlearning (RL) framework, we are the first to evaluate dynamic mediation effects\nin settings with infinite horizons. We decompose the average treatment effect\ninto an immediate direct effect, an immediate mediation effect, a delayed\ndirect effect, and a delayed mediation effect. Upon the identification of each\neffect component, we further develop robust and semi-parametrically efficient\nestimators under the RL framework to infer these causal effects. The superior\nperformance of the proposed method is demonstrated through extensive numerical\nstudies, theoretical results, and an analysis of a mobile health dataset.\n", "date": "2023-01-31", "categories": []}
{"id": "2308.11676", "title": "Does Misclassifying Non-confounding Covariates as Confounders Affect the\n  Causal Inference within the Potential Outcomes Framework?", "abstract": "  The Potential Outcome Framework (POF) plays a prominent role in the field of\ncausal inference. Most causal inference models based on the POF (CIMs-POF) are\ndesigned for eliminating confounding bias and default to an underlying\nassumption of Confounding Covariates. This assumption posits that the\ncovariates consist solely of confounders. However, the assumption of\nConfounding Covariates is challenging to maintain in practice, particularly\nwhen dealing with high-dimensional covariates. While certain methods have been\nproposed to differentiate the distinct components of covariates prior to\nconducting causal inference, the consequences of treating non-confounding\ncovariates as confounders remain unclear. This ambiguity poses a potential risk\nwhen conducting causal inference in practical scenarios. In this paper, we\npresent a unified graphical framework for the CIMs-POF, which greatly enhances\nthe comprehension of these models' underlying principles. Using this graphical\nframework, we quantitatively analyze the extent to which the inference\nperformance of CIMs-POF is influenced when incorporating various types of\nnon-confounding covariates, such as instrumental variables, mediators,\ncolliders, and adjustment variables. The key findings are: in the task of\neliminating confounding bias, the optimal scenario is for the covariates to\nexclusively encompass confounders; in the subsequent task of inferring\ncounterfactual outcomes, the adjustment variables contribute to more accurate\ninferences. Furthermore, extensive experiments conducted on synthetic datasets\nconsistently validate these theoretical conclusions.\n", "date": "2023-08-22", "categories": []}
{"id": "2309.00805", "title": "Fairness Implications of Heterogeneous Treatment Effect Estimation with\n  Machine Learning Methods in Policy-making", "abstract": "  Causal machine learning methods which flexibly generate heterogeneous\ntreatment effect estimates could be very useful tools for governments trying to\nmake and implement policy. However, as the critical artificial intelligence\nliterature has shown, governments must be very careful of unintended\nconsequences when using machine learning models. One way to try and protect\nagainst unintended bad outcomes is with AI Fairness methods which seek to\ncreate machine learning models where sensitive variables like race or gender do\nnot influence outcomes. In this paper we argue that standard AI Fairness\napproaches developed for predictive machine learning are not suitable for all\ncausal machine learning applications because causal machine learning generally\n(at least so far) uses modelling to inform a human who is the ultimate\ndecision-maker while AI Fairness approaches assume a model that is making\ndecisions directly. We define these scenarios as indirect and direct\ndecision-making respectively and suggest that policy-making is best seen as a\njoint decision where the causal machine learning model usually only has\nindirect power. We lay out a definition of fairness for this scenario - a model\nthat provides the information a decision-maker needs to accurately make a value\njudgement about just policy outcomes - and argue that the complexity of causal\nmachine learning models can make this difficult to achieve. The solution here\nis not traditional AI Fairness adjustments, but careful modelling and awareness\nof some of the decision-making biases that these methods might encourage which\nwe describe.\n", "date": "2023-09-02", "categories": []}
{"id": "2309.02287", "title": "Optimal Observation-Intervention Trade-Off in Optimisation Problems with\n  Causal Structure", "abstract": "  We consider the problem of optimising an expensive-to-evaluate grey-box\nobjective function, within a finite budget, where known side-information exists\nin the form of the causal structure between the design variables. Standard\nblack-box optimisation ignores the causal structure, often making it\ninefficient and expensive. The few existing methods that consider the causal\nstructure are myopic and do not fully accommodate the observation-intervention\ntrade-off that emerges when estimating causal effects. In this paper, we show\nthat the observation-intervention trade-off can be formulated as a non-myopic\noptimal stopping problem which permits an efficient solution. We give\ntheoretical results detailing the structure of the optimal stopping times and\ndemonstrate the generality of our approach by showing that it can be integrated\nwith existing causal Bayesian optimisation algorithms. Experimental results\nshow that our formulation can enhance existing algorithms on real and synthetic\nbenchmarks.\n", "date": "2023-09-05", "categories": []}
{"id": "2308.16879", "title": "Adaptation Speed Analysis for Fairness-aware Causal Models", "abstract": "  For example, in machine translation tasks, to achieve bidirectional\ntranslation between two languages, the source corpus is often used as the\ntarget corpus, which involves the training of two models with opposite\ndirections. The question of which one can adapt most quickly to a domain shift\nis of significant importance in many fields. Specifically, consider an original\ndistribution p that changes due to an unknown intervention, resulting in a\nmodified distribution p*. In aligning p with p*, several factors can affect the\nadaptation rate, including the causal dependencies between variables in p. In\nreal-life scenarios, however, we have to consider the fairness of the training\nprocess, and it is particularly crucial to involve a sensitive variable (bias)\npresent between a cause and an effect variable. To explore this scenario, we\nexamine a simple structural causal model (SCM) with a cause-bias-effect\nstructure, where variable A acts as a sensitive variable between cause (X) and\neffect (Y). The two models, respectively, exhibit consistent and contrary\ncause-effect directions in the cause-bias-effect SCM. After conducting unknown\ninterventions on variables within the SCM, we can simulate some kinds of domain\nshifts for analysis. We then compare the adaptation speeds of two models across\nfour shift scenarios. Additionally, we prove the connection between the\nadaptation speeds of the two models across all interventions.\n", "date": "2023-08-31", "categories": []}
{"id": "2308.15883", "title": "\"Would life be more interesting if I were in AI?\" Answering\n  Counterfactuals based on Probabilistic Inductive Logic Programming", "abstract": "  Probabilistic logic programs are logic programs where some facts hold with a\nspecified probability. Here, we investigate these programs with a causal\nframework that allows counterfactual queries. Learning the program structure\nfrom observational data is usually done through heuristic search relying on\nstatistical tests. However, these statistical tests lack information about the\ncausal mechanism generating the data, which makes it unfeasible to use the\nresulting programs for counterfactual reasoning. To address this, we propose a\nlanguage fragment that allows reconstructing a program from its induced\ndistribution. This further enables us to learn programs supporting\ncounterfactual queries.\n", "date": "2023-08-30", "categories": []}
{"id": "2308.14895", "title": "Conformal Meta-learners for Predictive Inference of Individual Treatment\n  Effects", "abstract": "  We investigate the problem of machine learning-based (ML) predictive\ninference on individual treatment effects (ITEs). Previous work has focused\nprimarily on developing ML-based meta-learners that can provide point estimates\nof the conditional average treatment effect (CATE); these are model-agnostic\napproaches for combining intermediate nuisance estimates to produce estimates\nof CATE. In this paper, we develop conformal meta-learners, a general framework\nfor issuing predictive intervals for ITEs by applying the standard conformal\nprediction (CP) procedure on top of CATE meta-learners. We focus on a broad\nclass of meta-learners based on two-stage pseudo-outcome regression and develop\na stochastic ordering framework to study their validity. We show that inference\nwith conformal meta-learners is marginally valid if their (pseudo outcome)\nconformity scores stochastically dominate oracle conformity scores evaluated on\nthe unobserved ITEs. Additionally, we prove that commonly used CATE\nmeta-learners, such as the doubly-robust learner, satisfy a model- and\ndistribution-free stochastic (or convex) dominance condition, making their\nconformal inferences valid for practically-relevant levels of target coverage.\nWhereas existing procedures conduct inference on nuisance parameters (i.e.,\npotential outcomes) via weighted CP, conformal meta-learners enable direct\ninference on the target parameter (ITE). Numerical experiments show that\nconformal meta-learners provide valid intervals with competitive efficiency\nwhile retaining the favorable point estimation properties of CATE\nmeta-learners.\n", "date": "2023-08-28", "categories": []}
{"id": "2308.13559", "title": "Machine Unlearning for Causal Inference", "abstract": "  Machine learning models play a vital role in making predictions and deriving\ninsights from data and are being increasingly used for causal inference. To\npreserve user privacy, it is important to enable the model to forget some of\nits learning/captured information about a given user (machine unlearning). This\npaper introduces the concept of machine unlearning for causal inference,\nparticularly propensity score matching and treatment effect estimation, which\naims to refine and improve the performance of machine learning models for\ncausal analysis given the above unlearning requirements. The paper presents a\nmethodology for machine unlearning using a neural network-based propensity\nscore model. The dataset used in the study is the Lalonde dataset, a widely\nused dataset for evaluating the effectiveness i.e. the treatment effect of job\ntraining programs. The methodology involves training an initial propensity\nscore model on the original dataset and then creating forget sets by\nselectively removing instances, as well as matched instance pairs. based on\npropensity score matching. These forget sets are used to evaluate the retrained\nmodel, allowing for the elimination of unwanted associations. The actual\nretraining of the model is performed using the retain set. The experimental\nresults demonstrate the effectiveness of the machine unlearning approach. The\ndistribution and histogram analysis of propensity scores before and after\nunlearning provide insights into the impact of the unlearning process on the\ndata. This study represents the first attempt to apply machine unlearning\ntechniques to causal inference.\n", "date": "2023-08-24", "categories": []}
{"id": "2308.13067", "title": "Causal Parrots: Large Language Models May Talk Causality But Are Not\n  Causal", "abstract": "  Some argue scale is all what is needed to achieve AI, covering even causal\nmodels. We make it clear that large language models (LLMs) cannot be causal and\ngive reason onto why sometimes we might feel otherwise. To this end, we define\nand exemplify a new subgroup of Structural Causal Model (SCM) that we call meta\nSCM which encode causal facts about other SCM within their variables. We\nconjecture that in the cases where LLM succeed in doing causal inference,\nunderlying was a respective meta SCM that exposed correlations between causal\nfacts in natural language on whose data the LLM was ultimately trained. If our\nhypothesis holds true, then this would imply that LLMs are like parrots in that\nthey simply recite the causal knowledge embedded in the data. Our empirical\nanalysis provides favoring evidence that current LLMs are even weak `causal\nparrots.'\n", "date": "2023-08-24", "categories": []}
{"id": "2308.12888", "title": "Inducing Causal Structure for Abstractive Text Summarization", "abstract": "  The mainstream of data-driven abstractive summarization models tends to\nexplore the correlations rather than the causal relationships. Among such\ncorrelations, there can be spurious ones which suffer from the language prior\nlearned from the training corpus and therefore undermine the overall\neffectiveness of the learned model. To tackle this issue, we introduce a\nStructural Causal Model (SCM) to induce the underlying causal structure of the\nsummarization data. We assume several latent causal factors and non-causal\nfactors, representing the content and style of the document and summary.\nTheoretically, we prove that the latent factors in our SCM can be identified by\nfitting the observed training data under certain conditions. On the basis of\nthis, we propose a Causality Inspired Sequence-to-Sequence model (CI-Seq2Seq)\nto learn the causal representations that can mimic the causal factors, guiding\nus to pursue causal information for summary generation. The key idea is to\nreformulate the Variational Auto-encoder (VAE) to fit the joint distribution of\nthe document and summary variables from the training corpus. Experimental\nresults on two widely used text summarization datasets demonstrate the\nadvantages of our approach.\n", "date": "2023-08-24", "categories": []}
{"id": "2308.11781", "title": "Addressing Dynamic and Sparse Qualitative Data: A Hilbert Space\n  Embedding of Categorical Variables", "abstract": "  We propose a novel framework for incorporating qualitative data into\nquantitative models for causal estimation. Previous methods use categorical\nvariables derived from qualitative data to build quantitative models. However,\nthis approach can lead to data-sparse categories and yield inconsistent\n(asymptotically biased) and imprecise (finite sample biased) estimates if the\nqualitative information is dynamic and intricate. We use functional analysis to\ncreate a more nuanced and flexible framework. We embed the observed categories\ninto a latent Baire space and introduce a continuous linear map -- a Hilbert\nspace embedding -- from the Baire space of categories to a Reproducing Kernel\nHilbert Space (RKHS) of representation functions. Through the Riesz\nrepresentation theorem, we establish that the canonical treatment of\ncategorical variables in causal models can be transformed into an identified\nstructure in the RKHS. Transfer learning acts as a catalyst to streamline\nestimation -- embeddings from traditional models are paired with the kernel\ntrick to form the Hilbert space embedding. We validate our model through\ncomprehensive simulation evidence and demonstrate its relevance in a real-world\nstudy that contrasts theoretical predictions from economics and psychology in\nan e-commerce marketplace. The results confirm the superior performance of our\nmodel, particularly in scenarios where qualitative information is nuanced and\ncomplex.\n", "date": "2023-08-22", "categories": []}
{"id": "2308.11043", "title": "Spurious Correlations and Where to Find Them", "abstract": "  Spurious correlations occur when a model learns unreliable features from the\ndata and are a well-known drawback of data-driven learning. Although there are\nseveral algorithms proposed to mitigate it, we are yet to jointly derive the\nindicators of spurious correlations. As a result, the solutions built upon\nstandalone hypotheses fail to beat simple ERM baselines. We collect some of the\ncommonly studied hypotheses behind the occurrence of spurious correlations and\ninvestigate their influence on standard ERM baselines using synthetic datasets\ngenerated from causal graphs. Subsequently, we observe patterns connecting\nthese hypotheses and model design choices.\n", "date": "2023-08-21", "categories": []}
{"id": "2209.03013", "title": "Quantitative probing: Validating causal models using quantitative domain\n  knowledge", "abstract": "  We present quantitative probing as a model-agnostic framework for validating\ncausal models in the presence of quantitative domain knowledge. The method is\nconstructed as an analogue of the train/test split in correlation-based machine\nlearning and as an enhancement of current causal validation strategies that are\nconsistent with the logic of scientific discovery. The effectiveness of the\nmethod is illustrated using Pearl's sprinkler example, before a thorough\nsimulation-based investigation is conducted. Limits of the technique are\nidentified by studying exemplary failing scenarios, which are furthermore used\nto propose a list of topics for future research and improvements of the\npresented version of quantitative probing. The code for integrating\nquantitative probing into causal analysis, as well as the code for the\npresented simulation-based studies of the effectiveness of quantitative probing\nis provided in two separate open-source Python packages.\n", "date": "2022-09-07", "categories": []}
{"id": "2308.08556", "title": "Causally Linking Health Application Data and Personal Information\n  Management Tools", "abstract": "  The proliferation of consumer health devices such as smart watches, sleep\nmonitors, smart scales, etc, in many countries, has not only led to growing\ninterest in health monitoring, but also to the development of a countless\nnumber of ``smart'' applications to support the exploration of such data by\nmembers of the general public, sometimes with integration into professional\nhealth services. While a variety of health data streams has been made available\nby such devices to users, these streams are often presented as separate\ntime-series visualizations, in which the potential relationships between health\nvariables are not explicitly made visible. Furthermore, despite the fact that\nother aspects of life, such as work and social connectivity, have become\nincreasingly digitised, health and well-being applications make little use of\nthe potentially useful contextual information provided by widely used personal\ninformation management tools, such as shared calendar and email systems. This\npaper presents a framework for the integration of these diverse data sources,\nanalytic and visualization tools, with inference methods and graphical user\ninterfaces to help users by highlighting causal connections among such\ntime-series.\n", "date": "2023-08-11", "categories": []}
{"id": "2308.09066", "title": "Uplift Modeling: from Causal Inference to Personalization", "abstract": "  Uplift modeling is a collection of machine learning techniques for estimating\ncausal effects of a treatment at the individual or subgroup levels. Over the\nlast years, causality and uplift modeling have become key trends in\npersonalization at online e-commerce platforms, enabling the selection of the\nbest treatment for each user in order to maximize the target business metric.\nUplift modeling can be particularly useful for personalized promotional\ncampaigns, where the potential benefit caused by a promotion needs to be\nweighed against the potential costs. In this tutorial we will cover basic\nconcepts of causality and introduce the audience to state-of-the-art techniques\nin uplift modeling. We will discuss the advantages and the limitations of\ndifferent approaches and dive into the unique setup of constrained uplift\nmodeling. Finally, we will present real-life applications and discuss\nchallenges in implementing these models in production.\n", "date": "2023-08-17", "categories": []}
{"id": "2209.04744", "title": "Active Learning for Optimal Intervention Design in Causal Models", "abstract": "  Sequential experimental design to discover interventions that achieve a\ndesired outcome is a key problem in various domains including science,\nengineering and public policy. When the space of possible interventions is\nlarge, making an exhaustive search infeasible, experimental design strategies\nare needed. In this context, encoding the causal relationships between the\nvariables, and thus the effect of interventions on the system, is critical for\nidentifying desirable interventions more efficiently. Here, we develop a causal\nactive learning strategy to identify interventions that are optimal, as\nmeasured by the discrepancy between the post-interventional mean of the\ndistribution and a desired target mean. The approach employs a Bayesian update\nfor the causal model and prioritizes interventions using a carefully designed,\ncausally informed acquisition function. This acquisition function is evaluated\nin closed form, allowing for fast optimization. The resulting algorithms are\ntheoretically grounded with information-theoretic bounds and provable\nconsistency results for linear causal models with known causal graph. We apply\nour approach to both synthetic data and single-cell transcriptomic data from\nPerturb-CITE-seq experiments to identify optimal perturbations that induce a\nspecific cell state transition. The causally informed acquisition function\ngenerally outperforms existing criteria allowing for optimal intervention\ndesign with fewer but carefully selected samples.\n", "date": "2022-09-10", "categories": []}
{"id": "2305.05890", "title": "CUTS+: High-dimensional Causal Discovery from Irregular Time-series", "abstract": "  Causal discovery in time-series is a fundamental problem in the machine\nlearning community, enabling causal reasoning and decision-making in complex\nscenarios. Recently, researchers successfully discover causality by combining\nneural networks with Granger causality, but their performances degrade largely\nwhen encountering high-dimensional data because of the highly redundant network\ndesign and huge causal graphs. Moreover, the missing entries in the\nobservations further hamper the causal structural learning. To overcome these\nlimitations, We propose CUTS+, which is built on the Granger-causality-based\ncausal discovery method CUTS and raises the scalability by introducing a\ntechnique called Coarse-to-fine-discovery (C2FD) and leveraging a\nmessage-passing-based graph neural network (MPGNN). Compared to previous\nmethods on simulated, quasi-real, and real datasets, we show that CUTS+ largely\nimproves the causal discovery performance on high-dimensional data with\ndifferent types of irregular sampling.\n", "date": "2023-05-10", "categories": []}
{"id": "2308.08148", "title": "Hierarchical Topological Ordering with Conditional Independence Test for\n  Limited Time Series", "abstract": "  Learning directed acyclic graphs (DAGs) to identify causal relations\nunderlying observational data is crucial but also poses significant challenges.\nRecently, topology-based methods have emerged as a two-step approach to\ndiscovering DAGs by first learning the topological ordering of variables and\nthen eliminating redundant edges, while ensuring that the graph remains\nacyclic. However, one limitation is that these methods would generate numerous\nspurious edges that require subsequent pruning. To overcome this limitation, in\nthis paper, we propose an improvement to topology-based methods by introducing\nlimited time series data, consisting of only two cross-sectional records that\nneed not be adjacent in time and are subject to flexible timing. By\nincorporating conditional instrumental variables as exogenous interventions, we\naim to identify descendant nodes for each variable. Following this line, we\npropose a hierarchical topological ordering algorithm with conditional\nindependence test (HT-CIT), which enables the efficient learning of sparse DAGs\nwith a smaller search space compared to other popular approaches. The HT-CIT\nalgorithm greatly reduces the number of edges that need to be pruned. Empirical\nresults from synthetic and real-world datasets demonstrate the superiority of\nthe proposed HT-CIT algorithm.\n", "date": "2023-08-16", "categories": []}
{"id": "2308.07346", "title": "Py-Tetrad and RPy-Tetrad: A New Python Interface with R Support for\n  Tetrad Causal Search", "abstract": "  We give novel Python and R interfaces for the (Java) Tetrad project for\ncausal modeling, search, and estimation. The Tetrad project is a mainstay in\nthe literature, having been under consistent development for over 30 years.\nSome of its algorithms are now classics, like PC and FCI; others are recent\ndevelopments. It is increasingly the case, however, that researchers need to\naccess the underlying Java code from Python or R. Existing methods for doing\nthis are inadequate. We provide new, up-to-date methods using the JPype\nPython-Java interface and the Reticulate Python-R interface, directly solving\nthese issues. With the addition of some simple tools and the provision of\nworking examples for both Python and R, using JPype and Reticulate to interface\nPython and R with Tetrad is straightforward and intuitive.\n", "date": "2023-08-13", "categories": []}
{"id": "2207.05302", "title": "Causal Conceptions of Fairness and their Consequences", "abstract": "  Recent work highlights the role of causality in designing equitable\ndecision-making algorithms. It is not immediately clear, however, how existing\ncausal conceptions of fairness relate to one another, or what the consequences\nare of using these definitions as design principles. Here, we first assemble\nand categorize popular causal definitions of algorithmic fairness into two\nbroad families: (1) those that constrain the effects of decisions on\ncounterfactual disparities; and (2) those that constrain the effects of legally\nprotected characteristics, like race and gender, on decisions. We then show,\nanalytically and empirically, that both families of definitions \\emph{almost\nalways} -- in a measure theoretic sense -- result in strongly Pareto dominated\ndecision policies, meaning there is an alternative, unconstrained policy\nfavored by every stakeholder with preferences drawn from a large, natural\nclass. For example, in the case of college admissions decisions, policies\nconstrained to satisfy causal fairness definitions would be disfavored by every\nstakeholder with neutral or positive preferences for both academic preparedness\nand diversity. Indeed, under a prominent definition of causal fairness, we\nprove the resulting policies require admitting all students with the same\nprobability, regardless of academic qualifications or group membership. Our\nresults highlight formal limitations and potential adverse consequences of\ncommon mathematical notions of causal fairness.\n", "date": "2022-07-12", "categories": []}
{"id": "2308.05707", "title": "Shadow Datasets, New challenging datasets for Causal Representation\n  Learning", "abstract": "  Discovering causal relations among semantic factors is an emergent topic in\nrepresentation learning. Most causal representation learning (CRL) methods are\nfully supervised, which is impractical due to costly labeling. To resolve this\nrestriction, weakly supervised CRL methods were introduced. To evaluate CRL\nperformance, four existing datasets, Pendulum, Flow, CelebA(BEARD) and\nCelebA(SMILE), are utilized. However, existing CRL datasets are limited to\nsimple graphs with few generative factors. Thus we propose two new datasets\nwith a larger number of diverse generative factors and more sophisticated\ncausal graphs. In addition, current real datasets, CelebA(BEARD) and\nCelebA(SMILE), the originally proposed causal graphs are not aligned with the\ndataset distributions. Thus, we propose modifications to them.\n", "date": "2023-08-10", "categories": []}
{"id": "2305.02640", "title": "Towards Causal Representation Learning and Deconfounding from Indefinite\n  Data", "abstract": "  Owing to the cross-pollination between causal discovery and deep learning,\nnon-statistical data (e.g., images, text, etc.) encounters significant\nconflicts in terms of properties and methods with traditional causal data. To\nunify these data types of varying forms, we redefine causal data from two novel\nperspectives and then propose three data paradigms. Among them, the indefinite\ndata (like dialogues or video sources) induce low sample utilization and\nincapability of the distribution assumption, both leading to the fact that\nlearning causal representation from indefinite data is, as of yet, largely\nunexplored. We design the causal strength variational model to settle down\nthese two problems. Specifically, we leverage the causal strength instead of\nindependent noise as the latent variable to construct evidence lower bound. By\nthis design ethos, The causal strengths of different structures are regarded as\na distribution and can be expressed as a 2D matrix. Moreover, considering the\nlatent confounders, we disentangle the causal graph G into two relation\nsubgraphs O and C. O contains pure relations between observed variables, while\nC represents the relations from latent variables to observed variables. We\nimplement the above designs as a dynamic variational inference model, tailored\nto learn causal representation from indefinite data under latent confounding.\nFinally, we conduct comprehensive experiments on synthetic and real-world data\nto demonstrate the effectiveness of our method.\n", "date": "2023-05-04", "categories": []}
{"id": "2308.06106", "title": "Hawkes Processes with Delayed Granger Causality", "abstract": "  We aim to explicitly model the delayed Granger causal effects based on\nmultivariate Hawkes processes. The idea is inspired by the fact that a causal\nevent usually takes some time to exert an effect. Studying this time lag itself\nis of interest. Given the proposed model, we first prove the identifiability of\nthe delay parameter under mild conditions. We further investigate a model\nestimation method under a complex setting, where we want to infer the posterior\ndistribution of the time lags and understand how this distribution varies\nacross different scenarios. We treat the time lags as latent variables and\nformulate a Variational Auto-Encoder (VAE) algorithm to approximate the\nposterior distribution of the time lags. By explicitly modeling the time lags\nin Hawkes processes, we add flexibility to the model. The inferred time-lag\nposterior distributions are of scientific meaning and help trace the original\ncausal time that supports the root cause analysis. We empirically evaluate our\nmodel's event prediction and time-lag inference accuracy on synthetic and real\ndata, achieving promising results.\n", "date": "2023-08-11", "categories": []}
{"id": "2306.13759", "title": "Incremental Profit per Conversion: a Response Transformation for Uplift\n  Modeling in E-Commerce Promotions", "abstract": "  Promotions play a crucial role in e-commerce platforms, and various cost\nstructures are employed to drive user engagement. This paper focuses on\npromotions with response-dependent costs, where expenses are incurred only when\na purchase is made. Such promotions include discounts and coupons. While\nexisting uplift model approaches aim to address this challenge, these\napproaches often necessitate training multiple models, like meta-learners, or\nencounter complications when estimating profit due to zero-inflated values\nstemming from non-converted individuals with zero cost and profit.\n  To address these challenges, we introduce Incremental Profit per Conversion\n(IPC), a novel uplift measure of promotional campaigns' efficiency in unit\neconomics. Through a proposed response transformation, we demonstrate that IPC\nrequires only converted data, its propensity, and a single model to be\nestimated. As a result, IPC resolves the issues mentioned above while\nmitigating the noise typically associated with the class imbalance in\nconversion datasets and biases arising from the many-to-one mapping between\nsearch and purchase data. Lastly, we validate the efficacy of our approach by\npresenting results obtained from a synthetic simulation of a discount coupon\ncampaign.\n", "date": "2023-06-23", "categories": []}
{"id": "2302.10331", "title": "Causal Razors", "abstract": "  When performing causal discovery, assumptions have to be made on how the true\ncausal mechanism corresponds to the underlying joint probability distribution.\nThese assumptions are labeled as causal razors in this work. We review numerous\ncausal razors that appeared in the literature, and offer a comprehensive\nlogical comparison of them. In particular, we scrutinize an unpopular causal\nrazor, namely parameter minimality, in multinomial causal models and its\nlogical relations with other well-studied causal razors. Our logical result\nposes a dilemma in selecting a reasonable scoring criterion for score-based\ncasual search algorithms.\n", "date": "2023-02-20", "categories": []}
{"id": "2308.04011", "title": "Generalization bound for estimating causal effects from observational\n  network data", "abstract": "  Estimating causal effects from observational network data is a significant\nbut challenging problem. Existing works in causal inference for observational\nnetwork data lack an analysis of the generalization bound, which can\ntheoretically provide support for alleviating the complex confounding bias and\npractically guide the design of learning objectives in a principled manner. To\nfill this gap, we derive a generalization bound for causal effect estimation in\nnetwork scenarios by exploiting 1) the reweighting schema based on joint\npropensity score and 2) the representation learning schema based on Integral\nProbability Metric (IPM). We provide two perspectives on the generalization\nbound in terms of reweighting and representation learning, respectively.\nMotivated by the analysis of the bound, we propose a weighting regression\nmethod based on the joint propensity score augmented with representation\nlearning. Extensive experimental studies on two real-world networks with\nsemi-synthetic data demonstrate the effectiveness of our algorithm.\n", "date": "2023-08-08", "categories": []}
{"id": "2303.05396", "title": "Bounding the Probabilities of Benefit and Harm Through Sensitivity\n  Parameters and Proxies", "abstract": "  We present two methods for bounding the probabilities of benefit and harm\nunder unmeasured confounding. The first method computes the (upper or lower)\nbound of either probability as a function of the observed data distribution and\ntwo intuitive sensitivity parameters which, then, can be presented to the\nanalyst as a 2-D plot to assist her in decision making. The second method\nassumes the existence of a measured nondifferential proxy (i.e., direct effect)\nof the unmeasured confounder. Using this proxy, tighter bounds than the\nexisting ones can be derived from just the observed data distribution.\n", "date": "2023-03-08", "categories": []}
{"id": "2308.02709", "title": "Scalable Computation of Causal Bounds", "abstract": "  We consider the problem of computing bounds for causal queries on causal\ngraphs with unobserved confounders and discrete valued observed variables,\nwhere identifiability does not hold. Existing non-parametric approaches for\ncomputing such bounds use linear programming (LP) formulations that quickly\nbecome intractable for existing solvers because the size of the LP grows\nexponentially in the number of edges in the causal graph. We show that this LP\ncan be significantly pruned, allowing us to compute bounds for significantly\nlarger causal inference problems compared to existing techniques. This pruning\nprocedure allows us to compute bounds in closed form for a special class of\nproblems, including a well-studied family of problems where multiple confounded\ntreatments influence an outcome. We extend our pruning methodology to\nfractional LPs which compute bounds for causal queries which incorporate\nadditional observations about the unit. We show that our methods provide\nsignificant runtime improvement compared to benchmarks in experiments and\nextend our results to the finite data setting. For causal inference without\nadditional observations, we propose an efficient greedy heuristic that produces\nhigh quality bounds, and scales to problems that are several orders of\nmagnitude larger than those for which the pruned LP can be solved.\n", "date": "2023-08-04", "categories": []}
{"id": "2308.02353", "title": "Adapting to Change: Robust Counterfactual Explanations in Dynamic Data\n  Landscapes", "abstract": "  We introduce a novel semi-supervised Graph Counterfactual Explainer (GCE)\nmethodology, Dynamic GRAph Counterfactual Explainer (DyGRACE). It leverages\ninitial knowledge about the data distribution to search for valid\ncounterfactuals while avoiding using information from potentially outdated\ndecision functions in subsequent time steps. Employing two graph autoencoders\n(GAEs), DyGRACE learns the representation of each class in a binary\nclassification scenario. The GAEs minimise the reconstruction error between the\noriginal graph and its learned representation during training. The method\ninvolves (i) optimising a parametric density function (implemented as a\nlogistic regression function) to identify counterfactuals by maximising the\nfactual autoencoder's reconstruction error, (ii) minimising the counterfactual\nautoencoder's error, and (iii) maximising the similarity between the factual\nand counterfactual graphs. This semi-supervised approach is independent of an\nunderlying black-box oracle. A logistic regression model is trained on a set of\ngraph pairs to learn weights that aid in finding counterfactuals. At inference,\nfor each unseen graph, the logistic regressor identifies the best\ncounterfactual candidate using these learned weights, while the GAEs can be\niteratively updated to represent the continual adaptation of the learned graph\nrepresentation over iterations. DyGRACE is quite effective and can act as a\ndrift detector, identifying distributional drift based on differences in\nreconstruction errors between iterations. It avoids reliance on the oracle's\npredictions in successive iterations, thereby increasing the efficiency of\ncounterfactual discovery. DyGRACE, with its capacity for contrastive learning\nand drift detection, will offer new avenues for semi-supervised learning and\nexplanation generation.\n", "date": "2023-08-04", "categories": []}
{"id": "2303.10112", "title": "Causal Discovery from Temporal Data: An Overview and New Perspectives", "abstract": "  Temporal data, representing chronological observations of complex systems,\nhas always been a typical data structure that can be widely generated by many\ndomains, such as industry, medicine and finance. Analyzing this type of data is\nextremely valuable for various applications. Thus, different temporal data\nanalysis tasks, eg, classification, clustering and prediction, have been\nproposed in the past decades. Among them, causal discovery, learning the causal\nrelations from temporal data, is considered an interesting yet critical task\nand has attracted much research attention. Existing causal discovery works can\nbe divided into two highly correlated categories according to whether the\ntemporal data is calibrated, ie, multivariate time series causal discovery, and\nevent sequence causal discovery. However, most previous surveys are only\nfocused on the time series causal discovery and ignore the second category. In\nthis paper, we specify the correlation between the two categories and provide a\nsystematical overview of existing solutions. Furthermore, we provide public\ndatasets, evaluation metrics and new perspectives for temporal data causal\ndiscovery.\n", "date": "2023-03-17", "categories": []}
{"id": "2305.05832", "title": "Causal Information Splitting: Engineering Proxy Features for Robustness\n  to Distribution Shifts", "abstract": "  Statistical prediction models are often trained on data from different\nprobability distributions than their eventual use cases. One approach to\nproactively prepare for these shifts harnesses the intuition that causal\nmechanisms should remain invariant between environments. Here we focus on a\nchallenging setting in which the causal and anticausal variables of the target\nare unobserved. Leaning on information theory, we develop feature selection and\nengineering techniques for the observed downstream variables that act as\nproxies. We identify proxies that help to build stable models and moreover\nutilize auxiliary training tasks to answer counterfactual questions that\nextract stability-enhancing information from proxies. We demonstrate the\neffectiveness of our techniques on synthetic and real data.\n", "date": "2023-05-10", "categories": []}
{"id": "2307.16625", "title": "Adversarial Causal Bayesian Optimization", "abstract": "  In Causal Bayesian Optimization (CBO), an agent intervenes on an unknown\nstructural causal model to maximize a downstream reward variable. In this\npaper, we consider the generalization where other agents or external events\nalso intervene on the system, which is key for enabling adaptiveness to\nnon-stationarities such as weather changes, market forces, or adversaries. We\nformalize this generalization of CBO as Adversarial Causal Bayesian\nOptimization (ACBO) and introduce the first algorithm for ACBO with bounded\nregret: Causal Bayesian Optimization with Multiplicative Weights (CBO-MW). Our\napproach combines a classical online learning strategy with causal modeling of\nthe rewards. To achieve this, it computes optimistic counterfactual reward\nestimates by propagating uncertainty through the causal graph. We derive regret\nbounds for CBO-MW that naturally depend on graph-related quantities. We further\npropose a scalable implementation for the case of combinatorial interventions\nand submodular rewards. Empirically, CBO-MW outperforms non-causal and\nnon-adversarial Bayesian optimization methods on synthetic environments and\nenvironments based on real-word data. Our experiments include a realistic\ndemonstration of how CBO-MW can be used to learn users' demand patterns in a\nshared mobility system and reposition vehicles in strategic areas.\n", "date": "2023-07-31", "categories": []}
{"id": "2307.04988", "title": "Benchmarking Bayesian Causal Discovery Methods for Downstream Treatment\n  Effect Estimation", "abstract": "  The practical utility of causality in decision-making is widespread and\nbrought about by the intertwining of causal discovery and causal inference.\nNevertheless, a notable gap exists in the evaluation of causal discovery\nmethods, where insufficient emphasis is placed on downstream inference. To\naddress this gap, we evaluate seven established baseline causal discovery\nmethods including a newly proposed method based on GFlowNets, on the downstream\ntask of treatment effect estimation. Through the implementation of a\ndistribution-level evaluation, we offer valuable and unique insights into the\nefficacy of these causal discovery methods for treatment effect estimation,\nconsidering both synthetic and real-world scenarios, as well as low-data\nscenarios. The results of our study demonstrate that some of the algorithms\nstudied are able to effectively capture a wide range of useful and diverse ATE\nmodes, while some tend to learn many low-probability modes which impacts the\n(unrelaxed) recall and precision.\n", "date": "2023-07-11", "categories": []}
{"id": "2307.13992", "title": "Causal reasoning in typical computer vision tasks", "abstract": "  Deep learning has revolutionized the field of artificial intelligence. Based\non the statistical correlations uncovered by deep learning-based methods,\ncomputer vision has contributed to tremendous growth in areas like autonomous\ndriving and robotics. Despite being the basis of deep learning, such\ncorrelation is not stable and is susceptible to uncontrolled factors. In the\nabsence of the guidance of prior knowledge, statistical correlations can easily\nturn into spurious correlations and cause confounders. As a result, researchers\nare now trying to enhance deep learning methods with causal theory. Causal\ntheory models the intrinsic causal structure unaffected by data bias and is\neffective in avoiding spurious correlations. This paper aims to comprehensively\nreview the existing causal methods in typical vision and vision-language tasks\nsuch as semantic segmentation, object detection, and image captioning. The\nadvantages of causality and the approaches for building causal paradigms will\nbe summarized. Future roadmaps are also proposed, including facilitating the\ndevelopment of causal theory and its application in other complex scenes and\nsystems.\n", "date": "2023-07-26", "categories": []}
{"id": "2307.16452", "title": "A continuous Structural Intervention Distance to compare Causal Graphs", "abstract": "  Understanding and adequately assessing the difference between a true and a\nlearnt causal graphs is crucial for causal inference under interventions. As an\nextension to the graph-based structural Hamming distance and structural\nintervention distance, we propose a novel continuous-measured metric that\nconsiders the underlying data in addition to the graph structure for its\ncalculation of the difference between a true and a learnt causal graph. The\ndistance is based on embedding intervention distributions over each pair of\nnodes as conditional mean embeddings into reproducing kernel Hilbert spaces and\nestimating their difference by the maximum (conditional) mean discrepancy. We\nshow theoretical results which we validate with numerical experiments on\nsynthetic data.\n", "date": "2023-07-31", "categories": []}
{"id": "2307.16577", "title": "Approximating Counterfactual Bounds while Fusing Observational, Biased\n  and Randomised Data Sources", "abstract": "  We address the problem of integrating data from multiple, possibly biased,\nobservational and interventional studies, to eventually compute counterfactuals\nin structural causal models. We start from the case of a single observational\ndataset affected by a selection bias. We show that the likelihood of the\navailable data has no local maxima. This enables us to use the causal\nexpectation-maximisation scheme to approximate the bounds for partially\nidentifiable counterfactual queries, which are the focus of this paper. We then\nshow how the same approach can address the general case of multiple datasets,\nno matter whether interventional or observational, biased or unbiased, by\nremapping it into the former one via graphical transformations. Systematic\nnumerical experiments and a case study on palliative care show the\neffectiveness of our approach, while hinting at the benefits of fusing\nheterogeneous data sources to get informative outcomes in case of partial\nidentifiability.\n", "date": "2023-07-31", "categories": []}
{"id": "2307.16851", "title": "Towards Trustworthy and Aligned Machine Learning: A Data-centric Survey\n  with Causality Perspectives", "abstract": "  The trustworthiness of machine learning has emerged as a critical topic in\nthe field, encompassing various applications and research areas such as\nrobustness, security, interpretability, and fairness. The last decade saw the\ndevelopment of numerous methods addressing these challenges. In this survey, we\nsystematically review these advancements from a data-centric perspective,\nhighlighting the shortcomings of traditional empirical risk minimization (ERM)\ntraining in handling challenges posed by the data.\n  Interestingly, we observe a convergence of these methods, despite being\ndeveloped independently across trustworthy machine learning subfields. Pearl's\nhierarchy of causality offers a unifying framework for these techniques.\nAccordingly, this survey presents the background of trustworthy machine\nlearning development using a unified set of concepts, connects this language to\nPearl's causal hierarchy, and finally discusses methods explicitly inspired by\ncausality literature. We provide a unified language with mathematical\nvocabulary to link these methods across robustness, adversarial robustness,\ninterpretability, and fairness, fostering a more cohesive understanding of the\nfield.\n  Further, we explore the trustworthiness of large pretrained models. After\nsummarizing dominant techniques like fine-tuning, parameter-efficient\nfine-tuning, prompting, and reinforcement learning with human feedback, we draw\nconnections between them and the standard ERM. This connection allows us to\nbuild upon the principled understanding of trustworthy methods, extending it to\nthese new techniques in large pretrained models, paving the way for future\nmethods. Existing methods under this perspective are also reviewed.\n  Lastly, we offer a brief summary of the applications of these methods and\ndiscuss potential future aspects related to our survey. For more information,\nplease visit http://trustai.one.\n", "date": "2023-07-31", "categories": []}
{"id": "2307.15678", "title": "Case Studies of Causal Discovery from IT Monitoring Time Series", "abstract": "  Information technology (IT) systems are vital for modern businesses, handling\ndata storage, communication, and process automation. Monitoring these systems\nis crucial for their proper functioning and efficiency, as it allows collecting\nextensive observational time series data for analysis. The interest in causal\ndiscovery is growing in IT monitoring systems as knowing causal relations\nbetween different components of the IT system helps in reducing downtime,\nenhancing system performance and identifying root causes of anomalies and\nincidents. It also allows proactive prediction of future issues through\nhistorical data analysis. Despite its potential benefits, applying causal\ndiscovery algorithms on IT monitoring data poses challenges, due to the\ncomplexity of the data. For instance, IT monitoring data often contains\nmisaligned time series, sleeping time series, timestamp errors and missing\nvalues. This paper presents case studies on applying causal discovery\nalgorithms to different IT monitoring datasets, highlighting benefits and\nongoing challenges.\n", "date": "2023-07-28", "categories": []}
{"id": "2302.01198", "title": "Causal Lifting and Link Prediction", "abstract": "  Existing causal models for link prediction assume an underlying set of\ninherent node factors -- an innate characteristic defined at the node's birth\n-- that governs the causal evolution of links in the graph. In some causal\ntasks, however, link formation is path-dependent: The outcome of link\ninterventions depends on existing links. Unfortunately, these existing causal\nmethods are not designed for path-dependent link formation, as the cascading\nfunctional dependencies between links (arising from path dependence) are either\nunidentifiable or require an impractical number of control variables. To\novercome this, we develop the first causal model capable of dealing with path\ndependencies in link prediction. In this work we introduce the concept of\ncausal lifting, an invariance in causal models of independent interest that, on\ngraphs, allows the identification of causal link prediction queries using\nlimited interventional data. Further, we show how structural pairwise\nembeddings exhibit lower bias and correctly represent the task's causal\nstructure, as opposed to existing node embeddings, e.g., graph neural network\nnode embeddings and matrix factorization. Finally, we validate our theoretical\nfindings on three scenarios for causal link prediction tasks: knowledge base\ncompletion, covariance matrix estimation and consumer-product recommendations.\n", "date": "2023-02-02", "categories": []}
{"id": "2304.06848", "title": "CAR-DESPOT: Causally-Informed Online POMDP Planning for Robots in\n  Confounded Environments", "abstract": "  Robots operating in real-world environments must reason about possible\noutcomes of stochastic actions and make decisions based on partial observations\nof the true world state. A major challenge for making accurate and robust\naction predictions is the problem of confounding, which if left untreated can\nlead to prediction errors. The partially observable Markov decision process\n(POMDP) is a widely-used framework to model these stochastic and\npartially-observable decision-making problems. However, due to a lack of\nexplicit causal semantics, POMDP planning methods are prone to confounding bias\nand thus in the presence of unobserved confounders may produce underperforming\npolicies. This paper presents a novel causally-informed extension of \"anytime\nregularized determinized sparse partially observable tree\" (AR-DESPOT), a\nmodern anytime online POMDP planner, using causal modelling and inference to\neliminate errors caused by unmeasured confounder variables. We further propose\na method to learn offline the partial parameterisation of the causal model for\nplanning, from ground truth model data. We evaluate our methods on a toy\nproblem with an unobserved confounder and show that the learned causal model is\nhighly accurate, while our planning method is more robust to confounding and\nproduces overall higher performing policies than AR-DESPOT.\n", "date": "2023-04-13", "categories": []}
{"id": "2307.13757", "title": "UPREVE: An End-to-End Causal Discovery Benchmarking System", "abstract": "  Discovering causal relationships in complex socio-behavioral systems is\nchallenging but essential for informed decision-making. We present Upload,\nPREprocess, Visualize, and Evaluate (UPREVE), a user-friendly web-based\ngraphical user interface (GUI) designed to simplify the process of causal\ndiscovery. UPREVE allows users to run multiple algorithms simultaneously,\nvisualize causal relationships, and evaluate the accuracy of learned causal\ngraphs. With its accessible interface and customizable features, UPREVE\nempowers researchers and practitioners in social computing and\nbehavioral-cultural modeling (among others) to explore and understand causal\nrelationships effectively. Our proposed solution aims to make causal discovery\nmore accessible and user-friendly, enabling users to gain valuable insights for\nbetter decision-making.\n", "date": "2023-07-25", "categories": []}
{"id": "2305.16165", "title": "A Conceptual Model for End-to-End Causal Discovery in Knowledge Tracing", "abstract": "  In this paper, we take a preliminary step towards solving the problem of\ncausal discovery in knowledge tracing, i.e., finding the underlying causal\nrelationship among different skills from real-world student response data. This\nproblem is important since it can potentially help us understand the causal\nrelationship between different skills without extensive A/B testing, which can\npotentially help educators to design better curricula according to skill\nprerequisite information. Specifically, we propose a conceptual solution, a\nnovel causal gated recurrent unit (GRU) module in a modified deep knowledge\ntracing model, which uses i) a learnable permutation matrix for causal ordering\namong skills and ii) an optionally learnable lower-triangular matrix for causal\nstructure among skills. We also detail how to learn the model parameters in an\nend-to-end, differentiable way. Our solution placed among the top entries in\nTask 3 of the NeurIPS 2022 Challenge on Causal Insights for Learning Paths in\nEducation. We detail preliminary experiments as evaluated on the challenge's\npublic leaderboard since the ground truth causal structure has not been\npublicly released, making detailed local evaluation impossible.\n", "date": "2023-05-11", "categories": []}
{"id": "2211.16596", "title": "Towards Dynamic Causal Discovery with Rare Events: A Nonparametric\n  Conditional Independence Test", "abstract": "  Causal phenomena associated with rare events occur across a wide range of\nengineering problems, such as risk-sensitive safety analysis, accident analysis\nand prevention, and extreme value theory. However, current methods for causal\ndiscovery are often unable to uncover causal links, between random variables in\na dynamic setting, that manifest only when the variables first experience\nlow-probability realizations. To address this issue, we introduce a novel\nstatistical independence test on data collected from time-invariant dynamical\nsystems in which rare but consequential events occur. In particular, we exploit\nthe time-invariance of the underlying data to construct a superimposed dataset\nof the system state before rare events happen at different timesteps. We then\ndesign a conditional independence test on the reorganized data. We provide\nnon-asymptotic sample complexity bounds for the consistency of our method, and\nvalidate its performance across various simulated and real-world datasets,\nincluding incident data collected from the Caltrans Performance Measurement\nSystem (PeMS). Code containing the datasets and experiments is publicly\navailable.\n", "date": "2022-11-29", "categories": []}
{"id": "2306.15764", "title": "High Fidelity Image Counterfactuals with Probabilistic Causal Models", "abstract": "  We present a general causal generative modelling framework for accurate\nestimation of high fidelity image counterfactuals with deep structural causal\nmodels. Estimation of interventional and counterfactual queries for\nhigh-dimensional structured variables, such as images, remains a challenging\ntask. We leverage ideas from causal mediation analysis and advances in\ngenerative modelling to design new deep causal mechanisms for structured\nvariables in causal models. Our experiments demonstrate that our proposed\nmechanisms are capable of accurate abduction and estimation of direct, indirect\nand total effects as measured by axiomatic soundness of counterfactuals.\n", "date": "2023-06-27", "categories": []}
{"id": "2307.08686", "title": "An R package for parametric estimation of causal effects", "abstract": "  This article explains the usage of R package CausalModels, which is publicly\navailable on the Comprehensive R Archive Network. While packages are available\nfor sufficiently estimating causal effects, there lacks a package that provides\na collection of structural models using the conventional statistical approach\ndeveloped by Hernan and Robins (2020). CausalModels addresses this deficiency\nof software in R concerning causal inference by offering tools for methods that\naccount for biases in observational data without requiring extensive\nstatistical knowledge. These methods should not be ignored and may be more\nappropriate or efficient in solving particular problems. While implementations\nof these statistical models are distributed among a number of causal packages,\nCausalModels introduces a simple and accessible framework for a consistent\nmodeling pipeline among a variety of statistical methods for estimating causal\neffects in a single R package. It consists of common methods including\nstandardization, IP weighting, G-estimation, outcome regression, instrumental\nvariables and propensity matching.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.07524", "title": "Reducing Causality to Functions with Structural Models", "abstract": "  The precise definition of causality is currently an open problem in\nphilosophy and statistics. We believe causality should be defined as functions\n(in mathematics) that map causes to effects. We propose a reductive definition\nof causality based on Structural Functional Model (SFM). Using delta\ncompression and contrastive forward inference, SFM can produce causal\nutterances like \"X causes Y\" and \"X is the cause of Y\" that match our\nintuitions. We compile a dataset of causal scenarios and use SFM in all of\nthem. SFM is compatible with but not reducible to probability theory. We also\ncompare SFM with other theories of causation and apply SFM to downstream\nproblems like free will, causal explanation, and mental causation.\n", "date": "2023-07-11", "categories": []}
{"id": "2307.07973", "title": "Heteroscedastic Causal Structure Learning", "abstract": "  Heretofore, learning the directed acyclic graphs (DAGs) that encode the\ncause-effect relationships embedded in observational data is a computationally\nchallenging problem. A recent trend of studies has shown that it is possible to\nrecover the DAGs with polynomial time complexity under the equal variances\nassumption. However, this prohibits the heteroscedasticity of the noise, which\nallows for more flexible modeling capabilities, but at the same time is\nsubstantially more challenging to handle. In this study, we tackle the\nheteroscedastic causal structure learning problem under Gaussian noises. By\nexploiting the normality of the causal mechanisms, we can recover a valid\ncausal ordering, which can uniquely identify the causal DAG using a series of\nconditional independence tests. The result is HOST (Heteroscedastic causal\nSTructure learning), a simple yet effective causal structure learning algorithm\nthat scales polynomially in both sample size and dimensionality. In addition,\nvia extensive empirical evaluations on a wide range of both controlled and real\ndatasets, we show that the proposed HOST method is competitive with\nstate-of-the-art approaches in both the causal order learning and structure\nlearning problems.\n", "date": "2023-07-16", "categories": []}
{"id": "2307.08232", "title": "Learning for Counterfactual Fairness from Observational Data", "abstract": "  Fairness-aware machine learning has attracted a surge of attention in many\ndomains, such as online advertising, personalized recommendation, and social\nmedia analysis in web applications. Fairness-aware machine learning aims to\neliminate biases of learning models against certain subgroups described by\ncertain protected (sensitive) attributes such as race, gender, and age. Among\nmany existing fairness notions, counterfactual fairness is a popular notion\ndefined from a causal perspective. It measures the fairness of a predictor by\ncomparing the prediction of each individual in the original world and that in\nthe counterfactual worlds in which the value of the sensitive attribute is\nmodified. A prerequisite for existing methods to achieve counterfactual\nfairness is the prior human knowledge of the causal model for the data.\nHowever, in real-world scenarios, the underlying causal model is often unknown,\nand acquiring such human knowledge could be very difficult. In these scenarios,\nit is risky to directly trust the causal models obtained from information\nsources with unknown reliability and even causal discovery methods, as\nincorrect causal models can consequently bring biases to the predictor and lead\nto unfair predictions. In this work, we address the problem of counterfactually\nfair prediction from observational data without given causal models by\nproposing a novel framework CLAIRE. Specifically, under certain general\nassumptions, CLAIRE effectively mitigates the biases from the sensitive\nattribute with a representation learning framework based on counterfactual data\naugmentation and an invariant penalty. Experiments conducted on both synthetic\nand real-world datasets validate the superiority of CLAIRE in both\ncounterfactual fairness and prediction performance.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.08237", "title": "A Look into Causal Effects under Entangled Treatment in Graphs:\n  Investigating the Impact of Contact on MRSA Infection", "abstract": "  Methicillin-resistant Staphylococcus aureus (MRSA) is a type of bacteria\nresistant to certain antibiotics, making it difficult to prevent MRSA\ninfections. Among decades of efforts to conquer infectious diseases caused by\nMRSA, many studies have been proposed to estimate the causal effects of close\ncontact (treatment) on MRSA infection (outcome) from observational data. In\nthis problem, the treatment assignment mechanism plays a key role as it\ndetermines the patterns of missing counterfactuals -- the fundamental challenge\nof causal effect estimation. Most existing observational studies for causal\neffect learning assume that the treatment is assigned individually for each\nunit. However, on many occasions, the treatments are pairwisely assigned for\nunits that are connected in graphs, i.e., the treatments of different units are\nentangled. Neglecting the entangled treatments can impede the causal effect\nestimation. In this paper, we study the problem of causal effect estimation\nwith treatment entangled in a graph. Despite a few explorations for entangled\ntreatments, this problem still remains challenging due to the following\nchallenges: (1) the entanglement brings difficulties in modeling and leveraging\nthe unknown treatment assignment mechanism; (2) there may exist hidden\nconfounders which lead to confounding biases in causal effect estimation; (3)\nthe observational data is often time-varying. To tackle these challenges, we\npropose a novel method NEAT, which explicitly leverages the graph structure to\nmodel the treatment assignment mechanism, and mitigates confounding biases\nbased on the treatment assignment modeling. We also extend our method into a\ndynamic setting to handle time-varying observational data. Experiments on both\nsynthetic datasets and a real-world MRSA dataset validate the effectiveness of\nthe proposed method, and provide insights for future applications.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.08519", "title": "Results on Counterfactual Invariance", "abstract": "  In this paper we provide a theoretical analysis of counterfactual invariance.\nWe present a variety of existing definitions, study how they relate to each\nother and what their graphical implications are. We then turn to the current\nmajor question surrounding counterfactual invariance, how does it relate to\nconditional independence? We show that whilst counterfactual invariance implies\nconditional independence, conditional independence does not give any\nimplications about the degree or likelihood of satisfying counterfactual\ninvariance. Furthermore, we show that for discrete causal models\ncounterfactually invariant functions are often constrained to be functions of\nparticular variables, or even constant.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.08617", "title": "Understanding the impacts of crop diversification in the context of\n  climate change: a machine learning approach", "abstract": "  The concept of sustainable intensification in agriculture necessitates the\nimplementation of management practices that prioritize sustainability without\ncompromising productivity. However, the effects of such practices are known to\ndepend on environmental conditions, and are therefore expected to change as a\nresult of a changing climate. We study the impact of crop diversification on\nproductivity in the context of climate change. We leverage heterogeneous Earth\nObservation data and contribute a data-driven approach based on causal machine\nlearning for understanding how crop diversification impacts may change in the\nfuture. We apply this method to the country of Cyprus throughout a 4-year\nperiod. We find that, on average, crop diversification significantly benefited\nthe net primary productivity of crops, increasing it by 2.8%. The effect\ngenerally synergized well with higher maximum temperatures and lower soil\nmoistures. In a warmer and more drought-prone climate, we conclude that crop\ndiversification exhibits promising adaptation potential and is thus a sensible\npolicy choice with regards to agricultural productivity for present and future.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.06825", "title": "A Causal Framework to Unify Common Domain Generalization Approaches", "abstract": "  Domain generalization (DG) is about learning models that generalize well to\nnew domains that are related to, but different from, the training domain(s). It\nis a fundamental problem in machine learning and has attracted much attention\nin recent years. A large number of approaches have been proposed. Different\napproaches are motivated from different perspectives, making it difficult to\ngain an overall understanding of the area. In this paper, we propose a causal\nframework for domain generalization and present an understanding of common DG\napproaches in the framework. Our work sheds new lights on the following\nquestions: (1) What are the key ideas behind each DG method? (2) Why is it\nexpected to improve generalization to new domains theoretically? (3) How are\ndifferent DG methods related to each other and what are relative advantages and\nlimitations? By providing a unified perspective on DG, we hope to help\nresearchers better understand the underlying principles and develop more\neffective approaches for this critical problem in machine learning.\n", "date": "2023-07-13", "categories": []}
{"id": "2307.05704", "title": "A Causal Ordering Prior for Unsupervised Representation Learning", "abstract": "  Unsupervised representation learning with variational inference relies\nheavily on independence assumptions over latent variables. Causal\nrepresentation learning (CRL), however, argues that factors of variation in a\ndataset are, in fact, causally related. Allowing latent variables to be\ncorrelated, as a consequence of causal relationships, is more realistic and\ngeneralisable. So far, provably identifiable methods rely on: auxiliary\ninformation, weak labels, and interventional or even counterfactual data.\nInspired by causal discovery with functional causal models, we propose a fully\nunsupervised representation learning method that considers a data generation\nprocess with a latent additive noise model (ANM). We encourage the latent space\nto follow a causal ordering via loss function based on the Hessian of the\nlatent distribution.\n", "date": "2023-07-11", "categories": []}
{"id": "2212.10515", "title": "CausalDialogue: Modeling Utterance-level Causality in Conversations", "abstract": "  Despite their widespread adoption, neural conversation models have yet to\nexhibit natural chat capabilities with humans. In this research, we examine\nuser utterances as causes and generated responses as effects, recognizing that\nchanges in a cause should produce a different effect. To further explore this\nconcept, we have compiled and expanded upon a new dataset called CausalDialogue\nthrough crowd-sourcing. This dataset includes multiple cause-effect pairs\nwithin a directed acyclic graph (DAG) structure. Our analysis reveals that\ntraditional loss functions struggle to effectively incorporate the DAG\nstructure, leading us to propose a causality-enhanced method called Exponential\nMaximum Average Treatment Effect (ExMATE) to enhance the impact of causality at\nthe utterance level in training neural conversation models. To evaluate the\nneeds of considering causality in dialogue generation, we built a comprehensive\nbenchmark on CausalDialogue dataset using different models, inference, and\ntraining methods. Through experiments, we find that a causality-inspired loss\nlike ExMATE can improve the diversity and agility of conventional loss function\nand there is still room for improvement to reach human-level quality on this\nnew dataset.\n", "date": "2022-12-20", "categories": []}
{"id": "2307.03315", "title": "Assisting Clinical Decisions for Scarcely Available Treatment via\n  Disentangled Latent Representation", "abstract": "  Extracorporeal membrane oxygenation (ECMO) is an essential life-supporting\nmodality for COVID-19 patients who are refractory to conventional therapies.\nHowever, the proper treatment decision has been the subject of significant\ndebate and it remains controversial about who benefits from this scarcely\navailable and technically complex treatment option. To support clinical\ndecisions, it is a critical need to predict the treatment need and the\npotential treatment and no-treatment responses. Targeting this clinical\nchallenge, we propose Treatment Variational AutoEncoder (TVAE), a novel\napproach for individualized treatment analysis. TVAE is specifically designed\nto address the modeling challenges like ECMO with strong treatment selection\nbias and scarce treatment cases. TVAE conceptualizes the treatment decision as\na multi-scale problem. We model a patient's potential treatment assignment and\nthe factual and counterfactual outcomes as part of their intrinsic\ncharacteristics that can be represented by a deep latent variable model. The\nfactual and counterfactual prediction errors are alleviated via a\nreconstruction regularization scheme together with semi-supervision, and the\nselection bias and the scarcity of treatment cases are mitigated by the\ndisentangled and distribution-matched latent space and the label-balancing\ngenerative strategy. We evaluate TVAE on two real-world COVID-19 datasets: an\ninternational dataset collected from 1651 hospitals across 63 countries, and a\ninstitutional dataset collected from 15 hospitals. The results show that TVAE\noutperforms state-of-the-art treatment effect models in predicting both the\npropensity scores and factual outcomes on heterogeneous COVID-19 datasets.\nAdditional experiments also show TVAE outperforms the best existing models in\nindividual treatment effect estimation on the synthesized IHDP benchmark\ndataset.\n", "date": "2023-07-06", "categories": []}
{"id": "2307.03750", "title": "When does the ID algorithm fail?", "abstract": "  The ID algorithm solves the problem of identification of interventional\ndistributions of the form p(Y | do(a)) in graphical causal models, and has been\nformulated in a number of ways [12, 9, 6]. The ID algorithm is sound (outputs\nthe correct functional of the observed data distribution whenever p(Y | do(a))\nis identified in the causal model represented by the input graph), and complete\n(explicitly flags as a failure any input p(Y | do(a)) whenever this\ndistribution is not identified in the causal model represented by the input\ngraph).\n  The reference [9] provides a result, the so called \"hedge criterion\"\n(Corollary 3), which aims to give a graphical characterization of situations\nwhen the ID algorithm fails to identify its input in terms of a structure in\nthe input graph called the hedge. While the ID algorithm is, indeed, a sound\nand complete algorithm, and the hedge structure does arise whenever the input\ndistribution is not identified, Corollary 3 presented in [9] is incorrect as\nstated. In this note, I outline the modern presentation of the ID algorithm,\ndiscuss a simple counterexample to Corollary 3, and provide a number of\ngraphical characterizations of the ID algorithm failing to identify its input\ndistribution.\n", "date": "2023-07-07", "categories": []}
{"id": "2206.04356", "title": "A Simple Unified Approach to Testing High-Dimensional Conditional\n  Independences for Categorical and Ordinal Data", "abstract": "  Conditional independence (CI) tests underlie many approaches to model testing\nand structure learning in causal inference. Most existing CI tests for\ncategorical and ordinal data stratify the sample by the conditioning variables,\nperform simple independence tests in each stratum, and combine the results.\nUnfortunately, the statistical power of this approach degrades rapidly as the\nnumber of conditioning variables increases. Here we propose a simple unified CI\ntest for ordinal and categorical data that maintains reasonable calibration and\npower in high dimensions. We show that our test outperforms existing baselines\nin model testing and structure learning for dense directed graphical models\nwhile being comparable for sparse models. Our approach could be attractive for\ncausal model testing because it is easy to implement, can be used with\nnon-parametric or parametric probability models, has the symmetry property, and\nhas reasonable computational requirements.\n", "date": "2022-06-09", "categories": []}
{"id": "2303.04209", "title": "Causal Dependence Plots", "abstract": "  Explaining artificial intelligence or machine learning models is increasingly\nimportant. To use such data-driven systems wisely we must understand how they\ninteract with the world, including how they depend causally on data inputs. In\nthis work we develop Causal Dependence Plots (CDPs) to visualize how one\nvariable--an outcome--depends on changes in another variable--a\npredictor--$\\textit{along with any consequent causal changes in other predictor\nvariables}$. Crucially, CDPs differ from standard methods based on holding\nother predictors constant or assuming they are independent. CDPs make use of an\nauxiliary causal model because causal conclusions require causal assumptions.\nWith simulations and real data experiments, we show CDPs can be combined in a\nmodular way with methods for causal learning or sensitivity analysis. Since\npeople often think causally about input-output dependence, CDPs can be powerful\ntools in the xAI or interpretable machine learning toolkit and contribute to\napplications like scientific machine learning and algorithmic fairness.\n", "date": "2023-03-07", "categories": []}
{"id": "2306.07024", "title": "DRCFS: Doubly Robust Causal Feature Selection", "abstract": "  Knowing the features of a complex system that are highly relevant to a\nparticular target variable is of fundamental interest in many areas of science.\nExisting approaches are often limited to linear settings, sometimes lack\nguarantees, and in most cases, do not scale to the problem at hand, in\nparticular to images. We propose DRCFS, a doubly robust feature selection\nmethod for identifying the causal features even in nonlinear and high\ndimensional settings. We provide theoretical guarantees, illustrate necessary\nconditions for our assumptions, and perform extensive experiments across a wide\nrange of simulated and semi-synthetic datasets. DRCFS significantly outperforms\nexisting state-of-the-art methods, selecting robust features even in\nchallenging highly non-linear and high-dimensional problems.\n", "date": "2023-06-12", "categories": []}
{"id": "2307.00165", "title": "Counterfactual Collaborative Reasoning", "abstract": "  Causal reasoning and logical reasoning are two important types of reasoning\nabilities for human intelligence. However, their relationship has not been\nextensively explored under machine intelligence context. In this paper, we\nexplore how the two reasoning abilities can be jointly modeled to enhance both\naccuracy and explainability of machine learning models. More specifically, by\nintegrating two important types of reasoning ability -- counterfactual\nreasoning and (neural) logical reasoning -- we propose Counterfactual\nCollaborative Reasoning (CCR), which conducts counterfactual logic reasoning to\nimprove the performance. In particular, we use recommender system as an example\nto show how CCR alleviate data scarcity, improve accuracy and enhance\ntransparency. Technically, we leverage counterfactual reasoning to generate\n\"difficult\" counterfactual training examples for data augmentation, which --\ntogether with the original training examples -- can enhance the model\nperformance. Since the augmented data is model irrelevant, they can be used to\nenhance any model, enabling the wide applicability of the technique. Besides,\nmost of the existing data augmentation methods focus on \"implicit data\naugmentation\" over users' implicit feedback, while our framework conducts\n\"explicit data augmentation\" over users explicit feedback based on\ncounterfactual logic reasoning. Experiments on three real-world datasets show\nthat CCR achieves better performance than non-augmented models and implicitly\naugmented models, and also improves model transparency by generating\ncounterfactual explanations.\n", "date": "2023-06-30", "categories": []}
{"id": "2307.01389", "title": "Identification of Causal Relationship between Amyloid-beta Accumulation\n  and Alzheimer's Disease Progression via Counterfactual Inference", "abstract": "  Alzheimer's disease (AD) is a neurodegenerative disorder that is beginning\nwith amyloidosis, followed by neuronal loss and deterioration in structure,\nfunction, and cognition. The accumulation of amyloid-beta in the brain,\nmeasured through 18F-florbetapir (AV45) positron emission tomography (PET)\nimaging, has been widely used for early diagnosis of AD. However, the\nrelationship between amyloid-beta accumulation and AD pathophysiology remains\nunclear, and causal inference approaches are needed to uncover how amyloid-beta\nlevels can impact AD development. In this paper, we propose a graph varying\ncoefficient neural network (GVCNet) for estimating the individual treatment\neffect with continuous treatment levels using a graph convolutional neural\nnetwork. We highlight the potential of causal inference approaches, including\nGVCNet, for measuring the regional causal connections between amyloid-beta\naccumulation and AD pathophysiology, which may serve as a robust tool for early\ndiagnosis and tailored care.\n", "date": "2023-07-03", "categories": []}
{"id": "2307.02390", "title": "Causal Discovery with Language Models as Imperfect Experts", "abstract": "  Understanding the causal relationships that underlie a system is a\nfundamental prerequisite to accurate decision-making. In this work, we explore\nhow expert knowledge can be used to improve the data-driven identification of\ncausal graphs, beyond Markov equivalence classes. In doing so, we consider a\nsetting where we can query an expert about the orientation of causal\nrelationships between variables, but where the expert may provide erroneous\ninformation. We propose strategies for amending such expert knowledge based on\nconsistency properties, e.g., acyclicity and conditional independencies in the\nequivalence class. We then report a case study, on real data, where a large\nlanguage model is used as an imperfect expert.\n", "date": "2023-07-05", "categories": []}
{"id": "2210.17283", "title": "CausalBench: A Large-scale Benchmark for Network Inference from\n  Single-cell Perturbation Data", "abstract": "  Causal inference is a vital aspect of multiple scientific disciplines and is\nroutinely applied to high-impact applications such as medicine. However,\nevaluating the performance of causal inference methods in real-world\nenvironments is challenging due to the need for observations under both\ninterventional and control conditions. Traditional evaluations conducted on\nsynthetic datasets do not reflect the performance in real-world systems. To\naddress this, we introduce CausalBench, a benchmark suite for evaluating\nnetwork inference methods on real-world interventional data from large-scale\nsingle-cell perturbation experiments. CausalBench incorporates\nbiologically-motivated performance metrics, including new distribution-based\ninterventional metrics. A systematic evaluation of state-of-the-art causal\ninference methods using our CausalBench suite highlights how poor scalability\nof current methods limits performance. Moreover, methods that use\ninterventional information do not outperform those that only use observational\ndata, contrary to what is observed on synthetic benchmarks. Thus, CausalBench\nopens new avenues in causal network inference research and provides a\nprincipled and reliable way to track progress in leveraging real-world\ninterventional data.\n", "date": "2022-10-31", "categories": []}
{"id": "2307.00227", "title": "Causal Structure Learning by Using Intersection of Markov Blankets", "abstract": "  In this paper, we introduce a novel causal structure learning algorithm\ncalled Endogenous and Exogenous Markov Blankets Intersection (EEMBI), which\ncombines the properties of Bayesian networks and Structural Causal Models\n(SCM). Furthermore, we propose an extended version of EEMBI, namely EEMBI-PC,\nwhich integrates the last step of the PC algorithm into EEMBI.\n", "date": "2023-07-01", "categories": []}
{"id": "2304.05294", "title": "Selecting Robust Features for Machine Learning Applications using\n  Multidata Causal Discovery", "abstract": "  Robust feature selection is vital for creating reliable and interpretable\nMachine Learning (ML) models. When designing statistical prediction models in\ncases where domain knowledge is limited and underlying interactions are\nunknown, choosing the optimal set of features is often difficult. To mitigate\nthis issue, we introduce a Multidata (M) causal feature selection approach that\nsimultaneously processes an ensemble of time series datasets and produces a\nsingle set of causal drivers. This approach uses the causal discovery\nalgorithms PC1 or PCMCI that are implemented in the Tigramite Python package.\nThese algorithms utilize conditional independence tests to infer parts of the\ncausal graph. Our causal feature selection approach filters out\ncausally-spurious links before passing the remaining causal features as inputs\nto ML models (Multiple linear regression, Random Forest) that predict the\ntargets. We apply our framework to the statistical intensity prediction of\nWestern Pacific Tropical Cyclones (TC), for which it is often difficult to\naccurately choose drivers and their dimensionality reduction (time lags,\nvertical levels, and area-averaging). Using more stringent significance\nthresholds in the conditional independence tests helps eliminate spurious\ncausal relationships, thus helping the ML model generalize better to unseen TC\ncases. M-PC1 with a reduced number of features outperforms M-PCMCI, non-causal\nML, and other feature selection methods (lagged correlation, random), even\nslightly outperforming feature selection based on eXplainable Artificial\nIntelligence. The optimal causal drivers obtained from our causal feature\nselection help improve our understanding of underlying relationships and\nsuggest new potential drivers of TC intensification.\n", "date": "2023-04-11", "categories": []}
{"id": "2306.06002", "title": "Causal Effect Estimation from Observational and Interventional Data\n  Through Matrix Weighted Linear Estimators", "abstract": "  We study causal effect estimation from a mixture of observational and\ninterventional data in a confounded linear regression model with multivariate\ntreatments. We show that the statistical efficiency in terms of expected\nsquared error can be improved by combining estimators arising from both the\nobservational and interventional setting. To this end, we derive methods based\non matrix weighted linear estimators and prove that our methods are\nasymptotically unbiased in the infinite sample limit. This is an important\nimprovement compared to the pooled estimator using the union of interventional\nand observational data, for which the bias only vanishes if the ratio of\nobservational to interventional data tends to zero. Studies on synthetic data\nconfirm our theoretical findings. In settings where confounding is substantial\nand the ratio of observational to interventional data is large, our estimators\noutperform a Stein-type estimator and various other baselines.\n", "date": "2023-06-09", "categories": []}
{"id": "2306.17378", "title": "Global Optimality in Bivariate Gradient-based DAG Learning", "abstract": "  Recently, a new class of non-convex optimization problems motivated by the\nstatistical problem of learning an acyclic directed graphical model from data\nhas attracted significant interest. While existing work uses standard\nfirst-order optimization schemes to solve this problem, proving the global\noptimality of such approaches has proven elusive. The difficulty lies in the\nfact that unlike other non-convex problems in the literature, this problem is\nnot \"benign\", and possesses multiple spurious solutions that standard\napproaches can easily get trapped in. In this paper, we prove that a simple\npath-following optimization scheme globally converges to the global minimum of\nthe population loss in the bivariate setting.\n", "date": "2023-06-30", "categories": []}
{"id": "2210.06201", "title": "Diffusion Models for Causal Discovery via Topological Ordering", "abstract": "  Discovering causal relations from observational data becomes possible with\nadditional assumptions such as considering the functional relations to be\nconstrained as nonlinear with additive noise (ANM). Even with strong\nassumptions, causal discovery involves an expensive search problem over the\nspace of directed acyclic graphs (DAGs). \\emph{Topological ordering} approaches\nreduce the optimisation space of causal discovery by searching over a\npermutation rather than graph space. For ANMs, the \\emph{Hessian} of the data\nlog-likelihood can be used for finding leaf nodes in a causal graph, allowing\nits topological ordering. However, existing computational methods for obtaining\nthe Hessian still do not scale as the number of variables and the number of\nsamples increase. Therefore, inspired by recent innovations in diffusion\nprobabilistic models (DPMs), we propose \\emph{DiffAN}\\footnote{Implementation\nis available at \\url{https://github.com/vios-s/DiffAN} .}, a topological\nordering algorithm that leverages DPMs for learning a Hessian function. We\nintroduce theory for updating the learned Hessian without re-training the\nneural network, and we show that computing with a subset of samples gives an\naccurate approximation of the ordering, which allows scaling to datasets with\nmore samples and variables. We show empirically that our method scales\nexceptionally well to datasets with up to $500$ nodes and up to $10^5$ samples\nwhile still performing on par over small datasets with state-of-the-art causal\ndiscovery methods. Implementation is available at\nhttps://github.com/vios-s/DiffAN .\n", "date": "2022-10-12", "categories": []}
{"id": "2301.12383", "title": "On Heterogeneous Treatment Effects in Heterogeneous Causal Graphs", "abstract": "  Heterogeneity and comorbidity are two interwoven challenges associated with\nvarious healthcare problems that greatly hampered research on developing\neffective treatment and understanding of the underlying neurobiological\nmechanism. Very few studies have been conducted to investigate heterogeneous\ncausal effects (HCEs) in graphical contexts due to the lack of statistical\nmethods. To characterize this heterogeneity, we first conceptualize\nheterogeneous causal graphs (HCGs) by generalizing the causal graphical model\nwith confounder-based interactions and multiple mediators. Such confounders\nwith an interaction with the treatment are known as moderators. This allows us\nto flexibly produce HCGs given different moderators and explicitly characterize\nHCEs from the treatment or potential mediators on the outcome. We establish the\ntheoretical forms of HCEs and derive their properties at the individual level\nin both linear and nonlinear models. An interactive structural learning is\ndeveloped to estimate the complex HCGs and HCEs with confidence intervals\nprovided. Our method is empirically justified by extensive simulations and its\npractical usefulness is illustrated by exploring causality among psychiatric\ndisorders for trauma survivors.\n", "date": "2023-01-29", "categories": []}
{"id": "2306.13932", "title": "Tuning structure learning algorithms with out-of-sample and resampling\n  strategies", "abstract": "  One of the challenges practitioners face when applying structure learning\nalgorithms to their data involves determining a set of hyperparameters;\notherwise, a set of hyperparameter defaults is assumed. The optimal\nhyperparameter configuration often depends on multiple factors, including the\nsize and density of the usually unknown underlying true graph, the sample size\nof the input data, and the structure learning algorithm. We propose a novel\nhyperparameter tuning method, called the Out-of-sample Tuning for Structure\nLearning (OTSL), that employs out-of-sample and resampling strategies to\nestimate the optimal hyperparameter configuration for structure learning, given\nthe input data set and structure learning algorithm. Synthetic experiments show\nthat employing OTSL as a means to tune the hyperparameters of hybrid and\nscore-based structure learning algorithms leads to improvements in graphical\naccuracy compared to the state-of-the-art. We also illustrate the applicability\nof this approach to real datasets from different disciplines.\n", "date": "2023-06-24", "categories": []}
{"id": "2306.13242", "title": "Approximate Causal Effect Identification under Weak Confounding", "abstract": "  Causal effect estimation has been studied by many researchers when only\nobservational data is available. Sound and complete algorithms have been\ndeveloped for pointwise estimation of identifiable causal queries. For\nnon-identifiable causal queries, researchers developed polynomial programs to\nestimate tight bounds on causal effect. However, these are computationally\ndifficult to optimize for variables with large support sizes. In this paper, we\nanalyze the effect of \"weak confounding\" on causal estimands. More\nspecifically, under the assumption that the unobserved confounders that render\na query non-identifiable have small entropy, we propose an efficient linear\nprogram to derive the upper and lower bounds of the causal effect. We show that\nour bounds are consistent in the sense that as the entropy of unobserved\nconfounders goes to zero, the gap between the upper and lower bound vanishes.\nFinally, we conduct synthetic and real data simulations to compare our bounds\nwith the bounds obtained by the existing work that cannot incorporate such\nentropy constraints and show that our bounds are tighter for the setting with\nweak confounders.\n", "date": "2023-06-22", "categories": []}
{"id": "2306.13271", "title": "Variational Counterfactual Prediction under Runtime Domain Corruption", "abstract": "  To date, various neural methods have been proposed for causal effect\nestimation based on observational data, where a default assumption is the same\ndistribution and availability of variables at both training and inference\n(i.e., runtime) stages. However, distribution shift (i.e., domain shift) could\nhappen during runtime, and bigger challenges arise from the impaired\naccessibility of variables. This is commonly caused by increasing privacy and\nethical concerns, which can make arbitrary variables unavailable in the entire\nruntime data and imputation impractical. We term the co-occurrence of domain\nshift and inaccessible variables runtime domain corruption, which seriously\nimpairs the generalizability of a trained counterfactual predictor. To counter\nruntime domain corruption, we subsume counterfactual prediction under the\nnotion of domain adaptation. Specifically, we upper-bound the error w.r.t. the\ntarget domain (i.e., runtime covariates) by the sum of source domain error and\ninter-domain distribution distance. In addition, we build an adversarially\nunified variational causal effect model, named VEGAN, with a novel two-stage\nadversarial domain adaptation scheme to reduce the latent distribution\ndisparity between treated and control groups first, and between training and\nruntime variables afterwards. We demonstrate that VEGAN outperforms other\nstate-of-the-art baselines on individual-level treatment effect estimation in\nthe presence of runtime domain corruption on benchmark datasets.\n", "date": "2023-06-23", "categories": []}
{"id": "2306.12453", "title": "Learning Conditional Instrumental Variable Representation for Causal\n  Effect Estimation", "abstract": "  One of the fundamental challenges in causal inference is to estimate the\ncausal effect of a treatment on its outcome of interest from observational\ndata. However, causal effect estimation often suffers from the impacts of\nconfounding bias caused by unmeasured confounders that affect both the\ntreatment and the outcome. The instrumental variable (IV) approach is a\npowerful way to eliminate the confounding bias from latent confounders.\nHowever, the existing IV-based estimators require a nominated IV, and for a\nconditional IV (CIV) the corresponding conditioning set too, for causal effect\nestimation. This limits the application of IV-based estimators. In this paper,\nby leveraging the advantage of disentangled representation learning, we propose\na novel method, named DVAE.CIV, for learning and disentangling the\nrepresentations of CIV and the representations of its conditioning set for\ncausal effect estimations from data with latent confounders. Extensive\nexperimental results on both synthetic and real-world datasets demonstrate the\nsuperiority of the proposed DVAE.CIV method against the existing causal effect\nestimators.\n", "date": "2023-06-21", "categories": []}
{"id": "2301.11214", "title": "Returning The Favour: When Regression Benefits From Probabilistic Causal\n  Knowledge", "abstract": "  A directed acyclic graph (DAG) provides valuable prior knowledge that is\noften discarded in regression tasks in machine learning. We show that the\nindependences arising from the presence of collider structures in DAGs provide\nmeaningful inductive biases, which constrain the regression hypothesis space\nand improve predictive performance. We introduce collider regression, a\nframework to incorporate probabilistic causal knowledge from a collider in a\nregression problem. When the hypothesis space is a reproducing kernel Hilbert\nspace, we prove a strictly positive generalisation benefit under mild\nassumptions and provide closed-form estimators of the empirical risk minimiser.\nExperiments on synthetic and climate model data demonstrate performance gains\nof the proposed methodology.\n", "date": "2023-01-26", "categories": []}
{"id": "2305.06789", "title": "Integrating Nearest Neighbors with Neural Network Models for Treatment\n  Effect Estimation", "abstract": "  Treatment effect estimation is of high-importance for both researchers and\npractitioners across many scientific and industrial domains. The abundance of\nobservational data makes them increasingly used by researchers for the\nestimation of causal effects. However, these data suffer from biases, from\nseveral weaknesses, leading to inaccurate causal effect estimations, if not\nhandled properly. Therefore, several machine learning techniques have been\nproposed, most of them focusing on leveraging the predictive power of neural\nnetwork models to attain more precise estimation of causal effects. In this\nwork, we propose a new methodology, named Nearest Neighboring Information for\nCausal Inference (NNCI), for integrating valuable nearest neighboring\ninformation on neural network-based models for estimating treatment effects.\nThe proposed NNCI methodology is applied to some of the most well established\nneural network-based models for treatment effect estimation with the use of\nobservational data. Numerical experiments and analysis provide empirical and\nstatistical evidence that the integration of NNCI with state-of-the-art neural\nnetwork models leads to considerably improved treatment effect estimations on a\nvariety of well-known challenging benchmarks.\n", "date": "2023-05-11", "categories": []}
{"id": "2305.16863", "title": "Controlling Learned Effects to Reduce Spurious Correlations in Text\n  Classifiers", "abstract": "  To address the problem of NLP classifiers learning spurious correlations\nbetween training features and target labels, a common approach is to make the\nmodel's predictions invariant to these features. However, this can be\ncounter-productive when the features have a non-zero causal effect on the\ntarget label and thus are important for prediction. Therefore, using methods\nfrom the causal inference literature, we propose an algorithm to regularize the\nlearnt effect of the features on the model's prediction to the estimated effect\nof feature on label. This results in an automated augmentation method that\nleverages the estimated effect of a feature to appropriately change the labels\nfor new augmented inputs. On toxicity and IMDB review datasets, the proposed\nalgorithm minimises spurious correlations and improves the minority group\n(i.e., samples breaking spurious correlations) accuracy, while also improving\nthe total accuracy compared to standard training.\n", "date": "2023-05-26", "categories": []}
{"id": "2306.11755", "title": "On Identifiability of Conditional Causal Effects", "abstract": "  We address the problem of identifiability of an arbitrary conditional causal\neffect given both the causal graph and a set of any observational and/or\ninterventional distributions of the form $Q[S]:=P(S|do(V\\setminus S))$, where\n$V$ denotes the set of all observed variables and $S\\subseteq V$. We call this\nproblem conditional generalized identifiability (c-gID in short) and prove the\ncompleteness of Pearl's $do$-calculus for the c-gID problem by providing sound\nand complete algorithm for the c-gID problem. This work revisited the c-gID\nproblem in Lee et al. [2020], Correa et al. [2021] by adding explicitly the\npositivity assumption which is crucial for identifiability. It extends the\nresults of [Lee et al., 2019, Kivva et al., 2022] on general identifiability\n(gID) which studied the problem for unconditional causal effects and Shpitser\nand Pearl [2006b] on identifiability of conditional causal effects given merely\nthe observational distribution $P(\\mathbf{V})$ as our algorithm generalizes the\nalgorithms proposed in [Kivva et al., 2022] and [Shpitser and Pearl, 2006b].\n", "date": "2023-06-19", "categories": []}
{"id": "2306.11846", "title": "Discovering Causality for Efficient Cooperation in Multi-Agent\n  Environments", "abstract": "  In cooperative Multi-Agent Reinforcement Learning (MARL) agents are required\nto learn behaviours as a team to achieve a common goal. However, while learning\na task, some agents may end up learning sub-optimal policies, not contributing\nto the objective of the team. Such agents are called lazy agents due to their\nnon-cooperative behaviours that may arise from failing to understand whether\nthey caused the rewards. As a consequence, we observe that the emergence of\ncooperative behaviours is not necessarily a byproduct of being able to solve a\ntask as a team. In this paper, we investigate the applications of causality in\nMARL and how it can be applied in MARL to penalise these lazy agents. We\nobserve that causality estimations can be used to improve the credit assignment\nto the agents and show how it can be leveraged to improve independent learning\nin MARL. Furthermore, we investigate how Amortized Causal Discovery can be used\nto automate causality detection within MARL environments. The results\ndemonstrate that causality relations between individual observations and the\nteam reward can be used to detect and punish lazy agents, making them develop\nmore intelligent behaviours. This results in improvements not only in the\noverall performances of the team but also in their individual capabilities. In\naddition, results show that Amortized Causal Discovery can be used efficiently\nto find causal relations in MARL.\n", "date": "2023-06-20", "categories": []}
{"id": "2209.04142", "title": "Causal Modeling of Policy Interventions From Sequences of Treatments and\n  Outcomes", "abstract": "  A treatment policy defines when and what treatments are applied to affect\nsome outcome of interest. Data-driven decision-making requires the ability to\npredict what happens if a policy is changed. Existing methods that predict how\nthe outcome evolves under different scenarios assume that the tentative\nsequences of future treatments are fixed in advance, while in practice the\ntreatments are determined stochastically by a policy and may depend, for\nexample, on the efficiency of previous treatments. Therefore, the current\nmethods are not applicable if the treatment policy is unknown or a\ncounterfactual analysis is needed. To handle these limitations, we model the\ntreatments and outcomes jointly in continuous time, by combining Gaussian\nprocesses and point processes. Our model enables the estimation of a treatment\npolicy from observational sequences of treatments and outcomes, and it can\npredict the interventional and counterfactual progression of the outcome after\nan intervention on the treatment policy (in contrast with the causal effect of\na single treatment). We show with real-world and semi-synthetic data on blood\nglucose progression that our method can answer causal queries more accurately\nthan existing alternatives.\n", "date": "2022-09-09", "categories": []}
{"id": "2209.06203", "title": "Normalizing Flows for Interventional Density Estimation", "abstract": "  Existing machine learning methods for causal inference usually estimate\nquantities expressed via the mean of potential outcomes (e.g., average\ntreatment effect). However, such quantities do not capture the full information\nabout the distribution of potential outcomes. In this work, we estimate the\ndensity of potential outcomes after interventions from observational data. For\nthis, we propose a novel, fully-parametric deep learning method called\nInterventional Normalizing Flows. Specifically, we combine two normalizing\nflows, namely (i) a nuisance flow for estimating nuisance parameters and (ii) a\ntarget flow for parametric estimation of the density of potential outcomes. We\nfurther develop a tractable optimization objective based on a one-step bias\ncorrection for efficient and doubly robust estimation of the target flow\nparameters. As a result, our Interventional Normalizing Flows offer a properly\nnormalized density estimator. Across various experiments, we demonstrate that\nour Interventional Normalizing Flows are expressive and highly effective, and\nscale well with both sample size and high-dimensional confounding. To the best\nof our knowledge, our Interventional Normalizing Flows are the first proper\nfully-parametric, deep learning method for density estimation of potential\noutcomes.\n", "date": "2022-09-13", "categories": []}
{"id": "2305.08950", "title": "Causal Analysis for Robust Interpretability of Neural Networks", "abstract": "  Interpreting the inner function of neural networks is crucial for the\ntrustworthy development and deployment of these black-box models. Prior\ninterpretability methods focus on correlation-based measures to attribute model\ndecisions to individual examples. However, these measures are susceptible to\nnoise and spurious correlations encoded in the model during the training phase\n(e.g., biased inputs, model overfitting, or misspecification). Moreover, this\nprocess has proven to result in noisy and unstable attributions that prevent\nany transparent understanding of the model's behavior. In this paper, we\ndevelop a robust interventional-based method grounded by causal analysis to\ncapture cause-effect mechanisms in pre-trained neural networks and their\nrelation to the prediction. Our novel approach relies on path interventions to\ninfer the causal mechanisms within hidden layers and isolate relevant and\nnecessary information (to model prediction), avoiding noisy ones. The result is\ntask-specific causal explanatory graphs that can audit model behavior and\nexpress the actual causes underlying its performance. We apply our method to\nvision models trained on classification tasks. On image classification tasks,\nwe provide extensive quantitative experiments to show that our approach can\ncapture more stable and faithful explanations than standard attribution-based\nmethods. Furthermore, the underlying causal graphs reveal the neural\ninteractions in the model, making it a valuable tool in other applications\n(e.g., model repair).\n", "date": "2023-05-15", "categories": []}
{"id": "2306.10551", "title": "Can predictive models be used for causal inference?", "abstract": "  Supervised machine learning (ML) and deep learning (DL) algorithms excel at\npredictive tasks, but it is commonly assumed that they often do so by\nexploiting non-causal correlations, which may limit both interpretability and\ngeneralizability. Here, we show that this trade-off between explanation and\nprediction is not as deep and fundamental as expected. Whereas ML and DL\nalgorithms will indeed tend to use non-causal features for prediction when fed\nindiscriminately with all data, it is possible to constrain the learning\nprocess of any ML and DL algorithm by selecting features according to Pearl's\nbackdoor adjustment criterion. In such a situation, some algorithms, in\nparticular deep neural networks, can provide near unbiased effect estimates\nunder feature collinearity. Remaining biases are explained by the specific\nalgorithmic structures as well as hyperparameter choice. Consequently, optimal\nhyperparameter settings are different when tuned for prediction or inference,\nconfirming the general expectation of a trade-off between prediction and\nexplanation. However, the effect of this trade-off is small compared to the\neffect of a causally constrained feature selection. Thus, once the causal\nrelationship between the features is accounted for, the difference between\nprediction and explanation may be much smaller than commonly assumed. We also\nshow that such causally constrained models generalize better to new data with\naltered collinearity structures, suggesting generalization failure may often be\ndue to a lack of causal learning. Our results not only provide a perspective\nfor using ML for inference of (causal) effects but also help to improve the\ngeneralizability of fitted ML and DL models to new data.\n", "date": "2023-06-18", "categories": []}
{"id": "2306.11008", "title": "Front-door Adjustment Beyond Markov Equivalence with Limited Graph\n  Knowledge", "abstract": "  Causal effect estimation from data typically requires assumptions about the\ncause-effect relations either explicitly in the form of a causal graph\nstructure within the Pearlian framework, or implicitly in terms of\n(conditional) independence statements between counterfactual variables within\nthe potential outcomes framework. When the treatment variable and the outcome\nvariable are confounded, front-door adjustment is an important special case\nwhere, given the graph, causal effect of the treatment on the target can be\nestimated using post-treatment variables. However, the exact formula for\nfront-door adjustment depends on the structure of the graph, which is difficult\nto learn in practice. In this work, we provide testable conditional\nindependence statements to compute the causal effect using front-door-like\nadjustment without knowing the graph under limited structural side information.\nWe show that our method is applicable in scenarios where knowing the Markov\nequivalence class is not sufficient for causal effect estimation. We\ndemonstrate the effectiveness of our method on a class of random graphs as well\nas real causal fairness benchmarks.\n", "date": "2023-06-19", "categories": []}
{"id": "2306.11216", "title": "CF-GODE: Continuous-Time Causal Inference for Multi-Agent Dynamical\n  Systems", "abstract": "  Multi-agent dynamical systems refer to scenarios where multiple units\ninteract with each other and evolve collectively over time. To make informed\ndecisions in multi-agent dynamical systems, such as determining the optimal\nvaccine distribution plan, it is essential for decision-makers to estimate the\ncontinuous-time counterfactual outcomes. However, existing studies of causal\ninference over time rely on the assumption that units are mutually independent,\nwhich is not valid for multi-agent dynamical systems. In this paper, we aim to\nbridge this gap and study how to estimate counterfactual outcomes in\nmulti-agent dynamical systems. Causal inference in a multi-agent dynamical\nsystem has unique challenges: 1) Confounders are time-varying and are present\nin both individual unit covariates and those of other units; 2) Units are\naffected by not only their own but also others' treatments; 3) The treatments\nare naturally dynamic, such as receiving vaccines and boosters in a seasonal\nmanner. We model a multi-agent dynamical system as a graph and propose\nCounterFactual GraphODE (CF-GODE), a causal model that estimates\ncontinuous-time counterfactual outcomes in the presence of inter-dependencies\nbetween units. To facilitate continuous-time estimation, we propose\nTreatment-Induced GraphODE, a novel ordinary differential equation based on\nGNN, which incorporates dynamical treatments as additional inputs to predict\npotential outcomes over time. To remove confounding bias, we propose two domain\nadversarial learning based objectives that learn balanced continuous\nrepresentation trajectories, which are not predictive of treatments and\ninterference. We further provide theoretical justification to prove their\neffectiveness. Experiments on two semi-synthetic datasets confirm that CF-GODE\noutperforms baselines on counterfactual estimation. We also provide extensive\nanalyses to understand how our model works.\n", "date": "2023-06-20", "categories": []}
{"id": "2306.11417", "title": "PyRCA: A Library for Metric-based Root Cause Analysis", "abstract": "  We introduce PyRCA, an open-source Python machine learning library of Root\nCause Analysis (RCA) for Artificial Intelligence for IT Operations (AIOps). It\nprovides a holistic framework to uncover the complicated metric causal\ndependencies and automatically locate root causes of incidents. It offers a\nunified interface for multiple commonly used RCA models, encompassing both\ngraph construction and scoring tasks. This library aims to provide IT\noperations staff, data scientists, and researchers a one-step solution to rapid\nmodel development, model evaluation and deployment to online applications. In\nparticular, our library includes various causal discovery methods to support\ncausal graph construction, and multiple types of root cause scoring methods\ninspired by Bayesian analysis, graph analysis and causal analysis, etc. Our GUI\ndashboard offers practitioners an intuitive point-and-click interface,\nempowering them to easily inject expert knowledge through human interaction.\nWith the ability to visualize causal graphs and the root cause of incidents,\npractitioners can quickly gain insights and improve their workflow efficiency.\nThis technical report introduces PyRCA's architecture and major\nfunctionalities, while also presenting benchmark performance numbers in\ncomparison to various baseline models. Additionally, we demonstrate PyRCA's\ncapabilities through several example use cases.\n", "date": "2023-06-20", "categories": []}
{"id": "2306.09643", "title": "BISCUIT: Causal Representation Learning from Binary Interactions", "abstract": "  Identifying the causal variables of an environment and how to intervene on\nthem is of core value in applications such as robotics and embodied AI. While\nan agent can commonly interact with the environment and may implicitly perturb\nthe behavior of some of these causal variables, often the targets it affects\nremain unknown. In this paper, we show that causal variables can still be\nidentified for many common setups, e.g., additive Gaussian noise models, if the\nagent's interactions with a causal variable can be described by an unknown\nbinary variable. This happens when each causal variable has two different\nmechanisms, e.g., an observational and an interventional one. Using this\nidentifiability result, we propose BISCUIT, a method for simultaneously\nlearning causal variables and their corresponding binary interaction variables.\nOn three robotic-inspired datasets, BISCUIT accurately identifies causal\nvariables and can even be scaled to complex, realistic environments for\nembodied AI.\n", "date": "2023-06-16", "categories": []}
{"id": "2306.09656", "title": "Temporal Causal Mediation through a Point Process: Direct and Indirect\n  Effects of Healthcare Interventions", "abstract": "  Deciding on an appropriate intervention requires a causal model of a\ntreatment, the outcome, and potential mediators. Causal mediation analysis lets\nus distinguish between direct and indirect effects of the intervention, but has\nmostly been studied in a static setting. In healthcare, data come in the form\nof complex, irregularly sampled time-series, with dynamic interdependencies\nbetween a treatment, outcomes, and mediators across time. Existing approaches\nto dynamic causal mediation analysis are limited to regular measurement\nintervals, simple parametric models, and disregard long-range mediator--outcome\ninteractions. To address these limitations, we propose a non-parametric\nmediator--outcome model where the mediator is assumed to be a temporal point\nprocess that interacts with the outcome process. With this model, we estimate\nthe direct and indirect effects of an external intervention on the outcome,\nshowing how each of these affects the whole future trajectory. We demonstrate\non semi-synthetic data that our method can accurately estimate direct and\nindirect effects. On real-world healthcare data, our model infers clinically\nmeaningful direct and indirect effect trajectories for blood glucose after a\nsurgery.\n", "date": "2023-06-16", "categories": []}
{"id": "2302.05052", "title": "Debiasing Recommendation by Learning Identifiable Latent Confounders", "abstract": "  Recommendation systems aim to predict users' feedback on items not exposed to\nthem.\n  Confounding bias arises due to the presence of unmeasured variables (e.g.,\nthe socio-economic status of a user) that can affect both a user's exposure and\nfeedback. Existing methods either (1) make untenable assumptions about these\nunmeasured variables or (2) directly infer latent confounders from users'\nexposure. However, they cannot guarantee the identification of counterfactual\nfeedback, which can lead to biased predictions. In this work, we propose a\nnovel method, i.e., identifiable deconfounder (iDCF), which leverages a set of\nproxy variables (e.g., observed user features) to resolve the aforementioned\nnon-identification issue. The proposed iDCF is a general deconfounded\nrecommendation framework that applies proximal causal inference to infer the\nunmeasured confounders and identify the counterfactual feedback with\ntheoretical guarantees. Extensive experiments on various real-world and\nsynthetic datasets verify the proposed method's effectiveness and robustness.\n", "date": "2023-02-10", "categories": []}
{"id": "2306.09302", "title": "Knowledge Guided Representation Learning and Causal Structure Learning\n  in Soil Science", "abstract": "  An improved understanding of soil can enable more sustainable land-use\npractices. Nevertheless, soil is called a complex, living medium due to the\ncomplex interaction of different soil processes that limit our understanding of\nsoil. Process-based models and analyzing observed data provide two avenues for\nimproving our understanding of soil processes. Collecting observed data is\ncost-prohibitive but reflects real-world behavior, while process-based models\ncan be used to generate ample synthetic data which may not be representative of\nreality. We propose a framework, knowledge-guided representation learning, and\ncausal structure learning (KGRCL), to accelerate scientific discoveries in soil\nscience. The framework improves representation learning for simulated soil\nprocesses via conditional distribution matching with observed soil processes.\nSimultaneously, the framework leverages both observed and simulated data to\nlearn a causal structure among the soil processes. The learned causal graph is\nmore representative of ground truth than other graphs generated from other\ncausal discovery methods. Furthermore, the learned causal graph is leveraged in\na supervised learning setup to predict the impact of fertilizer use and\nchanging weather on soil carbon. We present the results in five different\nlocations to show the improvement in the prediction performance in\nout-of-sample and few-shots setting.\n", "date": "2023-06-15", "categories": []}
{"id": "2212.03063", "title": "Causal Inference via Style Transfer for Out-of-distribution\n  Generalisation", "abstract": "  Out-of-distribution (OOD) generalisation aims to build a model that can\ngeneralise well on an unseen target domain using knowledge from multiple source\ndomains. To this end, the model should seek the causal dependence between\ninputs and labels, which may be determined by the semantics of inputs and\nremain invariant across domains. However, statistical or non-causal methods\noften cannot capture this dependence and perform poorly due to not considering\nspurious correlations learnt from model training via unobserved confounders. A\nwell-known existing causal inference method like back-door adjustment cannot be\napplied to remove spurious correlations as it requires the observation of\nconfounders. In this paper, we propose a novel method that effectively deals\nwith hidden confounders by successfully implementing front-door adjustment\n(FA). FA requires the choice of a mediator, which we regard as the semantic\ninformation of images that helps access the causal mechanism without the need\nfor observing confounders. Further, we propose to estimate the combination of\nthe mediator with other observed images in the front-door formula via style\ntransfer algorithms. Our use of style transfer to estimate FA is novel and\nsensible for OOD generalisation, which we justify by extensive experimental\nresults on widely used benchmark datasets.\n", "date": "2022-12-06", "categories": []}
{"id": "2305.20011", "title": "Constrained Causal Bayesian Optimization", "abstract": "  We propose constrained causal Bayesian optimization (cCBO), an approach for\nfinding interventions in a known causal graph that optimize a target variable\nunder some constraints. cCBO first reduces the search space by exploiting the\ngraph structure and, if available, an observational dataset; and then solves\nthe restricted optimization problem by modelling target and constraint\nquantities using Gaussian processes and by sequentially selecting interventions\nvia a constrained expected improvement acquisition function. We propose\ndifferent surrogate models that enable to integrate observational and\ninterventional data while capturing correlation among effects with increasing\nlevels of sophistication. We evaluate cCBO on artificial and real-world causal\ngraphs showing successful trade off between fast convergence and percentage of\nfeasible interventions.\n", "date": "2023-05-31", "categories": []}
{"id": "2306.06409", "title": "Functional Causal Bayesian Optimization", "abstract": "  We propose functional causal Bayesian optimization (fCBO), a method for\nfinding interventions that optimize a target variable in a known causal graph.\nfCBO extends the CBO family of methods to enable functional interventions,\nwhich set a variable to be a deterministic function of other variables in the\ngraph. fCBO models the unknown objectives with Gaussian processes whose inputs\nare defined in a reproducing kernel Hilbert space, thus allowing to compute\ndistances among vector-valued functions. In turn, this enables to sequentially\nselect functions to explore by maximizing an expected improvement acquisition\nfunctional while keeping the typical computational tractability of standard BO\nsettings. We introduce graphical criteria that establish when considering\nfunctional interventions allows attaining better target effects, and conditions\nunder which selected interventions are also optimal for conditional target\neffects. We demonstrate the benefits of the method in a synthetic and in a\nreal-world causal graph.\n", "date": "2023-06-10", "categories": []}
{"id": "2306.07858", "title": "Additive Causal Bandits with Unknown Graph", "abstract": "  We explore algorithms to select actions in the causal bandit setting where\nthe learner can choose to intervene on a set of random variables related by a\ncausal graph, and the learner sequentially chooses interventions and observes a\nsample from the interventional distribution. The learner's goal is to quickly\nfind the intervention, among all interventions on observable variables, that\nmaximizes the expectation of an outcome variable. We depart from previous\nliterature by assuming no knowledge of the causal graph except that latent\nconfounders between the outcome and its ancestors are not present. We first\nshow that the unknown graph problem can be exponentially hard in the parents of\nthe outcome. To remedy this, we adopt an additional additive assumption on the\noutcome which allows us to solve the problem by casting it as an additive\ncombinatorial linear bandit problem with full-bandit feedback. We propose a\nnovel action-elimination algorithm for this setting, show how to apply this\nalgorithm to the causal bandit problem, provide sample complexity bounds, and\nempirically validate our findings on a suite of randomly generated causal\nmodels, effectively showing that one does not need to explicitly learn the\nparents of the outcome to identify the best intervention.\n", "date": "2023-06-13", "categories": []}
{"id": "2306.07918", "title": "Causal Mediation Analysis with Multi-dimensional and Indirectly Observed\n  Mediators", "abstract": "  Causal mediation analysis (CMA) is a powerful method to dissect the total\neffect of a treatment into direct and mediated effects within the potential\noutcome framework. This is important in many scientific applications to\nidentify the underlying mechanisms of a treatment effect. However, in many\nscientific applications the mediator is unobserved, but there may exist related\nmeasurements. For example, we may want to identify how changes in brain\nactivity or structure mediate an antidepressant's effect on behavior, but we\nmay only have access to electrophysiological or imaging brain measurements. To\ndate, most CMA methods assume that the mediator is one-dimensional and\nobservable, which oversimplifies such real-world scenarios. To overcome this\nlimitation, we introduce a CMA framework that can handle complex and indirectly\nobserved mediators based on the identifiable variational autoencoder (iVAE)\narchitecture. We prove that the true joint distribution over observed and\nlatent variables is identifiable with the proposed method. Additionally, our\nframework captures a disentangled representation of the indirectly observed\nmediator and yields accurate estimation of the direct and mediated effects in\nsynthetic and semi-synthetic experiments, providing evidence of its potential\nutility in real-world applications.\n", "date": "2023-06-13", "categories": []}
{"id": "2206.06354", "title": "Differentiable and Transportable Structure Learning", "abstract": "  Directed acyclic graphs (DAGs) encode a lot of information about a particular\ndistribution in their structure. However, compute required to infer these\nstructures is typically super-exponential in the number of variables, as\ninference requires a sweep of a combinatorially large space of potential\nstructures. That is, until recent advances made it possible to search this\nspace using a differentiable metric, drastically reducing search time. While\nthis technique -- named NOTEARS -- is widely considered a seminal work in\nDAG-discovery, it concedes an important property in favour of\ndifferentiability: transportability. To be transportable, the structures\ndiscovered on one dataset must apply to another dataset from the same domain.\nWe introduce D-Struct which recovers transportability in the discovered\nstructures through a novel architecture and loss function while remaining fully\ndifferentiable. Because D-Struct remains differentiable, our method can be\neasily adopted in existing differentiable architectures, as was previously done\nwith NOTEARS. In our experiments, we empirically validate D-Struct with respect\nto edge accuracy and structural Hamming distance in a variety of settings.\n", "date": "2022-06-13", "categories": []}
{"id": "2211.16467", "title": "Linear Causal Disentanglement via Interventions", "abstract": "  Causal disentanglement seeks a representation of data involving latent\nvariables that relate to one another via a causal model. A representation is\nidentifiable if both the latent model and the transformation from latent to\nobserved variables are unique. In this paper, we study observed variables that\nare a linear transformation of a linear latent causal model. Data from\ninterventions are necessary for identifiability: if one latent variable is\nmissing an intervention, we show that there exist distinct models that cannot\nbe distinguished. Conversely, we show that a single intervention on each latent\nvariable is sufficient for identifiability. Our proof uses a generalization of\nthe RQ decomposition of a matrix that replaces the usual orthogonal and upper\ntriangular conditions with analogues depending on a partial order on the rows\nof the matrix, with partial order determined by a latent causal model. We\ncorroborate our theoretical results with a method for causal disentanglement\nthat accurately recovers a latent causal model.\n", "date": "2022-11-29", "categories": []}
{"id": "2306.00585", "title": "Causal Imitability Under Context-Specific Independence Relations", "abstract": "  Drawbacks of ignoring the causal mechanisms when performing imitation\nlearning have recently been acknowledged. Several approaches both to assess the\nfeasibility of imitation and to circumvent causal confounding and causal\nmisspecifications have been proposed in the literature. However, the potential\nbenefits of the incorporation of additional information about the underlying\ncausal structure are left unexplored. An example of such overlooked information\nis context-specific independence (CSI), i.e., independence that holds only in\ncertain contexts. We consider the problem of causal imitation learning when CSI\nrelations are known. We prove that the decision problem pertaining to the\nfeasibility of imitation in this setting is NP-hard. Further, we provide a\nnecessary graphical criterion for imitation learning under CSI and show that\nunder a structural assumption, this criterion is also sufficient. Finally, we\npropose a sound algorithmic approach for causal imitation learning which takes\nboth CSI relations and data into account.\n", "date": "2023-06-01", "categories": []}
{"id": "2211.15355", "title": "Causal Deep Reinforcement Learning Using Observational Data", "abstract": "  Deep reinforcement learning (DRL) requires the collection of interventional\ndata, which is sometimes expensive and even unethical in the real world, such\nas in the autonomous driving and the medical field. Offline reinforcement\nlearning promises to alleviate this issue by exploiting the vast amount of\nobservational data available in the real world. However, observational data may\nmislead the learning agent to undesirable outcomes if the behavior policy that\ngenerates the data depends on unobserved random variables (i.e., confounders).\nIn this paper, we propose two deconfounding methods in DRL to address this\nproblem. The methods first calculate the importance degree of different samples\nbased on the causal inference technique, and then adjust the impact of\ndifferent samples on the loss function by reweighting or resampling the offline\ndataset to ensure its unbiasedness. These deconfounding methods can be flexibly\ncombined with existing model-free DRL algorithms such as soft actor-critic and\ndeep Q-learning, provided that a weak condition can be satisfied by the loss\nfunctions of these algorithms. We prove the effectiveness of our deconfounding\nmethods and validate them experimentally.\n", "date": "2022-11-28", "categories": []}
{"id": "2302.06280", "title": "Causal Strategic Classification: A Tale of Two Shifts", "abstract": "  When users can benefit from certain predictive outcomes, they may be prone to\nact to achieve those outcome, e.g., by strategically modifying their features.\nThe goal in strategic classification is therefore to train predictive models\nthat are robust to such behavior. However, the conventional framework assumes\nthat changing features does not change actual outcomes, which depicts users as\n\"gaming\" the system. Here we remove this assumption, and study learning in a\ncausal strategic setting where true outcomes do change. Focusing on accuracy as\nour primary objective, we show how strategic behavior and causal effects\nunderlie two complementing forms of distribution shift. We characterize these\nshifts, and propose a learning algorithm that balances between these two forces\nand over time, and permits end-to-end training. Experiments on synthetic and\nsemi-synthetic data demonstrate the utility of our approach.\n", "date": "2023-02-13", "categories": []}
{"id": "2306.05484", "title": "Task-specific experimental design for treatment effect estimation", "abstract": "  Understanding causality should be a core requirement of any attempt to build\nreal impact through AI. Due to the inherent unobservability of counterfactuals,\nlarge randomised trials (RCTs) are the standard for causal inference. But large\nexperiments are generically expensive, and randomisation carries its own costs,\ne.g. when suboptimal decisions are trialed. Recent work has proposed more\nsample-efficient alternatives to RCTs, but these are not adaptable to the\ndownstream application for which the causal effect is sought. In this work, we\ndevelop a task-specific approach to experimental design and derive sampling\nstrategies customised to particular downstream applications. Across a range of\nimportant tasks, real-world datasets, and sample sizes, our method outperforms\nother benchmarks, e.g. requiring an order-of-magnitude less data to match RCT\nperformance on targeted marketing tasks.\n", "date": "2023-06-08", "categories": []}
{"id": "2306.05781", "title": "Adaptivity Complexity for Causal Graph Discovery", "abstract": "  Causal discovery from interventional data is an important problem, where the\ntask is to design an interventional strategy that learns the hidden ground\ntruth causal graph $G(V,E)$ on $|V| = n$ nodes while minimizing the number of\nperformed interventions. Most prior interventional strategies broadly fall into\ntwo categories: non-adaptive and adaptive. Non-adaptive strategies decide on a\nsingle fixed set of interventions to be performed while adaptive strategies can\ndecide on which nodes to intervene on sequentially based on past interventions.\nWhile adaptive algorithms may use exponentially fewer interventions than their\nnon-adaptive counterparts, there are practical concerns that constrain the\namount of adaptivity allowed. Motivated by this trade-off, we study the problem\nof $r$-adaptivity, where the algorithm designer recovers the causal graph under\na total of $r$ sequential rounds whilst trying to minimize the total number of\ninterventions. For this problem, we provide a $r$-adaptive algorithm that\nachieves $O(\\min\\{r,\\log n\\} \\cdot n^{1/\\min\\{r,\\log n\\}})$ approximation with\nrespect to the verification number, a well-known lower bound for adaptive\nalgorithms. Furthermore, for every $r$, we show that our approximation is\ntight. Our definition of $r$-adaptivity interpolates nicely between the\nnon-adaptive ($r=1$) and fully adaptive ($r=n$) settings where our\napproximation simplifies to $O(n)$ and $O(\\log n)$ respectively, matching the\nbest-known approximation guarantees for both extremes. Our results also extend\nnaturally to the bounded size interventions.\n", "date": "2023-06-09", "categories": []}
{"id": "2301.11401", "title": "Causal Bandits without Graph Learning", "abstract": "  We study the causal bandit problem when the causal graph is unknown and\ndevelop an efficient algorithm for finding the parent node of the reward node\nusing atomic interventions. We derive the exact equation for the expected\nnumber of interventions performed by the algorithm and show that under certain\ngraphical conditions it could perform either logarithmically fast or, under\nmore general assumptions, slower but still sublinearly in the number of\nvariables. We formally show that our algorithm is optimal as it meets the\nuniversal lower bound we establish for any algorithm that performs atomic\ninterventions. Finally, we extend our algorithm to the case when the reward\nnode has multiple parents. Using this algorithm together with a standard\nalgorithm from bandit literature leads to improved regret bounds.\n", "date": "2023-01-26", "categories": []}
{"id": "2306.04792", "title": "On the Use of Generative Models in Observational Causal Analysis", "abstract": "  The use of a hypothetical generative model was been suggested for causal\nanalysis of observational data. The very assumption of a particular model is a\ncommitment to a certain set of variables and therefore to a certain set of\npossible causes. Estimating the joint probability distribution of can be useful\nfor predicting values of variables in view of the observed values of others,\nbut it is not sufficient for inferring causal relationships. The model\ndescribes a single observable distribution and cannot a chain of effects of\nintervention that deviate from the observed distribution.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.05066", "title": "Causal Fairness for Outcome Control", "abstract": "  As society transitions towards an AI-based decision-making infrastructure, an\never-increasing number of decisions once under control of humans are now\ndelegated to automated systems. Even though such developments make various\nparts of society more efficient, a large body of evidence suggests that a great\ndeal of care needs to be taken to make such automated decision-making systems\nfair and equitable, namely, taking into account sensitive attributes such as\ngender, race, and religion. In this paper, we study a specific decision-making\ntask called outcome control in which an automated system aims to optimize an\noutcome variable $Y$ while being fair and equitable. The interest in such a\nsetting ranges from interventions related to criminal justice and welfare, all\nthe way to clinical decision-making and public health. In this paper, we first\nanalyze through causal lenses the notion of benefit, which captures how much a\nspecific individual would benefit from a positive decision, counterfactually\nspeaking, when contrasted with an alternative, negative one. We introduce the\nnotion of benefit fairness, which can be seen as the minimal fairness\nrequirement in decision-making, and develop an algorithm for satisfying it. We\nthen note that the benefit itself may be influenced by the protected attribute,\nand propose causal tools which can be used to analyze this. Finally, if some of\nthe variations of the protected attribute in the benefit are considered as\ndiscriminatory, the notion of benefit fairness may need to be strengthened,\nwhich leads us to articulating a notion of causal benefit fairness. Using this\nnotion, we develop a new optimization procedure capable of maximizing $Y$ while\nascertaining causal fairness in the decision process.\n", "date": "2023-06-08", "categories": []}
{"id": "2306.05071", "title": "A Causal Framework for Decomposing Spurious Variations", "abstract": "  One of the fundamental challenges found throughout the data sciences is to\nexplain why things happen in specific ways, or through which mechanisms a\ncertain variable $X$ exerts influences over another variable $Y$. In statistics\nand machine learning, significant efforts have been put into developing\nmachinery to estimate correlations across variables efficiently. In causal\ninference, a large body of literature is concerned with the decomposition of\ncausal effects under the rubric of mediation analysis. However, many variations\nare spurious in nature, including different phenomena throughout the applied\nsciences. Despite the statistical power to estimate correlations and the\nidentification power to decompose causal effects, there is still little\nunderstanding of the properties of spurious associations and how they can be\ndecomposed in terms of the underlying causal mechanisms. In this manuscript, we\ndevelop formal tools for decomposing spurious variations in both Markovian and\nSemi-Markovian models. We prove the first results that allow a non-parametric\ndecomposition of spurious effects and provide sufficient conditions for the\nidentification of such decompositions. The described approach has several\napplications, ranging from explainable and fair AI to questions in epidemiology\nand medicine, and we empirically demonstrate its use on a real-world dataset.\n", "date": "2023-06-08", "categories": []}
{"id": "2302.02228", "title": "Counterfactual Identifiability of Bijective Causal Models", "abstract": "  We study counterfactual identifiability in causal models with bijective\ngeneration mechanisms (BGM), a class that generalizes several widely-used\ncausal models in the literature. We establish their counterfactual\nidentifiability for three common causal structures with unobserved confounding,\nand propose a practical learning method that casts learning a BGM as structured\ngenerative modeling. Learned BGMs enable efficient counterfactual estimation\nand can be obtained using a variety of deep conditional generative models. We\nevaluate our techniques in a visual task and demonstrate its application in a\nreal-world video streaming simulation task.\n", "date": "2023-02-04", "categories": []}
{"id": "2306.04223", "title": "Causally Learning an Optimal Rework Policy", "abstract": "  In manufacturing, rework refers to an optional step of a production process\nwhich aims to eliminate errors or remedy products that do not meet the desired\nquality standards. Reworking a production lot involves repeating a previous\nproduction stage with adjustments to ensure that the final product meets the\nrequired specifications. While offering the chance to improve the yield and\nthus increase the revenue of a production lot, a rework step also incurs\nadditional costs. Additionally, the rework of parts that already meet the\ntarget specifications may damage them and decrease the yield. In this paper, we\napply double/debiased machine learning (DML) to estimate the conditional\ntreatment effect of a rework step during the color conversion process in\nopto-electronic semiconductor manufacturing on the final product yield. We\nutilize the implementation DoubleML to develop policies for the rework of\ncomponents and estimate their value empirically. From our causal machine\nlearning analysis we derive implications for the coating of monochromatic LEDs\nwith conversion layers.\n", "date": "2023-06-07", "categories": []}
{"id": "2306.04255", "title": "Accounting For Informative Sampling When Learning to Forecast Treatment\n  Outcomes Over Time", "abstract": "  Machine learning (ML) holds great potential for accurately forecasting\ntreatment outcomes over time, which could ultimately enable the adoption of\nmore individualized treatment strategies in many practical applications.\nHowever, a significant challenge that has been largely overlooked by the ML\nliterature on this topic is the presence of informative sampling in\nobservational data. When instances are observed irregularly over time, sampling\ntimes are typically not random, but rather informative -- depending on the\ninstance's characteristics, past outcomes, and administered treatments. In this\nwork, we formalize informative sampling as a covariate shift problem and show\nthat it can prohibit accurate estimation of treatment outcomes if not properly\naccounted for. To overcome this challenge, we present a general framework for\nlearning treatment outcomes in the presence of informative sampling using\ninverse intensity-weighting, and propose a novel method, TESAR-CDE, that\ninstantiates this framework using Neural CDEs. Using a simulation environment\nbased on a clinical use case, we demonstrate the effectiveness of our approach\nin learning under informative sampling.\n", "date": "2023-06-07", "categories": []}
{"id": "2210.00380", "title": "Transfer Learning for Individual Treatment Effect Estimation", "abstract": "  This work considers the problem of transferring causal knowledge between\ntasks for Individual Treatment Effect (ITE) estimation. To this end, we\ntheoretically assess the feasibility of transferring ITE knowledge and present\na practical framework for efficient transfer. A lower bound is introduced on\nthe ITE error of the target task to demonstrate that ITE knowledge transfer is\nchallenging due to the absence of counterfactual information. Nevertheless, we\nestablish generalization upper bounds on the counterfactual loss and ITE error\nof the target task, demonstrating the feasibility of ITE knowledge transfer.\nSubsequently, we introduce a framework with a new Causal Inference Task\nAffinity (CITA) measure for ITE knowledge transfer. Specifically, we use CITA\nto find the closest source task to the target task and utilize it for ITE\nknowledge transfer. Empirical studies are provided, demonstrating the efficacy\nof the proposed method. We observe that ITE knowledge transfer can\nsignificantly (up to 95%) reduce the amount of data required for ITE\nestimation.\n", "date": "2022-10-01", "categories": []}
{"id": "2301.08987", "title": "Tier Balancing: Towards Dynamic Fairness over Underlying Causal Factors", "abstract": "  The pursuit of long-term fairness involves the interplay between\ndecision-making and the underlying data generating process. In this paper,\nthrough causal modeling with a directed acyclic graph (DAG) on the\ndecision-distribution interplay, we investigate the possibility of achieving\nlong-term fairness from a dynamic perspective. We propose Tier Balancing, a\ntechnically more challenging but more natural notion to achieve in the context\nof long-term, dynamic fairness analysis. Different from previous fairness\nnotions that are defined purely on observed variables, our notion goes one step\nfurther, capturing behind-the-scenes situation changes on the unobserved latent\ncausal factors that directly carry out the influence from the current decision\nto the future data distribution. Under the specified dynamics, we prove that in\ngeneral one cannot achieve the long-term fairness goal only through one-step\ninterventions. Furthermore, in the effort of approaching long-term fairness, we\nconsider the mission of \"getting closer to\" the long-term fairness goal and\npresent possibility and impossibility results accordingly.\n", "date": "2023-01-21", "categories": []}
{"id": "2302.02923", "title": "In Search of Insights, Not Magic Bullets: Towards Demystification of the\n  Model Selection Dilemma in Heterogeneous Treatment Effect Estimation", "abstract": "  Personalized treatment effect estimates are often of interest in high-stakes\napplications -- thus, before deploying a model estimating such effects in\npractice, one needs to be sure that the best candidate from the ever-growing\nmachine learning toolbox for this task was chosen. Unfortunately, due to the\nabsence of counterfactual information in practice, it is usually not possible\nto rely on standard validation metrics for doing so, leading to a well-known\nmodel selection dilemma in the treatment effect estimation literature. While\nsome solutions have recently been investigated, systematic understanding of the\nstrengths and weaknesses of different model selection criteria is still\nlacking. In this paper, instead of attempting to declare a global `winner', we\ntherefore empirically investigate success- and failure modes of different\nselection criteria. We highlight that there is a complex interplay between\nselection strategies, candidate estimators and the data used for comparing\nthem, and provide interesting insights into the relative (dis)advantages of\ndifferent criteria alongside desiderata for the design of further illuminating\nempirical studies in this context.\n", "date": "2023-02-06", "categories": []}
{"id": "2302.14011", "title": "Causal isotonic calibration for heterogeneous treatment effects", "abstract": "  We propose causal isotonic calibration, a novel nonparametric method for\ncalibrating predictors of heterogeneous treatment effects. Furthermore, we\nintroduce cross-calibration, a data-efficient variant of calibration that\neliminates the need for hold-out calibration sets. Cross-calibration leverages\ncross-fitted predictors and generates a single calibrated predictor using all\navailable data. Under weak conditions that do not assume monotonicity, we\nestablish that both causal isotonic calibration and cross-calibration achieve\nfast doubly-robust calibration rates, as long as either the propensity score or\noutcome regression is estimated accurately in a suitable sense. The proposed\ncausal isotonic calibrator can be wrapped around any black-box learning\nalgorithm, providing robust and distribution-free calibration guarantees while\npreserving predictive performance.\n", "date": "2023-02-27", "categories": []}
{"id": "2211.05551", "title": "Causal Counterfactuals for Improving the Robustness of Reinforcement\n  Learning", "abstract": "  Reinforcement learning (RL) is used in various robotic applications. RL\nenables agents to learn tasks autonomously by interacting with the environment.\nThe more critical the tasks are, the higher the demand for the robustness of\nthe RL systems. Causal RL combines RL and causal inference to make RL more\nrobust. Causal RL agents use a causal representation to capture the invariant\ncausal mechanisms that can be transferred from one task to another. Currently,\nthere is limited research in Causal RL, and existing solutions are usually not\ncomplete or feasible for real-world applications. In this work, we propose\nCausalCF, the first complete Causal RL solution incorporating ideas from Causal\nCuriosity and CoPhy. Causal Curiosity provides an approach for using\ninterventions, and CoPhy is modified to enable the RL agent to perform\ncounterfactuals. Causal Curiosity has been applied to robotic grasping and\nmanipulation tasks in CausalWorld. CausalWorld provides a realistic simulation\nenvironment based on the TriFinger robot. We apply CausalCF to complex robotic\ntasks and show that it improves the RL agent's robustness using CausalWorld.\n", "date": "2022-11-02", "categories": []}
{"id": "2305.10638", "title": "Disentangled Causal Graph Learning for Online Unsupervised Root Cause\n  Analysis", "abstract": "  The task of root cause analysis (RCA) is to identify the root causes of\nsystem faults/failures by analyzing system monitoring data. Efficient RCA can\ngreatly accelerate system failure recovery and mitigate system damages or\nfinancial losses. However, previous research has mostly focused on developing\noffline RCA algorithms, which often require manually initiating the RCA\nprocess, a significant amount of time and data to train a robust model, and\nthen being retrained from scratch for a new system fault.\n  In this paper, we propose CORAL, a novel online RCA framework that can\nautomatically trigger the RCA process and incrementally update the RCA model.\nCORAL consists of Trigger Point Detection, Incremental Disentangled Causal\nGraph Learning, and Network Propagation-based Root Cause Localization. The\nTrigger Point Detection component aims to detect system state transitions\nautomatically and in near-real-time. To achieve this, we develop an online\ntrigger point detection approach based on multivariate singular spectrum\nanalysis and cumulative sum statistics. To efficiently update the RCA model, we\npropose an incremental disentangled causal graph learning approach to decouple\nthe state-invariant and state-dependent information. After that, CORAL applies\na random walk with restarts to the updated causal graph to accurately identify\nroot causes. The online RCA process terminates when the causal graph and the\ngenerated root cause list converge. Extensive experiments on three real-world\ndatasets with case studies demonstrate the effectiveness and superiority of the\nproposed framework.\n", "date": "2023-05-18", "categories": []}
{"id": "2306.02831", "title": "MM-DAG: Multi-task DAG Learning for Multi-modal Data -- with Application\n  for Traffic Congestion Analysis", "abstract": "  This paper proposes to learn Multi-task, Multi-modal Direct Acyclic Graphs\n(MM-DAGs), which are commonly observed in complex systems, e.g., traffic,\nmanufacturing, and weather systems, whose variables are multi-modal with\nscalars, vectors, and functions. This paper takes the traffic congestion\nanalysis as a concrete case, where a traffic intersection is usually regarded\nas a DAG. In a road network of multiple intersections, different intersections\ncan only have some overlapping and distinct variables observed. For example, a\nsignalized intersection has traffic light-related variables, whereas\nunsignalized ones do not. This encourages the multi-task design: with each DAG\nas a task, the MM-DAG tries to learn the multiple DAGs jointly so that their\nconsensus and consistency are maximized. To this end, we innovatively propose a\nmulti-modal regression for linear causal relationship description of different\nvariables. Then we develop a novel Causality Difference (CD) measure and its\ndifferentiable approximator. Compared with existing SOTA measures, CD can\npenalize the causal structural difference among DAGs with distinct nodes and\ncan better consider the uncertainty of causal orders. We rigidly prove our\ndesign's topological interpretation and consistency properties. We conduct\nthorough simulations and one case study to show the effectiveness of our\nMM-DAG. The code is available under https://github.com/Lantian72/MM-DAG\n", "date": "2023-06-05", "categories": []}
{"id": "2302.10607", "title": "Differentiable Multi-Target Causal Bayesian Experimental Design", "abstract": "  We introduce a gradient-based approach for the problem of Bayesian optimal\nexperimental design to learn causal models in a batch setting -- a critical\ncomponent for causal discovery from finite data where interventions can be\ncostly or risky. Existing methods rely on greedy approximations to construct a\nbatch of experiments while using black-box methods to optimize over a single\ntarget-state pair to intervene with. In this work, we completely dispose of the\nblack-box optimization techniques and greedy heuristics and instead propose a\nconceptually simple end-to-end gradient-based optimization procedure to acquire\na set of optimal intervention target-state pairs. Such a procedure enables\nparameterization of the design space to efficiently optimize over a batch of\nmulti-target-state interventions, a setting which has hitherto not been\nexplored due to its complexity. We demonstrate that our proposed method\noutperforms baselines and existing acquisition strategies in both single-target\nand multi-target settings across a number of synthetic datasets.\n", "date": "2023-02-21", "categories": []}
{"id": "2306.01117", "title": "Examining the Causal Effect of First Names on Language Models: The Case\n  of Social Commonsense Reasoning", "abstract": "  As language models continue to be integrated into applications of personal\nand societal relevance, ensuring these models' trustworthiness is crucial,\nparticularly with respect to producing consistent outputs regardless of\nsensitive attributes. Given that first names may serve as proxies for\n(intersectional) socio-demographic representations, it is imperative to examine\nthe impact of first names on commonsense reasoning capabilities. In this paper,\nwe study whether a model's reasoning given a specific input differs based on\nthe first names provided. Our underlying assumption is that the reasoning about\nAlice should not differ from the reasoning about James. We propose and\nimplement a controlled experimental framework to measure the causal effect of\nfirst names on commonsense reasoning, enabling us to distinguish between model\npredictions due to chance and caused by actual factors of interest. Our results\nindicate that the frequency of first names has a direct effect on model\nprediction, with less frequent names yielding divergent predictions compared to\nmore frequent names. To gain insights into the internal mechanisms of models\nthat are contributing to these behaviors, we also conduct an in-depth\nexplainable analysis. Overall, our findings suggest that to ensure model\nrobustness, it is essential to augment datasets with more diverse first names\nduring the configuration stage.\n", "date": "2023-06-01", "categories": []}
{"id": "2306.01638", "title": "Do we become wiser with time? On causal equivalence with tiered\n  background knowledge", "abstract": "  Equivalence classes of DAGs (represented by CPDAGs) may be too large to\nprovide useful causal information. Here, we address incorporating tiered\nbackground knowledge yielding restricted equivalence classes represented by\n'tiered MPDAGs'. Tiered knowledge leads to considerable gains in\ninformativeness and computational efficiency: We show that construction of\ntiered MPDAGs only requires application of Meek's 1st rule, and that tiered\nMPDAGs (unlike general MPDAGs) are chain graphs with chordal components. This\nentails simplifications e.g. of determining valid adjustment sets for causal\neffect estimation. Further, we characterise when one tiered ordering is more\ninformative than another, providing insights into useful aspects of background\nknowledge.\n", "date": "2023-06-02", "categories": []}
{"id": "2210.09054", "title": "On the Identifiability and Estimation of Causal Location-Scale Noise\n  Models", "abstract": "  We study the class of location-scale or heteroscedastic noise models (LSNMs),\nin which the effect $Y$ can be written as a function of the cause $X$ and a\nnoise source $N$ independent of $X$, which may be scaled by a positive function\n$g$ over the cause, i.e., $Y = f(X) + g(X)N$. Despite the generality of the\nmodel class, we show the causal direction is identifiable up to some\npathological cases. To empirically validate these theoretical findings, we\npropose two estimators for LSNMs: an estimator based on (non-linear) feature\nmaps, and one based on neural networks. Both model the conditional distribution\nof $Y$ given $X$ as a Gaussian parameterized by its natural parameters. When\nthe feature maps are correctly specified, we prove that our estimator is\njointly concave, and a consistent estimator for the cause-effect identification\ntask. Although the the neural network does not inherit those guarantees, it can\nfit functions of arbitrary complexity, and reaches state-of-the-art performance\nacross benchmarks.\n", "date": "2022-10-13", "categories": []}
{"id": "2302.05209", "title": "A Survey on Causal Reinforcement Learning", "abstract": "  While Reinforcement Learning (RL) achieves tremendous success in sequential\ndecision-making problems of many domains, it still faces key challenges of data\ninefficiency and the lack of interpretability. Interestingly, many researchers\nhave leveraged insights from the causality literature recently, bringing forth\nflourishing works to unify the merits of causality and address well the\nchallenges from RL. As such, it is of great necessity and significance to\ncollate these Causal Reinforcement Learning (CRL) works, offer a review of CRL\nmethods, and investigate the potential functionality from causality toward RL.\nIn particular, we divide existing CRL approaches into two categories according\nto whether their causality-based information is given in advance or not. We\nfurther analyze each category in terms of the formalization of different\nmodels, ranging from the Markov Decision Process (MDP), Partially Observed\nMarkov Decision Process (POMDP), Multi-Arm Bandits (MAB), and Dynamic Treatment\nRegime (DTR). Moreover, we summarize the evaluation matrices and open sources\nwhile we discuss emerging applications, along with promising prospects for the\nfuture development of CRL.\n", "date": "2023-02-10", "categories": []}
{"id": "2305.17574", "title": "Counterfactual Formulation of Patient-Specific Root Causes of Disease", "abstract": "  Root causes of disease intuitively correspond to root vertices that increase\nthe likelihood of a diagnosis. This description of a root cause nevertheless\nlacks the rigorous mathematical formulation needed for the development of\ncomputer algorithms designed to automatically detect root causes from data.\nPrior work defined patient-specific root causes of disease using an\ninterventionalist account that only climbs to the second rung of Pearl's Ladder\nof Causation. In this theoretical piece, we climb to the third rung by\nproposing a counterfactual definition matching clinical intuition based on\nfixed factual data alone. We then show how to assign a root causal contribution\nscore to each variable using Shapley values from explainable artificial\nintelligence. The proposed counterfactual formulation of patient-specific root\ncauses of disease accounts for noisy labels, adapts to disease prevalence and\nadmits fast computation without the need for counterfactual simulation.\n", "date": "2023-05-27", "categories": []}
{"id": "2306.00315", "title": "Explicit Feature Interaction-aware Uplift Network for Online Marketing", "abstract": "  As a key component in online marketing, uplift modeling aims to accurately\ncapture the degree to which different treatments motivate different users, such\nas coupons or discounts, also known as the estimation of individual treatment\neffect (ITE). In an actual business scenario, the options for treatment may be\nnumerous and complex, and there may be correlations between different\ntreatments. In addition, each marketing instance may also have rich user and\ncontextual features. However, existing methods still fall short in both fully\nexploiting treatment information and mining features that are sensitive to a\nparticular treatment. In this paper, we propose an explicit feature\ninteraction-aware uplift network (EFIN) to address these two problems. Our EFIN\nincludes four customized modules: 1) a feature encoding module encodes not only\nthe user and contextual features, but also the treatment features; 2) a\nself-interaction module aims to accurately model the user's natural response\nwith all but the treatment features; 3) a treatment-aware interaction module\naccurately models the degree to which a particular treatment motivates a user\nthrough interactions between the treatment features and other features, i.e.,\nITE; and 4) an intervention constraint module is used to balance the ITE\ndistribution of users between the control and treatment groups so that the\nmodel would still achieve a accurate uplift ranking on data collected from a\nnon-random intervention marketing scenario. We conduct extensive experiments on\ntwo public datasets and one product dataset to verify the effectiveness of our\nEFIN. In addition, our EFIN has been deployed in a credit card bill payment\nscenario of a large online financial platform with a significant improvement.\n", "date": "2023-06-01", "categories": []}
{"id": "2306.00485", "title": "Causal Estimation of User Learning in Personalized Systems", "abstract": "  In online platforms, the impact of a treatment on an observed outcome may\nchange over time as 1) users learn about the intervention, and 2) the system\npersonalization, such as individualized recommendations, change over time. We\nintroduce a non-parametric causal model of user actions in a personalized\nsystem. We show that the Cookie-Cookie-Day (CCD) experiment, designed for the\nmeasurement of the user learning effect, is biased when there is\npersonalization. We derive new experimental designs that intervene in the\npersonalization system to generate the variation necessary to separately\nidentify the causal effect mediated through user learning and personalization.\nMaking parametric assumptions allows for the estimation of long-term causal\neffects based on medium-term experiments. In simulations, we show that our new\ndesigns successfully recover the dynamic causal effects of interest.\n", "date": "2023-06-01", "categories": []}
{"id": "2306.00624", "title": "From Temporal to Contemporaneous Iterative Causal Discovery in the\n  Presence of Latent Confounders", "abstract": "  We present a constraint-based algorithm for learning causal structures from\nobservational time-series data, in the presence of latent confounders. We\nassume a discrete-time, stationary structural vector autoregressive process,\nwith both temporal and contemporaneous causal relations. One may ask if\ntemporal and contemporaneous relations should be treated differently. The\npresented algorithm gradually refines a causal graph by learning long-term\ntemporal relations before short-term ones, where contemporaneous relations are\nlearned last. This ordering of causal relations to be learnt leads to a\nreduction in the required number of statistical tests. We validate this\nreduction empirically and demonstrate that it leads to higher accuracy for\nsynthetic data and more plausible causal graphs for real-world data compared to\nstate-of-the-art algorithms.\n", "date": "2023-06-01", "categories": []}
{"id": "2305.19582", "title": "Causal Discovery with Latent Confounders Based on Higher-Order Cumulants", "abstract": "  Causal discovery with latent confounders is an important but challenging task\nin many scientific areas. Despite the success of some overcomplete independent\ncomponent analysis (OICA) based methods in certain domains, they are\ncomputationally expensive and can easily get stuck into local optima. We notice\nthat interestingly, by making use of higher-order cumulants, there exists a\nclosed-form solution to OICA in specific cases, e.g., when the mixing procedure\nfollows the One-Latent-Component structure. In light of the power of the\nclosed-form solution to OICA corresponding to the One-Latent-Component\nstructure, we formulate a way to estimate the mixing matrix using the\nhigher-order cumulants, and further propose the testable One-Latent-Component\ncondition to identify the latent variables and determine causal orders. By\niteratively removing the share identified latent components, we successfully\nextend the results on the One-Latent-Component structure to the\nMulti-Latent-Component structure and finally provide a practical and\nasymptotically correct algorithm to learn the causal structure with latent\nvariables. Experimental results illustrate the asymptotic correctness and\neffectiveness of the proposed method.\n", "date": "2023-05-31", "categories": []}
{"id": "2305.19588", "title": "Active causal structure learning with advice", "abstract": "  We introduce the problem of active causal structure learning with advice. In\nthe typical well-studied setting, the learning algorithm is given the essential\ngraph for the observational distribution and is asked to recover the underlying\ncausal directed acyclic graph (DAG) $G^*$ while minimizing the number of\ninterventions made. In our setting, we are additionally given side information\nabout $G^*$ as advice, e.g. a DAG $G$ purported to be $G^*$. We ask whether the\nlearning algorithm can benefit from the advice when it is close to being\ncorrect, while still having worst-case guarantees even when the advice is\narbitrarily bad. Our work is in the same space as the growing body of research\non algorithms with predictions. When the advice is a DAG $G$, we design an\nadaptive search algorithm to recover $G^*$ whose intervention cost is at most\n$O(\\max\\{1, \\log \\psi\\})$ times the cost for verifying $G^*$; here, $\\psi$ is a\ndistance measure between $G$ and $G^*$ that is upper bounded by the number of\nvariables $n$, and is exactly 0 when $G=G^*$. Our approximation factor matches\nthe state-of-the-art for the advice-less setting.\n", "date": "2023-05-31", "categories": []}
{"id": "2305.19695", "title": "Causal discovery for time series with constraint-based model and PMIME\n  measure", "abstract": "  Causality defines the relationship between cause and effect. In multivariate\ntime series field, this notion allows to characterize the links between several\ntime series considering temporal lags. These phenomena are particularly\nimportant in medicine to analyze the effect of a drug for example, in\nmanufacturing to detect the causes of an anomaly in a complex system or in\nsocial sciences... Most of the time, studying these complex systems is made\nthrough correlation only. But correlation can lead to spurious relationships.\nTo circumvent this problem, we present in this paper a novel approach for\ndiscovering causality in time series data that combines a causal discovery\nalgorithm with an information theoretic-based measure. Hence the proposed\nmethod allows inferring both linear and non-linear relationships and building\nthe underlying causal graph. We evaluate the performance of our approach on\nseveral simulated data sets, showing promising results.\n", "date": "2023-05-31", "categories": []}
{"id": "2305.19802", "title": "Neuro-Causal Factor Analysis", "abstract": "  Factor analysis (FA) is a statistical tool for studying how observed\nvariables with some mutual dependences can be expressed as functions of\nmutually independent unobserved factors, and it is widely applied throughout\nthe psychological, biological, and physical sciences. We revisit this classic\nmethod from the comparatively new perspective given by advancements in causal\ndiscovery and deep learning, introducing a framework for Neuro-Causal Factor\nAnalysis (NCFA). Our approach is fully nonparametric: it identifies factors via\nlatent causal discovery methods and then uses a variational autoencoder (VAE)\nthat is constrained to abide by the Markov factorization of the distribution\nwith respect to the learned graph. We evaluate NCFA on real and synthetic data\nsets, finding that it performs comparably to standard VAEs on data\nreconstruction tasks but with the advantages of sparser architecture, lower\nmodel complexity, and causal interpretability. Unlike traditional FA methods,\nour proposed NCFA method allows learning and reasoning about the latent factors\nunderlying observed data from a justifiably causal perspective, even when the\nrelations between factors and measurements are highly nonlinear.\n", "date": "2023-05-31", "categories": []}
{"id": "2305.20043", "title": "Deception by Omission: Using Adversarial Missingness to Poison Causal\n  Structure Learning", "abstract": "  Inference of causal structures from observational data is a key component of\ncausal machine learning; in practice, this data may be incompletely observed.\nPrior work has demonstrated that adversarial perturbations of completely\nobserved training data may be used to force the learning of inaccurate causal\nstructural models (SCMs). However, when the data can be audited for correctness\n(e.g., it is crytographically signed by its source), this adversarial mechanism\nis invalidated. This work introduces a novel attack methodology wherein the\nadversary deceptively omits a portion of the true training data to bias the\nlearned causal structures in a desired manner. Theoretically sound attack\nmechanisms are derived for the case of arbitrary SCMs, and a sample-efficient\nlearning-based heuristic is given for Gaussian SCMs. Experimental validation of\nthese approaches on real and synthetic data sets demonstrates the effectiveness\nof adversarial missingness attacks at deceiving popular causal structure\nlearning algorithms.\n", "date": "2023-05-31", "categories": []}
{"id": "2207.12718", "title": "XInsight: eXplainable Data Analysis Through The Lens of Causality", "abstract": "  In light of the growing popularity of Exploratory Data Analysis (EDA),\nunderstanding the underlying causes of the knowledge acquired by EDA is\ncrucial. However, it remains under-researched. This study promotes a\ntransparent and explicable perspective on data analysis, called eXplainable\nData Analysis (XDA). For this reason, we present XInsight, a general framework\nfor XDA. XInsight provides data analysis with qualitative and quantitative\nexplanations of causal and non-causal semantics. This way, it will\nsignificantly improve human understanding and confidence in the outcomes of\ndata analysis, facilitating accurate data interpretation and decision making in\nthe real world. XInsight is a three-module, end-to-end pipeline designed to\nextract causal graphs, translate causal primitives into XDA semantics, and\nquantify the quantitative contribution of each explanation to a data fact.\nXInsight uses a set of design concepts and optimizations to address the\ninherent difficulties associated with integrating causality into XDA.\nExperiments on synthetic and real-world datasets as well as a user study\ndemonstrate the highly promising capabilities of XInsight.\n", "date": "2022-07-26", "categories": []}
{"id": "2211.00472", "title": "Backtracking Counterfactuals", "abstract": "  Counterfactual reasoning -- envisioning hypothetical scenarios, or possible\nworlds, where some circumstances are different from what (f)actually occurred\n(counter-to-fact) -- is ubiquitous in human cognition. Conventionally,\ncounterfactually-altered circumstances have been treated as \"small miracles\"\nthat locally violate the laws of nature while sharing the same initial\nconditions. In Pearl's structural causal model (SCM) framework this is made\nmathematically rigorous via interventions that modify the causal laws while the\nvalues of exogenous variables are shared. In recent years, however, this purely\ninterventionist account of counterfactuals has increasingly come under scrutiny\nfrom both philosophers and psychologists. Instead, they suggest a backtracking\naccount of counterfactuals, according to which the causal laws remain unchanged\nin the counterfactual world; differences to the factual world are instead\n\"backtracked\" to altered initial conditions (exogenous variables). In the\npresent work, we explore and formalise this alternative mode of counterfactual\nreasoning within the SCM framework. Despite ample evidence that humans\nbacktrack, the present work constitutes, to the best of our knowledge, the\nfirst general account and algorithmisation of backtracking counterfactuals. We\ndiscuss our backtracking semantics in the context of related literature and\ndraw connections to recent developments in explainable artificial intelligence\n(XAI).\n", "date": "2022-11-01", "categories": []}
{"id": "2305.04445", "title": "New metrics and search algorithms for weighted causal DAGs", "abstract": "  Recovering causal relationships from data is an important problem. Using\nobservational data, one can typically only recover causal graphs up to a Markov\nequivalence class and additional assumptions or interventional data are needed\nfor complete recovery. In this work, under some standard assumptions, we study\ncausal graph discovery via adaptive interventions with node-dependent\ninterventional costs. For this setting, we show that no algorithm can achieve\nan approximation guarantee that is asymptotically better than linear in the\nnumber of vertices with respect to the verification number; a well-established\nbenchmark for adaptive search algorithms. Motivated by this negative result, we\ndefine a new benchmark that captures the worst-case interventional cost for any\nsearch algorithm. Furthermore, with respect to this new benchmark, we provide\nadaptive search algorithms that achieve logarithmic approximations under\nvarious settings: atomic, bounded size interventions and generalized cost\nobjectives.\n", "date": "2023-05-08", "categories": []}
{"id": "2305.18410", "title": "Understanding Breast Cancer Survival: Using Causality and Language\n  Models on Multi-omics Data", "abstract": "  The need for more usable and explainable machine learning models in\nhealthcare increases the importance of developing and utilizing causal\ndiscovery algorithms, which aim to discover causal relations by analyzing\nobservational data. Explainable approaches aid clinicians and biologists in\npredicting the prognosis of diseases and suggesting proper treatments. However,\nvery little research has been conducted at the crossroads between causal\ndiscovery, genomics, and breast cancer, and we aim to bridge this gap.\nMoreover, evaluation of causal discovery methods on real data is in general\nnotoriously difficult because ground-truth causal relations are usually\nunknown, and accordingly, in this paper, we also propose to address the\nevaluation problem with large language models. In particular, we exploit\nsuitable causal discovery algorithms to investigate how various perturbations\nin the genome can affect the survival of patients diagnosed with breast cancer.\nWe used three main causal discovery algorithms: PC, Greedy Equivalence Search\n(GES), and a Generalized Precision Matrix-based one. We experiment with a\nsubset of The Cancer Genome Atlas, which contains information about mutations,\ncopy number variations, protein levels, and gene expressions for 705 breast\ncancer patients. Our findings reveal important factors related to the vital\nstatus of patients using causal discovery algorithms. However, the reliability\nof these results remains a concern in the medical domain. Accordingly, as\nanother contribution of the work, the results are validated through language\nmodels trained on biomedical literature, such as BlueBERT and other large\nlanguage models trained on medical corpora. Our results profess proper\nutilization of causal discovery algorithms and language models for revealing\nreliable causal relations for clinical applications.\n", "date": "2023-05-28", "categories": []}
{"id": "2301.11351", "title": "Estimating Causal Effects using a Multi-task Deep Ensemble", "abstract": "  A number of methods have been proposed for causal effect estimation, yet few\nhave demonstrated efficacy in handling data with complex structures, such as\nimages. To fill this gap, we propose Causal Multi-task Deep Ensemble (CMDE), a\nnovel framework that learns both shared and group-specific information from the\nstudy population. We provide proofs demonstrating equivalency of CDME to a\nmulti-task Gaussian process (GP) with a coregionalization kernel a priori.\nCompared to multi-task GP, CMDE efficiently handles high-dimensional and\nmulti-modal covariates and provides pointwise uncertainty estimates of causal\neffects. We evaluate our method across various types of datasets and tasks and\nfind that CMDE outperforms state-of-the-art methods on a majority of these\ntasks.\n", "date": "2023-01-26", "categories": []}
{"id": "2305.17277", "title": "Optimizing NOTEARS Objectives via Topological Swaps", "abstract": "  Recently, an intriguing class of non-convex optimization problems has emerged\nin the context of learning directed acyclic graphs (DAGs). These problems\ninvolve minimizing a given loss or score function, subject to a non-convex\ncontinuous constraint that penalizes the presence of cycles in a graph. In this\nwork, we delve into the optimization challenges associated with this class of\nnon-convex programs. To address these challenges, we propose a bi-level\nalgorithm that leverages the non-convex constraint in a novel way. The outer\nlevel of the algorithm optimizes over topological orders by iteratively\nswapping pairs of nodes within the topological order of a DAG. A key innovation\nof our approach is the development of an effective method for generating a set\nof candidate swapping pairs for each iteration. At the inner level, given a\ntopological order, we utilize off-the-shelf solvers that can handle linear\nconstraints. The key advantage of our proposed algorithm is that it is\nguaranteed to find a local minimum or a KKT point under weaker conditions\ncompared to previous work and finds solutions with lower scores. Extensive\nexperiments demonstrate that our method outperforms state-of-the-art approaches\nin terms of achieving a better score. Additionally, our method can also be used\nas a post-processing algorithm to significantly improve the score of other\nalgorithms. Code implementing the proposed method is available at\nhttps://github.com/duntrain/topo.\n", "date": "2023-05-26", "categories": []}
{"id": "2305.18210", "title": "Learning Causal Graphs via Monotone Triangular Transport Maps", "abstract": "  We study the problem of causal structure learning from data using optimal\ntransport (OT). Specifically, we first provide a constraint-based method which\nbuilds upon lower-triangular monotone parametric transport maps to design\nconditional independence tests which are agnostic to the noise distribution. We\nprovide an algorithm for causal discovery up to Markov Equivalence with no\nassumptions on the structural equations/noise distributions, which allows for\nsettings with latent variables. Our approach also extends to score-based causal\ndiscovery by providing a novel means for defining scores. This allows us to\nuniquely recover the causal graph under additional identifiability and\nstructural assumptions, such as additive noise or post-nonlinear models. We\nprovide experimental results to compare the proposed approach with the state of\nthe art on both synthetic and real-world datasets.\n", "date": "2023-05-26", "categories": []}
{"id": "2206.06417", "title": "Image-based Treatment Effect Heterogeneity", "abstract": "  Randomized controlled trials (RCTs) are considered the gold standard for\nestimating the average treatment effect (ATE) of interventions. One use of RCTs\nis to study the causes of global poverty -- a subject explicitly cited in the\n2019 Nobel Memorial Prize awarded to Duflo, Banerjee, and Kremer \"for their\nexperimental approach to alleviating global poverty.\" Because the ATE is a\npopulation summary, anti-poverty experiments often seek to unpack the effect\nvariation around the ATE by conditioning (CATE) on tabular variables such as\nage and ethnicity that were measured during the RCT data collection. Although\nsuch variables are key to unpacking CATE, using only such variables may fail to\ncapture historical, geographical, or neighborhood-specific contributors to\neffect variation, as tabular RCT data are often only observed near the time of\nthe experiment. In global poverty research, when the location of the experiment\nunits is approximately known, satellite imagery can provide a window into such\nfactors important for understanding heterogeneity. However, there is no method\nthat specifically enables applied researchers to analyze CATE from images. In\nthis paper, using a deep probabilistic modeling framework, we develop such a\nmethod that estimates latent clusters of images by identifying images with\nsimilar treatment effects distributions. Our interpretable image CATE model\nalso includes a sensitivity factor that quantifies the importance of image\nsegments contributing to the effect cluster prediction. We compare the proposed\nmethods against alternatives in simulation; also, we show how the model works\nin an actual RCT, estimating the effects of an anti-poverty intervention in\nnorthern Uganda and obtaining a posterior predictive distribution over effects\nfor the rest of the country where no experimental data was collected. We make\nall models available in open-source software.\n", "date": "2022-06-13", "categories": []}
{"id": "2302.05684", "title": "Sequential Underspecified Instrument Selection for Cause-Effect\n  Estimation", "abstract": "  Instrumental variable (IV) methods are used to estimate causal effects in\nsettings with unobserved confounding, where we cannot directly experiment on\nthe treatment variable. Instruments are variables which only affect the outcome\nindirectly via the treatment variable(s). Most IV applications focus on\nlow-dimensional treatments and crucially require at least as many instruments\nas treatments. This assumption is restrictive: in the natural sciences we often\nseek to infer causal effects of high-dimensional treatments (e.g., the effect\nof gene expressions or microbiota on health and disease), but can only run few\nexperiments with a limited number of instruments (e.g., drugs or antibiotics).\nIn such underspecified problems, the full treatment effect is not identifiable\nin a single experiment even in the linear case. We show that one can still\nreliably recover the projection of the treatment effect onto the instrumented\nsubspace and develop techniques to consistently combine such partial estimates\nfrom different sets of instruments. We then leverage our combined estimators in\nan algorithm that iteratively proposes the most informative instruments at each\nround of experimentation to maximize the overall information about the full\ncausal effect.\n", "date": "2023-02-11", "categories": []}
{"id": "2305.13341", "title": "Discovering Causal Relations and Equations from Data", "abstract": "  Physics is a field of science that has traditionally used the scientific\nmethod to answer questions about why natural phenomena occur and to make\ntestable models that explain the phenomena. Discovering equations, laws and\nprinciples that are invariant, robust and causal explanations of the world has\nbeen fundamental in physical sciences throughout the centuries. Discoveries\nemerge from observing the world and, when possible, performing interventional\nstudies in the system under study. With the advent of big data and the use of\ndata-driven methods, causal and equation discovery fields have grown and made\nprogress in computer science, physics, statistics, philosophy, and many applied\nfields. All these domains are intertwined and can be used to discover causal\nrelations, physical laws, and equations from observational data. This paper\nreviews the concepts, methods, and relevant works on causal and equation\ndiscovery in the broad field of Physics and outlines the most important\nchallenges and promising future lines of research. We also provide a taxonomy\nfor observational causal and equation discovery, point out connections, and\nshowcase a complete set of case studies in Earth and climate sciences, fluid\ndynamics and mechanics, and the neurosciences. This review demonstrates that\ndiscovering fundamental laws and causal relations by observing natural\nphenomena is being revolutionised with the efficient exploitation of\nobservational data, modern machine learning algorithms and the interaction with\ndomain knowledge. Exciting times are ahead with many challenges and\nopportunities to improve our understanding of complex systems.\n", "date": "2023-05-21", "categories": []}
{"id": "2305.13715", "title": "Covariate balancing using the integral probability metric for causal\n  inference", "abstract": "  Weighting methods in causal inference have been widely used to achieve a\ndesirable level of covariate balancing. However, the existing weighting methods\nhave desirable theoretical properties only when a certain model, either the\npropensity score or outcome regression model, is correctly specified. In\naddition, the corresponding estimators do not behave well for finite samples\ndue to large variance even when the model is correctly specified. In this\npaper, we consider to use the integral probability metric (IPM), which is a\nmetric between two probability measures, for covariate balancing. Optimal\nweights are determined so that weighted empirical distributions for the treated\nand control groups have the smallest IPM value for a given set of\ndiscriminators. We prove that the corresponding estimator can be consistent\nwithout correctly specifying any model (neither the propensity score nor the\noutcome regression model). In addition, we empirically show that our proposed\nmethod outperforms existing weighting methods with large margins for finite\nsamples.\n", "date": "2023-05-23", "categories": []}
{"id": "2209.01321", "title": "Deep Stable Representation Learning on Electronic Health Records", "abstract": "  Deep learning models have achieved promising disease prediction performance\nof the Electronic Health Records (EHR) of patients. However, most models\ndeveloped under the I.I.D. hypothesis fail to consider the agnostic\ndistribution shifts, diminishing the generalization ability of deep learning\nmodels to Out-Of-Distribution (OOD) data. In this setting, spurious statistical\ncorrelations that may change in different environments will be exploited, which\ncan cause sub-optimal performances of deep learning models. The unstable\ncorrelation between procedures and diagnoses existed in the training\ndistribution can cause spurious correlation between historical EHR and future\ndiagnosis. To address this problem, we propose to use a causal representation\nlearning method called Causal Healthcare Embedding (CHE). CHE aims at\neliminating the spurious statistical relationship by removing the dependencies\nbetween diagnoses and procedures. We introduce the Hilbert-Schmidt Independence\nCriterion (HSIC) to measure the degree of independence between the embedded\ndiagnosis and procedure features. Based on causal view analyses, we perform the\nsample weighting technique to get rid of such spurious relationship for the\nstable learning of EHR across different environments. Moreover, our proposed\nCHE method can be used as a flexible plug-and-play module that can enhance\nexisting deep learning models on EHR. Extensive experiments on two public\ndatasets and five state-of-the-art baselines unequivocally show that CHE can\nimprove the prediction accuracy of deep learning models on out-of-distribution\ndata by a large margin. In addition, the interpretability study shows that CHE\ncould successfully leverage causal structures to reflect a more reasonable\ncontribution of historical records for predictions.\n", "date": "2022-09-03", "categories": []}
{"id": "2305.12072", "title": "Chest X-ray Image Classification: A Causal Perspective", "abstract": "  The chest X-ray (CXR) is one of the most common and easy-to-get medical tests\nused to diagnose common diseases of the chest. Recently, many deep\nlearning-based methods have been proposed that are capable of effectively\nclassifying CXRs. Even though these techniques have worked quite well, it is\ndifficult to establish whether what these algorithms actually learn is the\ncause-and-effect link between diseases and their causes or just how to map\nlabels to photos.In this paper, we propose a causal approach to address the CXR\nclassification problem, which constructs a structural causal model (SCM) and\nuses the backdoor adjustment to select effective visual information for CXR\nclassification. Specially, we design different probability optimization\nfunctions to eliminate the influence of confounders on the learning of real\ncausality. Experimental results demonstrate that our proposed method\noutperforms the open-source NIH ChestX-ray14 in terms of classification\nperformance.\n", "date": "2023-05-20", "categories": []}
{"id": "2211.13447", "title": "On the Complexity of Counterfactual Reasoning", "abstract": "  We study the computational complexity of counterfactual reasoning in relation\nto the complexity of associational and interventional reasoning on structural\ncausal models (SCMs). We show that counterfactual reasoning is no harder than\nassociational or interventional reasoning on fully specified SCMs in the\ncontext of two computational frameworks. The first framework is based on the\nnotion of treewidth and includes the classical variable elimination and\njointree algorithms. The second framework is based on the more recent and\nrefined notion of causal treewidth which is directed towards models with\nfunctional dependencies such as SCMs. Our results are constructive and based on\nbounding the (causal) treewidth of twin networks -- used in standard\ncounterfactual reasoning that contemplates two worlds, real and imaginary -- to\nthe (causal) treewidth of the underlying SCM structure. In particular, we show\nthat the latter (causal) treewidth is no more than twice the former plus one.\nHence, if associational or interventional reasoning is tractable on a fully\nspecified SCM then counterfactual reasoning is tractable too. We extend our\nresults to general counterfactual reasoning that requires contemplating more\nthan two worlds and discuss applications of our results to counterfactual\nreasoning with a partially specified SCM that is coupled with data. We finally\npresent empirical results that measure the gap between the complexities of\ncounterfactual reasoning and associational/interventional reasoning on random\nSCMs.\n", "date": "2022-11-24", "categories": []}
{"id": "2305.11353", "title": "Meta-learning for heterogeneous treatment effect estimation with\n  closed-form solvers", "abstract": "  This article proposes a meta-learning method for estimating the conditional\naverage treatment effect (CATE) from a few observational data. The proposed\nmethod learns how to estimate CATEs from multiple tasks and uses the knowledge\nfor unseen tasks. In the proposed method, based on the meta-learner framework,\nwe decompose the CATE estimation problem into sub-problems. For each\nsub-problem, we formulate our estimation models using neural networks with\ntask-shared and task-specific parameters. With our formulation, we can obtain\noptimal task-specific parameters in a closed form that are differentiable with\nrespect to task-shared parameters, making it possible to perform effective\nmeta-learning. The task-shared parameters are trained such that the expected\nCATE estimation performance in few-shot settings is improved by minimizing the\ndifference between a CATE estimated with a large amount of data and one\nestimated with just a few data. Our experimental results demonstrate that our\nmethod outperforms the existing meta-learning approaches and CATE estimation\nmethods.\n", "date": "2023-05-19", "categories": []}
{"id": "2302.11121", "title": "Counterfactual Prediction Under Outcome Measurement Error", "abstract": "  Across domains such as medicine, employment, and criminal justice, predictive\nmodels often target labels that imperfectly reflect the outcomes of interest to\nexperts and policymakers. For example, clinical risk assessments deployed to\ninform physician decision-making often predict measures of healthcare\nutilization (e.g., costs, hospitalization) as a proxy for patient medical need.\nThese proxies can be subject to outcome measurement error when they\nsystematically differ from the target outcome they are intended to measure.\nHowever, prior modeling efforts to characterize and mitigate outcome\nmeasurement error overlook the fact that the decision being informed by a model\noften serves as a risk-mitigating intervention that impacts the target outcome\nof interest and its recorded proxy. Thus, in these settings, addressing\nmeasurement error requires counterfactual modeling of treatment effects on\noutcomes. In this work, we study intersectional threats to model reliability\nintroduced by outcome measurement error, treatment effects, and selection bias\nfrom historical decision-making policies. We develop an unbiased risk\nminimization method which, given knowledge of proxy measurement error\nproperties, corrects for the combined effects of these challenges. We also\ndevelop a method for estimating treatment-dependent measurement error\nparameters when these are unknown in advance. We demonstrate the utility of our\napproach theoretically and via experiments on real-world data from randomized\ncontrolled trials conducted in healthcare and employment domains. As\nimportantly, we demonstrate that models correcting for outcome measurement\nerror or treatment effects alone suffer from considerable reliability\nlimitations. Our work underscores the importance of considering intersectional\nthreats to model validity during the design and evaluation of predictive models\nfor decision support.\n", "date": "2023-02-22", "categories": []}
{"id": "2305.10736", "title": "Counterfactual Debiasing for Generating Factually Consistent Text\n  Summaries", "abstract": "  Despite substantial progress in abstractive text summarization to generate\nfluent and informative texts, the factual inconsistency in the generated\nsummaries remains an important yet challenging problem to be solved. In this\npaper, we construct causal graphs for abstractive text summarization and\nidentify the intrinsic causes of the factual inconsistency, i.e., the language\nbias and irrelevancy bias, and further propose a debiasing framework, named\nCoFactSum, to alleviate the causal effects of these biases by counterfactual\nestimation. Specifically, the proposed CoFactSum provides two counterfactual\nestimation strategies, i.e., Explicit Counterfactual Masking with an explicit\ndynamic masking strategy, and Implicit Counterfactual Training with an implicit\ndiscriminative cross-attention mechanism. Meanwhile, we design a Debiasing\nDegree Adjustment mechanism to dynamically adapt the debiasing degree at each\ndecoding step. Extensive experiments on two widely-used summarization datasets\ndemonstrate the effectiveness of CoFactSum in enhancing the factual consistency\nof generated summaries compared with several baselines.\n", "date": "2023-05-18", "categories": []}
{"id": "2207.09081", "title": "Generalizing Goal-Conditioned Reinforcement Learning with Variational\n  Causal Reasoning", "abstract": "  As a pivotal component to attaining generalizable solutions in human\nintelligence, reasoning provides great potential for reinforcement learning\n(RL) agents' generalization towards varied goals by summarizing part-to-whole\narguments and discovering cause-and-effect relations. However, how to discover\nand represent causalities remains a huge gap that hinders the development of\ncausal RL. In this paper, we augment Goal-Conditioned RL (GCRL) with Causal\nGraph (CG), a structure built upon the relation between objects and events. We\nnovelly formulate the GCRL problem into variational likelihood maximization\nwith CG as latent variables. To optimize the derived objective, we propose a\nframework with theoretical performance guarantees that alternates between two\nsteps: using interventional data to estimate the posterior of CG; using CG to\nlearn generalizable models and interpretable policies. Due to the lack of\npublic benchmarks that verify generalization capability under reasoning, we\ndesign nine tasks and then empirically show the effectiveness of the proposed\nmethod against five baselines on these tasks. Further theoretical analysis\nshows that our performance improvement is attributed to the virtuous cycle of\ncausal discovery, transition modeling, and policy training, which aligns with\nthe experimental evidence in extensive ablation studies.\n", "date": "2022-07-19", "categories": []}
{"id": "2305.10041", "title": "Risk Assessment of Lymph Node Metastases in Endometrial Cancer Patients:\n  A Causal Approach", "abstract": "  Assessing the pre-operative risk of lymph node metastases in endometrial\ncancer patients is a complex and challenging task. In principle, machine\nlearning and deep learning models are flexible and expressive enough to capture\nthe dynamics of clinical risk assessment. However, in this setting we are\nlimited to observational data with quality issues, missing values, small sample\nsize and high dimensionality: we cannot reliably learn such models from limited\nobservational data with these sources of bias. Instead, we choose to learn a\ncausal Bayesian network to mitigate the issues above and to leverage the prior\nknowledge on endometrial cancer available from clinicians and physicians. We\nintroduce a causal discovery algorithm for causal Bayesian networks based on\nbootstrap resampling, as opposed to the single imputation used in related\nworks. Moreover, we include a context variable to evaluate whether selection\nbias results in learning spurious associations. Finally, we discuss the\nstrengths and limitations of our findings in light of the presence of missing\ndata that may be missing-not-at-random, which is common in real-world clinical\nsettings.\n", "date": "2023-05-17", "categories": []}
{"id": "2305.10187", "title": "Evaluating Dynamic Conditional Quantile Treatment Effects with\n  Applications in Ridesharing", "abstract": "  Many modern tech companies, such as Google, Uber, and Didi, utilize online\nexperiments (also known as A/B testing) to evaluate new policies against\nexisting ones. While most studies concentrate on average treatment effects,\nsituations with skewed and heavy-tailed outcome distributions may benefit from\nalternative criteria, such as quantiles. However, assessing dynamic quantile\ntreatment effects (QTE) remains a challenge, particularly when dealing with\ndata from ride-sourcing platforms that involve sequential decision-making\nacross time and space. In this paper, we establish a formal framework to\ncalculate QTE conditional on characteristics independent of the treatment.\nUnder specific model assumptions, we demonstrate that the dynamic conditional\nQTE (CQTE) equals the sum of individual CQTEs across time, even though the\nconditional quantile of cumulative rewards may not necessarily equate to the\nsum of conditional quantiles of individual rewards. This crucial insight\nsignificantly streamlines the estimation and inference processes for our target\ncausal estimand. We then introduce two varying coefficient decision process\n(VCDP) models and devise an innovative method to test the dynamic CQTE.\nMoreover, we expand our approach to accommodate data from spatiotemporal\ndependent experiments and examine both conditional quantile direct and indirect\neffects. To showcase the practical utility of our method, we apply it to three\nreal-world datasets from a ride-sourcing platform. Theoretical findings and\ncomprehensive simulation studies further substantiate our proposal.\n", "date": "2023-05-17", "categories": []}
{"id": "2208.09104", "title": "A Causality-Based Learning Approach for Discovering the Underlying\n  Dynamics of Complex Systems from Partial Observations with Stochastic\n  Parameterization", "abstract": "  Discovering the underlying dynamics of complex systems from data is an\nimportant practical topic. Constrained optimization algorithms are widely\nutilized and lead to many successes. Yet, such purely data-driven methods may\nbring about incorrect physics in the presence of random noise and cannot easily\nhandle the situation with incomplete data. In this paper, a new iterative\nlearning algorithm for complex turbulent systems with partial observations is\ndeveloped that alternates between identifying model structures, recovering\nunobserved variables, and estimating parameters. First, a causality-based\nlearning approach is utilized for the sparse identification of model\nstructures, which takes into account certain physics knowledge that is\npre-learned from data. It has unique advantages in coping with indirect\ncoupling between features and is robust to the stochastic noise. A practical\nalgorithm is designed to facilitate the causal inference for high-dimensional\nsystems. Next, a systematic nonlinear stochastic parameterization is built to\ncharacterize the time evolution of the unobserved variables. Closed analytic\nformula via an efficient nonlinear data assimilation is exploited to sample the\ntrajectories of the unobserved variables, which are then treated as synthetic\nobservations to advance a rapid parameter estimation. Furthermore, the\nlocalization of the state variable dependence and the physics constraints are\nincorporated into the learning procedure, which mitigate the curse of\ndimensionality and prevent the finite time blow-up issue. Numerical experiments\nshow that the new algorithm succeeds in identifying the model structure and\nproviding suitable stochastic parameterizations for many complex nonlinear\nsystems with chaotic dynamics, spatiotemporal multiscale structures,\nintermittency, and extreme events.\n", "date": "2022-08-19", "categories": []}
{"id": "2302.00370", "title": "How to select predictive models for causal inference?", "abstract": "  As predictive models -- e.g., from machine learning -- give likely outcomes,\nthey may be used to reason on the effect of an intervention, a causal-inference\ntask. The increasing complexity of health data has opened the door to a\nplethora of models, but also the Pandora box of model selection: which of these\nmodels yield the most valid causal estimates? Here we highlight that classic\nmachine-learning model selection does not select the best outcome models for\ncausal inference. Indeed, causal model selection should control both outcome\nerrors for each individual, treated or not treated, whereas only one outcome is\nobserved. Theoretically, simple risks used in machine learning do not control\ncausal effects when treated and non-treated population differ too much. More\nelaborate risks build proxies of the causal error using ``nuisance''\nre-weighting to compute it on the observed data. But does computing these\nnuisance adds noise to model selection? Drawing from an extensive empirical\nstudy, we outline a good causal model-selection procedure: using the so-called\n$R\\text{-risk}$; using flexible estimators to compute the nuisance models on\nthe train set; and splitting out 10\\% of the data to compute risks.\n", "date": "2023-02-01", "categories": []}
{"id": "2305.09207", "title": "Counterfactual Outcome Prediction using Structured State Space Model", "abstract": "  Counterfactual outcome prediction in longitudinal data has recently gained\nattention due to its potential applications in healthcare and social sciences.\nIn this paper, we explore the use of the state space model, a popular sequence\nmodel, for this task. Specifically, we compare the performance of two models:\nTreatment Effect Neural Controlled Differential Equation (TE-CDE) and\nstructured state space model (S4Model). While TE-CDE uses controlled\ndifferential equations to address time-dependent confounding, it suffers from\noptimization issues and slow training. In contrast, S4Model is more efficient\nat modeling long-range dependencies and easier to train. We evaluate the models\non a simulated lung tumor growth dataset and find that S4Model outperforms\nTE-CDE with 1.63x reduction in per epoch training time and 10x better\nnormalized mean squared error. Additionally, S4Model is more stable during\ntraining and less sensitive to weight initialization than TE-CDE. Our results\nsuggest that the state space model may be a promising approach for\ncounterfactual outcome prediction in longitudinal data, with S4Model offering a\nmore efficient and effective alternative to TE-CDE.\n", "date": "2023-05-16", "categories": []}
{"id": "2305.06894", "title": "Reinterpreting causal discovery as the task of predicting unobserved\n  joint statistics", "abstract": "  If $X,Y,Z$ denote sets of random variables, two different data sources may\ncontain samples from $P_{X,Y}$ and $P_{Y,Z}$, respectively. We argue that\ncausal discovery can help inferring properties of the `unobserved joint\ndistributions' $P_{X,Y,Z}$ or $P_{X,Z}$. The properties may be conditional\nindependences (as in `integrative causal inference') or also quantitative\nstatements about dependences.\n  More generally, we define a learning scenario where the input is a subset of\nvariables and the label is some statistical property of that subset. Sets of\njointly observed variables define the training points, while unobserved sets\nare possible test points. To solve this learning task, we infer, as an\nintermediate step, a causal model from the observations that then entails\nproperties of unobserved sets. Accordingly, we can define the VC dimension of a\nclass of causal models and derive generalization bounds for the predictions.\n  Here, causal discovery becomes more modest and better accessible to empirical\ntests than usual: rather than trying to find a causal hypothesis that is `true'\na causal hypothesis is {\\it useful} whenever it correctly predicts statistical\nproperties of unobserved joint distributions. This way, a sparse causal graph\nthat omits weak influences may be more useful than a dense one (despite being\nless accurate) because it is able to reconstruct the full joint distribution\nfrom marginal distributions of smaller subsets.\n  Within such a `pragmatic' application of causal discovery, some popular\nheuristic approaches become justified in retrospect. It is, for instance,\nallowed to infer DAGs from partial correlations instead of conditional\nindependences if the DAGs are only used to predict partial correlations.\n", "date": "2023-05-11", "categories": []}
{"id": "2305.05986", "title": "Structural Hawkes Processes for Learning Causal Structure from\n  Discrete-Time Event Sequences", "abstract": "  Learning causal structure among event types from discrete-time event\nsequences is a particularly important but challenging task. Existing methods,\nsuch as the multivariate Hawkes processes based methods, mostly boil down to\nlearning the so-called Granger causality which assumes that the cause event\nhappens strictly prior to its effect event. Such an assumption is often\nuntenable beyond applications, especially when dealing with discrete-time event\nsequences in low-resolution; and typical discrete Hawkes processes mainly\nsuffer from identifiability issues raised by the instantaneous effect, i.e.,\nthe causal relationship that occurred simultaneously due to the low-resolution\ndata will not be captured by Granger causality. In this work, we propose\nStructure Hawkes Processes (SHPs) that leverage the instantaneous effect for\nlearning the causal structure among events type in discrete-time event\nsequence. The proposed method is featured with the minorization-maximization of\nthe likelihood function and a sparse optimization scheme. Theoretical results\nshow that the instantaneous effect is a blessing rather than a curse, and the\ncausal structure is identifiable under the existence of the instantaneous\neffect. Experiments on synthetic and real-world data verify the effectiveness\nof the proposed method.\n", "date": "2023-05-10", "categories": []}
{"id": "2301.10913", "title": "Proximal Causal Learning of Conditional Average Treatment Effects", "abstract": "  Efficiently and flexibly estimating treatment effect heterogeneity is an\nimportant task in a wide variety of settings ranging from medicine to\nmarketing, and there are a considerable number of promising conditional average\ntreatment effect estimators currently available. These, however, typically rely\non the assumption that the measured covariates are enough to justify\nconditional exchangeability. We propose the P-learner, motivated by the R- and\nDR-learner, a tailored two-stage loss function for learning heterogeneous\ntreatment effects in settings where exchangeability given observed covariates\nis an implausible assumption, and we wish to rely on proxy variables for causal\ninference. Our proposed estimator can be implemented by off-the-shelf\nloss-minimizing machine learning methods, which in the case of kernel\nregression satisfies an oracle bound on the estimated error as long as the\nnuisance components are estimated reasonably well.\n", "date": "2023-01-26", "categories": []}
{"id": "2305.05078", "title": "SECRETS: Subject-Efficient Clinical Randomized Controlled Trials using\n  Synthetic Intervention", "abstract": "  The randomized controlled trial (RCT) is the gold standard for estimating the\naverage treatment effect (ATE) of a medical intervention but requires\n100s-1000s of subjects, making it expensive and difficult to implement. While a\ncross-over trial can reduce sample size requirements by measuring the treatment\neffect per individual, it is only applicable to chronic conditions and\ninterventions whose effects dissipate rapidly. Another approach is to replace\nor augment data collected from an RCT with external data from prospective\nstudies or prior RCTs, but it is vulnerable to confounders in the external or\naugmented data. We propose to simulate the cross-over trial to overcome its\npractical limitations while exploiting its strengths. We propose a novel\nframework, SECRETS, which, for the first time, estimates the individual\ntreatment effect (ITE) per patient in the RCT study without using any external\ndata by leveraging a state-of-the-art counterfactual estimation algorithm,\ncalled synthetic intervention. It also uses a new hypothesis testing strategy\nto determine whether the treatment has a clinically significant ATE based on\nthe estimated ITEs. We show that SECRETS can improve the power of an RCT while\nmaintaining comparable significance levels; in particular, on three real-world\nclinical RCTs (Phase-3 trials), SECRETS increases power over the baseline\nmethod by $\\boldsymbol{6}$-$\\boldsymbol{54\\%}$ (average: 21.5%, standard\ndeviation: 15.8%).\n", "date": "2023-05-08", "categories": []}
{"id": "2305.05191", "title": "COLA: Contextualized Commonsense Causal Reasoning from the Causal\n  Inference Perspective", "abstract": "  Detecting commonsense causal relations (causation) between events has long\nbeen an essential yet challenging task. Given that events are complicated, an\nevent may have different causes under various contexts. Thus, exploiting\ncontext plays an essential role in detecting causal relations. Meanwhile,\nprevious works about commonsense causation only consider two events and ignore\ntheir context, simplifying the task formulation. This paper proposes a new task\nto detect commonsense causation between two events in an event sequence (i.e.,\ncontext), called contextualized commonsense causal reasoning. We also design a\nzero-shot framework: COLA (Contextualized Commonsense Causality Reasoner) to\nsolve the task from the causal inference perspective. This framework obtains\nrich incidental supervision from temporality and balances covariates from\nmultiple timestamps to remove confounding effects. Our extensive experiments\nshow that COLA can detect commonsense causality more accurately than baselines.\n", "date": "2023-05-09", "categories": []}
{"id": "2301.05893", "title": "Jointly Learning Consistent Causal Abstractions Over Multiple\n  Interventional Distributions", "abstract": "  An abstraction can be used to relate two structural causal models\nrepresenting the same system at different levels of resolution. Learning\nabstractions which guarantee consistency with respect to interventional\ndistributions would allow one to jointly reason about evidence across multiple\nlevels of granularity while respecting the underlying cause-effect\nrelationships. In this paper, we introduce a first framework for causal\nabstraction learning between SCMs based on the formalization of abstraction\nrecently proposed by Rischel (2020). Based on that, we propose a differentiable\nprogramming solution that jointly solves a number of combinatorial\nsub-problems, and we study its performance and benefits against independent and\nsequential approaches on synthetic settings and on a challenging real-world\nproblem related to electric vehicle battery manufacturing.\n", "date": "2023-01-14", "categories": []}
{"id": "2305.04357", "title": "Quantifying Consistency and Information Loss for Causal Abstraction\n  Learning", "abstract": "  Structural causal models provide a formalism to express causal relations\nbetween variables of interest. Models and variables can represent a system at\ndifferent levels of abstraction, whereby relations may be coarsened and refined\naccording to the need of a modeller. However, switching between different\nlevels of abstraction requires evaluating a trade-off between the consistency\nand the information loss among different models. In this paper we introduce a\nfamily of interventional measures that an agent may use to evaluate such a\ntrade-off. We consider four measures suited for different tasks, analyze their\nproperties, and propose algorithms to evaluate and learn causal abstractions.\nFinally, we illustrate the flexibility of our setup by empirically showing how\ndifferent measures and algorithmic choices may lead to different abstractions.\n", "date": "2023-05-07", "categories": []}
{"id": "2305.04638", "title": "Learning Good Interventions in Causal Graphs via Covering", "abstract": "  We study the causal bandit problem that entails identifying a near-optimal\nintervention from a specified set $A$ of (possibly non-atomic) interventions\nover a given causal graph. Here, an optimal intervention in ${A}$ is one that\nmaximizes the expected value for a designated reward variable in the graph, and\nwe use the standard notion of simple regret to quantify near optimality.\nConsidering Bernoulli random variables and for causal graphs on $N$ vertices\nwith constant in-degree, prior work has achieved a worst case guarantee of\n$\\widetilde{O} (N/\\sqrt{T})$ for simple regret. The current work utilizes the\nidea of covering interventions (which are not necessarily contained within\n${A}$) and establishes a simple regret guarantee of\n$\\widetilde{O}(\\sqrt{N/T})$. Notably, and in contrast to prior work, our simple\nregret bound depends only on explicit parameters of the problem instance. We\nalso go beyond prior work and achieve a simple regret guarantee for causal\ngraphs with unobserved variables. Further, we perform experiments to show\nimprovements over baselines in this setting.\n", "date": "2023-05-08", "categories": []}
{"id": "2305.03662", "title": "Causal Discovery with Stage Variables for Health Time Series", "abstract": "  Using observational data to learn causal relationships is essential when\nrandomized experiments are not possible, such as in healthcare. Discovering\ncausal relationships in time-series health data is even more challenging when\nrelationships change over the course of a disease, such as medications that are\nmost effective early on or for individuals with severe disease. Stage variables\nsuch as weeks of pregnancy, disease stages, or biomarkers like HbA1c, can\ninfluence what causal relationships are true for a patient. However, causal\ninference within each stage is often not possible due to limited amounts of\ndata, and combining all data risks incorrect or missed inferences. To address\nthis, we propose Causal Discovery with Stage Variables (CDSV), which uses stage\nvariables to reweight data from multiple time-series while accounting for\ndifferent causal relationships in each stage. In simulated data, CDSV discovers\nmore causes with fewer false discoveries compared to baselines, in eICU it has\na lower FDR than baselines, and in MIMIC-III it discovers more clinically\nrelevant causes of high blood pressure.\n", "date": "2023-05-05", "categories": []}
{"id": "2305.02542", "title": "Correcting for Interference in Experiments: A Case Study at Douyin", "abstract": "  Interference is a ubiquitous problem in experiments conducted on two-sided\ncontent marketplaces, such as Douyin (China's analog of TikTok). In many cases,\ncreators are the natural unit of experimentation, but creators interfere with\neach other through competition for viewers' limited time and attention. \"Naive\"\nestimators currently used in practice simply ignore the interference, but in\ndoing so incur bias on the order of the treatment effect. We formalize the\nproblem of inference in such experiments as one of policy evaluation.\nOff-policy estimators, while unbiased, are impractically high variance. We\nintroduce a novel Monte-Carlo estimator, based on \"Differences-in-Qs\" (DQ)\ntechniques, which achieves bias that is second-order in the treatment effect,\nwhile remaining sample-efficient to estimate. On the theoretical side, our\ncontribution is to develop a generalized theory of Taylor expansions for policy\nevaluation, which extends DQ theory to all major MDP formulations. On the\npractical side, we implement our estimator on Douyin's experimentation\nplatform, and in the process develop DQ into a truly \"plug-and-play\" estimator\nfor interference in real-world settings: one which provides robust, low-bias,\nlow-variance treatment effect estimates; admits computationally cheap,\nasymptotically exact uncertainty quantification; and reduces MSE by 99\\%\ncompared to the best existing alternatives in our applications.\n", "date": "2023-05-04", "categories": []}
{"id": "2302.02672", "title": "Identifiability of latent-variable and structural-equation models: from\n  linear to nonlinear", "abstract": "  An old problem in multivariate statistics is that linear Gaussian models are\noften unidentifiable, i.e. some parameters cannot be uniquely estimated. In\nfactor (component) analysis, an orthogonal rotation of the factors is\nunidentifiable, while in linear regression, the direction of effect cannot be\nidentified. For such linear models, non-Gaussianity of the (latent) variables\nhas been shown to provide identifiability. In the case of factor analysis, this\nleads to independent component analysis, while in the case of the direction of\neffect, non-Gaussian versions of structural equation modelling solve the\nproblem. More recently, we have shown how even general nonparametric nonlinear\nversions of such models can be estimated. Non-Gaussianity is not enough in this\ncase, but assuming we have time series, or that the distributions are suitably\nmodulated by some observed auxiliary variables, the models are identifiable.\nThis paper reviews the identifiability theory for the linear and nonlinear\ncases, considering both factor analytic models and structural equation models.\n", "date": "2023-02-06", "categories": []}
{"id": "2305.01334", "title": "Validation of massively-parallel adaptive testing using dynamic control\n  matching", "abstract": "  A/B testing is a widely-used paradigm within marketing optimization because\nit promises identification of causal effects and because it is implemented out\nof the box in most messaging delivery software platforms. Modern businesses,\nhowever, often run many A/B/n tests at the same time and in parallel, and\npackage many content variations into the same messages, not all of which are\npart of an explicit test. Whether as the result of many teams testing at the\nsame time, or as part of a more sophisticated reinforcement learning (RL)\napproach that continuously adapts tests and test condition assignment based on\nprevious results, dynamic parallel testing cannot be evaluated the same way\ntraditional A/B tests are evaluated. This paper presents a method for\ndisentangling the causal effects of the various tests under conditions of\ncontinuous test adaptation, using a matched-synthetic control group that adapts\nalongside the tests.\n", "date": "2023-05-02", "categories": []}
{"id": "2301.08230", "title": "Score-based Causal Representation Learning with Interventions", "abstract": "  This paper studies the causal representation learning problem when the latent\ncausal variables are observed indirectly through an unknown linear\ntransformation. The objectives are: (i) recovering the unknown linear\ntransformation (up to scaling) and (ii) determining the directed acyclic graph\n(DAG) underlying the latent variables. Sufficient conditions for DAG recovery\nare established, and it is shown that a large class of non-linear models in the\nlatent space (e.g., causal mechanisms parameterized by two-layer neural\nnetworks) satisfy these conditions. These sufficient conditions ensure that the\neffect of an intervention can be detected correctly from changes in the score.\nCapitalizing on this property, recovering a valid transformation is facilitated\nby the following key property: any valid transformation renders latent\nvariables' score function to necessarily have the minimal variations across\ndifferent interventional environments. This property is leveraged for perfect\nrecovery of the latent DAG structure using only \\emph{soft} interventions. For\nthe special case of stochastic \\emph{hard} interventions, with an additional\nhypothesis testing step, one can also uniquely recover the linear\ntransformation up to scaling and a valid causal ordering.\n", "date": "2023-01-19", "categories": []}
{"id": "2304.14109", "title": "The Structurally Complex with Additive Parent Causality (SCARY) Dataset", "abstract": "  Causal datasets play a critical role in advancing the field of causality.\nHowever, existing datasets often lack the complexity of real-world issues such\nas selection bias, unfaithful data, and confounding. To address this gap, we\npropose a new synthetic causal dataset, the Structurally Complex with Additive\npaRent causalitY (SCARY) dataset, which includes the following features. The\ndataset comprises 40 scenarios, each generated with three different seeds,\nallowing researchers to leverage relevant subsets of the dataset. Additionally,\nwe use two different data generation mechanisms for generating the causal\nrelationship between parents and child nodes, including linear and mixed causal\nmechanisms with multiple sub-types. Our dataset generator is inspired by the\nCausal Discovery Toolbox and generates only additive models. The dataset has a\nVarsortability of 0.5. Our SCARY dataset provides a valuable resource for\nresearchers to explore causal discovery under more realistic scenarios. The\ndataset is available at https://github.com/JayJayc/SCARY.\n", "date": "2023-04-27", "categories": []}
{"id": "2302.08635", "title": "Generative Causal Representation Learning for Out-of-Distribution Motion\n  Forecasting", "abstract": "  Conventional supervised learning methods typically assume i.i.d samples and\nare found to be sensitive to out-of-distribution (OOD) data. We propose\nGenerative Causal Representation Learning (GCRL) which leverages causality to\nfacilitate knowledge transfer under distribution shifts. While we evaluate the\neffectiveness of our proposed method in human trajectory prediction models,\nGCRL can be applied to other domains as well. First, we propose a novel causal\nmodel that explains the generative factors in motion forecasting datasets using\nfeatures that are common across all environments and with features that are\nspecific to each environment. Selection variables are used to determine which\nparts of the model can be directly transferred to a new environment without\nfine-tuning. Second, we propose an end-to-end variational learning paradigm to\nlearn the causal mechanisms that generate observations from features. GCRL is\nsupported by strong theoretical results that imply identifiability of the\ncausal model under certain assumptions. Experimental results on synthetic and\nreal-world motion forecasting datasets show the robustness and effectiveness of\nour proposed method for knowledge transfer under zero-shot and low-shot\nsettings by substantially outperforming the prior motion forecasting models on\nout-of-distribution prediction. Our code is available at\nhttps://github.com/sshirahmad/GCRL.\n", "date": "2023-02-17", "categories": []}
{"id": "2304.02813", "title": "Causal Repair of Learning-enabled Cyber-physical Systems", "abstract": "  Models of actual causality leverage domain knowledge to generate convincing\ndiagnoses of events that caused an outcome. It is promising to apply these\nmodels to diagnose and repair run-time property violations in cyber-physical\nsystems (CPS) with learning-enabled components (LEC). However, given the high\ndiversity and complexity of LECs, it is challenging to encode domain knowledge\n(e.g., the CPS dynamics) in a scalable actual causality model that could\ngenerate useful repair suggestions. In this paper, we focus causal diagnosis on\nthe input/output behaviors of LECs. Specifically, we aim to identify which\nsubset of I/O behaviors of the LEC is an actual cause for a property violation.\nAn important by-product is a counterfactual version of the LEC that repairs the\nrun-time property by fixing the identified problematic behaviors. Based on this\ninsights, we design a two-step diagnostic pipeline: (1) construct and\nHalpern-Pearl causality model that reflects the dependency of property outcome\non the component's I/O behaviors, and (2) perform a search for an actual cause\nand corresponding repair on the model. We prove that our pipeline has the\nfollowing guarantee: if an actual cause is found, the system is guaranteed to\nbe repaired; otherwise, we have high probabilistic confidence that the LEC\nunder analysis did not cause the property violation. We demonstrate that our\napproach successfully repairs learned controllers on a standard OpenAI Gym\nbenchmark.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.12523", "title": "CIMLA: Interpretable AI for inference of differential causal networks", "abstract": "  The discovery of causal relationships from high-dimensional data is a major\nopen problem in bioinformatics. Machine learning and feature attribution models\nhave shown great promise in this context but lack causal interpretation. Here,\nwe show that a popular feature attribution model estimates a causal quantity\nreflecting the influence of one variable on another, under certain assumptions.\nWe leverage this insight to implement a new tool, CIMLA, for discovering\ncondition-dependent changes in causal relationships. We then use CIMLA to\nidentify differences in gene regulatory networks between biological conditions,\na problem that has received great attention in recent years. Using extensive\nbenchmarking on simulated data sets, we show that CIMLA is more robust to\nconfounding variables and is more accurate than leading methods. Finally, we\nemploy CIMLA to analyze a previously published single-cell RNA-seq data set\ncollected from subjects with and without Alzheimer's disease (AD), discovering\nseveral potential regulators of AD.\n", "date": "2023-04-25", "categories": []}
{"id": "2304.11503", "title": "Improved Churn Causal Analysis Through Restrained High-Dimensional\n  Feature Space Effects in Financial Institutions", "abstract": "  Customer churn describes terminating a relationship with a business or\nreducing customer engagement over a specific period. Customer acquisition cost\ncan be five to six times that of customer retention, hence investing in\ncustomers with churn risk is wise. Causal analysis of the churn model can\npredict whether a customer will churn in the foreseeable future and identify\neffects and possible causes for churn. In general, this study presents a\nconceptual framework to discover the confounding features that correlate with\nindependent variables and are causally related to those dependent variables\nthat impact churn. We combine different algorithms including the SMOTE,\nensemble ANN, and Bayesian networks to address churn prediction problems on a\nmassive and high-dimensional finance data that is usually generated in\nfinancial institutions due to employing interval-based features used in\nCustomer Relationship Management systems. The effects of the curse and blessing\nof dimensionality assessed by utilising the Recursive Feature Elimination\nmethod to overcome the high dimension feature space problem. Moreover, a causal\ndiscovery performed to find possible interpretation methods to describe cause\nprobabilities that lead to customer churn. Evaluation metrics on validation\ndata confirm the random forest and our ensemble ANN model, with %86 accuracy,\noutperformed other approaches. Causal analysis results confirm that some\nindependent causal variables representing the level of super guarantee\ncontribution, account growth, and account balance amount were identified as\nconfounding variables that cause customer churn with a high degree of belief.\nThis article provides a real-world customer churn analysis from current status\ninference to future directions in local superannuation funds.\n", "date": "2023-04-23", "categories": []}
{"id": "2304.11969", "title": "Causal Effect Estimation with Variational AutoEncoder and the Front Door\n  Criterion", "abstract": "  An essential problem in causal inference is estimating causal effects from\nobservational data. The problem becomes more challenging with the presence of\nunobserved confounders. When there are unobserved confounders, the commonly\nused back-door adjustment is not applicable. Although the instrumental variable\n(IV) methods can deal with unobserved confounders, they all assume that the\ntreatment directly affects the outcome, and there is no mediator between the\ntreatment and the outcome. This paper aims to use the front-door criterion to\naddress the challenging problem with the presence of unobserved confounders and\nmediators. In practice, it is often difficult to identify the set of variables\nused for front-door adjustment from data. By leveraging the ability of deep\ngenerative models in representation learning, we propose FDVAE to learn the\nrepresentation of a Front-Door adjustment set with a Variational AutoEncoder,\ninstead of trying to search for a set of variables for front-door adjustment.\nExtensive experiments on synthetic datasets validate the effectiveness of FDVAE\nand its superiority over existing methods. The experiments also show that the\nperformance of FDVAE is not sensitive to the causal strength of unobserved\nconfounders and is feasible in the case of dimensionality mismatch between\nlearned representations and the ground truth. We further apply the method to\nthree real-world datasets to demonstrate its potential applications.\n", "date": "2023-04-24", "categories": []}
{"id": "2304.10604", "title": "Causal Analysis of Customer Churn Using Deep Learning", "abstract": "  Customer churn describes terminating a relationship with a business or\nreducing customer engagement over a specific period. Two main business\nmarketing strategies play vital roles to increase market share dollar-value:\ngaining new and preserving existing customers. Customer acquisition cost can be\nfive to six times that for customer retention, hence investing in customers\nwith churn risk is smart. Causal analysis of the churn model can predict\nwhether a customer will churn in the foreseeable future and assist enterprises\nto identify effects and possible causes for churn and subsequently use that\nknowledge to apply tailored incentives. This paper proposes a framework using a\ndeep feedforward neural network for classification accompanied by a sequential\npattern mining method on high-dimensional sparse data. We also propose a causal\nBayesian network to predict cause probabilities that lead to customer churn.\nEvaluation metrics on test data confirm the XGBoost and our deep learning model\noutperformed previous techniques. Experimental analysis confirms that some\nindependent causal variables representing the level of super guarantee\ncontribution, account growth, and customer tenure were identified as\nconfounding factors for customer churn with a high degree of belief. This paper\nprovides a real-world customer churn analysis from current status inference to\nfuture directions in local superannuation funds.\n", "date": "2023-04-20", "categories": []}
{"id": "2304.10706", "title": "TC-GAT: Graph Attention Network for Temporal Causality Discovery", "abstract": "  The present study explores the intricacies of causal relationship extraction,\na vital component in the pursuit of causality knowledge. Causality is\nfrequently intertwined with temporal elements, as the progression from cause to\neffect is not instantaneous but rather ensconced in a temporal dimension. Thus,\nthe extraction of temporal causality holds paramount significance in the field.\nIn light of this, we propose a method for extracting causality from the text\nthat integrates both temporal and causal relations, with a particular focus on\nthe time aspect. To this end, we first compile a dataset that encompasses\ntemporal relationships. Subsequently, we present a novel model, TC-GAT, which\nemploys a graph attention mechanism to assign weights to the temporal\nrelationships and leverages a causal knowledge graph to determine the adjacency\nmatrix. Additionally, we implement an equilibrium mechanism to regulate the\ninterplay between temporal and causal relations. Our experiments demonstrate\nthat our proposed method significantly surpasses baseline models in the task of\ncausality extraction.\n", "date": "2023-04-21", "categories": []}
{"id": "2304.10900", "title": "A Common Misassumption in Online Experiments with Machine Learning\n  Models", "abstract": "  Online experiments such as Randomised Controlled Trials (RCTs) or A/B-tests\nare the bread and butter of modern platforms on the web. They are conducted\ncontinuously to allow platforms to estimate the causal effect of replacing\nsystem variant \"A\" with variant \"B\", on some metric of interest. These variants\ncan differ in many aspects. In this paper, we focus on the common use-case\nwhere they correspond to machine learning models. The online experiment then\nserves as the final arbiter to decide which model is superior, and should thus\nbe shipped.\n  The statistical literature on causal effect estimation from RCTs has a\nsubstantial history, which contributes deservedly to the level of trust\nresearchers and practitioners have in this \"gold standard\" of evaluation\npractices. Nevertheless, in the particular case of machine learning\nexperiments, we remark that certain critical issues remain. Specifically, the\nassumptions that are required to ascertain that A/B-tests yield unbiased\nestimates of the causal effect, are seldom met in practical applications. We\nargue that, because variants typically learn using pooled data, a lack of model\ninterference cannot be guaranteed. This undermines the conclusions we can draw\nfrom online experiments with machine learning models. We discuss the\nimplications this has for practitioners, and for the research literature.\n", "date": "2023-04-21", "categories": []}
{"id": "2211.03046", "title": "Knowledge is Power: Understanding Causality Makes Legal judgment\n  Prediction Models More Generalizable and Robust", "abstract": "  Legal Judgment Prediction (LJP), aiming to predict a judgment based on fact\ndescriptions according to rule of law, serves as legal assistance to mitigate\nthe great work burden of limited legal practitioners. Most existing methods\napply various large-scale pre-trained language models (PLMs) finetuned in LJP\ntasks to obtain consistent improvements. However, we discover the fact that the\nstate-of-the-art (SOTA) model makes judgment predictions according to\nirrelevant (or non-casual) information. The violation of rule of law not only\nweakens the robustness and generalization ability of models but also results in\nsevere social problems like discrimination. In this paper, we use causal\nstructural models (SCMs) to theoretically analyze how LJP models learn to make\ndecisions and why they can succeed in passing the traditional testing paradigm\nwithout learning causality. According to our analysis, we provide two solutions\nintervening on data and model by causality, respectively. In detail, we first\ndistinguish non-causal information by applying the open information extraction\n(OIE) technique. Then, we propose a method named the Causal Information\nEnhanced SAmpling Method (CIESAM) to eliminate the non-causal information from\ndata. To validate our theoretical analysis, we further propose another method\nusing our proposed Causality-Aware Self-Attention Mechanism (CASAM) to guide\nthe model to learn the underlying causality knowledge in legal texts. The\nconfidence of CASAM in learning causal information is higher than that of\nCIESAM. The extensive experimental results show that both our proposed methods\nachieve state-of-the-art (SOTA) performance on three commonly used\nlegal-specific datasets. The stronger performance of CASAM further demonstrates\nthat causality is the key to the robustness and generalization ability of\nmodels.\n", "date": "2022-11-06", "categories": []}
{"id": "2304.09085", "title": "Balancing Unobserved Confounding with a Few Unbiased Ratings in Debiased\n  Recommendations", "abstract": "  Recommender systems are seen as an effective tool to address information\noverload, but it is widely known that the presence of various biases makes\ndirect training on large-scale observational data result in sub-optimal\nprediction performance. In contrast, unbiased ratings obtained from randomized\ncontrolled trials or A/B tests are considered to be the golden standard, but\nare costly and small in scale in reality. To exploit both types of data, recent\nworks proposed to use unbiased ratings to correct the parameters of the\npropensity or imputation models trained on the biased dataset. However, the\nexisting methods fail to obtain accurate predictions in the presence of\nunobserved confounding or model misspecification. In this paper, we propose a\ntheoretically guaranteed model-agnostic balancing approach that can be applied\nto any existing debiasing method with the aim of combating unobserved\nconfounding and model misspecification. The proposed approach makes full use of\nunbiased data by alternatively correcting model parameters learned with biased\ndata, and adaptively learning balance coefficients of biased samples for\nfurther debiasing. Extensive real-world experiments are conducted along with\nthe deployment of our proposal on four representative debiasing methods to\ndemonstrate the effectiveness.\n", "date": "2023-04-17", "categories": []}
{"id": "2301.02324", "title": "Reasoning about Causality in Games", "abstract": "  Causal reasoning and game-theoretic reasoning are fundamental topics in\nartificial intelligence, among many other disciplines: this paper is concerned\nwith their intersection. Despite their importance, a formal framework that\nsupports both these forms of reasoning has, until now, been lacking. We offer a\nsolution in the form of (structural) causal games, which can be seen as\nextending Pearl's causal hierarchy to the game-theoretic domain, or as\nextending Koller and Milch's multi-agent influence diagrams to the causal\ndomain. We then consider three key questions: i) How can the (causal)\ndependencies in games - either between variables, or between strategies - be\nmodelled in a uniform, principled manner? ii) How may causal queries be\ncomputed in causal games, and what assumptions does this require? iii) How do\ncausal games compare to existing formalisms? To address question i), we\nintroduce mechanised games, which encode dependencies between agents' decision\nrules and the distributions governing the game. In response to question ii), we\npresent definitions of predictions, interventions, and counterfactuals, and\ndiscuss the assumptions required for each. Regarding question iii), we describe\ncorrespondences between causal games and other formalisms, and explain how\ncausal games can be used to answer queries that other causal or game-theoretic\nmodels do not support. Finally, we highlight possible applications of causal\ngames, aided by an extensive open-source Python library.\n", "date": "2023-01-05", "categories": []}
{"id": "2301.11336", "title": "Causal Structural Learning from Time Series: A Convex Optimization\n  Approach", "abstract": "  Structural learning, which aims to learn directed acyclic graphs (DAGs) from\nobservational data, is foundational to causal reasoning and scientific\ndiscovery. Recent advancements formulate structural learning into a continuous\noptimization problem; however, DAG learning remains a highly non-convex\nproblem, and there has not been much work on leveraging well-developed convex\noptimization techniques for causal structural learning. We fill this gap by\nproposing a data-adaptive linear approach for causal structural learning from\ntime series data, which can be conveniently cast into a convex optimization\nproblem using a recently developed monotone operator variational inequality\n(VI) formulation. Furthermore, we establish non-asymptotic recovery guarantee\nof the VI-based approach and show the superior performance of our proposed\nmethod on structure recovery over existing methods via extensive numerical\nexperiments.\n", "date": "2023-01-26", "categories": []}
{"id": "2304.07638", "title": "Causal models in string diagrams", "abstract": "  The framework of causal models provides a principled approach to causal\nreasoning, applied today across many scientific domains. Here we present this\nframework in the language of string diagrams, interpreted formally using\ncategory theory. A class of string diagrams, called network diagrams, are in\n1-to-1 correspondence with directed acyclic graphs. A causal model is given by\nsuch a diagram with its components interpreted as stochastic maps, functions,\nor general channels in a symmetric monoidal category with a 'copy-discard'\nstructure (cd-category), turning a model into a single mathematical object that\ncan be reasoned with intuitively and yet rigorously. Building on prior works by\nFong and Jacobs, Kissinger and Zanasi, as well as Fritz and Klingler, we\npresent diagrammatic definitions of causal models and functional causal models\nin a cd-category, generalising causal Bayesian networks and structural causal\nmodels, respectively. We formalise general interventions on a model, including\nbut beyond do-interventions, and present the natural notion of an open causal\nmodel with inputs. We also give an approach to conditioning based on a\nnormalisation box, allowing for causal inference calculations to be done fully\ndiagrammatically. We define counterfactuals in this setup, and treat the\nproblems of the identifiability of causal effects and counterfactuals fully\ndiagrammatically. The benefits of such a presentation of causal models lie in\nfoundational questions in causal reasoning and in their clarificatory role and\npedagogical value. This work aims to be accessible to different communities,\nfrom causal model practitioners to researchers in applied category theory, and\ndiscusses many examples from the literature for illustration. Overall, we argue\nand demonstrate that causal reasoning according to the causal model framework\nis most naturally and intuitively done as diagrammatic reasoning.\n", "date": "2023-04-15", "categories": []}
{"id": "2304.07922", "title": "Causal Disentangled Variational Auto-Encoder for Preference\n  Understanding in Recommendation", "abstract": "  Recommendation models are typically trained on observational user interaction\ndata, but the interactions between latent factors in users' decision-making\nprocesses lead to complex and entangled data. Disentangling these latent\nfactors to uncover their underlying representation can improve the robustness,\ninterpretability, and controllability of recommendation models. This paper\nintroduces the Causal Disentangled Variational Auto-Encoder (CaD-VAE), a novel\napproach for learning causal disentangled representations from interaction data\nin recommender systems. The CaD-VAE method considers the causal relationships\nbetween semantically related factors in real-world recommendation scenarios,\nrather than enforcing independence as in existing disentanglement methods. The\napproach utilizes structural causal models to generate causal representations\nthat describe the causal relationship between latent factors. The results\ndemonstrate that CaD-VAE outperforms existing methods, offering a promising\nsolution for disentangling complex user behavior data in recommendation\nsystems.\n", "date": "2023-04-17", "categories": []}
{"id": "2304.08278", "title": "Compositional Probabilistic and Causal Inference using Tractable Circuit\n  Models", "abstract": "  Probabilistic circuits (PCs) are a class of tractable probabilistic models,\nwhich admit efficient inference routines depending on their structural\nproperties. In this paper, we introduce md-vtrees, a novel structural\nformulation of (marginal) determinism in structured decomposable PCs, which\ngeneralizes previously proposed classes such as probabilistic sentential\ndecision diagrams. Crucially, we show how mdvtrees can be used to derive\ntractability conditions and efficient algorithms for advanced inference queries\nexpressed as arbitrary compositions of basic probabilistic operations, such as\nmarginalization, multiplication and reciprocals, in a sound and generalizable\nmanner. In particular, we derive the first polytime algorithms for causal\ninference queries such as backdoor adjustment on PCs. As a practical\ninstantiation of the framework, we propose MDNets, a novel PC architecture\nusing md-vtrees, and empirically demonstrate their application to causal\ninference.\n", "date": "2023-04-17", "categories": []}
{"id": "2304.05524", "title": "Understanding Causality with Large Language Models: Feasibility and\n  Opportunities", "abstract": "  We assess the ability of large language models (LLMs) to answer causal\nquestions by analyzing their strengths and weaknesses against three types of\ncausal question. We believe that current LLMs can answer causal questions with\nexisting causal knowledge as combined domain experts. However, they are not yet\nable to provide satisfactory answers for discovering new knowledge or for\nhigh-stakes decision-making tasks with high precision. We discuss possible\nfuture directions and opportunities, such as enabling explicit and implicit\ncausal modules as well as deep causal-aware LLMs. These will not only enable\nLLMs to answer many different types of causal questions for greater impact but\nalso enable LLMs to be more trustworthy and efficient in general.\n", "date": "2023-04-11", "categories": []}
{"id": "2304.05976", "title": "Bayesian Causal Inference in Doubly Gaussian DAG-probit Models", "abstract": "  We consider modeling a binary response variable together with a set of\ncovariates for two groups under observational data. The grouping variable can\nbe the confounding variable (the common cause of treatment and outcome),\ngender, case/control, ethnicity, etc. Given the covariates and a binary latent\nvariable, the goal is to construct two directed acyclic graphs (DAGs), while\nsharing some common parameters. The set of nodes, which represent the\nvariables, are the same for both groups but the directed edges between nodes,\nwhich represent the causal relationships between the variables, can be\npotentially different. For each group, we also estimate the effect size for\neach node. We assume that each group follows a Gaussian distribution under its\nDAG. Given the parent nodes, the joint distribution of DAG is conditionally\nindependent due to the Markov property of DAGs. We introduce the concept of\nGaussian DAG-probit model under two groups and hence doubly Gaussian DAG-probit\nmodel. To estimate the skeleton of the DAGs and the model parameters, we took\nsamples from the posterior distribution of doubly Gaussian DAG-probit model via\nMCMC method. We validated the proposed method using a comprehensive simulation\nexperiment and applied it on two real datasets. Furthermore, we validated the\nresults of the real data analysis using well-known experimental studies to show\nthe value of the proposed grouping variable in the causality domain.\n", "date": "2023-04-12", "categories": []}
{"id": "2304.04755", "title": "A Novel Two-level Causal Inference Framework for On-road Vehicle Quality\n  Issues Diagnosis", "abstract": "  In the automotive industry, the full cycle of managing in-use vehicle quality\nissues can take weeks to investigate. The process involves isolating root\ncauses, defining and implementing appropriate treatments, and refining\ntreatments if needed. The main pain-point is the lack of a systematic method to\nidentify causal relationships, evaluate treatment effectiveness, and direct the\nnext actionable treatment if the current treatment was deemed ineffective. This\npaper will show how we leverage causal Machine Learning (ML) to speed up such\nprocesses. A real-word data set collected from on-road vehicles will be used to\ndemonstrate the proposed framework. Open challenges for vehicle quality\napplications will also be discussed.\n", "date": "2023-03-31", "categories": []}
{"id": "2301.01026", "title": "Continual Causal Effect Estimation: Challenges and Opportunities", "abstract": "  A further understanding of cause and effect within observational data is\ncritical across many domains, such as economics, health care, public policy,\nweb mining, online advertising, and marketing campaigns. Although significant\nadvances have been made to overcome the challenges in causal effect estimation\nwith observational data, such as missing counterfactual outcomes and selection\nbias between treatment and control groups, the existing methods mainly focus on\nsource-specific and stationary observational data. Such learning strategies\nassume that all observational data are already available during the training\nphase and from only one source. This practical concern of accessibility is\nubiquitous in various academic and industrial applications. That's what it\nboiled down to: in the era of big data, we face new challenges in causal\ninference with observational data, i.e., the extensibility for incrementally\navailable observational data, the adaptability for extra domain adaptation\nproblem except for the imbalance between treatment and control groups, and the\naccessibility for an enormous amount of data. In this position paper, we\nformally define the problem of continual treatment effect estimation, describe\nits research challenges, and then present possible solutions to this problem.\nMoreover, we will discuss future research directions on this topic.\n", "date": "2023-01-03", "categories": []}
{"id": "2304.04566", "title": "Linking a predictive model to causal effect estimation", "abstract": "  A predictive model makes outcome predictions based on some given features,\ni.e., it estimates the conditional probability of the outcome given a feature\nvector. In general, a predictive model cannot estimate the causal effect of a\nfeature on the outcome, i.e., how the outcome will change if the feature is\nchanged while keeping the values of other features unchanged. This is because\ncausal effect estimation requires interventional probabilities. However, many\nreal world problems such as personalised decision making, recommendation, and\nfairness computing, need to know the causal effect of any feature on the\noutcome for a given instance. This is different from the traditional causal\neffect estimation problem with a fixed treatment variable. This paper first\ntackles the challenge of estimating the causal effect of any feature (as the\ntreatment) on the outcome w.r.t. a given instance. The theoretical results\nnaturally link a predictive model to causal effect estimations and imply that a\npredictive model is causally interpretable when the conditions identified in\nthe paper are satisfied. The paper also reveals the robust property of a\ncausally interpretable model. We use experiments to demonstrate that various\ntypes of predictive models, when satisfying the conditions identified in this\npaper, can estimate the causal effects of features as accurately as\nstate-of-the-art causal effect estimation methods. We also show the potential\nof such causally interpretable predictive models for robust predictions and\npersonalised decision making.\n", "date": "2023-04-10", "categories": []}
{"id": "2304.01237", "title": "A Guide for Practical Use of ADMG Causal Data Augmentation", "abstract": "  Data augmentation is essential when applying Machine Learning in small-data\nregimes. It generates new samples following the observed data distribution\nwhile increasing their diversity and variability to help researchers and\npractitioners improve their models' robustness and, thus, deploy them in the\nreal world. Nevertheless, its usage in tabular data still needs to be improved,\nas prior knowledge about the underlying data mechanism is seldom considered,\nlimiting the fidelity and diversity of the generated data. Causal data\naugmentation strategies have been pointed out as a solution to handle these\nchallenges by relying on conditional independence encoded in a causal graph. In\nthis context, this paper experimentally analyzed the ADMG causal augmentation\nmethod considering different settings to support researchers and practitioners\nin understanding under which conditions prior knowledge helps generate new data\npoints and, consequently, enhances the robustness of their models. The results\nhighlighted that the studied method (a) is independent of the underlying model\nmechanism, (b) requires a minimal number of observations that may be\nchallenging in a small-data regime to improve an ML model's accuracy, (c)\npropagates outliers to the augmented set degrading the performance of the\nmodel, and (d) is sensitive to its hyperparameter's value.\n", "date": "2023-04-03", "categories": []}
{"id": "2304.03382", "title": "Scalable Causal Discovery with Score Matching", "abstract": "  This paper demonstrates how to discover the whole causal graph from the\nsecond derivative of the log-likelihood in observational non-linear additive\nGaussian noise models. Leveraging scalable machine learning approaches to\napproximate the score function $\\nabla \\log p(\\mathbf{X})$, we extend the work\nof Rolland et al. (2022) that only recovers the topological order from the\nscore and requires an expensive pruning step removing spurious edges among\nthose admitted by the ordering. Our analysis leads to DAS (acronym for\nDiscovery At Scale), a practical algorithm that reduces the complexity of the\npruning by a factor proportional to the graph size. In practice, DAS achieves\ncompetitive accuracy with current state-of-the-art while being over an order of\nmagnitude faster. Overall, our approach enables principled and scalable causal\ndiscovery, significantly lowering the compute bar.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.03265", "title": "Causal Discovery with Score Matching on Additive Models with Arbitrary\n  Noise", "abstract": "  Causal discovery methods are intrinsically constrained by the set of\nassumptions needed to ensure structure identifiability. Moreover additional\nrestrictions are often imposed in order to simplify the inference task: this is\nthe case for the Gaussian noise assumption on additive non-linear models, which\nis common to many causal discovery approaches. In this paper we show the\nshortcomings of inference under this hypothesis, analyzing the risk of edge\ninversion under violation of Gaussianity of the noise terms. Then, we propose a\nnovel method for inferring the topological ordering of the variables in the\ncausal graph, from data generated according to an additive non-linear model\nwith a generic noise distribution. This leads to NoGAM (Not only Gaussian\nAdditive noise Models), a causal discovery algorithm with a minimal set of\nassumptions and state of the art performance, experimentally benchmarked on\nsynthetic data.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.02286", "title": "A step towards the applicability of algorithms based on invariant causal\n  learning on observational data", "abstract": "  Machine learning can benefit from causal discovery for interpretation and\nfrom causal inference for generalization. In this line of research, a few\ninvariant learning algorithms for out-of-distribution (OOD) generalization have\nbeen proposed by using multiple training environments to find invariant\nrelationships. Some of them are focused on causal discovery as Invariant Causal\nPrediction (ICP), which finds causal parents of a variable of interest, and\nsome directly provide a causal optimal predictor that generalizes well in OOD\nenvironments as Invariant Risk Minimization (IRM). This group of algorithms\nworks under the assumption of multiple environments that represent different\ninterventions in the causal inference context. Those environments are not\nnormally available when working with observational data and real-world\napplications. Here we propose a method to generate them in an efficient way. We\nassess the performance of this unsupervised learning problem by implementing\nICP on simulated data. We also show how to apply ICP efficiently integrated\nwith our method for causal discovery. Finally, we proposed an improved version\nof our method in combination with ICP for datasets with multiple covariates\nwhere ICP and other causal discovery methods normally degrade in performance.\n", "date": "2023-04-05", "categories": []}
{"id": "2208.12764", "title": "Causal Bandits for Linear Structural Equation Models", "abstract": "  This paper studies the problem of designing an optimal sequence of\ninterventions in a causal graphical model to minimize cumulative regret with\nrespect to the best intervention in hindsight. This is, naturally, posed as a\ncausal bandit problem. The focus is on causal bandits for linear structural\nequation models (SEMs) and soft interventions. It is assumed that the graph's\nstructure is known and has $N$ nodes. Two linear mechanisms, one soft\nintervention and one observational, are assumed for each node, giving rise to\n$2^N$ possible interventions. Majority of the existing causal bandit algorithms\nassume that at least the interventional distributions of the reward node's\nparents are fully specified. However, there are $2^N$ such distributions (one\ncorresponding to each intervention), acquiring which becomes prohibitive even\nin moderate-sized graphs. This paper dispenses with the assumption of knowing\nthese distributions or their marginals. Two algorithms are proposed for the\nfrequentist (UCB-based) and Bayesian (Thompson Sampling-based) settings. The\nkey idea of these algorithms is to avoid directly estimating the $2^N$ reward\ndistributions and instead estimate the parameters that fully specify the SEMs\n(linear in $N$) and use them to compute the rewards. In both algorithms, under\nboundedness assumptions on noise and the parameter space, the cumulative\nregrets scale as $\\tilde{\\cal O} (d^{L+\\frac{1}{2}} \\sqrt{NT})$, where $d$ is\nthe graph's maximum degree, and $L$ is the length of its longest causal path.\nAdditionally, a minimax lower of $\\Omega(d^{\\frac{L}{2}-2}\\sqrt{T})$ is\npresented, which suggests that the achievable and lower bounds conform in their\nscaling behavior with respect to the horizon $T$ and graph parameters $d$ and\n$L$.\n", "date": "2022-08-26", "categories": []}
{"id": "2301.05169", "title": "Causal Triplet: An Open Challenge for Intervention-centric Causal\n  Representation Learning", "abstract": "  Recent years have seen a surge of interest in learning high-level causal\nrepresentations from low-level image pairs under interventions. Yet, existing\nefforts are largely limited to simple synthetic settings that are far away from\nreal-world problems. In this paper, we present Causal Triplet, a causal\nrepresentation learning benchmark featuring not only visually more complex\nscenes, but also two crucial desiderata commonly overlooked in previous works:\n(i) an actionable counterfactual setting, where only certain object-level\nvariables allow for counterfactual observations whereas others do not; (ii) an\ninterventional downstream task with an emphasis on out-of-distribution\nrobustness from the independent causal mechanisms principle. Through extensive\nexperiments, we find that models built with the knowledge of disentangled or\nobject-centric representations significantly outperform their distributed\ncounterparts. However, recent causal representation learning methods still\nstruggle to identify such latent structures, indicating substantial challenges\nand opportunities for future work. Our code and datasets will be available at\nhttps://sites.google.com/view/causaltriplet.\n", "date": "2023-01-12", "categories": []}
{"id": "2303.17648", "title": "Practical Policy Optimization with Personalized Experimentation", "abstract": "  Many organizations measure treatment effects via an experimentation platform\nto evaluate the casual effect of product variations prior to full-scale\ndeployment. However, standard experimentation platforms do not perform\noptimally for end user populations that exhibit heterogeneous treatment effects\n(HTEs). Here we present a personalized experimentation framework, Personalized\nExperiments (PEX), which optimizes treatment group assignment at the user level\nvia HTE modeling and sequential decision policy optimization to optimize\nmultiple short-term and long-term outcomes simultaneously. We describe an\nend-to-end workflow that has proven to be successful in practice and can be\nreadily implemented using open-source software.\n", "date": "2023-03-30", "categories": []}
{"id": "2303.14349", "title": "Causal Image Synthesis of Brain MR in 3D", "abstract": "  Clinical decision making requires counterfactual reasoning based on a factual\nmedical image and thus necessitates causal image synthesis. To this end, we\npresent a novel method for modeling the causality between demographic\nvariables, clinical indices and brain MR images for Alzheimer's Diseases.\nSpecifically, we leverage a structural causal model to depict the causality and\na styled generator to synthesize the image. Furthermore, as a crucial step to\nreduce modeling complexity and make learning tractable, we propose the use of\nlow dimensional latent feature representation of a high-dimensional 3D image,\ntogether with exogenous noise, to build causal relationship between the image\nand non image variables. We experiment the proposed method based on 1586\nsubjects and 3683 3D images and synthesize counterfactual brain MR images\nintervened on certain attributes, such as age, brain volume and cognitive test\nscore. Quantitative metrics and qualitative evaluation of counterfactual images\ndemonstrates the superiority of our generated images.\n", "date": "2023-03-25", "categories": []}
{"id": "2303.14665", "title": "Achieving Counterfactual Fairness with Imperfect Structural Causal Model", "abstract": "  Counterfactual fairness alleviates the discrimination between the model\nprediction toward an individual in the actual world (observational data) and\nthat in counterfactual world (i.e., what if the individual belongs to other\nsensitive groups). The existing studies need to pre-define the structural\ncausal model that captures the correlations among variables for counterfactual\ninference; however, the underlying causal model is usually unknown and\ndifficult to be validated in real-world scenarios. Moreover, the\nmisspecification of the causal model potentially leads to poor performance in\nmodel prediction and thus makes unfair decisions. In this research, we propose\na novel minimax game-theoretic model for counterfactual fairness that can\nproduce accurate results meanwhile achieve a counterfactually fair decision\nwith the relaxation of strong assumptions of structural causal models. In\naddition, we also theoretically prove the error bound of the proposed minimax\nmodel. Empirical experiments on multiple real-world datasets illustrate our\nsuperior performance in both accuracy and fairness. Source code is available at\n\\url{https://github.com/tridungduong16/counterfactual_fairness_game_theoretic}.\n", "date": "2023-03-26", "categories": []}
{"id": "2303.14084", "title": "Differentially Private Synthetic Control", "abstract": "  Synthetic control is a causal inference tool used to estimate the treatment\neffects of an intervention by creating synthetic counterfactual data. This\napproach combines measurements from other similar observations (i.e., donor\npool ) to predict a counterfactual time series of interest (i.e., target unit)\nby analyzing the relationship between the target and the donor pool before the\nintervention. As synthetic control tools are increasingly applied to sensitive\nor proprietary data, formal privacy protections are often required. In this\nwork, we provide the first algorithms for differentially private synthetic\ncontrol with explicit error bounds. Our approach builds upon tools from\nnon-private synthetic control and differentially private empirical risk\nminimization. We provide upper and lower bounds on the sensitivity of the\nsynthetic control query and provide explicit error bounds on the accuracy of\nour private synthetic control algorithms. We show that our algorithms produce\naccurate predictions for the target unit, and that the cost of privacy is\nsmall. Finally, we empirically evaluate the performance of our algorithm, and\nshow favorable performance in a variety of parameter regimes, as well as\nproviding guidance to practitioners for hyperparameter tuning.\n", "date": "2023-03-24", "categories": []}
{"id": "2303.12703", "title": "Causal Reasoning in the Presence of Latent Confounders via Neural ADMG\n  Learning", "abstract": "  Latent confounding has been a long-standing obstacle for causal reasoning\nfrom observational data. One popular approach is to model the data using\nacyclic directed mixed graphs (ADMGs), which describe ancestral relations\nbetween variables using directed and bidirected edges. However, existing\nmethods using ADMGs are based on either linear functional assumptions or a\ndiscrete search that is complicated to use and lacks computational tractability\nfor large datasets. In this work, we further extend the existing body of work\nand develop a novel gradient-based approach to learning an ADMG with non-linear\nfunctional relations from observational data. We first show that the presence\nof latent confounding is identifiable under the assumptions of bow-free ADMGs\nwith non-linear additive noise models. With this insight, we propose a novel\nneural causal model based on autoregressive flows for ADMG learning. This not\nonly enables us to determine complex causal structural relationships behind the\ndata in the presence of latent confounding, but also estimate their functional\nrelationships (hence treatment effects) simultaneously. We further validate our\napproach via experiments on both synthetic and real-world datasets, and\ndemonstrate the competitive performance against relevant baselines.\n", "date": "2023-03-22", "categories": []}
{"id": "2208.12210", "title": "Learning Relational Causal Models with Cycles through Relational\n  Acyclification", "abstract": "  In real-world phenomena which involve mutual influence or causal effects\nbetween interconnected units, equilibrium states are typically represented with\ncycles in graphical models. An expressive class of graphical models, relational\ncausal models, can represent and reason about complex dynamic systems\nexhibiting such cycles or feedback loops. Existing cyclic causal discovery\nalgorithms for learning causal models from observational data assume that the\ndata instances are independent and identically distributed which makes them\nunsuitable for relational causal models. At the same time, causal discovery\nalgorithms for relational causal models assume acyclicity. In this work, we\nexamine the necessary and sufficient conditions under which a constraint-based\nrelational causal discovery algorithm is sound and complete for cyclic\nrelational causal models. We introduce relational acyclification, an operation\nspecifically designed for relational models that enables reasoning about the\nidentifiability of cyclic relational causal models. We show that under the\nassumptions of relational acyclification and $\\sigma$-faithfulness, the\nrelational causal discovery algorithm RCD (Maier et al. 2013) is sound and\ncomplete for cyclic models. We present experimental results to support our\nclaim.\n", "date": "2022-08-25", "categories": []}
{"id": "2212.05925", "title": "CausalEGM: a general causal inference framework by encoding generative\n  modeling", "abstract": "  Although understanding and characterizing causal effects have become\nessential in observational studies, it is challenging when the confounders are\nhigh-dimensional. In this article, we develop a general framework\n$\\textit{CausalEGM}$ for estimating causal effects by encoding generative\nmodeling, which can be applied in both binary and continuous treatment\nsettings. Under the potential outcome framework with unconfoundedness, we\nestablish a bidirectional transformation between the high-dimensional\nconfounders space and a low-dimensional latent space where the density is known\n(e.g., multivariate normal distribution). Through this, CausalEGM\nsimultaneously decouples the dependencies of confounders on both treatment and\noutcome and maps the confounders to the low-dimensional latent space. By\nconditioning on the low-dimensional latent features, CausalEGM can estimate the\ncausal effect for each individual or the average causal effect within a\npopulation. Our theoretical analysis shows that the excess risk for CausalEGM\ncan be bounded through empirical process theory. Under an assumption on\nencoder-decoder networks, the consistency of the estimate can be guaranteed. In\na series of experiments, CausalEGM demonstrates superior performance over\nexisting methods for both binary and continuous treatments. Specifically, we\nfind CausalEGM to be substantially more powerful than competing methods in the\npresence of large sample sizes and high dimensional confounders. The software\nof CausalEGM is freely available at https://github.com/SUwonglab/CausalEGM.\n", "date": "2022-12-08", "categories": []}
{"id": "2212.02932", "title": "Learning to Bound Counterfactual Inference from Observational, Biased\n  and Randomised Data", "abstract": "  We address the problem of integrating data from multiple, possibly biased,\nobservational and interventional studies, to eventually compute counterfactuals\nin structural causal models. We start from the case of a single observational\ndataset affected by a selection bias. We show that the likelihood of the\navailable data has no local maxima. This enables us to use the causal\nexpectation-maximisation scheme to compute approximate bounds for partially\nidentifiable counterfactual queries, which are the focus of this paper. We then\nshow how the same approach can solve the general case of multiple datasets, no\nmatter whether interventional or observational, biased or unbiased, by\nremapping it into the former one via graphical transformations. Systematic\nnumerical experiments and a case study on palliative care show the\neffectiveness and accuracy of our approach, while hinting at the benefits of\nintegrating heterogeneous data to get informative bounds in case of partial\nidentifiability.\n", "date": "2022-12-06", "categories": []}
{"id": "2303.08572", "title": "Distinguishing Cause from Effect on Categorical Data: The Uniform\n  Channel Model", "abstract": "  Distinguishing cause from effect using observations of a pair of random\nvariables is a core problem in causal discovery. Most approaches proposed for\nthis task, namely additive noise models (ANM), are only adequate for\nquantitative data. We propose a criterion to address the cause-effect problem\nwith categorical variables (living in sets with no meaningful order), inspired\nby seeing a conditional probability mass function (pmf) as a discrete\nmemoryless channel. We select as the most likely causal direction the one in\nwhich the conditional pmf is closer to a uniform channel (UC). The rationale is\nthat, in a UC, as in an ANM, the conditional entropy (of the effect given the\ncause) is independent of the cause distribution, in agreement with the\nprinciple of independence of cause and mechanism. Our approach, which we call\nthe uniform channel model (UCM), thus extends the ANM rationale to categorical\nvariables. To assess how close a conditional pmf (estimated from data) is to a\nUC, we use statistical testing, supported by a closed-form estimate of a UC\nchannel. On the theoretical front, we prove identifiability of the UCM and show\nits equivalence with a structural causal model with a low-cardinality exogenous\nvariable. Finally, the proposed method compares favorably with recent\nstate-of-the-art alternatives in experiments on synthetic, benchmark, and real\ndata.\n", "date": "2023-03-14", "categories": []}
{"id": "2206.07883", "title": "Combinatorial Pure Exploration of Causal Bandits", "abstract": "  The combinatorial pure exploration of causal bandits is the following online\nlearning task: given a causal graph with unknown causal inference\ndistributions, in each round we choose a subset of variables to intervene or do\nno intervention, and observe the random outcomes of all random variables, with\nthe goal that using as few rounds as possible, we can output an intervention\nthat gives the best (or almost best) expected outcome on the reward variable\n$Y$ with probability at least $1-\\delta$, where $\\delta$ is a given confidence\nlevel. We provide the first gap-dependent and fully adaptive pure exploration\nalgorithms on two types of causal models -- the binary generalized linear model\n(BGLM) and general graphs. For BGLM, our algorithm is the first to be designed\nspecifically for this setting and achieves polynomial sample complexity, while\nall existing algorithms for general graphs have either sample complexity\nexponential to the graph size or some unreasonable assumptions. For general\ngraphs, our algorithm provides a significant improvement on sample complexity,\nand it nearly matches the lower bound we prove. Our algorithms achieve such\nimprovement by a novel integration of prior causal bandit algorithms and prior\nadaptive pure exploration algorithms, the former of which utilize the rich\nobservational feedback in causal bandits but are not adaptive to reward gaps,\nwhile the latter of which have the issue in reverse.\n", "date": "2022-06-16", "categories": []}
{"id": "2303.07774", "title": "Testing Causality for High Dimensional Data", "abstract": "  Determining causal relationship between high dimensional observations are\namong the most important tasks in scientific discoveries. In this paper, we\nrevisited the \\emph{linear trace method}, a technique proposed\nin~\\citep{janzing2009telling,zscheischler2011testing} to infer the causal\ndirection between two random variables of high dimensions. We strengthen the\nexisting results significantly by providing an improved tail analysis in\naddition to extending the results to nonlinear trace functionals with sharper\nconfidence bounds under certain distributional assumptions. We obtain our\nresults by interpreting the trace estimator in the causal regime as a function\nover random orthogonal matrices, where the concentration of Lipschitz functions\nover such space could be applied. We additionally propose a novel\nridge-regularized variant of the estimator in \\cite{zscheischler2011testing},\nand give provable bounds relating the ridge-estimated terms to their\nground-truth counterparts. We support our theoretical results with encouraging\nexperiments on synthetic datasets, more prominently, under high-dimension low\nsample size regime.\n", "date": "2023-03-14", "categories": []}
{"id": "2301.05975", "title": "Generalized Invariant Matching Property via LASSO", "abstract": "  Learning under distribution shifts is a challenging task. One principled\napproach is to exploit the invariance principle via the structural causal\nmodels. However, the invariance principle is violated when the response is\nintervened, making it a difficult setting. In a recent work, the invariant\nmatching property has been developed to shed light on this scenario and shows\npromising performance. In this work, by formulating a high-dimensional problem\nwith intrinsic sparsity, we generalize the invariant matching property for an\nimportant setting when only the target is intervened. We propose a more robust\nand computation-efficient algorithm by leveraging a variant of Lasso, improving\nupon the existing algorithms.\n", "date": "2023-01-14", "categories": []}
{"id": "2303.06423", "title": "Learning interpretable causal networks from very large datasets,\n  application to 400,000 medical records of breast cancer patients", "abstract": "  Discovering causal effects is at the core of scientific investigation but\nremains challenging when only observational data is available. In practice,\ncausal networks are difficult to learn and interpret, and limited to relatively\nsmall datasets. We report a more reliable and scalable causal discovery method\n(iMIIC), based on a general mutual information supremum principle, which\ngreatly improves the precision of inferred causal relations while\ndistinguishing genuine causes from putative and latent causal effects. We\nshowcase iMIIC on synthetic and real-life healthcare data from 396,179 breast\ncancer patients from the US Surveillance, Epidemiology, and End Results\nprogram. More than 90\\% of predicted causal effects appear correct, while the\nremaining unexpected direct and indirect causal effects can be interpreted in\nterms of diagnostic procedures, therapeutic timing, patient preference or\nsocio-economic disparity. iMIIC's unique capabilities open up new avenues to\ndiscover reliable and interpretable causal networks across a range of research\nfields.\n", "date": "2023-03-11", "categories": []}
{"id": "2210.08185", "title": "GFlowCausal: Generative Flow Networks for Causal Discovery", "abstract": "  Causal discovery aims to uncover causal structure among a set of variables.\nScore-based approaches mainly focus on searching for the best Directed Acyclic\nGraph (DAG) based on a predefined score function. However, most of them are not\napplicable on a large scale due to the limited searchability. Inspired by the\nactive learning in generative flow networks, we propose a novel approach to\nlearning a DAG from observational data called GFlowCausal. It converts the\ngraph search problem to a generation problem, in which direct edges are added\ngradually. GFlowCausal aims to learn the best policy to generate high-reward\nDAGs by sequential actions with probabilities proportional to predefined\nrewards. We propose a plug-and-play module based on transitive closure to\nensure efficient sampling. Theoretical analysis shows that this module could\nguarantee acyclicity properties effectively and the consistency between final\nstates and fully-connected graphs. We conduct extensive experiments on both\nsynthetic and real datasets, and results show the proposed approach to be\nsuperior and also performs well in a large-scale setting.\n", "date": "2022-10-15", "categories": []}
{"id": "2211.10257", "title": "Model-based Causal Bayesian Optimization", "abstract": "  How should we intervene on an unknown structural equation model to maximize a\ndownstream variable of interest? This setting, also known as causal Bayesian\noptimization (CBO), has important applications in medicine, ecology, and\nmanufacturing. Standard Bayesian optimization algorithms fail to effectively\nleverage the underlying causal structure. Existing CBO approaches assume\nnoiseless measurements and do not come with guarantees. We propose the\nmodel-based causal Bayesian optimization algorithm (MCBO) that learns a full\nsystem model instead of only modeling intervention-reward pairs. MCBO\npropagates epistemic uncertainty about the causal mechanisms through the graph\nand trades off exploration and exploitation via the optimism principle. We\nbound its cumulative regret, and obtain the first non-asymptotic bounds for\nCBO. Unlike in standard Bayesian optimization, our acquisition function cannot\nbe evaluated in closed form, so we show how the reparameterization trick can be\nused to apply gradient-based optimizers. The resulting practical implementation\nof MCBO compares favorably with state-of-the-art approaches empirically.\n", "date": "2022-11-18", "categories": []}
{"id": "2302.05380", "title": "On the Interventional Kullback-Leibler Divergence", "abstract": "  Modern machine learning approaches excel in static settings where a large\namount of i.i.d. training data are available for a given task. In a dynamic\nenvironment, though, an intelligent agent needs to be able to transfer\nknowledge and re-use learned components across domains. It has been argued that\nthis may be possible through causal models, aiming to mirror the modularity of\nthe real world in terms of independent causal mechanisms. However, the true\ncausal structure underlying a given set of data is generally not identifiable,\nso it is desirable to have means to quantify differences between models (e.g.,\nbetween the ground truth and an estimate), on both the observational and\ninterventional level.\n  In the present work, we introduce the Interventional Kullback-Leibler (IKL)\ndivergence to quantify both structural and distributional differences between\nmodels based on a finite set of multi-environment distributions generated by\ninterventions from the ground truth. Since we generally cannot quantify all\ndifferences between causal models for every finite set of interventional\ndistributions, we propose a sufficient condition on the intervention targets to\nidentify subsets of observed variables on which the models provably agree or\ndisagree.\n", "date": "2023-02-10", "categories": []}
{"id": "2303.04874", "title": "Bayesian Causal Forests for Multivariate Outcomes: Application to Irish\n  Data From an International Large Scale Education Assessment", "abstract": "  Bayesian Causal Forests (BCF) is a causal inference machine learning model\nbased on a highly flexible non-parametric regression and classification tool\ncalled Bayesian Additive Regression Trees (BART). Motivated by data from the\nTrends in International Mathematics and Science Study (TIMSS), which includes\ndata on student achievement in both mathematics and science, we present a\nmultivariate extension of the BCF algorithm. With the help of simulation\nstudies we show that our approach can accurately estimate causal effects for\nmultiple outcomes subject to the same treatment. We also apply our model to\nIrish data from TIMSS 2019. Our findings reveal the positive effects of having\naccess to a study desk at home (Mathematics ATE 95% CI: [0.20, 11.67]) while\nalso highlighting the negative consequences of students often feeling hungry at\nschool (Mathematics ATE 95% CI: [-11.15, -2.78] , Science ATE 95% CI:\n[-10.82,-1.72]) or often being absent (Mathematics ATE 95% CI: [-12.47,\n-1.55]).\n", "date": "2023-03-08", "categories": []}
{"id": "2206.06169", "title": "Causal Representation Learning for Instantaneous and Temporal Effects in\n  Interactive Systems", "abstract": "  Causal representation learning is the task of identifying the underlying\ncausal variables and their relations from high-dimensional observations, such\nas images. Recent work has shown that one can reconstruct the causal variables\nfrom temporal sequences of observations under the assumption that there are no\ninstantaneous causal relations between them. In practical applications,\nhowever, our measurement or frame rate might be slower than many of the causal\neffects. This effectively creates \"instantaneous\" effects and invalidates\nprevious identifiability results. To address this issue, we propose iCITRIS, a\ncausal representation learning method that allows for instantaneous effects in\nintervened temporal sequences when intervention targets can be observed, e.g.,\nas actions of an agent. iCITRIS identifies the potentially multidimensional\ncausal variables from temporal observations, while simultaneously using a\ndifferentiable causal discovery method to learn their causal graph. In\nexperiments on three datasets of interactive systems, iCITRIS accurately\nidentifies the causal variables and their causal graph.\n", "date": "2022-06-13", "categories": []}
{"id": "2303.04579", "title": "\"How to make them stay?\" -- Diverse Counterfactual Explanations of\n  Employee Attrition", "abstract": "  Employee attrition is an important and complex problem that can directly\naffect an organisation's competitiveness and performance. Explaining the\nreasons why employees leave an organisation is a key human resource management\nchallenge due to the high costs and time required to attract and keep talented\nemployees. Businesses therefore aim to increase employee retention rates to\nminimise their costs and maximise their performance. Machine learning (ML) has\nbeen applied in various aspects of human resource management including\nattrition prediction to provide businesses with insights on proactive measures\non how to prevent talented employees from quitting. Among these ML methods, the\nbest performance has been reported by ensemble or deep neural networks, which\nby nature constitute black box techniques and thus cannot be easily\ninterpreted. To enable the understanding of these models' reasoning several\nexplainability frameworks have been proposed. Counterfactual explanation\nmethods have attracted considerable attention in recent years since they can be\nused to explain and recommend actions to be performed to obtain the desired\noutcome. However current counterfactual explanations methods focus on\noptimising the changes to be made on individual cases to achieve the desired\noutcome. In the attrition problem it is important to be able to foresee what\nwould be the effect of an organisation's action to a group of employees where\nthe goal is to prevent them from leaving the company. Therefore, in this paper\nwe propose the use of counterfactual explanations focusing on multiple\nattrition cases from historical data, to identify the optimum interventions\nthat an organisation needs to make to its practices/policies to prevent or\nminimise attrition probability for these cases.\n", "date": "2023-03-08", "categories": []}
{"id": "2209.13708", "title": "Falsification before Extrapolation in Causal Effect Estimation", "abstract": "  Randomized Controlled Trials (RCTs) represent a gold standard when developing\npolicy guidelines. However, RCTs are often narrow, and lack data on broader\npopulations of interest. Causal effects in these populations are often\nestimated using observational datasets, which may suffer from unobserved\nconfounding and selection bias. Given a set of observational estimates (e.g.\nfrom multiple studies), we propose a meta-algorithm that attempts to reject\nobservational estimates that are biased. We do so using validation effects,\ncausal effects that can be inferred from both RCT and observational data. After\nrejecting estimators that do not pass this test, we generate conservative\nconfidence intervals on the extrapolated causal effects for subgroups not\nobserved in the RCT. Under the assumption that at least one observational\nestimator is asymptotically normal and consistent for both the validation and\nextrapolated effects, we provide guarantees on the coverage probability of the\nintervals output by our algorithm. To facilitate hypothesis testing in settings\nwhere causal effect transportation across datasets is necessary, we give\nconditions under which a doubly-robust estimator of group average treatment\neffects is asymptotically normal, even when flexible machine learning methods\nare used for estimation of nuisance parameters. We illustrate the properties of\nour approach on semi-synthetic and real world datasets, and show that it\ncompares favorably to standard meta-analysis techniques.\n", "date": "2022-09-27", "categories": []}
{"id": "2301.13133", "title": "Falsification of Internal and External Validity in Observational Studies\n  via Conditional Moment Restrictions", "abstract": "  Randomized Controlled Trials (RCT)s are relied upon to assess new treatments,\nbut suffer from limited power to guide personalized treatment decisions. On the\nother hand, observational (i.e., non-experimental) studies have large and\ndiverse populations, but are prone to various biases (e.g. residual\nconfounding). To safely leverage the strengths of observational studies, we\nfocus on the problem of falsification, whereby RCTs are used to validate causal\neffect estimates learned from observational data. In particular, we show that,\ngiven data from both an RCT and an observational study, assumptions on internal\nand external validity have an observable, testable implication in the form of a\nset of Conditional Moment Restrictions (CMRs). Further, we show that expressing\nthese CMRs with respect to the causal effect, or \"causal contrast\", as opposed\nto individual counterfactual means, provides a more reliable falsification\ntest. In addition to giving guarantees on the asymptotic properties of our\ntest, we demonstrate superior power and type I error of our approach on\nsemi-synthetic and real world datasets. Our approach is interpretable, allowing\na practitioner to visualize which subgroups in the population lead to\nfalsification of an observational study.\n", "date": "2023-01-30", "categories": []}
{"id": "2303.02320", "title": "Estimating Treatment Effects from Irregular Time Series Observations\n  with Hidden Confounders", "abstract": "  Causal analysis for time series data, in particular estimating individualized\ntreatment effect (ITE), is a key task in many real-world applications, such as\nfinance, retail, healthcare, etc. Real-world time series can include\nlarge-scale, irregular, and intermittent time series observations, raising\nsignificant challenges to existing work attempting to estimate treatment\neffects. Specifically, the existence of hidden confounders can lead to biased\ntreatment estimates and complicate the causal inference process. In particular,\nanomaly hidden confounders which exceed the typical range can lead to high\nvariance estimates. Moreover, in continuous time settings with irregular\nsamples, it is challenging to directly handle the dynamics of causality. In\nthis paper, we leverage recent advances in Lipschitz regularization and neural\ncontrolled differential equations (CDE) to develop an effective and scalable\nsolution, namely LipCDE, to address the above challenges. LipCDE can directly\nmodel the dynamic causal relationships between historical data and outcomes\nwith irregular samples by considering the boundary of hidden confounders given\nby Lipschitz-constrained neural networks. Furthermore, we conduct extensive\nexperiments on both synthetic and real-world datasets to demonstrate the\neffectiveness and scalability of LipCDE.\n", "date": "2023-03-04", "categories": []}
{"id": "2303.02980", "title": "KDSM: An uplift modeling framework based on knowledge distillation and\n  sample matching", "abstract": "  Uplift modeling aims to estimate the treatment effect on individuals, widely\napplied in the e-commerce platform to target persuadable customers and maximize\nthe return of marketing activities. Among the existing uplift modeling methods,\ntree-based methods are adept at fitting increment and generalization, while\nneural-network-based models excel at predicting absolute value and precision,\nand these advantages have not been fully explored and combined. Also, the lack\nof counterfactual sample pairs is the root challenge in uplift modeling. In\nthis paper, we proposed an uplift modeling framework based on Knowledge\nDistillation and Sample Matching (KDSM). The teacher model is the uplift\ndecision tree (UpliftDT), whose structure is exploited to construct\ncounterfactual sample pairs, and the pairwise incremental prediction is treated\nas another objective for the student model. Under the idea of multitask\nlearning, the student model can achieve better performance on generalization\nand even surpass the teacher. Extensive offline experiments validate the\nuniversality of different combinations of teachers and student models and the\nsuperiority of KDSM measured against the baselines. In online A/B testing, the\ncost of each incremental room night is reduced by 6.5\\%.\n", "date": "2023-03-06", "categories": []}
{"id": "2303.03187", "title": "Boosting Differentiable Causal Discovery via Adaptive Sample Reweighting", "abstract": "  Under stringent model type and variable distribution assumptions,\ndifferentiable score-based causal discovery methods learn a directed acyclic\ngraph (DAG) from observational data by evaluating candidate graphs over an\naverage score function. Despite great success in low-dimensional linear\nsystems, it has been observed that these approaches overly exploit\neasier-to-fit samples, thus inevitably learning spurious edges. Worse still,\ninherent mostly in these methods the common homogeneity assumption can be\neasily violated, due to the widespread existence of heterogeneous data in the\nreal world, resulting in performance vulnerability when noise distributions\nvary. We propose a simple yet effective model-agnostic framework to boost\ncausal discovery performance by dynamically learning the adaptive weights for\nthe Reweighted Score function, ReScore for short, where the weights tailor\nquantitatively to the importance degree of each sample. Intuitively, we\nleverage the bilevel optimization scheme to \\wx{alternately train a standard\nDAG learner and reweight samples -- that is, upweight the samples the learner\nfails to fit and downweight the samples that the learner easily extracts the\nspurious information from. Extensive experiments on both synthetic and\nreal-world datasets are carried out to validate the effectiveness of ReScore.\nWe observe consistent and significant boosts in structure learning performance.\nFurthermore, we visualize that ReScore concurrently mitigates the influence of\nspurious edges and generalizes to heterogeneous data. Finally, we perform the\ntheoretical analysis to guarantee the structure identifiability and the weight\nadaptive properties of ReScore in linear systems. Our codes are available at\nhttps://github.com/anzhang314/ReScore.\n", "date": "2023-03-06", "categories": []}
{"id": "2303.01775", "title": "Continual Causal Inference with Incremental Observational Data", "abstract": "  The era of big data has witnessed an increasing availability of observational\ndata from mobile and social networking, online advertising, web mining,\nhealthcare, education, public policy, marketing campaigns, and so on, which\nfacilitates the development of causal effect estimation. Although significant\nadvances have been made to overcome the challenges in the academic area, such\nas missing counterfactual outcomes and selection bias, they only focus on\nsource-specific and stationary observational data, which is unrealistic in most\nindustrial applications. In this paper, we investigate a new industrial problem\nof causal effect estimation from incrementally available observational data and\npresent three new evaluation criteria accordingly, including extensibility,\nadaptability, and accessibility. We propose a Continual Causal Effect\nRepresentation Learning method for estimating causal effects with observational\ndata, which are incrementally available from non-stationary data distributions.\nInstead of having access to all seen observational data, our method only stores\na limited subset of feature representations learned from previous data.\nCombining selective and balanced representation learning, feature\nrepresentation distillation, and feature transformation, our method achieves\nthe continual causal effect estimation for new data without compromising the\nestimation capability for original data. Extensive experiments demonstrate the\nsignificance of continual causal effect estimation and the effectiveness of our\nmethod.\n", "date": "2023-03-03", "categories": []}
{"id": "2303.00821", "title": "Learning high-dimensional causal effect", "abstract": "  The scarcity of high-dimensional causal inference datasets restricts the\nexploration of complex deep models. In this work, we propose a method to\ngenerate a synthetic causal dataset that is high-dimensional. The synthetic\ndata simulates a causal effect using the MNIST dataset with Bernoulli treatment\nvalues. This provides an opportunity to study varieties of models for causal\neffect estimation. We experiment on this dataset using Dragonnet architecture\n(Shi et al. (2019)) and modified architectures. We use the modified\narchitectures to explore different types of initial Neural Network layers and\nobserve that the modified architectures perform better in estimations. We\nobserve that residual and transformer models estimate treatment effect very\nclosely without the need for targeted regularization, introduced by Shi et al.\n(2019).\n", "date": "2023-03-01", "categories": []}
{"id": "2303.01274", "title": "Measuring axiomatic soundness of counterfactual image models", "abstract": "  We present a general framework for evaluating image counterfactuals. The\npower and flexibility of deep generative models make them valuable tools for\nlearning mechanisms in structural causal models. However, their flexibility\nmakes counterfactual identifiability impossible in the general case. Motivated\nby these issues, we revisit Pearl's axiomatic definition of counterfactuals to\ndetermine the necessary constraints of any counterfactual inference model:\ncomposition, reversibility, and effectiveness. We frame counterfactuals as\nfunctions of an input variable, its parents, and counterfactual parents and use\nthe axiomatic constraints to restrict the set of functions that could represent\nthe counterfactual, thus deriving distance metrics between the approximate and\nideal functions. We demonstrate how these metrics can be used to compare and\nchoose between different approximate counterfactual inference models and to\nprovide insight into a model's shortcomings and trade-offs.\n", "date": "2023-03-02", "categories": []}
{"id": "2302.14386", "title": "Practical Algorithms for Orientations of Partially Directed Graphical\n  Models", "abstract": "  In observational studies, the true causal model is typically unknown and\nneeds to be estimated from available observational and limited experimental\ndata. In such cases, the learned causal model is commonly represented as a\npartially directed acyclic graph (PDAG), which contains both directed and\nundirected edges indicating uncertainty of causal relations between random\nvariables. The main focus of this paper is on the maximal orientation task,\nwhich, for a given PDAG, aims to orient the undirected edges maximally such\nthat the resulting graph represents the same Markov equivalent DAGs as the\ninput PDAG. This task is a subroutine used frequently in causal discovery, e.\ng., as the final step of the celebrated PC algorithm. Utilizing connections to\nthe problem of finding a consistent DAG extension of a PDAG, we derive faster\nalgorithms for computing the maximal orientation by proposing two novel\napproaches for extending PDAGs, both constructed with an emphasis on simplicity\nand practical effectiveness.\n", "date": "2023-02-28", "categories": []}
{"id": "2302.14412", "title": "An Algorithm and Complexity Results for Causal Unit Selection", "abstract": "  The unit selection problem aims to identify objects, called units, that are\nmost likely to exhibit a desired mode of behavior when subjected to stimuli\n(e.g., customers who are about to churn but would change their mind if\nencouraged). Unit selection with counterfactual objective functions was\nintroduced relatively recently with existing work focusing on bounding a\nspecific class of objective functions, called the benefit functions, based on\nobservational and interventional data -- assuming a fully specified model is\nnot available to evaluate these functions. We complement this line of work by\nproposing the first exact algorithm for finding optimal units given a broad\nclass of causal objective functions and a fully specified structural causal\nmodel (SCM). We show that unit selection under this class of objective\nfunctions is $\\text{NP}^\\text{PP}$-complete but is $\\text{NP}$-complete when\nunit variables correspond to all exogenous variables in the SCM. We also\nprovide treewidth-based complexity bounds on our proposed algorithm while\nrelating it to a well-known algorithm for Maximum a Posteriori (MAP) inference.\n", "date": "2023-02-28", "categories": []}
{"id": "2302.13240", "title": "Q-Cogni: An Integrated Causal Reinforcement Learning Framework", "abstract": "  We present Q-Cogni, an algorithmically integrated causal reinforcement\nlearning framework that redesigns Q-Learning with an autonomous causal\nstructure discovery method to improve the learning process with causal\ninference. Q-Cogni achieves optimal learning with a pre-learned structural\ncausal model of the environment that can be queried during the learning process\nto infer cause-and-effect relationships embedded in a state-action space. We\nleverage on the sample efficient techniques of reinforcement learning, enable\nreasoning about a broader set of policies and bring higher degrees of\ninterpretability to decisions made by the reinforcement learning agent. We\napply Q-Cogni on the Vehicle Routing Problem (VRP) and compare against\nstate-of-the-art reinforcement learning algorithms. We report results that\ndemonstrate better policies, improved learning efficiency and superior\ninterpretability of the agent's decision making. We also compare this approach\nwith traditional shortest-path search algorithms and demonstrate the benefits\nof our causal reinforcement learning framework to high dimensional problems.\nFinally, we apply Q-Cogni to derive optimal routing decisions for taxis in New\nYork City using the Taxi & Limousine Commission trip record data and compare\nwith shortest-path search, reporting results that show 85% of the cases with an\nequal or better policy derived from Q-Cogni in a real-world domain.\n", "date": "2023-02-26", "categories": []}
{"id": "2302.12504", "title": "Recovering Sparse and Interpretable Subgroups with Heterogeneous\n  Treatment Effects with Censored Time-to-Event Outcomes", "abstract": "  Studies involving both randomized experiments as well as observational data\ntypically involve time-to-event outcomes such as time-to-failure, death or\nonset of an adverse condition. Such outcomes are typically subject to censoring\ndue to loss of follow-up and established statistical practice involves\ncomparing treatment efficacy in terms of hazard ratios between the treated and\ncontrol groups. In this paper we propose a statistical approach to recovering\nsparse phenogroups (or subtypes) that demonstrate differential treatment\neffects as compared to the study population. Our approach involves modelling\nthe data as a mixture while enforcing parameter shrinkage through structured\nsparsity regularization. We propose a novel inference procedure for the\nproposed model and demonstrate its efficacy in recovering sparse phenotypes\nacross large landmark real world clinical studies in cardiovascular health.\n", "date": "2023-02-24", "categories": []}
{"id": "2302.12670", "title": "Personalized Pricing with Invalid Instrumental Variables:\n  Identification, Estimation, and Policy Learning", "abstract": "  Pricing based on individual customer characteristics is widely used to\nmaximize sellers' revenues. This work studies offline personalized pricing\nunder endogeneity using an instrumental variable approach. Standard\ninstrumental variable methods in causal inference/econometrics either focus on\na discrete treatment space or require the exclusion restriction of instruments\nfrom having a direct effect on the outcome, which limits their applicability in\npersonalized pricing. In this paper, we propose a new policy learning method\nfor Personalized pRicing using Invalid iNsTrumental variables (PRINT) for\ncontinuous treatment that allow direct effects on the outcome. Specifically,\nrelying on the structural models of revenue and price, we establish the\nidentifiability condition of an optimal pricing strategy under endogeneity with\nthe help of invalid instrumental variables. Based on this new identification,\nwhich leads to solving conditional moment restrictions with generalized\nresidual functions, we construct an adversarial min-max estimator and learn an\noptimal pricing strategy. Furthermore, we establish an asymptotic regret bound\nto find an optimal pricing strategy. Finally, we demonstrate the effectiveness\nof the proposed method via extensive simulation studies as well as a real data\napplication from an US online auto loan company.\n", "date": "2023-02-24", "categories": []}
{"id": "2302.12718", "title": "Understanding the Impact of Competing Events on Heterogeneous Treatment\n  Effect Estimation from Time-to-Event Data", "abstract": "  We study the problem of inferring heterogeneous treatment effects (HTEs) from\ntime-to-event data in the presence of competing events. Albeit its great\npractical relevance, this problem has received little attention compared to its\ncounterparts studying HTE estimation without time-to-event data or competing\nevents. We take an outcome modeling approach to estimating HTEs, and consider\nhow and when existing prediction models for time-to-event data can be used as\nplug-in estimators for potential outcomes. We then investigate whether\ncompeting events present new challenges for HTE estimation -- in addition to\nthe standard confounding problem --, and find that, because there are multiple\ndefinitions of causal effects in this setting -- namely total, direct and\nseparable effects --, competing events can act as an additional source of\ncovariate shift depending on the desired treatment effect interpretation and\nassociated estimand. We theoretically analyze and empirically illustrate when\nand how these challenges play a role when using generic machine learning\nprediction models for the estimation of HTEs.\n", "date": "2023-02-23", "categories": []}
{"id": "2206.01665", "title": "BaCaDI: Bayesian Causal Discovery with Unknown Interventions", "abstract": "  Inferring causal structures from experimentation is a central task in many\ndomains. For example, in biology, recent advances allow us to obtain\nsingle-cell expression data under multiple interventions such as drugs or gene\nknockouts. However, the targets of the interventions are often uncertain or\nunknown and the number of observations limited. As a result, standard causal\ndiscovery methods can no longer be reliably used. To fill this gap, we propose\na Bayesian framework (BaCaDI) for discovering and reasoning about the causal\nstructure that underlies data generated under various unknown experimental or\ninterventional conditions. BaCaDI is fully differentiable, which allows us to\ninfer the complex joint posterior over the intervention targets and the causal\nstructure via efficient gradient-based variational inference. In experiments on\nsynthetic causal discovery tasks and simulated gene-expression data, BaCaDI\noutperforms related methods in identifying causal structures and intervention\ntargets.\n", "date": "2022-06-03", "categories": []}
{"id": "2207.08457", "title": "A Meta-Reinforcement Learning Algorithm for Causal Discovery", "abstract": "  Causal discovery is a major task with the utmost importance for machine\nlearning since causal structures can enable models to go beyond pure\ncorrelation-based inference and significantly boost their performance. However,\nfinding causal structures from data poses a significant challenge both in\ncomputational effort and accuracy, let alone its impossibility without\ninterventions in general. In this paper, we develop a meta-reinforcement\nlearning algorithm that performs causal discovery by learning to perform\ninterventions such that it can construct an explicit causal graph. Apart from\nbeing useful for possible downstream applications, the estimated causal graph\nalso provides an explanation for the data-generating process. In this article,\nwe show that our algorithm estimates a good graph compared to the SOTA\napproaches, even in environments whose underlying causal structure is\npreviously unseen. Further, we make an ablation study that shows how learning\ninterventions contribute to the overall performance of our approach. We\nconclude that interventions indeed help boost the performance, efficiently\nyielding an accurate estimate of the causal structure of a possibly unseen\nenvironment.\n", "date": "2022-07-18", "categories": []}
{"id": "2302.06804", "title": "Discovering Optimal Scoring Mechanisms in Causal Strategic Prediction", "abstract": "  Faced with data-driven policies, individuals will manipulate their features\nto obtain favorable decisions. While earlier works cast these manipulations as\nundesirable gaming, recent works have adopted a more nuanced causal framing in\nwhich manipulations can improve outcomes of interest, and setting coherent\nmechanisms requires accounting for both predictive accuracy and improvement of\nthe outcome. Typically, these works focus on known causal graphs, consisting\nonly of an outcome and its parents. In this paper, we introduce a general\nframework in which an outcome and n observed features are related by an\narbitrary unknown graph and manipulations are restricted by a fixed budget and\ncost structure. We develop algorithms that leverage strategic responses to\ndiscover the causal graph in a finite number of steps. Given this graph\nstructure, we can then derive mechanisms that trade off between accuracy and\nimprovement. Altogether, our work deepens links between causal discovery and\nincentive design and provides a more nuanced view of learning under causal\nstrategic prediction.\n", "date": "2023-02-14", "categories": []}
{"id": "2302.09446", "title": "Estimating Treatment Effects in Continuous Time with Hidden Confounders", "abstract": "  Estimating treatment effects plays a crucial role in causal inference, having\nmany real-world applications like policy analysis and decision making.\nNevertheless, estimating treatment effects in the longitudinal setting in the\npresence of hidden confounders remains an extremely challenging problem.\nRecently, there is a growing body of work attempting to obtain unbiased ITE\nestimates from time-dynamic observational data by ignoring the possible\nexistence of hidden confounders. Additionally, many existing works handling\nhidden confounders are not applicable for continuous-time settings. In this\npaper, we extend the line of work focusing on deconfounding in the dynamic time\nsetting in the presence of hidden confounders. We leverage recent advancements\nin neural differential equations to build a latent factor model using a\nstochastic controlled differential equation and Lipschitz constrained\nconvolutional operation in order to continuously incorporate information about\nongoing interventions and irregularly sampled observations. Experiments on both\nsynthetic and real-world datasets highlight the promise of continuous time\nmethods for estimating treatment effects in the presence of hidden confounders.\n", "date": "2023-02-19", "categories": []}
{"id": "2302.10625", "title": "Estimating long-term causal effects from short-term experiments and\n  long-term observational data with unobserved confounding", "abstract": "  Understanding and quantifying cause and effect is an important problem in\nmany domains. The generally-agreed solution to this problem is to perform a\nrandomised controlled trial. However, even when randomised controlled trials\ncan be performed, they usually have relatively short duration's due to cost\nconsiderations. This makes learning long-term causal effects a very challenging\ntask in practice, since the long-term outcome is only observed after a long\ndelay. In this paper, we study the identification and estimation of long-term\ntreatment effects when both experimental and observational data are available.\nPrevious work provided an estimation strategy to determine long-term causal\neffects from such data regimes. However, this strategy only works if one\nassumes there are no unobserved confounders in the observational data. In this\npaper, we specifically address the challenging case where unmeasured\nconfounders are present in the observational data. Our long-term causal effect\nestimator is obtained by combining regression residuals with short-term\nexperimental outcomes in a specific manner to create an instrumental variable,\nwhich is then used to quantify the long-term causal effect through instrumental\nvariable regression. We prove this estimator is unbiased, and analytically\nstudy its variance. In the context of the front-door causal structure, this\nprovides a new causal estimator, which may be of independent interest. Finally,\nwe empirically test our approach on synthetic-data, as well as real-data from\nthe International Stroke Trial.\n", "date": "2023-02-21", "categories": []}
{"id": "2302.10804", "title": "GDBN: a Graph Neural Network Approach to Dynamic Bayesian Network", "abstract": "  Identifying causal relations among multi-variate time series is one of the\nmost important elements towards understanding the complex mechanisms underlying\nthe dynamic system. It provides critical tools for forecasting, simulations and\ninterventions in science and business analytics. In this paper, we proposed a\ngraph neural network approach with score-based method aiming at learning a\nsparse DAG that captures the causal dependencies in a discretized time temporal\ngraph. We demonstrate methods with graph neural network significantly\noutperformed other state-of-the-art methods with dynamic bayesian networking\ninference. In addition, from the experiments, the structural causal model can\nbe more accurate than a linear SCM discovered by the methods such as Notears.\n", "date": "2023-01-28", "categories": []}
{"id": "2210.16535", "title": "Causal Discovery of Dynamic Models for Predicting Human Spatial\n  Interactions", "abstract": "  Exploiting robots for activities in human-shared environments, whether\nwarehouses, shopping centres or hospitals, calls for such robots to understand\nthe underlying physical interactions between nearby agents and objects. In\nparticular, modelling cause-and-effect relations between the latter can help to\npredict unobserved human behaviours and anticipate the outcome of specific\nrobot interventions. In this paper, we propose an application of causal\ndiscovery methods to model human-robot spatial interactions, trying to\nunderstand human behaviours from real-world sensor data in two possible\nscenarios: humans interacting with the environment, and humans interacting with\nobstacles. New methods and practical solutions are discussed to exploit, for\nthe first time, a state-of-the-art causal discovery algorithm in some\nchallenging human environments, with potential application in many service\nrobotics scenarios. To demonstrate the utility of the causal models obtained\nfrom real-world datasets, we present a comparison between causal and non-causal\nprediction approaches. Our results show that the causal model correctly\ncaptures the underlying interactions of the considered scenarios and improves\nits prediction accuracy.\n", "date": "2022-10-29", "categories": []}
{"id": "2302.09067", "title": "Causal Confirmation Measures: From Simpson's Paradox to COVID-19", "abstract": "  When we compare the influences of two causes on an outcome, if the conclusion\nfrom every group is against that from the conflation, we think there is\nSimpson's Paradox. The Existing Causal Inference Theory (ECIT) can make the\noverall conclusion consistent with the grouping conclusion by removing the\nconfounder's influence to eliminate the paradox. The ECIT uses relative risk\ndifference Pd = max(0, (R - 1)/R) (R denotes the risk ratio) as the probability\nof causation. In contrast, Philosopher Fitelson uses confirmation measure D\n(posterior probability minus prior probability) to measure the strength of\ncausation. Fitelson concludes that from the perspective of Bayesian\nconfirmation, we should directly accept the overall conclusion without\nconsidering the paradox. The author proposed a Bayesian confirmation measure b*\nsimilar to Pd before. To overcome the contradiction between the ECIT and\nBayesian confirmation, the author uses the semantic information method with the\nminimum cross-entropy criterion to deduce causal confirmation measure Cc = (R\n-1)/max(R, 1). Cc is like Pd but has normalizing property (between -1 and 1)\nand cause symmetry. It especially fits cases where a cause restrains an\noutcome, such as the COVID-19 vaccine controlling the infection. Some examples\n(about kidney stone treatments and COVID-19) reveal that Pd and Cc are more\nreasonable than D; Cc is more useful than Pd.\n", "date": "2023-02-03", "categories": []}
{"id": "2209.02946", "title": "On the Sparse DAG Structure Learning Based on Adaptive Lasso", "abstract": "  Learning the underlying Bayesian Networks (BNs), represented by directed\nacyclic graphs (DAGs), of the concerned events from purely-observational data\nis a crucial part of evidential reasoning. This task remains challenging due to\nthe large and discrete search space. A recent flurry of developments followed\nNOTEARS[1] recast this combinatorial problem into a continuous optimization\nproblem by leveraging an algebraic equality characterization of acyclicity.\nHowever, the continuous optimization methods suffer from obtaining non-spare\ngraphs after the numerical optimization, which leads to the inflexibility to\nrule out the potentially cycle-inducing edges or false discovery edges with\nsmall values. To address this issue, in this paper, we develop a completely\ndata-driven DAG structure learning method without a predefined value to\npost-threshold small values. We name our method NOTEARS with adaptive Lasso\n(NOTEARS-AL), which is achieved by applying the adaptive penalty method to\nensure the sparsity of the estimated DAG. Moreover, we show that NOTEARS-AL\nalso inherits the oracle properties under some specific conditions. Extensive\nexperiments on both synthetic and a real-world dataset demonstrate that our\nmethod consistently outperforms NOTEARS.\n", "date": "2022-09-07", "categories": []}
{"id": "2211.03553", "title": "Learning Causal Representations of Single Cells via Sparse Mechanism\n  Shift Modeling", "abstract": "  Latent variable models such as the Variational Auto-Encoder (VAE) have become\na go-to tool for analyzing biological data, especially in the field of\nsingle-cell genomics. One remaining challenge is the interpretability of latent\nvariables as biological processes that define a cell's identity. Outside of\nbiological applications, this problem is commonly referred to as learning\ndisentangled representations. Although several disentanglement-promoting\nvariants of the VAE were introduced, and applied to single-cell genomics data,\nthis task has been shown to be infeasible from independent and identically\ndistributed measurements, without additional structure. Instead, recent methods\npropose to leverage non-stationary data, as well as the sparse mechanism shift\nassumption in order to learn disentangled representations with a causal\nsemantic. Here, we extend the application of these methodological advances to\nthe analysis of single-cell genomics data with genetic or chemical\nperturbations. More precisely, we propose a deep generative model of\nsingle-cell gene expression data for which each perturbation is treated as a\nstochastic intervention targeting an unknown, but sparse, subset of latent\nvariables. We benchmark these methods on simulated single-cell data to evaluate\ntheir performance at latent units recovery, causal target identification and\nout-of-domain generalization. Finally, we apply those approaches to two\nreal-world large-scale gene perturbation data sets and find that models that\nexploit the sparse mechanism shift hypothesis surpass contemporary methods on a\ntransfer learning task. We implement our new model and benchmarks using the\nscvi-tools library, and release it as open-source software at\nhttps://github.com/Genentech/sVAE.\n", "date": "2022-11-07", "categories": []}
{"id": "2206.06410", "title": "Estimating Causal Effects Under Image Confounding Bias with an\n  Application to Poverty in Africa", "abstract": "  Observational studies of causal effects require adjustment for confounding\nfactors. In the tabular setting, where these factors are well-defined, separate\nrandom variables, the effect of confounding is well understood. However, in\npublic policy, ecology, and in medicine, decisions are often made in\nnon-tabular settings, informed by patterns or objects detected in images (e.g.,\nmaps, satellite or tomography imagery). Using such imagery for causal inference\npresents an opportunity because objects in the image may be related to the\ntreatment and outcome of interest. In these cases, we rely on the images to\nadjust for confounding but observed data do not directly label the existence of\nthe important objects. Motivated by real-world applications, we formalize this\nchallenge, how it can be handled, and what conditions are sufficient to\nidentify and estimate causal effects. We analyze finite-sample performance\nusing simulation experiments, estimating effects using a propensity adjustment\nalgorithm that employs a machine learning model to estimate the image\nconfounding. Our experiments also examine sensitivity to misspecification of\nthe image pattern mechanism. Finally, we use our methodology to estimate the\neffects of policy interventions on poverty in African communities from\nsatellite imagery.\n", "date": "2022-06-13", "categories": []}
{"id": "2302.07458", "title": "CUTS: Neural Causal Discovery from Irregular Time-Series Data", "abstract": "  Causal discovery from time-series data has been a central task in machine\nlearning. Recently, Granger causality inference is gaining momentum due to its\ngood explainability and high compatibility with emerging deep neural networks.\nHowever, most existing methods assume structured input data and degenerate\ngreatly when encountering data with randomly missing entries or non-uniform\nsampling frequencies, which hampers their applications in real scenarios. To\naddress this issue, here we present CUTS, a neural Granger causal discovery\nalgorithm to jointly impute unobserved data points and build causal graphs, via\nplugging in two mutually boosting modules in an iterative framework: (i) Latent\ndata prediction stage: designs a Delayed Supervision Graph Neural Network\n(DSGNN) to hallucinate and register unstructured data which might be of high\ndimension and with complex distribution; (ii) Causal graph fitting stage:\nbuilds a causal adjacency matrix with imputed data under sparse penalty.\nExperiments show that CUTS effectively infers causal graphs from unstructured\ntime-series data, with significantly superior performance to existing methods.\nOur approach constitutes a promising step towards applying causal discovery to\nreal applications with non-ideal observations.\n", "date": "2023-02-15", "categories": []}
{"id": "2302.06975", "title": "A Review of the Role of Causality in Developing Trustworthy AI Systems", "abstract": "  State-of-the-art AI models largely lack an understanding of the cause-effect\nrelationship that governs human understanding of the real world. Consequently,\nthese models do not generalize to unseen data, often produce unfair results,\nand are difficult to interpret. This has led to efforts to improve the\ntrustworthiness aspects of AI models. Recently, causal modeling and inference\nmethods have emerged as powerful tools. This review aims to provide the reader\nwith an overview of causal methods that have been developed to improve the\ntrustworthiness of AI models. We hope that our contribution will motivate\nfuture research on causality-based solutions for trustworthy AI.\n", "date": "2023-02-14", "categories": []}
{"id": "2207.01575", "title": "De-Biasing Generative Models using Counterfactual Methods", "abstract": "  Variational autoencoders (VAEs) and other generative methods have garnered\ngrowing interest not just for their generative properties but also for the\nability to dis-entangle a low-dimensional latent variable space. However, few\nexisting generative models take causality into account. We propose a new\ndecoder based framework named the Causal Counterfactual Generative Model\n(CCGM), which includes a partially trainable causal layer in which a part of a\ncausal model can be learned without significantly impacting reconstruction\nfidelity. By learning the causal relationships between image semantic labels or\ntabular variables, we can analyze biases, intervene on the generative model,\nand simulate new scenarios. Furthermore, by modifying the causal structure, we\ncan generate samples outside the domain of the original training data and use\nsuch counterfactual models to de-bias datasets. Thus, datasets with known\nbiases can still be used to train the causal generative model and learn the\ncausal relationships, but we can produce de-biased datasets on the generative\nside. Our proposed method combines a causal latent space VAE model with\nspecific modification to emphasize causal fidelity, enabling finer control over\nthe causal layer and the ability to learn a robust intervention framework. We\nexplore how better disentanglement of causal learning and encoding/decoding\ngenerates higher causal intervention quality. We also compare our model against\nsimilar research to demonstrate the need for explicit generative de-biasing\nbeyond interventions. Our initial experiments show that our model can generate\nimages and tabular data with high fidelity to the causal framework and\naccommodate explicit de-biasing to ignore undesired relationships in the causal\ndata compared to the baseline.\n", "date": "2022-07-04", "categories": []}
{"id": "2208.08584", "title": "Robust Causal Graph Representation Learning against Confounding Effects", "abstract": "  The prevailing graph neural network models have achieved significant progress\nin graph representation learning. However, in this paper, we uncover an\never-overlooked phenomenon: the pre-trained graph representation learning model\ntested with full graphs underperforms the model tested with well-pruned graphs.\nThis observation reveals that there exist confounders in graphs, which may\ninterfere with the model learning semantic information, and current graph\nrepresentation learning methods have not eliminated their influence. To tackle\nthis issue, we propose Robust Causal Graph Representation Learning (RCGRL) to\nlearn robust graph representations against confounding effects. RCGRL\nintroduces an active approach to generate instrumental variables under\nunconditional moment restrictions, which empowers the graph representation\nlearning model to eliminate confounders, thereby capturing discriminative\ninformation that is causally related to downstream predictions. We offer\ntheorems and proofs to guarantee the theoretical effectiveness of the proposed\napproach. Empirically, we conduct extensive experiments on a synthetic dataset\nand multiple benchmark datasets. The results demonstrate that compared with\nstate-of-the-art methods, RCGRL achieves better prediction performance and\ngeneralization ability.\n", "date": "2022-08-18", "categories": []}
{"id": "2301.11898", "title": "DAG Learning on the Permutahedron", "abstract": "  We propose a continuous optimization framework for discovering a latent\ndirected acyclic graph (DAG) from observational data. Our approach optimizes\nover the polytope of permutation vectors, the so-called Permutahedron, to learn\na topological ordering. Edges can be optimized jointly, or learned conditional\non the ordering via a non-differentiable subroutine. Compared to existing\ncontinuous optimization approaches our formulation has a number of advantages\nincluding: 1. validity: optimizes over exact DAGs as opposed to other\nrelaxations optimizing approximate DAGs; 2. modularity: accommodates any\nedge-optimization procedure, edge structural parameterization, and optimization\nloss; 3. end-to-end: either alternately iterates between node-ordering and\nedge-optimization, or optimizes them jointly. We demonstrate, on real-world\ndata problems in protein-signaling and transcriptional network discovery, that\nour approach lies on the Pareto frontier of two key metrics, the SID and SHD.\n", "date": "2023-01-27", "categories": []}
{"id": "2302.03465", "title": "Robustness Implies Fairness in Causal Algorithmic Recourse", "abstract": "  Algorithmic recourse aims to disclose the inner workings of the black-box\ndecision process in situations where decisions have significant consequences,\nby providing recommendations to empower beneficiaries to achieve a more\nfavorable outcome. To ensure an effective remedy, suggested interventions must\nnot only be low-cost but also robust and fair. This goal is accomplished by\nproviding similar explanations to individuals who are alike. This study\nexplores the concept of individual fairness and adversarial robustness in\ncausal algorithmic recourse and addresses the challenge of achieving both. To\nresolve the challenges, we propose a new framework for defining adversarially\nrobust recourse. The new setting views the protected feature as a pseudometric\nand demonstrates that individual fairness is a special case of adversarial\nrobustness. Finally, we introduce the fair robust recourse problem to achieve\nboth desirable properties and show how it can be satisfied both theoretically\nand empirically.\n", "date": "2023-02-07", "categories": []}
{"id": "2302.04989", "title": "Causal Inference out of Control: Estimating the Steerability of\n  Consumption", "abstract": "  Regulators and academics are increasingly interested in the causal effect\nthat algorithmic actions of a digital platform have on consumption. We\nintroduce a general causal inference problem we call the steerability of\nconsumption that abstracts many settings of interest. Focusing on observational\ndesigns and exploiting the structure of the problem, we exhibit a set of\nassumptions for causal identifiability that significantly weaken the often\nunrealistic overlap assumptions of standard designs. The key novelty of our\napproach is to explicitly model the dynamics of consumption over time, viewing\nthe platform as a controller acting on a dynamical system. From this dynamical\nsystems perspective, we are able to show that exogenous variation in\nconsumption and appropriately responsive algorithmic control actions are\nsufficient for identifying steerability of consumption. Our results illustrate\nthe fruitful interplay of control theory and causal inference, which we\nillustrate with examples from econometrics, macroeconomics, and machine\nlearning.\n", "date": "2023-02-10", "categories": []}
{"id": "2302.05404", "title": "Minimax Instrumental Variable Regression and $L_2$ Convergence\n  Guarantees without Identification or Closedness", "abstract": "  In this paper, we study nonparametric estimation of instrumental variable\n(IV) regressions. Recently, many flexible machine learning methods have been\ndeveloped for instrumental variable estimation. However, these methods have at\nleast one of the following limitations: (1) restricting the IV regression to be\nuniquely identified; (2) only obtaining estimation error rates in terms of\npseudometrics (\\emph{e.g.,} projected norm) rather than valid metrics\n(\\emph{e.g.,} $L_2$ norm); or (3) imposing the so-called closedness condition\nthat requires a certain conditional expectation operator to be sufficiently\nsmooth. In this paper, we present the first method and analysis that can avoid\nall three limitations, while still permitting general function approximation.\nSpecifically, we propose a new penalized minimax estimator that can converge to\na fixed IV solution even when there are multiple solutions, and we derive a\nstrong $L_2$ error rate for our estimator under lax conditions. Notably, this\nguarantee only needs a widely-used source condition and realizability\nassumptions, but not the so-called closedness condition. We argue that the\nsource condition and the closedness condition are inherently conflicting, so\nrelaxing the latter significantly improves upon the existing literature that\nrequires both conditions. Our estimator can achieve this improvement because it\nbuilds on a novel formulation of the IV estimation problem as a constrained\noptimization problem.\n", "date": "2023-02-10", "categories": []}
{"id": "2208.06987", "title": "The Causal Structure of Domain Invariant Supervised Representation\n  Learning", "abstract": "  Machine learning methods can be unreliable when deployed in domains that\ndiffer from the domains on which they were trained. There are a wide range of\nproposals for mitigating this problem by learning representations that are\n``invariant'' in some sense.However, these methods generally contradict each\nother, and none of them consistently improve performance on real-world domain\nshift benchmarks. There are two main questions that must be addressed to\nunderstand when, if ever, we should use each method. First, how does each ad\nhoc notion of ``invariance'' relate to the structure of real-world problems?\nAnd, second, when does learning invariant representations actually yield robust\nmodels? To address these issues, we introduce a broad formal notion of what it\nmeans for a real-world domain shift to admit invariant structure. Then, we\ncharacterize the causal structures that are compatible with this notion of\ninvariance.With this in hand, we find conditions under which method-specific\ninvariance notions correspond to real-world invariant structure, and we clarify\nthe relationship between invariant structure and robustness to domain shifts.\nFor both questions, we find that the true underlying causal structure of the\ndata plays a critical role.\n", "date": "2022-08-15", "categories": []}
{"id": "2210.00079", "title": "Causal Estimation for Text Data with (Apparent) Overlap Violations", "abstract": "  Consider the problem of estimating the causal effect of some attribute of a\ntext document; for example: what effect does writing a polite vs. rude email\nhave on response time? To estimate a causal effect from observational data, we\nneed to adjust for confounding aspects of the text that affect both the\ntreatment and outcome -- e.g., the topic or writing level of the text. These\nconfounding aspects are unknown a priori, so it seems natural to adjust for the\nentirety of the text (e.g., using a transformer). However, causal\nidentification and estimation procedures rely on the assumption of overlap: for\nall levels of the adjustment variables, there is randomness leftover so that\nevery unit could have (not) received treatment. Since the treatment here is\nitself an attribute of the text, it is perfectly determined, and overlap is\napparently violated. The purpose of this paper is to show how to handle causal\nidentification and obtain robust causal estimation in the presence of apparent\noverlap violations. In brief, the idea is to use supervised representation\nlearning to produce a data representation that preserves confounding\ninformation while eliminating information that is only predictive of the\ntreatment. This representation then suffices for adjustment and can satisfy\noverlap. Adapting results on non-parametric estimation, we find that this\nprocedure is robust to conditional outcome misestimation, yielding a low-bias\nestimator with valid uncertainty quantification under weak conditions.\nEmpirical results show strong improvements in bias and uncertainty\nquantification relative to the natural baseline.\n", "date": "2022-09-30", "categories": []}
{"id": "2301.13819", "title": "Causal-Discovery Performance of ChatGPT in the context of Neuropathic\n  Pain Diagnosis", "abstract": "  ChatGPT has demonstrated exceptional proficiency in natural language\nconversation, e.g., it can answer a wide range of questions while no previous\nlarge language models can. Thus, we would like to push its limit and explore\nits ability to answer causal discovery questions by using a medical benchmark\n(Tu et al. 2019) in causal discovery.\n", "date": "2023-01-24", "categories": []}
{"id": "2302.01987", "title": "Hierarchical Graph Neural Networks for Causal Discovery and Root Cause\n  Localization", "abstract": "  In this paper, we propose REASON, a novel framework that enables the\nautomatic discovery of both intra-level (i.e., within-network) and inter-level\n(i.e., across-network) causal relationships for root cause localization. REASON\nconsists of Topological Causal Discovery and Individual Causal Discovery. The\nTopological Causal Discovery component aims to model the fault propagation in\norder to trace back to the root causes. To achieve this, we propose novel\nhierarchical graph neural networks to construct interdependent causal networks\nby modeling both intra-level and inter-level non-linear causal relations. Based\non the learned interdependent causal networks, we then leverage random walks\nwith restarts to model the network propagation of a system fault. The\nIndividual Causal Discovery component focuses on capturing abrupt change\npatterns of a single system entity. This component examines the temporal\npatterns of each entity's metric data (i.e., time series), and estimates its\nlikelihood of being a root cause based on the Extreme Value theory. Combining\nthe topological and individual causal scores, the top K system entities are\nidentified as root causes. Extensive experiments on three real-world datasets\nwith case studies demonstrate the effectiveness and superiority of the proposed\nframework.\n", "date": "2023-02-03", "categories": []}
{"id": "2302.02160", "title": "Directed Acyclic Graphs With Tears", "abstract": "  Bayesian network is a frequently-used method for fault detection and\ndiagnosis in industrial processes. The basis of Bayesian network is structure\nlearning which learns a directed acyclic graph (DAG) from data. However, the\nsearch space will scale super-exponentially with the increase of process\nvariables, which makes the data-driven structure learning a challenging\nproblem. To this end, the DAGs with NOTEARs methods are being well studied not\nonly for their conversion of the discrete optimization into continuous\noptimization problem but also their compatibility with deep learning framework.\nNevertheless, there still remain challenges for NOTEAR-based methods: 1) the\ninfeasible solution results from the gradient descent-based optimization\nparadigm; 2) the truncation operation to promise the learned graph acyclic. In\nthis work, the reason for challenge 1) is analyzed theoretically, and a novel\nmethod named DAGs with Tears method is proposed based on mix-integer\nprogramming to alleviate challenge 2). In addition, prior knowledge is able to\nincorporate into the new proposed method, making structure learning more\npractical and useful in industrial processes. Finally, a numerical example and\nan industrial example are adopted as case studies to demonstrate the\nsuperiority of the developed method.\n", "date": "2023-02-04", "categories": []}
{"id": "2302.02570", "title": "Improved Policy Evaluation for Randomized Trials of Algorithmic Resource\n  Allocation", "abstract": "  We consider the task of evaluating policies of algorithmic resource\nallocation through randomized controlled trials (RCTs). Such policies are\ntasked with optimizing the utilization of limited intervention resources, with\nthe goal of maximizing the benefits derived. Evaluation of such allocation\npolicies through RCTs proves difficult, notwithstanding the scale of the trial,\nbecause the individuals' outcomes are inextricably interlinked through resource\nconstraints controlling the policy decisions. Our key contribution is to\npresent a new estimator leveraging our proposed novel concept, that involves\nretrospective reshuffling of participants across experimental arms at the end\nof an RCT. We identify conditions under which such reassignments are\npermissible and can be leveraged to construct counterfactual trials, whose\noutcomes can be accurately ascertained, for free. We prove theoretically that\nsuch an estimator is more accurate than common estimators based on sample means\n-- we show that it returns an unbiased estimate and simultaneously reduces\nvariance. We demonstrate the value of our approach through empirical\nexperiments on synthetic, semi-synthetic as well as real case study data and\nshow improved estimation accuracy across the board.\n", "date": "2023-02-06", "categories": []}
{"id": "2210.08326", "title": "Distributionally Robust Causal Inference with Observational Data", "abstract": "  We consider the estimation of average treatment effects in observational\nstudies and propose a new framework of robust causal inference with unobserved\nconfounders. Our approach is based on distributionally robust optimization and\nproceeds in two steps. We first specify the maximal degree to which the\ndistribution of unobserved potential outcomes may deviate from that of observed\noutcomes. We then derive sharp bounds on the average treatment effects under\nthis assumption. Our framework encompasses the popular marginal sensitivity\nmodel as a special case, and we demonstrate how the proposed methodology can\naddress a primary challenge of the marginal sensitivity model that it produces\nuninformative results when unobserved confounders substantially affect\ntreatment and outcome. Specifically, we develop an alternative sensitivity\nmodel, called the distributional sensitivity model, under the assumption that\nheterogeneity of treatment effect due to unobserved variables is relatively\nsmall. Unlike the marginal sensitivity model, the distributional sensitivity\nmodel allows for potential lack of overlap and often produces informative\nbounds even when unobserved variables substantially affect both treatment and\noutcome. Finally, we show how to extend the distributional sensitivity model to\ndifference-in-differences designs and settings with instrumental variables.\nThrough simulation and empirical studies, we demonstrate the applicability of\nthe proposed methodology.\n", "date": "2022-10-15", "categories": []}
{"id": "2302.01367", "title": "Augmented Learning of Heterogeneous Treatment Effects via Gradient\n  Boosting Trees", "abstract": "  Heterogeneous treatment effects (HTE) based on patients' genetic or clinical\nfactors are of significant interest to precision medicine. Simultaneously\nmodeling HTE and corresponding main effects for randomized clinical trials with\nhigh-dimensional predictive markers is challenging. Motivated by the modified\ncovariates approach, we propose a two-stage statistical learning procedure for\nestimating HTE with optimal efficiency augmentation, generalizing to arbitrary\ninteraction model and exploiting powerful extreme gradient boosting trees\n(XGBoost). Target estimands for HTE are defined in the scale of mean difference\nfor quantitative outcomes, or risk ratio for binary outcomes, which are the\nminimizers of specialized loss functions. The first stage is to estimate the\nmain-effect equivalency of the baseline markers on the outcome, which is then\nused as an augmentation term in the second stage estimation for HTE. The\nproposed two-stage procedure is robust to model mis-specification of main\neffects and improves efficiency for estimating HTE through nonparametric\nfunction estimation, e.g., XGBoost. A permutation test is proposed for global\nassessment of evidence for HTE. An analysis of a genetic study in Prostate\nCancer Prevention Trial led by the SWOG Cancer Research Network, is conducted\nto showcase the properties and the utilities of the two-stage method.\n", "date": "2023-02-02", "categories": []}
{"id": "2302.01563", "title": "Causal Inference Based Single-branch Ensemble Trees For Uplift Modeling", "abstract": "  In this manuscript (ms), we propose causal inference based single-branch\nensemble trees for uplift modeling, namely CIET. Different from standard\nclassification methods for predictive probability modeling, CIET aims to\nachieve the change in the predictive probability of outcome caused by an action\nor a treatment. According to our CIET, two partition criteria are specifically\ndesigned to maximize the difference in outcome distribution between the\ntreatment and control groups. Next, a novel single-branch tree is built by\ntaking a top-down node partition approach, and the remaining samples are\ncensored since they are not covered by the upper node partition logic.\nRepeating the tree-building process on the censored data, single-branch\nensemble trees with a set of inference rules are thus formed. Moreover, CIET is\nexperimentally demonstrated to outperform previous approaches for uplift\nmodeling in terms of both area under uplift curve (AUUC) and Qini coefficient\nsignificantly. At present, CIET has already been applied to online personal\nloans in a national financial holdings group in China. CIET will also be of use\nto analysts applying machine learning techniques to causal inference in broader\nbusiness domains such as web advertising, medicine and economics.\n", "date": "2023-02-03", "categories": []}
{"id": "2302.00848", "title": "Causal Effect Estimation: Recent Advances, Challenges, and Opportunities", "abstract": "  Causal inference has numerous real-world applications in many domains, such\nas health care, marketing, political science, and online advertising. Treatment\neffect estimation, a fundamental problem in causal inference, has been\nextensively studied in statistics for decades. However, traditional treatment\neffect estimation methods may not well handle large-scale and high-dimensional\nheterogeneous data. In recent years, an emerging research direction has\nattracted increasing attention in the broad artificial intelligence field,\nwhich combines the advantages of traditional treatment effect estimation\napproaches (e.g., propensity score, matching, and reweighing) and advanced\nmachine learning approaches (e.g., representation learning, adversarial\nlearning, and graph neural networks). Although the advanced machine learning\napproaches have shown extraordinary performance in treatment effect estimation,\nit also comes with a lot of new topics and new research questions. In view of\nthe latest research efforts in the causal inference field, we provide a\ncomprehensive discussion of challenges and opportunities for the three core\ncomponents of the treatment effect estimation task, i.e., treatment,\ncovariates, and outcome. In addition, we showcase the promising research\ndirections of this topic from multiple perspectives.\n", "date": "2023-02-02", "categories": []}
{"id": "2212.14580", "title": "Heterogeneous Synthetic Learner for Panel Data", "abstract": "  In the new era of personalization, learning the heterogeneous treatment\neffect (HTE) becomes an inevitable trend with numerous applications. Yet, most\nexisting HTE estimation methods focus on independently and identically\ndistributed observations and cannot handle the non-stationarity and temporal\ndependency in the common panel data setting. The treatment evaluators developed\nfor panel data, on the other hand, typically ignore the individualized\ninformation. To fill the gap, in this paper, we initialize the study of HTE\nestimation in panel data. Under different assumptions for HTE identifiability,\nwe propose the corresponding heterogeneous one-side and two-side synthetic\nlearner, namely H1SL and H2SL, by leveraging the state-of-the-art HTE estimator\nfor non-panel data and generalizing the synthetic control method that allows\nflexible data generating process. We establish the convergence rates of the\nproposed estimators. The superior performance of the proposed methods over\nexisting ones is demonstrated by extensive numerical studies.\n", "date": "2022-12-30", "categories": []}
{"id": "2301.06199", "title": "Doubly Robust Counterfactual Classification", "abstract": "  We study counterfactual classification as a new tool for decision-making\nunder hypothetical (contrary to fact) scenarios. We propose a doubly-robust\nnonparametric estimator for a general counterfactual classifier, where we can\nincorporate flexible constraints by casting the classification problem as a\nnonlinear mathematical program involving counterfactuals. We go on to analyze\nthe rates of convergence of the estimator and provide a closed-form expression\nfor its asymptotic distribution. Our analysis shows that the proposed estimator\nis robust against nuisance model misspecification, and can attain fast\n$\\sqrt{n}$ rates with tractable inference even when using nonparametric machine\nlearning approaches. We study the empirical performance of our methods by\nsimulation and apply them for recidivism risk prediction.\n", "date": "2023-01-15", "categories": []}
{"id": "2301.11197", "title": "Causal Graph Discovery from Self and Mutually Exciting Time Series", "abstract": "  We present a generalized linear structural causal model, coupled with a novel\ndata-adaptive linear regularization, to recover causal directed acyclic graphs\n(DAGs) from time series. By leveraging a recently developed stochastic monotone\nVariational Inequality (VI) formulation, we cast the causal discovery problem\nas a general convex optimization. Furthermore, we develop a non-asymptotic\nrecovery guarantee and quantifiable uncertainty by solving a linear program to\nestablish confidence intervals for a wide range of non-linear monotone link\nfunctions. We validate our theoretical results and show the competitive\nperformance of our method via extensive numerical experiments. Most\nimportantly, we demonstrate the effectiveness of our approach in recovering\nhighly interpretable causal DAGs over Sepsis Associated Derangements (SADs)\nwhile achieving comparable prediction performance to powerful ``black-box''\nmodels such as XGBoost. Thus, the future adoption of our proposed method to\nconduct continuous surveillance of high-risk patients by clinicians is much\nmore likely.\n", "date": "2023-01-26", "categories": []}
{"id": "2301.11921", "title": "Using uncertainty-aware machine learning models to study aerosol-cloud\n  interactions", "abstract": "  Aerosol-cloud interactions (ACI) include various effects that result from\naerosols entering a cloud, and affecting cloud properties. In general, an\nincrease in aerosol concentration results in smaller droplet sizes which leads\nto larger, brighter, longer-lasting clouds that reflect more sunlight and cool\nthe Earth. The strength of the effect is however heterogeneous, meaning it\ndepends on the surrounding environment, making ACI one of the most uncertain\neffects in our current climate models. In our work, we use causal machine\nlearning to estimate ACI from satellite observations by reframing the problem\nas a treatment (aerosol) and outcome (change in droplet radius). We predict the\ncausal effect of aerosol on clouds with uncertainty bounds depending on the\nunknown factors that may be influencing the impact of aerosol. Of the three\nclimate models evaluated, we find that only one plausibly recreates the trend,\nlending more credence to its estimate cooling due to ACI.\n", "date": "2022-11-30", "categories": []}
{"id": "2301.12022", "title": "Epsilon-Identifiability of Causal Quantities", "abstract": "  Identifying the effects of causes and causes of effects is vital in virtually\nevery scientific field. Often, however, the needed probabilities may not be\nfully identifiable from the data sources available. This paper shows how\npartial identifiability is still possible for several probabilities of\ncausation. We term this epsilon-identifiability and demonstrate its usefulness\nin cases where the behavior of certain subpopulations can be restricted to\nwithin some narrow bounds. In particular, we show how unidentifiable causal\neffects and counterfactual probabilities can be narrowly bounded when such\nallowances are made. Often those allowances are easily measured and reasonably\nassumed. Finally, epsilon-identifiability is applied to the unit selection\nproblem.\n", "date": "2023-01-27", "categories": []}
{"id": "2301.12278", "title": "Pragmatic Fairness: Developing Policies with Outcome Disparity Control", "abstract": "  We introduce a causal framework for designing optimal policies that satisfy\nfairness constraints. We take a pragmatic approach asking what we can do with\nan action space available to us and only with access to historical data. We\npropose two different fairness constraints: a moderation breaking constraint\nwhich aims at blocking moderation paths from the action and sensitive attribute\nto the outcome, and by that at reducing disparity in outcome levels as much as\nthe provided action space permits; and an equal benefit constraint which aims\nat distributing gain from the new and maximized policy equally across sensitive\nattribute levels, and thus at keeping pre-existing preferential treatment in\nplace or avoiding the introduction of new disparity. We introduce practical\nmethods for implementing the constraints and illustrate their uses on\nexperiments with semi-synthetic models.\n", "date": "2023-01-28", "categories": []}
{"id": "2301.12985", "title": "Integrating Earth Observation Data into Causal Inference: Challenges and\n  Opportunities", "abstract": "  Observational studies require adjustment for confounding factors that are\ncorrelated with both the treatment and outcome. In the setting where the\nobserved variables are tabular quantities such as average income in a\nneighborhood, tools have been developed for addressing such confounding.\nHowever, in many parts of the developing world, features about local\ncommunities may be scarce. In this context, satellite imagery can play an\nimportant role, serving as a proxy for the confounding variables otherwise\nunobserved. In this paper, we study confounder adjustment in this non-tabular\nsetting, where patterns or objects found in satellite images contribute to the\nconfounder bias. Using the evaluation of anti-poverty aid programs in Africa as\nour running example, we formalize the challenge of performing causal adjustment\nwith such unstructured data -- what conditions are sufficient to identify\ncausal effects, how to perform estimation, and how to quantify the ways in\nwhich certain aspects of the unstructured image object are most predictive of\nthe treatment decision. Via simulation, we also explore the sensitivity of\nsatellite image-based observational inference to image resolution and to\nmisspecification of the image-associated confounder. Finally, we apply these\ntools in estimating the effect of anti-poverty interventions in African\ncommunities from satellite imagery.\n", "date": "2023-01-30", "categories": []}
{"id": "2210.10530", "title": "Adversarial De-confounding in Individualised Treatment Effects\n  Estimation", "abstract": "  Observational studies have recently received significant attention from the\nmachine learning community due to the increasingly available non-experimental\nobservational data and the limitations of the experimental studies, such as\nconsiderable cost, impracticality, small and less representative sample sizes,\netc. In observational studies, de-confounding is a fundamental problem of\nindividualised treatment effects (ITE) estimation. This paper proposes\ndisentangled representations with adversarial training to selectively balance\nthe confounders in the binary treatment setting for the ITE estimation. The\nadversarial training of treatment policy selectively encourages\ntreatment-agnostic balanced representations for the confounders and helps to\nestimate the ITE in the observational studies via counterfactual inference.\nEmpirical results on synthetic and real-world datasets, with varying degrees of\nconfounding, prove that our proposed approach improves the state-of-the-art\nmethods in achieving lower error in the ITE estimation.\n", "date": "2022-10-19", "categories": []}
{"id": "2208.08544", "title": "Estimating individual treatment effects under unobserved confounding\n  using binary instruments", "abstract": "  Estimating conditional average treatment effects (CATEs) from observational\ndata is relevant in many fields such as personalized medicine. However, in\npractice, the treatment assignment is usually confounded by unobserved\nvariables and thus introduces bias. A remedy to remove the bias is the use of\ninstrumental variables (IVs). Such settings are widespread in medicine (e.g.,\ntrials where the treatment assignment is used as binary IV). In this paper, we\npropose a novel, multiply robust machine learning framework, called MRIV, for\nestimating CATEs using binary IVs and thus yield an unbiased CATE estimator.\nDifferent from previous work for binary IVs, our framework estimates the CATE\ndirectly via a pseudo outcome regression. (1)~We provide a theoretical analysis\nwhere we show that our framework yields multiple robust convergence rates: our\nCATE estimator achieves fast convergence even if several nuisance estimators\nconverge slowly. (2)~We further show that our framework asymptotically\noutperforms state-of-the-art plug-in IV methods for CATE estimation, in the\nsense that it achieves a faster rate of convergence if the CATE is smoother\nthan the individual outcome surfaces. (3)~We build upon our theoretical results\nand propose a tailored deep neural network architecture called MRIV-Net for\nCATE estimation using binary IVs. Across various computational experiments, we\ndemonstrate empirically that our MRIV-Net achieves state-of-the-art\nperformance. To the best of our knowledge, our MRIV is the first multiply\nrobust machine learning framework tailored to estimating CATEs in the binary IV\nsetting.\n", "date": "2022-08-17", "categories": []}
{"id": "2301.08788", "title": "Causal Inference under Data Restrictions", "abstract": "  This dissertation focuses on modern causal inference under uncertainty and\ndata restrictions, with applications to neoadjuvant clinical trials,\ndistributed data networks, and robust individualized decision making.\n  In the first project, we propose a method under the principal stratification\nframework to identify and estimate the average treatment effects on a binary\noutcome, conditional on the counterfactual status of a post-treatment\nintermediate response. Under mild assumptions, the treatment effect of interest\ncan be identified. We extend the approach to address censored outcome data. The\nproposed method is applied to a neoadjuvant clinical trial and its performance\nis evaluated via simulation studies.\n  In the second project, we propose a tree-based model averaging approach to\nimprove the estimation accuracy of conditional average treatment effects at a\ntarget site by leveraging models derived from other potentially heterogeneous\nsites, without them sharing subject-level data. The performance of this\napproach is demonstrated by a study of the causal effects of oxygen therapy on\nhospital survival rates and backed up by comprehensive simulations.\n  In the third project, we propose a robust individualized decision learning\nframework with sensitive variables to improve the worst-case outcomes of\nindividuals caused by sensitive variables that are unavailable at the time of\ndecision. Unlike most existing work that uses mean-optimal objectives, we\npropose a robust learning framework by finding a newly defined quantile- or\ninfimum-optimal decision rule. From a causal perspective, we also generalize\nthe classic notion of (average) fairness to conditional fairness for individual\nsubjects. The reliable performance of the proposed method is demonstrated\nthrough synthetic experiments and three real-data applications.\n", "date": "2023-01-20", "categories": []}
{"id": "2301.09031", "title": "Counterfactual (Non-)identifiability of Learned Structural Causal Models", "abstract": "  Recent advances in probabilistic generative modeling have motivated learning\nStructural Causal Models (SCM) from observational datasets using deep\nconditional generative models, also known as Deep Structural Causal Models\n(DSCM). If successful, DSCMs can be utilized for causal estimation tasks, e.g.,\nfor answering counterfactual queries. In this work, we warn practitioners about\nnon-identifiability of counterfactual inference from observational data, even\nin the absence of unobserved confounding and assuming known causal structure.\nWe prove counterfactual identifiability of monotonic generation mechanisms with\nsingle dimensional exogenous variables. For general generation mechanisms with\nmulti-dimensional exogenous variables, we provide an impossibility result for\ncounterfactual identifiability, motivating the need for parametric assumptions.\nAs a practical approach, we propose a method for estimating worst-case errors\nof learned DSCMs' counterfactual predictions. The size of this error can be an\nessential metric for deciding whether or not DSCMs are a viable approach for\ncounterfactual inference in a specific problem setting. In evaluation, our\nmethod confirms negligible counterfactual errors for an identifiable SCM from\nprior work, and also provides informative error bounds on counterfactual errors\nfor a non-identifiable synthetic SCM.\n", "date": "2023-01-22", "categories": []}
{"id": "2301.07843", "title": "Spatio-temporal neural structural causal models for bike flow prediction", "abstract": "  As a representative of public transportation, the fundamental issue of\nmanaging bike-sharing systems is bike flow prediction. Recent methods\noveremphasize the spatio-temporal correlations in the data, ignoring the\neffects of contextual conditions on the transportation system and the\ninter-regional timevarying causality. In addition, due to the disturbance of\nincomplete observations in the data, random contextual conditions lead to\nspurious correlations between data and features, making the prediction of the\nmodel ineffective in special scenarios. To overcome this issue, we propose a\nSpatio-temporal Neural Structure Causal Model(STNSCM) from the perspective of\ncausality. First, we build a causal graph to describe the traffic prediction,\nand further analyze the causal relationship between the input data, contextual\nconditions, spatiotemporal states, and prediction results. Second, we propose\nto apply the frontdoor criterion to eliminate confounding biases in the feature\nextraction process. Finally, we propose a counterfactual representation\nreasoning module to extrapolate the spatio-temporal state under the factual\nscenario to future counterfactual scenarios to improve the prediction\nperformance. Experiments on real-world datasets demonstrate the superior\nperformance of our model, especially its resistance to fluctuations caused by\nthe external environment. The source code and data will be released.\n", "date": "2023-01-19", "categories": []}
{"id": "2301.07656", "title": "Non-parametric identifiability and sensitivity analysis of synthetic\n  control models", "abstract": "  Quantifying cause and effect relationships is an important problem in many\ndomains. The gold standard solution is to conduct a randomised controlled\ntrial. However, in many situations such trials cannot be performed. In the\nabsence of such trials, many methods have been devised to quantify the causal\nimpact of an intervention from observational data given certain assumptions.\nOne widely used method are synthetic control models. While identifiability of\nthe causal estimand in such models has been obtained from a range of\nassumptions, it is widely and implicitly assumed that the underlying\nassumptions are satisfied for all time periods both pre- and post-intervention.\nThis is a strong assumption, as synthetic control models can only be learned in\npre-intervention period. In this paper we address this challenge, and prove\nidentifiability can be obtained without the need for this assumption, by\nshowing it follows from the principle of invariant causal mechanisms. Moreover,\nfor the first time, we formulate and study synthetic control models in Pearl's\nstructural causal model framework. Importantly, we provide a general framework\nfor sensitivity analysis of synthetic control causal inference to violations of\nthe assumptions underlying non-parametric identifiability. We end by providing\nan empirical demonstration of our sensitivity analysis framework on simulated\nand real data in the widely-used linear synthetic control framework.\n", "date": "2023-01-18", "categories": []}
{"id": "2209.08037", "title": "DAGMA: Learning DAGs via M-matrices and a Log-Determinant Acyclicity\n  Characterization", "abstract": "  The combinatorial problem of learning directed acyclic graphs (DAGs) from\ndata was recently framed as a purely continuous optimization problem by\nleveraging a differentiable acyclicity characterization of DAGs based on the\ntrace of a matrix exponential function. Existing acyclicity characterizations\nare based on the idea that powers of an adjacency matrix contain information\nabout walks and cycles. In this work, we propose a new acyclicity\ncharacterization based on the log-determinant (log-det) function, which\nleverages the nilpotency property of DAGs. To deal with the inherent\nasymmetries of a DAG, we relate the domain of our log-det characterization to\nthe set of $\\textit{M-matrices}$, which is a key difference to the classical\nlog-det function defined over the cone of positive definite matrices. Similar\nto acyclicity functions previously proposed, our characterization is also exact\nand differentiable. However, when compared to existing characterizations, our\nlog-det function: (1) Is better at detecting large cycles; (2) Has\nbetter-behaved gradients; and (3) Its runtime is in practice about an order of\nmagnitude faster. From the optimization side, we drop the typically used\naugmented Lagrangian scheme and propose DAGMA ($\\textit{DAGs via M-matrices for\nAcyclicity}$), a method that resembles the central path for barrier methods.\nEach point in the central path of DAGMA is a solution to an unconstrained\nproblem regularized by our log-det function, then we show that at the limit of\nthe central path the solution is guaranteed to be a DAG. Finally, we provide\nextensive experiments for $\\textit{linear}$ and $\\textit{nonlinear}$ SEMs and\nshow that our approach can reach large speed-ups and smaller structural Hamming\ndistances against state-of-the-art methods. Code implementing the proposed\nmethod is open-source and publicly available at\nhttps://github.com/kevinsbello/dagma.\n", "date": "2022-09-16", "categories": []}
{"id": "2210.05811", "title": "Deep Counterfactual Estimation with Categorical Background Variables", "abstract": "  Referred to as the third rung of the causal inference ladder, counterfactual\nqueries typically ask the \"What if ?\" question retrospectively. The standard\napproach to estimate counterfactuals resides in using a structural equation\nmodel that accurately reflects the underlying data generating process. However,\nsuch models are seldom available in practice and one usually wishes to infer\nthem from observational data alone. Unfortunately, the correct structural\nequation model is in general not identifiable from the observed factual\ndistribution. Nevertheless, in this work, we show that under the assumption\nthat the main latent contributors to the treatment responses are categorical,\nthe counterfactuals can be still reliably predicted. Building upon this\nassumption, we introduce CounterFactual Query Prediction (CFQP), a novel method\nto infer counterfactuals from continuous observations when the background\nvariables are categorical. We show that our method significantly outperforms\npreviously available deep-learning-based counterfactual methods, both\ntheoretically and empirically on time series and image data. Our code is\navailable at https://github.com/edebrouwer/cfqp.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.13005", "title": "Towards Out-of-Distribution Sequential Event Prediction: A Causal\n  Treatment", "abstract": "  The goal of sequential event prediction is to estimate the next event based\non a sequence of historical events, with applications to sequential\nrecommendation, user behavior analysis and clinical treatment. In practice, the\nnext-event prediction models are trained with sequential data collected at one\ntime and need to generalize to newly arrived sequences in remote future, which\nrequires models to handle temporal distribution shift from training to testing.\nIn this paper, we first take a data-generating perspective to reveal a negative\nresult that existing approaches with maximum likelihood estimation would fail\nfor distribution shift due to the latent context confounder, i.e., the common\ncause for the historical events and the next event. Then we devise a new\nlearning objective based on backdoor adjustment and further harness variational\ninference to make it tractable for sequence learning problems. On top of that,\nwe propose a framework with hierarchical branching structures for learning\ncontext-specific representations. Comprehensive experiments on diverse tasks\n(e.g., sequential recommendation) demonstrate the effectiveness, applicability\nand scalability of our method with various off-the-shelf models as backbones.\n", "date": "2022-10-24", "categories": []}
{"id": "2301.06300", "title": "Sensor data-driven analysis for identification of causal relationships\n  between exposure to air pollution and respiratory rate in asthmatics", "abstract": "  According to the Lancet report on the global burden of disease published in\nOctober 2020, air pollution is among the five highest risk factors for global\nhealth, reducing life expectancy on average by 20 months. This paper describes\na data-driven method for establishing causal relationships within and between\ntwo multivariate time series data streams derived from wearable sensors:\npersonal exposure to airborne particulate matter of aerodynamic sizes less than\n2.5um (PM2.5) gathered from the Airspeck monitor worn on the person and\ncontinuous respiratory rate (breaths per minute) measured by the Respeck\nmonitor worn as a plaster on the chest. Results are presented for a cohort of\n113 asthmatic adolescents using the PCMCI+ algorithm to learn the short-term\ncausal relationships between lags of \\pm exposure and respiratory rate. We\nconsider causal effects up to a maximum delay of 8 hours, using data at both a\n1 minute and 15 minute resolution in different experiments. For the first time\na personalised exposure-response relationship between PM2.5 exposure and\nrespiratory rate has been demonstrated to exist for short-term effects in\nasthmatic adolescents during their everyday lives. Our results lead to\nrecommendations for work on specific open problems in causal discovery, to\nincrease the feasibility of this approach for similar epidemiology studies in\nthe future.\n", "date": "2023-01-16", "categories": []}
{"id": "2301.06845", "title": "Causal Models with Constraints", "abstract": "  Causal models have proven extremely useful in offering formal representations\nof causal relationships between a set of variables. Yet in many situations,\nthere are non-causal relationships among variables. For example, we may want\nvariables $LDL$, $HDL$, and $TOT$ that represent the level of low-density\nlipoprotein cholesterol, the level of lipoprotein high-density lipoprotein\ncholesterol, and total cholesterol level, with the relation $LDL+HDL=TOT$. This\ncannot be done in standard causal models, because we can intervene\nsimultaneously on all three variables. The goal of this paper is to extend\nstandard causal models to allow for constraints on settings of variables.\nAlthough the extension is relatively straightforward, to make it useful we have\nto define a new intervention operation that $disconnects$ a variable from a\ncausal equation. We give examples showing the usefulness of this extension, and\nprovide a sound and complete axiomatization for causal models with constraints.\n", "date": "2023-01-17", "categories": []}
{"id": "2212.12575", "title": "Continual Causal Abstractions", "abstract": "  This short paper discusses continually updated causal abstractions as a\npotential direction of future research. The key idea is to revise the existing\nlevel of causal abstraction to a different level of detail that is both\nconsistent with the history of observed data and more effective in solving a\ngiven task.\n", "date": "2022-12-23", "categories": []}
{"id": "2301.02499", "title": "Evaluating counterfactual explanations using Pearl's counterfactual\n  method", "abstract": "  Counterfactual explanations (CEs) are methods for generating an alternative\nscenario that produces a different desirable outcome. For example, if a student\nis predicted to fail a course, then counterfactual explanations can provide the\nstudent with alternate ways so that they would be predicted to pass. The\napplications are many. However, CEs are currently generated from machine\nlearning models that do not necessarily take into account the true causal\nstructure in the data. By doing this, bias can be introduced into the CE\nquantities. I propose in this study to test the CEs using Judea Pearl's method\nof computing counterfactuals which has thus far, surprisingly, not been seen in\nthe counterfactual explanation (CE) literature. I furthermore evaluate these\nCEs on three different causal structures to show how the true underlying causal\nstructure affects the CEs that are generated. This study presented a method of\nevaluating CEs using Pearl's method and it showed, (although using a limited\nsample size), that thirty percent of the CEs conflicted with those computed by\nPearl's method. This shows that we cannot simply trust CEs and it is vital for\nus to know the true causal structure before we blindly compute counterfactuals\nusing the original machine learning model.\n", "date": "2023-01-06", "categories": []}
{"id": "2301.01817", "title": "Evaluation of Induced Expert Knowledge in Causal Structure Learning by\n  NOTEARS", "abstract": "  Causal modeling provides us with powerful counterfactual reasoning and\ninterventional mechanism to generate predictions and reason under various\nwhat-if scenarios. However, causal discovery using observation data remains a\nnontrivial task due to unobserved confounding factors, finite sampling, and\nchanges in the data distribution. These can lead to spurious cause-effect\nrelationships. To mitigate these challenges in practice, researchers augment\ncausal learning with known causal relations. The goal of the paper is to study\nthe impact of expert knowledge on causal relations in the form of additional\nconstraints used in the formulation of the nonparametric NOTEARS. We provide a\ncomprehensive set of comparative analyses of biasing the model using different\ntypes of knowledge. We found that (i) knowledge that corrects the mistakes of\nthe NOTEARS model can lead to statistically significant improvements, (ii)\nconstraints on active edges have a larger positive impact on causal discovery\nthan inactive edges, and surprisingly, (iii) the induced knowledge does not\ncorrect on average more incorrect active and/or inactive edges than expected.\nWe also demonstrate the behavior of the model and the effectiveness of domain\nknowledge on a real-world dataset.\n", "date": "2023-01-04", "categories": []}
{"id": "2301.01849", "title": "NODAGS-Flow: Nonlinear Cyclic Causal Structure Learning", "abstract": "  Learning causal relationships between variables is a well-studied problem in\nstatistics, with many important applications in science. However, modeling\nreal-world systems remain challenging, as most existing algorithms assume that\nthe underlying causal graph is acyclic. While this is a convenient framework\nfor developing theoretical developments about causal reasoning and inference,\nthe underlying modeling assumption is likely to be violated in real systems,\nbecause feedback loops are common (e.g., in biological systems). Although a few\nmethods search for cyclic causal models, they usually rely on some form of\nlinearity, which is also limiting, or lack a clear underlying probabilistic\nmodel. In this work, we propose a novel framework for learning nonlinear cyclic\ncausal graphical models from interventional data, called NODAGS-Flow. We\nperform inference via direct likelihood optimization, employing techniques from\nresidual normalizing flows for likelihood estimation. Through synthetic\nexperiments and an application to single-cell high-content perturbation\nscreening data, we show significant performance improvements with our approach\ncompared to state-of-the-art methods with respect to structure recovery and\npredictive performance.\n", "date": "2023-01-04", "categories": []}
{"id": "2301.01109", "title": "On the causality-preservation capabilities of generative modelling", "abstract": "  Modeling lies at the core of both the financial and the insurance industry\nfor a wide variety of tasks. The rise and development of machine learning and\ndeep learning models have created many opportunities to improve our modeling\ntoolbox. Breakthroughs in these fields often come with the requirement of large\namounts of data. Such large datasets are often not publicly available in\nfinance and insurance, mainly due to privacy and ethics concerns. This lack of\ndata is currently one of the main hurdles in developing better models. One\npossible option to alleviating this issue is generative modeling. Generative\nmodels are capable of simulating fake but realistic-looking data, also referred\nto as synthetic data, that can be shared more freely. Generative Adversarial\nNetworks (GANs) is such a model that increases our capacity to fit very\nhigh-dimensional distributions of data. While research on GANs is an active\ntopic in fields like computer vision, they have found limited adoption within\nthe human sciences, like economics and insurance. Reason for this is that in\nthese fields, most questions are inherently about identification of causal\neffects, while to this day neural networks, which are at the center of the GAN\nframework, focus mostly on high-dimensional correlations. In this paper we\nstudy the causal preservation capabilities of GANs and whether the produced\nsynthetic data can reliably be used to answer causal questions. This is done by\nperforming causal analyses on the synthetic data, produced by a GAN, with\nincreasingly more lenient assumptions. We consider the cross-sectional case,\nthe time series case and the case with a complete structural model. It is shown\nthat in the simple cross-sectional scenario where correlation equals causation\nthe GAN preserves causality, but that challenges arise for more advanced\nanalyses.\n", "date": "2023-01-03", "categories": []}
{"id": "2301.01110", "title": "Causal Discovery for Gene Regulatory Network Prediction", "abstract": "  Biological systems and processes are networks of complex nonlinear regulatory\ninteractions between nucleic acids, proteins, and metabolites. A natural way in\nwhich to represent these interaction networks is through the use of a graph. In\nthis formulation, each node represents a nucleic acid, protein, or metabolite\nand edges represent intermolecular interactions (inhibition, regulation,\npromotion, coexpression, etc.). In this work, a novel algorithm for the\ndiscovery of latent graph structures given experimental data is presented.\n", "date": "2023-01-03", "categories": []}
{"id": "2301.00346", "title": "An Adaptive Kernel Approach to Federated Learning of Heterogeneous\n  Causal Effects", "abstract": "  We propose a new causal inference framework to learn causal effects from\nmultiple, decentralized data sources in a federated setting. We introduce an\nadaptive transfer algorithm that learns the similarities among the data sources\nby utilizing Random Fourier Features to disentangle the loss function into\nmultiple components, each of which is associated with a data source. The data\nsources may have different distributions; the causal effects are independently\nand systematically incorporated. The proposed method estimates the similarities\namong the sources through transfer coefficients, and hence requiring no prior\ninformation about the similarity measures. The heterogeneous causal effects can\nbe estimated with no sharing of the raw training data among the sources, thus\nminimizing the risk of privacy leak. We also provide minimax lower bounds to\nassess the quality of the parameters learned from the disparate sources. The\nproposed method is empirically shown to outperform the baselines on\ndecentralized data sources with dissimilar distributions.\n", "date": "2023-01-01", "categories": []}
{"id": "2301.00629", "title": "Learning and interpreting asymmetry-labeled DAGs: a case study on\n  COVID-19 fear", "abstract": "  Bayesian networks are widely used to learn and reason about the dependence\nstructure of discrete variables. However, they are only capable of formally\nencoding symmetric conditional independence, which in practice is often too\nstrict to hold. Asymmetry-labeled DAGs have been recently proposed to both\nextend the class of Bayesian networks by relaxing the symmetric assumption of\nindependence and denote the type of dependence existing between the variables\nof interest. Here, we introduce novel structural learning algorithms for this\nclass of models which, whilst being efficient, allow for a straightforward\ninterpretation of the underlying dependence structure. A comprehensive\ncomputational study highlights the efficiency of the algorithms. A real-world\ndata application using data from the Fear of COVID-19 Scale collected in Italy\nshowcases their use in practice.\n", "date": "2023-01-02", "categories": []}
{"id": "2301.00793", "title": "Causal Inference (C-inf) -- closed form worst case typical phase\n  transitions", "abstract": "  In this paper we establish a mathematically rigorous connection between\nCausal inference (C-inf) and the low-rank recovery (LRR). Using Random Duality\nTheory (RDT) concepts developed in [46,48,50] and novel mathematical strategies\nrelated to free probability theory, we obtain the exact explicit typical (and\nachievable) worst case phase transitions (PT). These PT precisely separate\nscenarios where causal inference via LRR is possible from those where it is\nnot. We supplement our mathematical analysis with numerical experiments that\nconfirm the theoretical predictions of PT phenomena, and further show that the\ntwo closely match for fairly small sample sizes. We obtain simple closed form\nrepresentations for the resulting PTs, which highlight direct relations between\nthe low rankness of the target C-inf matrix and the time of the treatment.\nHence, our results can be used to determine the range of C-inf's typical\napplicability.\n", "date": "2023-01-02", "categories": []}
{"id": "2301.00801", "title": "Causal Inference (C-inf) -- asymmetric scenario of typical phase\n  transitions", "abstract": "  In this paper, we revisit and further explore a mathematically rigorous\nconnection between Causal inference (C-inf) and the Low-rank recovery (LRR)\nestablished in [10]. Leveraging the Random duality - Free probability theory\n(RDT-FPT) connection, we obtain the exact explicit typical C-inf asymmetric\nphase transitions (PT). We uncover a doubling low-rankness phenomenon, which\nmeans that exactly two times larger low rankness is allowed in asymmetric\nscenarios compared to the symmetric worst case ones considered in [10].\nConsequently, the final PT mathematical expressions are as elegant as those\nobtained in [10], and highlight direct relations between the targeted C-inf\nmatrix low rankness and the time of treatment. Our results have strong\nimplications for applications, where C-inf matrices are not necessarily\nsymmetric.\n", "date": "2023-01-02", "categories": []}
{"id": "2212.14133", "title": "Investigating Sindy As a Tool For Causal Discovery In Time Series\n  Signals", "abstract": "  The SINDy algorithm has been successfully used to identify the governing\nequations of dynamical systems from time series data. In this paper, we argue\nthat this makes SINDy a potentially useful tool for causal discovery and that\nexisting tools for causal discovery can be used to dramatically improve the\nperformance of SINDy as tool for robust sparse modeling and system\nidentification. We then demonstrate empirically that augmenting the SINDy\nalgorithm with tools from causal discovery can provides engineers with a tool\nfor learning causally robust governing equations.\n", "date": "2022-12-29", "categories": []}
{"id": "2210.16525", "title": "Spectral Representation Learning for Conditional Moment Models", "abstract": "  Many problems in causal inference and economics can be formulated in the\nframework of conditional moment models, which characterize the target function\nthrough a collection of conditional moment restrictions. For nonparametric\nconditional moment models, efficient estimation often relies on preimposed\nconditions on various measures of ill-posedness of the hypothesis space, which\nare hard to validate when flexible models are used. In this work, we address\nthis issue by proposing a procedure that automatically learns representations\nwith controlled measures of ill-posedness. Our method approximates a linear\nrepresentation defined by the spectral decomposition of a conditional\nexpectation operator, which can be used for kernelized estimators and is known\nto facilitate minimax optimal estimation in certain settings. We show this\nrepresentation can be efficiently estimated from data, and establish L2\nconsistency for the resulting estimator. We evaluate the proposed method on\nproximal causal inference tasks, exhibiting promising performance on\nhigh-dimensional, semi-synthetic data.\n", "date": "2022-10-29", "categories": []}
{"id": "2210.04389", "title": "DeepMed: Semiparametric Causal Mediation Analysis with Debiased Deep\n  Learning", "abstract": "  Causal mediation analysis can unpack the black box of causality and is\ntherefore a powerful tool for disentangling causal pathways in biomedical and\nsocial sciences, and also for evaluating machine learning fairness. To reduce\nbias for estimating Natural Direct and Indirect Effects in mediation analysis,\nwe propose a new method called DeepMed that uses deep neural networks (DNNs) to\ncross-fit the infinite-dimensional nuisance functions in the efficient\ninfluence functions. We obtain novel theoretical results that our DeepMed\nmethod (1) can achieve semiparametric efficiency bound without imposing\nsparsity constraints on the DNN architecture and (2) can adapt to certain low\ndimensional structures of the nuisance functions, significantly advancing the\nexisting literature on DNN-based semiparametric causal inference. Extensive\nsynthetic experiments are conducted to support our findings and also expose the\ngap between theory and practice. As a proof of concept, we apply DeepMed to\nanalyze two real datasets on machine learning fairness and reach conclusions\nconsistent with previous findings.\n", "date": "2022-10-10", "categories": []}
{"id": "2212.12560", "title": "On How AI Needs to Change to Advance the Science of Drug Discovery", "abstract": "  Research around AI for Science has seen significant success since the rise of\ndeep learning models over the past decade, even with longstanding challenges\nsuch as protein structure prediction. However, this fast development inevitably\nmade their flaws apparent -- especially in domains of reasoning where\nunderstanding the cause-effect relationship is important. One such domain is\ndrug discovery, in which such understanding is required to make sense of data\notherwise plagued by spurious correlations. Said spuriousness only becomes\nworse with the ongoing trend of ever-increasing amounts of data in the life\nsciences and thereby restricts researchers in their ability to understand\ndisease biology and create better therapeutics. Therefore, to advance the\nscience of drug discovery with AI it is becoming necessary to formulate the key\nproblems in the language of causality, which allows the explication of\nmodelling assumptions needed for identifying true cause-effect relationships.\n  In this attention paper, we present causal drug discovery as the craft of\ncreating models that ground the process of drug discovery in causal reasoning.\n", "date": "2022-12-23", "categories": []}
{"id": "2212.12570", "title": "Pearl Causal Hierarchy on Image Data: Intricacies & Challenges", "abstract": "  Many researchers have voiced their support towards Pearl's counterfactual\ntheory of causation as a stepping stone for AI/ML research's ultimate goal of\nintelligent systems. As in any other growing subfield, patience seems to be a\nvirtue since significant progress on integrating notions from both fields takes\ntime, yet, major challenges such as the lack of ground truth benchmarks or a\nunified perspective on classical problems such as computer vision seem to\nhinder the momentum of the research movement. This present work exemplifies how\nthe Pearl Causal Hierarchy (PCH) can be understood on image data by providing\ninsights on several intricacies but also challenges that naturally arise when\napplying key concepts from Pearlian causality to the study of image data.\n", "date": "2022-12-23", "categories": []}
{"id": "2206.10591", "title": "Can Foundation Models Talk Causality?", "abstract": "  Foundation models are subject to an ongoing heated debate, leaving open the\nquestion of progress towards AGI and dividing the community into two camps: the\nones who see the arguably impressive results as evidence to the scaling\nhypothesis, and the others who are worried about the lack of interpretability\nand reasoning capabilities. By investigating to which extent causal\nrepresentations might be captured by these large scale language models, we make\na humble efforts towards resolving the ongoing philosophical conflicts.\n", "date": "2022-06-14", "categories": []}
{"id": "2208.14571", "title": "Truncated Matrix Power Iteration for Differentiable DAG Learning", "abstract": "  Recovering underlying Directed Acyclic Graph (DAG) structures from\nobservational data is highly challenging due to the combinatorial nature of the\nDAG-constrained optimization problem. Recently, DAG learning has been cast as a\ncontinuous optimization problem by characterizing the DAG constraint as a\nsmooth equality one, generally based on polynomials over adjacency matrices.\nExisting methods place very small coefficients on high-order polynomial terms\nfor stabilization, since they argue that large coefficients on the higher-order\nterms are harmful due to numeric exploding. On the contrary, we discover that\nlarge coefficients on higher-order terms are beneficial for DAG learning, when\nthe spectral radiuses of the adjacency matrices are small, and that larger\ncoefficients for higher-order terms can approximate the DAG constraints much\nbetter than the small counterparts. Based on this, we propose a novel DAG\nlearning method with efficient truncated matrix power iteration to approximate\ngeometric series based DAG constraints. Empirically, our DAG learning method\noutperforms the previous state-of-the-arts in various settings, often by a\nfactor of $3$ or more in terms of structural Hamming distance.\n", "date": "2022-08-30", "categories": []}
{"id": "2212.11322", "title": "Debiased machine learning for estimating the causal effect of urban\n  traffic on pedestrian crossing behaviour", "abstract": "  Before the transition of AVs to urban roads and subsequently unprecedented\nchanges in traffic conditions, evaluation of transportation policies and\nfuturistic road design related to pedestrian crossing behavior is of vital\nimportance. Recent studies analyzed the non-causal impact of various variables\non pedestrian waiting time in the presence of AVs. However, we mainly\ninvestigate the causal effect of traffic density on pedestrian waiting time. We\ndevelop a Double/Debiased Machine Learning (DML) model in which the impact of\nconfounders variable influencing both a policy and an outcome of interest is\naddressed, resulting in unbiased policy evaluation. Furthermore, we try to\nanalyze the effect of traffic density by developing a copula-based joint model\nof two main components of pedestrian crossing behavior, pedestrian stress level\nand waiting time. The copula approach has been widely used in the literature,\nfor addressing self-selection problems, which can be classified as a causality\nanalysis in travel behavior modeling. The results obtained from copula approach\nand DML are compared based on the effect of traffic density. In DML model\nstructure, the standard error term of density parameter is lower than copula\napproach and the confidence interval is considerably more reliable. In\naddition, despite the similar sign of effect, the copula approach estimates the\neffect of traffic density lower than DML, due to the spurious effect of\nconfounders. In short, the DML model structure can flexibly adjust the impact\nof confounders by using machine learning algorithms and is more reliable for\nplanning future policies.\n", "date": "2022-12-21", "categories": []}
{"id": "2212.10718", "title": "Interpretability and causal discovery of the machine learning models to\n  predict the production of CBM wells after hydraulic fracturing", "abstract": "  Machine learning approaches are widely studied in the production prediction\nof CBM wells after hydraulic fracturing, but merely used in practice due to the\nlow generalization ability and the lack of interpretability. A novel\nmethodology is proposed in this article to discover the latent causality from\nobserved data, which is aimed at finding an indirect way to interpret the\nmachine learning results. Based on the theory of causal discovery, a causal\ngraph is derived with explicit input, output, treatment and confounding\nvariables. Then, SHAP is employed to analyze the influence of the factors on\nthe production capability, which indirectly interprets the machine learning\nmodels. The proposed method can capture the underlying nonlinear relationship\nbetween the factors and the output, which remedies the limitation of the\ntraditional machine learning routines based on the correlation analysis of\nfactors. The experiment on the data of CBM shows that the detected relationship\nbetween the production and the geological/engineering factors by the presented\nmethod, is coincident with the actual physical mechanism. Meanwhile, compared\nwith traditional methods, the interpretable machine learning models have better\nperformance in forecasting production capability, averaging 20% improvement in\naccuracy.\n", "date": "2022-12-21", "categories": []}
{"id": "2206.01995", "title": "Combinatorial Causal Bandits", "abstract": "  In combinatorial causal bandits (CCB), the learning agent chooses at most $K$\nvariables in each round to intervene, collects feedback from the observed\nvariables, with the goal of minimizing expected regret on the target variable\n$Y$. We study under the context of binary generalized linear models (BGLMs)\nwith a succinct parametric representation of the causal models. We present the\nalgorithm BGLM-OFU for Markovian BGLMs (i.e. no hidden variables) based on the\nmaximum likelihood estimation method, and show that it achieves $O(\\sqrt{T}\\log\nT)$ regret, where $T$ is the time horizon. For the special case of linear\nmodels with hidden variables, we apply causal inference techniques such as the\ndo-calculus to convert the original model into a Markovian model, and then show\nthat our BGLM-OFU algorithm and another algorithm based on the linear\nregression both solve such linear models with hidden variables. Our novelty\nincludes (a) considering the combinatorial intervention action space and the\ngeneral causal models including ones with hidden variables, (b) integrating and\nadapting techniques from diverse studies such as generalized linear bandits and\nonline influence maximization, and (c) avoiding unrealistic assumptions (such\nas knowing the joint distribution of the parents of $Y$ under all\ninterventions) and regret factors exponential to causal graph size in prior\nstudies.\n", "date": "2022-06-04", "categories": []}
{"id": "2212.10076", "title": "Out-of-sample scoring and automatic selection of causal estimators", "abstract": "  Recently, many causal estimators for Conditional Average Treatment Effect\n(CATE) and instrumental variable (IV) problems have been published and open\nsourced, allowing to estimate granular impact of both randomized treatments\n(such as A/B tests) and of user choices on the outcomes of interest. However,\nthe practical application of such models has ben hampered by the lack of a\nvalid way to score the performance of such models out of sample, in order to\nselect the best one for a given application. We address that gap by proposing\nnovel scoring approaches for both the CATE case and an important subset of\ninstrumental variable problems, namely those where the instrumental variable is\ncustomer acces to a product feature, and the treatment is the customer's choice\nto use that feature. Being able to score model performance out of sample allows\nus to apply hyperparameter optimization methods to causal model selection and\ntuning. We implement that in an open source package that relies on DoWhy and\nEconML libraries for implementation of causal inference models (and also\nincludes a Transformed Outcome model implementation), and on FLAML for\nhyperparameter optimization and for component models used in the causal models.\nWe demonstrate on synthetic data that optimizing the proposed scores is a\nreliable method for choosing the model and its hyperparameter values, whose\nestimates are close to the true impact, in the randomized CATE and IV cases.\nFurther, we provide examles of applying these methods to real customer data\nfrom Wise.\n", "date": "2022-12-20", "categories": []}
{"id": "2206.01152", "title": "Causal Structure Learning: a Combinatorial Perspective", "abstract": "  In this review, we discuss approaches for learning causal structure from\ndata, also called causal discovery. In particular, we focus on approaches for\nlearning directed acyclic graphs (DAGs) and various generalizations which allow\nfor some variables to be unobserved in the available data. We devote special\nattention to two fundamental combinatorial aspects of causal structure\nlearning. First, we discuss the structure of the search space over causal\ngraphs. Second, we discuss the structure of equivalence classes over causal\ngraphs, i.e., sets of graphs which represent what can be learned from\nobservational data alone, and how these equivalence classes can be refined by\nadding interventional data.\n", "date": "2022-06-02", "categories": []}
{"id": "2212.08890", "title": "TCFimt: Temporal Counterfactual Forecasting from Individual Multiple\n  Treatment Perspective", "abstract": "  Determining causal effects of temporal multi-intervention assists\ndecision-making. Restricted by time-varying bias, selection bias, and\ninteractions of multiple interventions, the disentanglement and estimation of\nmultiple treatment effects from individual temporal data is still rare. To\ntackle these challenges, we propose a comprehensive framework of temporal\ncounterfactual forecasting from an individual multiple treatment perspective\n(TCFimt). TCFimt constructs adversarial tasks in a seq2seq framework to\nalleviate selection and time-varying bias and designs a contrastive\nlearning-based block to decouple a mixed treatment effect into separated main\ntreatment effects and causal interactions which further improves estimation\naccuracy. Through implementing experiments on two real-world datasets from\ndistinct fields, the proposed method shows satisfactory performance in\npredicting future outcomes with specific treatments and in choosing optimal\ntreatment type and timing than state-of-the-art methods.\n", "date": "2022-12-17", "categories": []}
{"id": "2212.08981", "title": "A Layered Architecture for Universal Causality", "abstract": "  We propose a layered hierarchical architecture called UCLA (Universal\nCausality Layered Architecture), which combines multiple levels of categorical\nabstraction for causal inference. At the top-most level, causal interventions\nare modeled combinatorially using a simplicial category of ordinal numbers. At\nthe second layer, causal models are defined by a graph-type category. The\nnon-random ``surgical\" operations on causal structures, such as edge deletion,\nare captured using degeneracy and face operators from the simplicial layer\nabove. The third categorical abstraction layer corresponds to the data layer in\ncausal inference. The fourth homotopy layer comprises of additional structure\nimposed on the instance layer above, such as a topological space, which enables\nevaluating causal models on datasets. Functors map between every pair of layers\nin UCLA. Each functor between layers is characterized by a universal arrow,\nwhich defines an isomorphism between every pair of categorical layers. These\nuniversal arrows define universal elements and representations through the\nYoneda Lemma, and in turn lead to a new category of elements based on a\nconstruction introduced by Grothendieck. Causal inference between each pair of\nlayers is defined as a lifting problem, a commutative diagram whose objects are\ncategories, and whose morphisms are functors that are characterized as\ndifferent types of fibrations. We illustrate the UCLA architecture using a\nrange of examples, including integer-valued multisets that represent a\nnon-graphical framework for conditional independence, and causal models based\non graphs and string diagrams using symmetric monoidal categories. We define\ncausal effect in terms of the homotopy colimit of the nerve of the category of\nelements.\n", "date": "2022-12-18", "categories": []}
{"id": "2212.08498", "title": "Evaluating vaccine allocation strategies using simulation-assisted\n  causal modelling", "abstract": "  Early on during a pandemic, vaccine availability is limited, requiring\nprioritisation of different population groups. Evaluating vaccine allocation is\ntherefore a crucial element of pandemics response. In the present work, we\ndevelop a model to retrospectively evaluate age-dependent counterfactual\nvaccine allocation strategies against the COVID-19 pandemic. To estimate the\neffect of allocation on the expected severe-case incidence, we employ a\nsimulation-assisted causal modelling approach which combines a compartmental\ninfection-dynamics simulation, a coarse-grained, data-driven causal model and\nliterature estimates for immunity waning. We compare Israel's implemented\nvaccine allocation strategy in 2021 to counterfactual strategies such as no\nprioritisation, prioritisation of younger age groups or a strict risk-ranked\napproach; we find that Israel's implemented strategy was indeed highly\neffective. We also study the marginal impact of increasing vaccine uptake for a\ngiven age group and find that increasing vaccinations in the elderly is most\neffective at preventing severe cases, whereas additional vaccinations for\nmiddle-aged groups reduce infections most effectively. Due to its modular\nstructure, our model can easily be adapted to study future pandemics. We\ndemonstrate this flexibility by investigating vaccine allocation strategies for\na pandemic with characteristics of the Spanish Flu. Our approach thus helps\nevaluate vaccination strategies under the complex interplay of core epidemic\nfactors, including age-dependent risk profiles, immunity waning, vaccine\navailability and spreading rates.\n", "date": "2022-12-14", "categories": []}
{"id": "2209.08579", "title": "Bivariate Causal Discovery for Categorical Data via Classification with\n  Optimal Label Permutation", "abstract": "  Causal discovery for quantitative data has been extensively studied but less\nis known for categorical data. We propose a novel causal model for categorical\ndata based on a new classification model, termed classification with optimal\nlabel permutation (COLP). By design, COLP is a parsimonious classifier, which\ngives rise to a provably identifiable causal model. A simple learning algorithm\nvia comparing likelihood functions of causal and anti-causal models suffices to\nlearn the causal direction. Through experiments with synthetic and real data,\nwe demonstrate the favorable performance of the proposed COLP-based causal\nmodel compared to state-of-the-art methods. We also make available an\naccompanying R package COLP, which contains the proposed causal discovery\nalgorithm and a benchmark dataset of categorical cause-effect pairs.\n", "date": "2022-09-18", "categories": []}
{"id": "2209.12316", "title": "Weather2vec: Representation Learning for Causal Inference with Non-Local\n  Confounding in Air Pollution and Climate Studies", "abstract": "  Estimating the causal effects of a spatially-varying intervention on a\nspatially-varying outcome may be subject to non-local confounding (NLC), a\nphenomenon that can bias estimates when the treatments and outcomes of a given\nunit are dictated in part by the covariates of other nearby units. In\nparticular, NLC is a challenge for evaluating the effects of environmental\npolicies and climate events on health-related outcomes such as air pollution\nexposure. This paper first formalizes NLC using the potential outcomes\nframework, providing a comparison with the related phenomenon of causal\ninterference. Then, it proposes a broadly applicable framework, termed\n\"weather2vec\", that uses the theory of balancing scores to learn\nrepresentations of non-local information into a scalar or vector defined for\neach observational unit, which is subsequently used to adjust for confounding\nin conjunction with causal inference methods. The framework is evaluated in a\nsimulation study and two case studies on air pollution where the weather is an\n(inherently regional) known confounder.\n", "date": "2022-09-25", "categories": []}
{"id": "2210.12368", "title": "Counterfactual Generation Under Confounding", "abstract": "  A machine learning model, under the influence of observed or unobserved\nconfounders in the training data, can learn spurious correlations and fail to\ngeneralize when deployed. For image classifiers, augmenting a training dataset\nusing counterfactual examples has been empirically shown to break spurious\ncorrelations. However, the counterfactual generation task itself becomes more\ndifficult as the level of confounding increases. Existing methods for\ncounterfactual generation under confounding consider a fixed set of\ninterventions (e.g., texture, rotation) and are not flexible enough to capture\ndiverse data-generating processes. Given a causal generative process, we\nformally characterize the adverse effects of confounding on any downstream\ntasks and show that the correlation between generative factors (attributes) can\nbe used to quantitatively measure confounding between generative factors. To\nminimize such correlation, we propose a counterfactual generation method that\nlearns to modify the value of any attribute in an image and generate new images\ngiven a set of observed attributes, even when the dataset is highly confounded.\nThese counterfactual images are then used to regularize the downstream\nclassifier such that the learned representations are the same across various\ngenerative factors conditioned on the class label. Our method is\ncomputationally efficient, simple to implement, and works well for any number\nof generative factors and confounding variables. Our experimental results on\nboth synthetic (MNIST variants) and real-world (CelebA) datasets show the\nusefulness of our approach.\n", "date": "2022-10-22", "categories": []}
{"id": "2212.05778", "title": "Instrumental Variables in Causal Inference and Machine Learning: A\n  Survey", "abstract": "  Causal inference is the process of using assumptions, study designs, and\nestimation strategies to draw conclusions about the causal relationships\nbetween variables based on data. This allows researchers to better understand\nthe underlying mechanisms at work in complex systems and make more informed\ndecisions. In many settings, we may not fully observe all the confounders that\naffect both the treatment and outcome variables, complicating the estimation of\ncausal effects. To address this problem, a growing literature in both causal\ninference and machine learning proposes to use Instrumental Variables (IV).\nThis paper serves as the first effort to systematically and comprehensively\nintroduce and discuss the IV methods and their applications in both causal\ninference and machine learning. First, we provide the formal definition of IVs\nand discuss the identification problem of IV regression methods under different\nassumptions. Second, we categorize the existing work on IV methods into three\nstreams according to the focus on the proposed methods, including two-stage\nleast squares with IVs, control function with IVs, and evaluation of IVs. For\neach stream, we present both the classical causal inference methods, and recent\ndevelopments in the machine learning literature. Then, we introduce a variety\nof applications of IV methods in real-world scenarios and provide a summary of\nthe available datasets and algorithms. Finally, we summarize the literature,\ndiscuss the open problems and suggest promising future research directions for\nIV methods and their applications. We also develop a toolkit of IVs methods\nreviewed in this survey at https://github.com/causal-machine-learning-lab/mliv.\n", "date": "2022-12-12", "categories": []}
{"id": "2212.04866", "title": "Deep Learning of Causal Structures in High Dimensions", "abstract": "  Recent years have seen rapid progress at the intersection between causality\nand machine learning. Motivated by scientific applications involving\nhigh-dimensional data, in particular in biomedicine, we propose a deep neural\narchitecture for learning causal relationships between variables from a\ncombination of empirical data and prior causal knowledge. We combine\nconvolutional and graph neural networks within a causal risk framework to\nprovide a flexible and scalable approach. Empirical results include linear and\nnonlinear simulations (where the underlying causal structures are known and can\nbe directly compared against), as well as a real biological example where the\nmodels are applied to high-dimensional molecular data and their output compared\nagainst entirely unseen validation experiments. These results demonstrate the\nfeasibility of using deep learning approaches to learn causal networks in\nlarge-scale problems spanning thousands of variables.\n", "date": "2022-12-09", "categories": []}
{"id": "2212.03974", "title": "Counterfactuals for the Future", "abstract": "  Counterfactuals are often described as 'retrospective,' focusing on\nhypothetical alternatives to a realized past. This description relates to an\noften implicit assumption about the structure and stability of exogenous\nvariables in the system being modeled -- an assumption that is reasonable in\nmany settings where counterfactuals are used. In this work, we consider cases\nwhere we might reasonably make a different assumption about exogenous\nvariables, namely, that the exogenous noise terms of each unit do exhibit some\nunit-specific structure and/or stability. This leads us to a different use of\ncounterfactuals -- a 'forward-looking' rather than 'retrospective'\ncounterfactual. We introduce \"counterfactual treatment choice,\" a type of\ntreatment choice problem that motivates using forward-looking counterfactuals.\nWe then explore how mismatches between interventional versus forward-looking\ncounterfactual approaches to treatment choice, consistent with different\nassumptions about exogenous noise, can lead to counterintuitive results.\n", "date": "2022-12-07", "categories": []}
{"id": "2212.04235", "title": "A probabilistic autoencoder for causal discovery", "abstract": "  The paper addresses the problem of finding the causal direction between two\nassociated variables. The proposed solution is to build an autoencoder of their\njoint distribution and to maximize its estimation capacity relative to both the\nmarginal distributions. It is shown that the resulting two capacities cannot,\nin general, be equal. This leads to a new criterion for causal discovery: the\nhigher capacity is consistent with the unconstrained choice of a distribution\nrepresenting the cause while the lower capacity reflects the constraints\nimposed by the mechanism on the distribution of the effect. Estimation capacity\nis defined as the ability of the auto-encoder to represent arbitrary datasets.\nA regularization term forces it to decide which one of the variables to model\nin a more generic way i.e., while maintaining higher model capacity. The causal\ndirection is revealed by the constraints encountered while encoding the data\ninstead of being measured as a property of the data itself. The idea is\nimplemented and tested using a restricted Boltzmann machine.\n", "date": "2022-12-08", "categories": []}
{"id": "2208.10912", "title": "Learning Instrumental Variable from Data Fusion for Treatment Effect\n  Estimation", "abstract": "  The advent of the big data era brought new opportunities and challenges to\ndraw treatment effect in data fusion, that is, a mixed dataset collected from\nmultiple sources (each source with an independent treatment assignment\nmechanism). Due to possibly omitted source labels and unmeasured confounders,\ntraditional methods cannot estimate individual treatment assignment probability\nand infer treatment effect effectively. Therefore, we propose to reconstruct\nthe source label and model it as a Group Instrumental Variable (GIV) to\nimplement IV-based Regression for treatment effect estimation. In this paper,\nwe conceptualize this line of thought and develop a unified framework (Meta-EM)\nto (1) map the raw data into a representation space to construct Linear Mixed\nModels for the assigned treatment variable; (2) estimate the distribution\ndifferences and model the GIV for the different treatment assignment\nmechanisms; and (3) adopt an alternating training strategy to iteratively\noptimize the representations and the joint distribution to model GIV for IV\nregression. Empirical results demonstrate the advantages of our Meta-EM\ncompared with state-of-the-art methods.\n", "date": "2022-08-23", "categories": []}
{"id": "2212.01437", "title": "Identifying Heterogeneous Treatment Effects in Multiple Outcomes using\n  Joint Confidence Intervals", "abstract": "  Heterogeneous treatment effects (HTEs) are commonly identified during\nrandomized controlled trials (RCTs). Identifying subgroups of patients with\nsimilar treatment effects is of high interest in clinical research to advance\nprecision medicine. Often, multiple clinical outcomes are measured during an\nRCT, each having a potentially heterogeneous effect. Recently there has been\nhigh interest in identifying subgroups from HTEs, however, there has been less\nfocus on developing tools in settings where there are multiple outcomes. In\nthis work, we propose a framework for partitioning the covariate space to\nidentify subgroups across multiple outcomes based on the joint CIs. We test our\nalgorithm on synthetic and semi-synthetic data where there are two outcomes,\nand demonstrate that our algorithm is able to capture the HTE in both outcomes\nsimultaneously.\n", "date": "2022-12-02", "categories": []}
{"id": "2212.02435", "title": "Observational and Interventional Causal Learning for Regret-Minimizing\n  Control", "abstract": "  We explore how observational and interventional causal discovery methods can\nbe combined. A state-of-the-art observational causal discovery algorithm for\ntime series capable of handling latent confounders and contemporaneous effects,\ncalled LPCMCI, is extended to profit from casual constraints found through\nrandomized control trials. Numerical results show that, given perfect\ninterventional constraints, the reconstructed structural causal models (SCMs)\nof the extended LPCMCI allow 84.6% of the time for the optimal prediction of\nthe target variable. The implementation of interventional and observational\ncausal discovery is modular, allowing causal constraints from other sources.\n  The second part of this thesis investigates the question of regret minimizing\ncontrol by simultaneously learning a causal model and planning actions through\nthe causal model. The idea is that an agent to optimize a measured variable\nfirst learns the system's mechanics through observational causal discovery. The\nagent then intervenes on the most promising variable with randomized values\nallowing for the exploitation and generation of new interventional data. The\nagent then uses the interventional data to enhance the causal model further,\nallowing improved actions the next time.\n  The extended LPCMCI can be favorable compared to the original LPCMCI\nalgorithm. The numerical results show that detecting and using interventional\nconstraints leads to reconstructed SCMs that allow 60.9% of the time for the\noptimal prediction of the target variable in contrast to the baseline of 53.6%\nwhen using the original LPCMCI algorithm. Furthermore, the induced average\nregret decreases from 1.2 when using the original LPCMCI algorithm to 1.0 when\nusing the extended LPCMCI algorithm with interventional discovery.\n", "date": "2022-12-05", "categories": []}
{"id": "2212.00911", "title": "Navigating causal deep learning", "abstract": "  Causal deep learning (CDL) is a new and important research area in the larger\nfield of machine learning. With CDL, researchers aim to structure and encode\ncausal knowledge in the extremely flexible representation space of deep\nlearning models. Doing so will lead to more informed, robust, and general\npredictions and inference -- which is important! However, CDL is still in its\ninfancy. For example, it is not clear how we ought to compare different methods\nas they are so different in their output, the way they encode causal knowledge,\nor even how they represent this knowledge. This is a living paper that\ncategorises methods in causal deep learning beyond Pearl's ladder of causation.\nWe refine the rungs in Pearl's ladder, while also adding a separate dimension\nthat categorises the parametric assumptions of both input and representation,\narriving at the map of causal deep learning. Our map covers machine learning\ndisciplines such as supervised learning, reinforcement learning, generative\nmodelling and beyond. Our paradigm is a tool which helps researchers to: find\nbenchmarks, compare methods, and most importantly: identify research gaps. With\nthis work we aim to structure the avalanche of papers being published on causal\ndeep learning. While papers on the topic are being published daily, our map\nremains fixed. We open-source our map for others to use as they see fit:\nperhaps to offer guidance in a related works section, or to better highlight\nthe contribution of their paper.\n", "date": "2022-12-01", "categories": []}
{"id": "2212.01279", "title": "Initial Results for Pairwise Causal Discovery Using Quantitative\n  Information Flow", "abstract": "  Pairwise Causal Discovery is the task of determining causal, anticausal,\nconfounded or independence relationships from pairs of variables. Over the last\nfew years, this challenging task has promoted not only the discovery of novel\nmachine learning models aimed at solving the task, but also discussions on how\nlearning the causal direction of variables may benefit machine learning\noverall. In this paper, we show that Quantitative Information Flow (QIF), a\nmeasure usually employed for measuring leakages of information from a system to\nan attacker, shows promising results as features for the task. In particular,\nexperiments with real-world datasets indicate that QIF is statistically tied to\nthe state of the art. Our initial results motivate further inquiries on how QIF\nrelates to causality and what are its limitations.\n", "date": "2022-12-02", "categories": []}
{"id": "2211.15728", "title": "Direct Heterogeneous Causal Learning for Resource Allocation Problems in\n  Marketing", "abstract": "  Marketing is an important mechanism to increase user engagement and improve\nplatform revenue, and heterogeneous causal learning can help develop more\neffective strategies. Most decision-making problems in marketing can be\nformulated as resource allocation problems and have been studied for decades.\nExisting works usually divide the solution procedure into two fully decoupled\nstages, i.e., machine learning (ML) and operation research (OR) -- the first\nstage predicts the model parameters and they are fed to the optimization in the\nsecond stage. However, the error of the predicted parameters in ML cannot be\nrespected and a series of complex mathematical operations in OR lead to the\nincreased accumulative errors. Essentially, the improved precision on the\nprediction parameters may not have a positive correlation on the final solution\ndue to the side-effect from the decoupled design.\n  In this paper, we propose a novel approach for solving resource allocation\nproblems to mitigate the side-effects. Our key intuition is that we introduce\nthe decision factor to establish a bridge between ML and OR such that the\nsolution can be directly obtained in OR by only performing the sorting or\ncomparison operations on the decision factor. Furthermore, we design a\ncustomized loss function that can conduct direct heterogeneous causal learning\non the decision factor, an unbiased estimation of which can be guaranteed when\nthe loss converges. As a case study, we apply our approach to two crucial\nproblems in marketing: the binary treatment assignment problem and the budget\nallocation problem with multiple treatments. Both large-scale simulations and\nonline A/B Tests demonstrate that our approach achieves significant improvement\ncompared with state-of-the-art.\n", "date": "2022-11-28", "categories": []}
{"id": "2211.16938", "title": "Evaluating Digital Agriculture Recommendations with Causal Inference", "abstract": "  In contrast to the rapid digitalization of several industries, agriculture\nsuffers from low adoption of smart farming tools. While AI-driven digital\nagriculture tools can offer high-performing predictive functionalities, they\nlack tangible quantitative evidence on their benefits to the farmers. Field\nexperiments can derive such evidence, but are often costly, time consuming and\nhence limited in scope and scale of application. To this end, we propose an\nobservational causal inference framework for the empirical evaluation of the\nimpact of digital tools on target farm performance indicators (e.g., yield in\nthis case). This way, we can increase farmers' trust via enhancing the\ntransparency of the digital agriculture market and accelerate the adoption of\ntechnologies that aim to secure farmer income resilience and global\nagricultural sustainability. As a case study, we designed and implemented a\nrecommendation system for the optimal sowing time of cotton based on numerical\nweather predictions, which was used by a farmers' cooperative during the\ngrowing season of 2021. We then leverage agricultural knowledge, collected\nyield data, and environmental information to develop a causal graph of the farm\nsystem. Using the back-door criterion, we identify the impact of sowing\nrecommendations on the yield and subsequently estimate it using linear\nregression, matching, inverse propensity score weighting and meta-learners. The\nresults reveal that a field sown according to our recommendations exhibited a\nstatistically significant yield increase that ranged from 12% to 17%, depending\non the method. The effect estimates were robust, as indicated by the agreement\namong the estimation methods and four successful refutation tests. We argue\nthat this approach can be implemented for decision support systems of other\nfields, extending their evaluation beyond a performance assessment of internal\nfunctionalities.\n", "date": "2022-11-30", "categories": []}
{"id": "2211.16246", "title": "Causal Inference with Conditional Instruments using Deep Generative\n  Models", "abstract": "  The instrumental variable (IV) approach is a widely used way to estimate the\ncausal effects of a treatment on an outcome of interest from observational data\nwith latent confounders. A standard IV is expected to be related to the\ntreatment variable and independent of all other variables in the system.\nHowever, it is challenging to search for a standard IV from data directly due\nto the strict conditions. The conditional IV (CIV) method has been proposed to\nallow a variable to be an instrument conditioning on a set of variables,\nallowing a wider choice of possible IVs and enabling broader practical\napplications of the IV approach. Nevertheless, there is not a data-driven\nmethod to discover a CIV and its conditioning set directly from data. To fill\nthis gap, in this paper, we propose to learn the representations of the\ninformation of a CIV and its conditioning set from data with latent confounders\nfor average causal effect estimation. By taking advantage of deep generative\nmodels, we develop a novel data-driven approach for simultaneously learning the\nrepresentation of a CIV from measured variables and generating the\nrepresentation of its conditioning set given measured variables. Extensive\nexperiments on synthetic and real-world datasets show that our method\noutperforms the existing IV methods.\n", "date": "2022-11-29", "categories": []}
{"id": "2207.04208", "title": "SCouT: Synthetic Counterfactuals via Spatiotemporal Transformers for\n  Actionable Healthcare", "abstract": "  The Synthetic Control method has pioneered a class of powerful data-driven\ntechniques to estimate the counterfactual reality of a unit from donor units.\nAt its core, the technique involves a linear model fitted on the\npre-intervention period that combines donor outcomes to yield the\ncounterfactual. However, linearly combining spatial information at each time\ninstance using time-agnostic weights fails to capture important inter-unit and\nintra-unit temporal contexts and complex nonlinear dynamics of real data. We\ninstead propose an approach to use local spatiotemporal information before the\nonset of the intervention as a promising way to estimate the counterfactual\nsequence. To this end, we suggest a Transformer model that leverages particular\npositional embeddings, a modified decoder attention mask, and a novel\npre-training task to perform spatiotemporal sequence-to-sequence modeling. Our\nexperiments on synthetic data demonstrate the efficacy of our method in the\ntypical small donor pool setting and its robustness against noise. We also\ngenerate actionable healthcare insights at the population and patient levels by\nsimulating a state-wide public health policy to evaluate its effectiveness, an\nin silico trial for asthma medications to support randomized controlled trials,\nand a medical intervention for patients with Friedreich's ataxia to improve\nclinical decision-making and promote personalized therapy.\n", "date": "2022-07-09", "categories": []}
{"id": "2211.13800", "title": "Estimation of a Causal Directed Acyclic Graph Process using\n  Non-Gaussianity", "abstract": "  Numerous approaches have been proposed to discover causal dependencies in\nmachine learning and data mining; among them, the state-of-the-art VAR-LiNGAM\n(short for Vector Auto-Regressive Linear Non-Gaussian Acyclic Model) is a\ndesirable approach to reveal both the instantaneous and time-lagged\nrelationships. However, all the obtained VAR matrices need to be analyzed to\ninfer the final causal graph, leading to a rise in the number of parameters. To\naddress this issue, we propose the CGP-LiNGAM (short for Causal Graph\nProcess-LiNGAM), which has significantly fewer model parameters and deals with\nonly one causal graph for interpreting the causal relations by exploiting Graph\nSignal Processing (GSP).\n", "date": "2022-11-24", "categories": []}
{"id": "2211.12983", "title": "Causal Analysis of the TOPCAT Trial: Spironolactone for Preserved\n  Cardiac Function Heart Failure", "abstract": "  We describe the results of applying causal discovery methods on the data from\na multi-site clinical trial, on the Treatment of Preserved Cardiac Function\nHeart Failure with an Aldosterone Antagonist (TOPCAT). The trial was\ninconclusive, with no clear benefits consistently shown for the whole cohort.\nHowever, there were questions regarding the reliability of the diagnosis and\ntreatment protocol for a geographic subgroup of the cohort. With the inclusion\nof medical context in the form of domain knowledge, causal discovery is used to\ndemonstrate regional discrepancies and to frame the regional transportability\nof the results. Furthermore, we show that, globally and especially for some\nsubgroups, the treatment has significant causal effects, thus offering a more\nrefined view of the trial results.\n", "date": "2022-11-23", "categories": []}
{"id": "2211.12016", "title": "Variation-based Cause Effect Identification", "abstract": "  Mining genuine mechanisms underlying the complex data generation process in\nreal-world systems is a fundamental step in promoting interpretability of, and\nthus trust in, data-driven models. Therefore, we propose a variation-based\ncause effect identification (VCEI) framework for causal discovery in bivariate\nsystems from a single observational setting. Our framework relies on the\nprinciple of independence of cause and mechanism (ICM) under the assumption of\nan existing acyclic causal link, and offers a practical realization of this\nprinciple. Principally, we artificially construct two settings in which the\nmarginal distributions of one covariate, claimed to be the cause, are\nguaranteed to have non-negligible variations. This is achieved by re-weighting\nsamples of the marginal so that the resultant distribution is notably distinct\nfrom this marginal according to some discrepancy measure. In the causal\ndirection, such variations are expected to have no impact on the effect\ngeneration mechanism. Therefore, quantifying the impact of these variations on\nthe conditionals reveals the genuine causal direction. Moreover, we formulate\nour approach in the kernel-based maximum mean discrepancy, lifting all\nconstraints on the data types of cause-and-effect covariates, and rendering\nsuch artificial interventions a convex optimization problem. We provide a\nseries of experiments on real and synthetic data showing that VCEI is, in\nprinciple, competitive to other cause effect identification frameworks.\n", "date": "2022-11-22", "categories": []}
{"id": "2211.12151", "title": "Reinforcement Causal Structure Learning on Order Graph", "abstract": "  Learning directed acyclic graph (DAG) that describes the causality of\nobserved data is a very challenging but important task. Due to the limited\nquantity and quality of observed data, and non-identifiability of causal graph,\nit is almost impossible to infer a single precise DAG. Some methods approximate\nthe posterior distribution of DAGs to explore the DAG space via Markov chain\nMonte Carlo (MCMC), but the DAG space is over the nature of super-exponential\ngrowth, accurately characterizing the whole distribution over DAGs is very\nintractable. In this paper, we propose {Reinforcement Causal Structure Learning\non Order Graph} (RCL-OG) that uses order graph instead of MCMC to model\ndifferent DAG topological orderings and to reduce the problem size. RCL-OG\nfirst defines reinforcement learning with a new reward mechanism to approximate\nthe posterior distribution of orderings in an efficacy way, and uses deep\nQ-learning to update and transfer rewards between nodes. Next, it obtains the\nprobability transition model of nodes on order graph, and computes the\nposterior probability of different orderings. In this way, we can sample on\nthis model to obtain the ordering with high probability. Experiments on\nsynthetic and benchmark datasets show that RCL-OG provides accurate posterior\nprobability approximation and achieves better results than competitive causal\ndiscovery algorithms.\n", "date": "2022-11-22", "categories": []}
{"id": "2211.12270", "title": "Causal Abstraction with Soft Interventions", "abstract": "  Causal abstraction provides a theory describing how several causal models can\nrepresent the same system at different levels of detail. Existing theoretical\nproposals limit the analysis of abstract models to \"hard\" interventions fixing\ncausal variables to be constant values. In this work, we extend causal\nabstraction to \"soft\" interventions, which assign possibly non-constant\nfunctions to variables without adding new causal connections. Specifically, (i)\nwe generalize $\\tau$-abstraction from Beckers and Halpern (2019) to soft\ninterventions, (ii) we propose a further definition of soft abstraction to\nensure a unique map $\\omega$ between soft interventions, and (iii) we prove\nthat our constructive definition of soft abstraction guarantees the\nintervention map $\\omega$ has a specific and necessary explicit form.\n", "date": "2022-11-22", "categories": []}
{"id": "2211.10793", "title": "BENK: The Beran Estimator with Neural Kernels for Estimating the\n  Heterogeneous Treatment Effect", "abstract": "  A method for estimating the conditional average treatment effect under\ncondition of censored time-to-event data called BENK (the Beran Estimator with\nNeural Kernels) is proposed. The main idea behind the method is to apply the\nBeran estimator for estimating the survival functions of controls and\ntreatments. Instead of typical kernel functions in the Beran estimator, it is\nproposed to implement kernels in the form of neural networks of a specific form\ncalled the neural kernels. The conditional average treatment effect is\nestimated by using the survival functions as outcomes of the control and\ntreatment neural networks which consists of a set of neural kernels with shared\nparameters. The neural kernels are more flexible and can accurately model a\ncomplex location structure of feature vectors. Various numerical simulation\nexperiments illustrate BENK and compare it with the well-known T-learner,\nS-learner and X-learner for several types of the control and treatment outcome\nfunctions based on the Cox models, the random survival forest and the\nNadaraya-Watson regression with Gaussian kernels. The code of proposed\nalgorithms implementing BENK is available in https://github.com/Stasychbr/BENK.\n", "date": "2022-11-19", "categories": []}
{"id": "2211.11119", "title": "Counterfactual Learning with Multioutput Deep Kernels", "abstract": "  In this paper, we address the challenge of performing counterfactual\ninference with observational data via Bayesian nonparametric regression\nadjustment, with a focus on high-dimensional settings featuring multiple\nactions and multiple correlated outcomes. We present a general class of\ncounterfactual multi-task deep kernels models that estimate causal effects and\nlearn policies proficiently thanks to their sample efficiency gains, while\nscaling well with high dimensions. In the first part of the work, we rely on\nStructural Causal Models (SCM) to formally introduce the setup and the problem\nof identifying counterfactual quantities under observed confounding. We then\ndiscuss the benefits of tackling the task of causal effects estimation via\nstacked coregionalized Gaussian Processes and Deep Kernels. Finally, we\ndemonstrate the use of the proposed methods on simulated experiments that span\nindividual causal effects estimation, off-policy evaluation and optimization.\n", "date": "2022-11-20", "categories": []}
{"id": "2209.11497", "title": "Time Series Causal Link Estimation under Hidden Confounding using\n  Knockoff Interventions", "abstract": "  Latent variables often mask cause-effect relationships in observational data\nwhich provokes spurious links that may be misinterpreted as causal. This\nproblem sparks great interest in the fields such as climate science and\neconomics. We propose to estimate confounded causal links of time series using\nSequential Causal Effect Variational Autoencoder (SCEVAE) while applying\nKnockoff interventions. Knockoff variables have the same distribution as the\noriginals and preserve the correlation to other variables. This allows for\ncounterfactuals that are more faithful to the observational distribution. We\nshow the advantage of Knockoff interventions by applying SCEVAE to synthetic\ndatasets with both linear and nonlinear causal links. Moreover, we apply SCEVAE\nwith Knockoffs to real aerosol-cloud-climate observational time series data. We\ncompare our results on synthetic data to those of a time series deconfounding\nmethod both with and without estimated confounders. We show that our method\noutperforms this benchmark by comparing both methods to the ground truth. For\nthe real data analysis, we rely on expert knowledge of causal links and\ndemonstrate how using suitable proxy variables improves the causal link\nestimation in the presence of hidden confounders.\n", "date": "2022-09-23", "categories": []}
{"id": "2211.08883", "title": "Identifying the Causes of Pyrocumulonimbus (PyroCb)", "abstract": "  A first causal discovery analysis from observational data of pyroCb (storm\nclouds generated from extreme wildfires) is presented. Invariant Causal\nPrediction was used to develop tools to understand the causal drivers of pyroCb\nformation. This includes a conditional independence test for testing $Y$\nconditionally independent of $E$ given $X$ for binary variable $Y$ and\nmultivariate, continuous variables $X$ and $E$, and a greedy-ICP search\nalgorithm that relies on fewer conditional independence tests to obtain a\nsmaller more manageable set of causal predictors. With these tools, we\nidentified a subset of seven causal predictors which are plausible when\ncontrasted with domain knowledge: surface sensible heat flux, relative humidity\nat $850$ hPa, a component of wind at $250$ hPa, $13.3$ micro-meters, thermal\nemissions, convective available potential energy, and altitude.\n", "date": "2022-11-16", "categories": []}
{"id": "2211.10008", "title": "Confounder Balancing for Instrumental Variable Regression with Latent\n  Variable", "abstract": "  This paper studies the confounding effects from the unmeasured confounders\nand the imbalance of observed confounders in IV regression and aims at unbiased\ncausal effect estimation. Recently, nonlinear IV estimators were proposed to\nallow for nonlinear model in both stages. However, the observed confounders may\nbe imbalanced in stage 2, which could still lead to biased treatment effect\nestimation in certain cases. To this end, we propose a Confounder Balanced IV\nRegression (CB-IV) algorithm to jointly remove the bias from the unmeasured\nconfounders and the imbalance of observed confounders. Theoretically, by\nredefining and solving an inverse problem for potential outcome function, we\nshow that our CB-IV algorithm can unbiasedly estimate treatment effects and\nachieve lower variance. The IV methods have a major disadvantage in that little\nprior or theory is currently available to pre-define a valid IV in real-world\nscenarios. Thus, we study two more challenging settings without pre-defined\nvalid IVs: (1) indistinguishable IVs implicitly present in observations, i.e.,\nmixed-variable challenge, and (2) latent IVs don't appear in observations,\ni.e., latent-variable challenge. To address these two challenges, we extend our\nCB-IV by a latent-variable module, namely CB-IV-L algorithm. Extensive\nexperiments demonstrate that our CB-IV(-L) outperforms the existing approaches.\n", "date": "2022-11-18", "categories": []}
{"id": "2211.09024", "title": "Phenomenological Causality", "abstract": "  Discussions on causal relations in real life often consider variables for\nwhich the definition of causality is unclear since the notion of interventions\non the respective variables is obscure. Asking 'what qualifies an action for\nbeing an intervention on the variable X' raises the question whether the action\nimpacted all other variables only through X or directly, which implicitly\nrefers to a causal model.\n  To avoid this known circularity, we instead suggest a notion of\n'phenomenological causality' whose basic concept is a set of elementary\nactions. Then the causal structure is defined such that elementary actions\nchange only the causal mechanism at one node (e.g. one of the causal\nconditionals in the Markov factorization). This way, the Principle of\nIndependent Mechanisms becomes the defining property of causal structure in\ndomains where causality is a more abstract phenomenon rather than being an\nobjective fact relying on hard-wired causal links between tangible objects. We\ndescribe this phenomenological approach to causality for toy and hypothetical\nreal-world examples and argue that it is consistent with the causal Markov\ncondition when the system under consideration interacts with other variables\nthat control the elementary actions.\n", "date": "2022-11-15", "categories": []}
{"id": "2206.01825", "title": "Debiased Machine Learning without Sample-Splitting for Stable Estimators", "abstract": "  Estimation and inference on causal parameters is typically reduced to a\ngeneralized method of moments problem, which involves auxiliary functions that\ncorrespond to solutions to a regression or classification problem. Recent line\nof work on debiased machine learning shows how one can use generic machine\nlearning estimators for these auxiliary problems, while maintaining asymptotic\nnormality and root-$n$ consistency of the target parameter of interest, while\nonly requiring mean-squared-error guarantees from the auxiliary estimation\nalgorithms. The literature typically requires that these auxiliary problems are\nfitted on a separate sample or in a cross-fitting manner. We show that when\nthese auxiliary estimation algorithms satisfy natural leave-one-out stability\nproperties, then sample splitting is not required. This allows for sample\nre-use, which can be beneficial in moderately sized sample regimes. For\ninstance, we show that the stability properties that we propose are satisfied\nfor ensemble bagged estimators, built via sub-sampling without replacement, a\npopular technique in machine learning practice.\n", "date": "2022-06-03", "categories": []}
{"id": "2211.08243", "title": "Neural Bayesian Network Understudy", "abstract": "  Bayesian Networks may be appealing for clinical decision-making due to their\ninclusion of causal knowledge, but their practical adoption remains limited as\na result of their inability to deal with unstructured data. While neural\nnetworks do not have this limitation, they are not interpretable and are\ninherently unable to deal with causal structure in the input space. Our goal is\nto build neural networks that combine the advantages of both approaches.\nMotivated by the perspective to inject causal knowledge while training such\nneural networks, this work presents initial steps in that direction. We\ndemonstrate how a neural network can be trained to output conditional\nprobabilities, providing approximately the same functionality as a Bayesian\nNetwork. Additionally, we propose two training strategies that allow encoding\nthe independence relations inferred from a given causal structure into the\nneural network. We present initial results in a proof-of-concept setting,\nshowing that the neural model acts as an understudy to its Bayesian Network\ncounterpart, approximating its probabilistic and causal properties.\n", "date": "2022-11-15", "categories": []}
{"id": "2211.06536", "title": "Improving the Efficiency of the PC Algorithm by Using Model-Based\n  Conditional Independence Tests", "abstract": "  Learning causal structure is useful in many areas of artificial intelligence,\nincluding planning, robotics, and explanation. Constraint-based structure\nlearning algorithms such as PC use conditional independence (CI) tests to infer\ncausal structure. Traditionally, constraint-based algorithms perform CI tests\nwith a preference for smaller-sized conditioning sets, partially because the\nstatistical power of conventional CI tests declines rapidly as the size of the\nconditioning set increases. However, many modern conditional independence tests\nare model-based, and these tests use well-regularized models that maintain\nstatistical power even with very large conditioning sets. This suggests an\nintriguing new strategy for constraint-based algorithms which may result in a\nreduction of the total number of CI tests performed: Test variable pairs with\nlarge conditioning sets first, as a pre-processing step that finds some\nconditional independencies quickly, before moving on to the more conventional\nstrategy that favors small conditioning sets. We propose such a pre-processing\nstep for the PC algorithm which relies on performing CI tests on a few randomly\nselected large conditioning sets. We perform an empirical analysis on directed\nacyclic graphs (DAGs) that correspond to real-world systems and both empirical\nand theoretical analyses for Erd\\H{o}s-Renyi DAGs. Our results show that\nPre-Processing Plus PC (P3PC) performs far fewer CI tests than the original PC\nalgorithm, between 0.5% to 36%, and often less than 10%, of the CI tests that\nthe PC algorithm alone performs. The efficiency gains are particularly\nsignificant for the DAGs corresponding to real-world systems.\n", "date": "2022-11-12", "categories": []}
{"id": "2211.06569", "title": "RISE: Robust Individualized Decision Learning with Sensitive Variables", "abstract": "  This paper introduces RISE, a robust individualized decision learning\nframework with sensitive variables, where sensitive variables are collectible\ndata and important to the intervention decision, but their inclusion in\ndecision making is prohibited due to reasons such as delayed availability or\nfairness concerns. A naive baseline is to ignore these sensitive variables in\nlearning decision rules, leading to significant uncertainty and bias. To\naddress this, we propose a decision learning framework to incorporate sensitive\nvariables during offline training but not include them in the input of the\nlearned decision rule during model deployment. Specifically, from a causal\nperspective, the proposed framework intends to improve the worst-case outcomes\nof individuals caused by sensitive variables that are unavailable at the time\nof decision. Unlike most existing literature that uses mean-optimal objectives,\nwe propose a robust learning framework by finding a newly defined quantile- or\ninfimum-optimal decision rule. The reliable performance of the proposed method\nis demonstrated through synthetic experiments and three real-world\napplications.\n", "date": "2022-11-12", "categories": []}
{"id": "2211.06684", "title": "A Generalized Doubly Robust Learning Framework for Debiasing Post-Click\n  Conversion Rate Prediction", "abstract": "  Post-click conversion rate (CVR) prediction is an essential task for\ndiscovering user interests and increasing platform revenues in a range of\nindustrial applications. One of the most challenging problems of this task is\nthe existence of severe selection bias caused by the inherent self-selection\nbehavior of users and the item selection process of systems. Currently, doubly\nrobust (DR) learning approaches achieve the state-of-the-art performance for\ndebiasing CVR prediction. However, in this paper, by theoretically analyzing\nthe bias, variance and generalization bounds of DR methods, we find that\nexisting DR approaches may have poor generalization caused by inaccurate\nestimation of propensity scores and imputation errors, which often occur in\npractice. Motivated by such analysis, we propose a generalized learning\nframework that not only unifies existing DR methods, but also provides a\nvaluable opportunity to develop a series of new debiasing techniques to\naccommodate different application scenarios. Based on the framework, we propose\ntwo new DR methods, namely DR-BIAS and DR-MSE. DR-BIAS directly controls the\nbias of DR loss, while DR-MSE balances the bias and variance flexibly, which\nachieves better generalization performance. In addition, we propose a novel\ntri-level joint learning optimization method for DR-MSE in CVR prediction, and\nan efficient training algorithm correspondingly. We conduct extensive\nexperiments on both real-world and semi-synthetic datasets, which validate the\neffectiveness of our proposed methods.\n", "date": "2022-11-12", "categories": []}
{"id": "2211.05613", "title": "Adjustment formulas for learning causal steady-state models from\n  closed-loop operational data", "abstract": "  Steady-state models which have been learned from historical operational data\nmay be unfit for model-based optimization unless correlations in the training\ndata which are introduced by control are accounted for. Using recent results\nfrom work on structural dynamical causal models, we derive a formula for\nadjusting for this control confounding, enabling the estimation of a causal\nsteady-state model from closed-loop steady-state data. The formula assumes that\nthe available data have been gathered under some fixed control law. It works by\nestimating and taking into account the disturbance which the controller is\ntrying to counteract, and enables learning from data gathered under both\nfeedforward and feedback control.\n", "date": "2022-11-10", "categories": []}
{"id": "2211.05675", "title": "Causal Modeling of Soil Processes for Improved Generalization", "abstract": "  Measuring and monitoring soil organic carbon is critical for agricultural\nproductivity and for addressing critical environmental problems. Soil organic\ncarbon not only enriches nutrition in soil, but also has a gamut of co-benefits\nsuch as improving water storage and limiting physical erosion. Despite a litany\nof work in soil organic carbon estimation, current approaches do not generalize\nwell across soil conditions and management practices. We empirically show that\nexplicit modeling of cause-and-effect relationships among the soil processes\nimproves the out-of-distribution generalizability of prediction models. We\nprovide a comparative analysis of soil organic carbon estimation models where\nthe skeleton is estimated using causal discovery methods. Our framework provide\nan average improvement of 81% in test mean squared error and 52% in test mean\nabsolute error.\n", "date": "2022-11-10", "categories": []}
{"id": "2206.01802", "title": "Do-Operation Guided Causal Representation Learning with Reduced\n  Supervision Strength", "abstract": "  Causal representation learning has been proposed to encode relationships\nbetween factors presented in the high dimensional data. However, existing\nmethods suffer from merely using a large amount of labeled data and ignore the\nfact that samples generated by the same causal mechanism follow the same causal\nrelationships. In this paper, we seek to explore such information by leveraging\ndo-operation to reduce supervision strength. We propose a framework that\nimplements do-operation by swapping latent cause and effect factors encoded\nfrom a pair of inputs. Moreover, we also identify the inadequacy of existing\ncausal representation metrics empirically and theoretically and introduce new\nmetrics for better evaluation. Experiments conducted on both synthetic and real\ndatasets demonstrate the superiorities of our method compared with\nstate-of-the-art methods.\n", "date": "2022-06-03", "categories": []}
{"id": "2211.03984", "title": "Causal Discovery in Linear Latent Variable Models Subject to Measurement\n  Error", "abstract": "  We focus on causal discovery in the presence of measurement error in linear\nsystems where the mixing matrix, i.e., the matrix indicating the independent\nexogenous noise terms pertaining to the observed variables, is identified up to\npermutation and scaling of the columns. We demonstrate a somewhat surprising\nconnection between this problem and causal discovery in the presence of\nunobserved parentless causes, in the sense that there is a mapping, given by\nthe mixing matrix, between the underlying models to be inferred in these\nproblems. Consequently, any identifiability result based on the mixing matrix\nfor one model translates to an identifiability result for the other model. We\ncharacterize to what extent the causal models can be identified under a\ntwo-part faithfulness assumption. Under only the first part of the assumption\n(corresponding to the conventional definition of faithfulness), the structure\ncan be learned up to the causal ordering among an ordered grouping of the\nvariables but not all the edges across the groups can be identified. We further\nshow that if both parts of the faithfulness assumption are imposed, the\nstructure can be learned up to a more refined ordered grouping. As a result of\nthis refinement, for the latent variable model with unobserved parentless\ncauses, the structure can be identified. Based on our theoretical results, we\npropose causal structure learning methods for both models, and evaluate their\nperformance on synthetic data.\n", "date": "2022-11-08", "categories": []}
{"id": "2211.04312", "title": "From Causal Pairs to Causal Graphs", "abstract": "  Causal structure learning from observational data remains a non-trivial task\ndue to various factors such as finite sampling, unobserved confounding factors,\nand measurement errors. Constraint-based and score-based methods tend to suffer\nfrom high computational complexity due to the combinatorial nature of\nestimating the directed acyclic graph (DAG). Motivated by the `Cause-Effect\nPair' NIPS 2013 Workshop on Causality Challenge, in this paper, we take a\ndifferent approach and generate a probability distribution over all possible\ngraphs informed by the cause-effect pair features proposed in response to the\nworkshop challenge. The goal of the paper is to propose new methods based on\nthis probabilistic information and compare their performance with traditional\nand state-of-the-art approaches. Our experiments, on both synthetic and real\ndatasets, show that our proposed methods not only have statistically similar or\nbetter performances than some traditional approaches but also are\ncomputationally faster.\n", "date": "2022-11-08", "categories": []}
{"id": "2209.15177", "title": "Domain Generalization -- A Causal Perspective", "abstract": "  Machine learning models rely on various assumptions to attain high accuracy.\nOne of the preliminary assumptions of these models is the independent and\nidentical distribution, which suggests that the train and test data are sampled\nfrom the same distribution. However, this assumption seldom holds in the real\nworld due to distribution shifts. As a result models that rely on this\nassumption exhibit poor generalization capabilities. Over the recent years,\ndedicated efforts have been made to improve the generalization capabilities of\nthese models collectively known as -- \\textit{domain generalization methods}.\nThe primary idea behind these methods is to identify stable features or\nmechanisms that remain invariant across the different distributions. Many\ngeneralization approaches employ causal theories to describe invariance since\ncausality and invariance are inextricably intertwined. However, current surveys\ndeal with the causality-aware domain generalization methods on a very\nhigh-level. Furthermore, we argue that it is possible to categorize the methods\nbased on how causality is leveraged in that method and in which part of the\nmodel pipeline is it used. To this end, we categorize the causal domain\ngeneralization methods into three categories, namely, (i) Invariance via Causal\nData Augmentation methods which are applied during the data pre-processing\nstage, (ii) Invariance via Causal representation learning methods that are\nutilized during the representation learning stage, and (iii) Invariance via\nTransferring Causal mechanisms methods that are applied during the\nclassification stage of the pipeline. Furthermore, this survey includes\nin-depth insights into benchmark datasets and code repositories for domain\ngeneralization methods. We conclude the survey with insights and discussions on\nfuture directions.\n", "date": "2022-09-30", "categories": []}
{"id": "2210.11275", "title": "Causal Structural Hypothesis Testing and Data Generation Models", "abstract": "  A vast amount of expert and domain knowledge is captured by causal structural\npriors, yet there has been little research on testing such priors for\ngeneralization and data synthesis purposes. We propose a novel model\narchitecture, Causal Structural Hypothesis Testing, that can use nonparametric,\nstructural causal knowledge and approximate a causal model's functional\nrelationships using deep neural networks. We use these architectures for\ncomparing structural priors, akin to hypothesis testing, using a deliberate\n(non-random) split of training and testing data. Extensive simulations\ndemonstrate the effectiveness of out-of-distribution generalization error as a\nproxy for causal structural prior hypothesis testing and offers a statistical\nbaseline for interpreting results. We show that the variational version of the\narchitecture, Causal Structural Variational Hypothesis Testing can improve\nperformance in low SNR regimes. Due to the simplicity and low parameter count\nof the models, practitioners can test and compare structural prior hypotheses\non small dataset and use the priors with the best generalization capacity to\nsynthesize much larger, causally-informed datasets. Finally, we validate our\nmethods on a synthetic pendulum dataset, and show a use-case on a real-world\ntrauma surgery ground-level falls dataset.\n", "date": "2022-10-20", "categories": []}
{"id": "2211.03179", "title": "Personalizing Sustainable Agriculture with Causal Machine Learning", "abstract": "  To fight climate change and accommodate the increasing population, global\ncrop production has to be strengthened. To achieve the \"sustainable\nintensification\" of agriculture, transforming it from carbon emitter to carbon\nsink is a priority, and understanding the environmental impact of agricultural\nmanagement practices is a fundamental prerequisite to that. At the same time,\nthe global agricultural landscape is deeply heterogeneous, with differences in\nclimate, soil, and land use inducing variations in how agricultural systems\nrespond to farmer actions. The \"personalization\" of sustainable agriculture\nwith the provision of locally adapted management advice is thus a necessary\ncondition for the efficient uplift of green metrics, and an integral\ndevelopment in imminent policies. Here, we formulate personalized sustainable\nagriculture as a Conditional Average Treatment Effect estimation task and use\nCausal Machine Learning for tackling it. Leveraging climate data, land use\ninformation and employing Double Machine Learning, we estimate the\nheterogeneous effect of sustainable practices on the field-level Soil Organic\nCarbon content in Lithuania. We thus provide a data-driven perspective for\ntargeting sustainable practices and effectively expanding the global carbon\nsink.\n", "date": "2022-11-06", "categories": []}
{"id": "2211.03195", "title": "Evaluating Digital Tools for Sustainable Agriculture using Causal\n  Inference", "abstract": "  In contrast to the rapid digitalization of several industries, agriculture\nsuffers from low adoption of climate-smart farming tools. Even though AI-driven\ndigital agriculture can offer high-performing predictive functionalities, it\nlacks tangible quantitative evidence on its benefits to the farmers. Field\nexperiments can derive such evidence, but are often costly and time consuming.\nTo this end, we propose an observational causal inference framework for the\nempirical evaluation of the impact of digital tools on target farm performance\nindicators. This way, we can increase farmers' trust by enhancing the\ntransparency of the digital agriculture market, and in turn accelerate the\nadoption of technologies that aim to increase productivity and secure a\nsustainable and resilient agriculture against a changing climate. As a case\nstudy, we perform an empirical evaluation of a recommendation system for\noptimal cotton sowing, which was used by a farmers' cooperative during the\ngrowing season of 2021. We leverage agricultural knowledge to develop a causal\ngraph of the farm system, we use the back-door criterion to identify the impact\nof recommendations on the yield and subsequently estimate it using several\nmethods on observational data. The results show that a field sown according to\nour recommendations enjoyed a significant increase in yield (12% to 17%).\n", "date": "2022-11-06", "categories": []}
{"id": "2211.01518", "title": "Bayesian Counterfactual Mean Embeddings and Off-Policy Evaluation", "abstract": "  The counterfactual distribution models the effect of the treatment in the\nuntreated group. While most of the work focuses on the expected values of the\ntreatment effect, one may be interested in the whole counterfactual\ndistribution or other quantities associated to it. Building on the framework of\nBayesian conditional mean embeddings, we propose a Bayesian approach for\nmodeling the counterfactual distribution, which leads to quantifying the\nepistemic uncertainty about the distribution. The framework naturally extends\nto the setting where one observes multiple treatment effects (e.g. an\nintermediate effect after an interim period, and an ultimate treatment effect\nwhich is of main interest) and allows for additionally modelling uncertainty\nabout the relationship of these effects. For such goal, we present three novel\nBayesian methods to estimate the expectation of the ultimate treatment effect,\nwhen only noisy samples of the dependence between intermediate and ultimate\neffects are provided. These methods differ on the source of uncertainty\nconsidered and allow for combining two sources of data. Moreover, we generalize\nthese ideas to the off-policy evaluation framework, which can be seen as an\nextension of the counterfactual estimation problem. We empirically explore the\ncalibration of the algorithms in two different experimental settings which\nrequire data fusion, and illustrate the value of considering the uncertainty\nstemming from the two sources of data.\n", "date": "2022-11-02", "categories": []}
{"id": "2211.01575", "title": "Are Synthetic Control Weights Balancing Score?", "abstract": "  In this short note, I outline conditions under which conditioning on\nSynthetic Control (SC) weights emulates a randomized control trial where the\ntreatment status is independent of potential outcomes. Specifically, I\ndemonstrate that if there exist SC weights such that (i) the treatment effects\nare exactly identified and (ii) these weights are uniformly and cumulatively\nbounded, then SC weights are balancing scores.\n", "date": "2022-11-03", "categories": []}
{"id": "2211.01903", "title": "A Consistent Estimator for Confounding Strength", "abstract": "  Regression on observational data can fail to capture a causal relationship in\nthe presence of unobserved confounding. Confounding strength measures this\nmismatch, but estimating it requires itself additional assumptions. A common\nassumption is the independence of causal mechanisms, which relies on\nconcentration phenomena in high dimensions. While high dimensions enable the\nestimation of confounding strength, they also necessitate adapted estimators.\nIn this paper, we derive the asymptotic behavior of the confounding strength\nestimator by Janzing and Sch\\\"olkopf (2018) and show that it is generally not\nconsistent. We then use tools from random matrix theory to derive an adapted,\nconsistent estimator.\n", "date": "2022-11-03", "categories": []}
{"id": "2211.00249", "title": "Robust Direct Learning for Causal Data Fusion", "abstract": "  In the era of big data, the explosive growth of multi-source heterogeneous\ndata offers many exciting challenges and opportunities for improving the\ninference of conditional average treatment effects. In this paper, we\ninvestigate homogeneous and heterogeneous causal data fusion problems under a\ngeneral setting that allows for the presence of source-specific covariates. We\nprovide a direct learning framework for integrating multi-source data that\nseparates the treatment effect from other nuisance functions, and achieves\ndouble robustness against certain misspecification. To improve estimation\nprecision and stability, we propose a causal information-aware weighting\nfunction motivated by theoretical insights from the semiparametric efficiency\ntheory; it assigns larger weights to samples containing more causal information\nwith high interpretability. We introduce a two-step algorithm, the weighted\nmulti-source direct learner, based on constructing a pseudo-outcome and\nregressing it on covariates under a weighted least square criterion; it offers\nus a powerful tool for causal data fusion, enjoying the advantages of easy\nimplementation, double robustness and model flexibility. In simulation studies,\nwe demonstrate the effectiveness of our proposed methods in both homogeneous\nand heterogeneous causal data fusion scenarios.\n", "date": "2022-11-01", "categories": []}
{"id": "2211.00486", "title": "Causal DAG extraction from a library of books or videos/movies", "abstract": "  Determining a causal DAG (directed acyclic graph) for a problem under\nconsideration, is a major roadblock when doing Judea Pearl's Causal Inference\n(CI) in Statistics. The same problem arises when doing CI in Artificial\nIntelligence (AI) and Machine Learning (ML). As with many problems in Science,\nwe think Nature has found an effective solution to this problem. We argue that\nhuman and animal brains contain an explicit engine for doing CI, and that such\nan engine uses as input an atlas (i.e., collection) of causal DAGs. We propose\na simple algorithm for constructing such an atlas from a library of books or\nvideos/movies. We illustrate our method by applying it to a database of\nrandomly generated Tic-Tac-Toe games. The software used to generate this\nTic-Tac-Toe example is open source and available at GitHub.\n", "date": "2022-10-29", "categories": []}
{"id": "2210.16775", "title": "Nonlinear Causal Discovery via Kernel Anchor Regression", "abstract": "  Learning causal relationships is a fundamental problem in science. Anchor\nregression has been developed to address this problem for a large class of\ncausal graphical models, though the relationships between the variables are\nassumed to be linear. In this work, we tackle the nonlinear setting by\nproposing kernel anchor regression (KAR). Beyond the natural formulation using\na classic two-stage least square estimator, we also study an improved variant\nthat involves nonparametric regression in three separate stages. We provide\nconvergence results for the proposed KAR estimators and the identifiability\nconditions for KAR to learn the nonlinear structural equation models (SEM).\nExperimental results demonstrate the superior performances of the proposed KAR\nestimators over existing baselines.\n", "date": "2022-10-30", "categories": []}
{"id": "2210.09232", "title": "Confound-leakage: Confound Removal in Machine Learning Leads to Leakage", "abstract": "  Machine learning (ML) approaches to data analysis are now widely adopted in\nmany fields including epidemiology and medicine. To apply these approaches,\nconfounds must first be removed as is commonly done by featurewise removal of\ntheir variance by linear regression before applying ML. Here, we show this\ncommon approach to confound removal biases ML models, leading to misleading\nresults. Specifically, this common deconfounding approach can leak information\nsuch that what are null or moderate effects become amplified to near-perfect\nprediction when nonlinear ML approaches are subsequently applied. We identify\nand evaluate possible mechanisms for such confound-leakage and provide\npractical guidance to mitigate its negative impact. We demonstrate the\nreal-world importance of confound-leakage by analyzing a clinical dataset where\naccuracy is overestimated for predicting attention deficit hyperactivity\ndisorder (ADHD) with depression as a confound. Our results have wide-reaching\nimplications for implementation and deployment of ML workflows and beg caution\nagainst na\\\"ive use of standard confound removal approaches.\n", "date": "2022-10-17", "categories": []}
{"id": "2210.16034", "title": "A Survey on Causal Representation Learning and Future Work for Medical\n  Image Analysis", "abstract": "  Statistical machine learning algorithms have achieved state-of-the-art\nresults on benchmark datasets, outperforming humans in many tasks. However, the\nout-of-distribution data and confounder, which have an unpredictable causal\nrelationship, significantly degrade the performance of the existing models.\nCausal Representation Learning (CRL) has recently been a promising direction to\naddress the causal relationship problem in vision understanding. This survey\npresents recent advances in CRL in vision. Firstly, we introduce the basic\nconcept of causal inference. Secondly, we analyze the CRL theoretical work,\nespecially in invariant risk minimization, and the practical work in feature\nunderstanding and transfer learning. Finally, we propose a future research\ndirection in medical image analysis and CRL general theory.\n", "date": "2022-10-28", "categories": []}
{"id": "2210.15340", "title": "Sample-Specific Root Causal Inference with Latent Variables", "abstract": "  Root causal analysis seeks to identify the set of initial perturbations that\ninduce an unwanted outcome. In prior work, we defined sample-specific root\ncauses of disease using exogenous error terms that predict a diagnosis in a\nstructural equation model. We rigorously quantified predictivity using Shapley\nvalues. However, the associated algorithms for inferring root causes assume no\nlatent confounding. We relax this assumption by permitting confounding among\nthe predictors. We then introduce a corresponding procedure called Extract\nErrors with Latents (EEL) for recovering the error terms up to contamination by\nvertices on certain paths under the linear non-Gaussian acyclic model. EEL also\nidentifies the smallest sets of dependent errors for fast computation of the\nShapley values. The algorithm bypasses the hard problem of estimating the\nunderlying causal graph in both cases. Experiments highlight the superior\naccuracy and robustness of EEL relative to its predecessors.\n", "date": "2022-10-27", "categories": []}
{"id": "2210.15353", "title": "Learning Discrete Directed Acyclic Graphs via Backpropagation", "abstract": "  Recently continuous relaxations have been proposed in order to learn Directed\nAcyclic Graphs (DAGs) from data by backpropagation, instead of using\ncombinatorial optimization. However, a number of techniques for fully discrete\nbackpropagation could instead be applied. In this paper, we explore that\ndirection and propose DAG-DB, a framework for learning DAGs by Discrete\nBackpropagation. Based on the architecture of Implicit Maximum Likelihood\nEstimation [I-MLE, arXiv:2106.01798], DAG-DB adopts a probabilistic approach to\nthe problem, sampling binary adjacency matrices from an implicit probability\ndistribution. DAG-DB learns a parameter for the distribution from the loss\nincurred by each sample, performing competitively using either of two fully\ndiscrete backpropagation techniques, namely I-MLE and Straight-Through\nEstimation.\n", "date": "2022-10-27", "categories": []}
{"id": "2210.15375", "title": "Grasping Causality for the Explanation of Criticality for Automated\n  Driving", "abstract": "  The verification and validation of automated driving systems at SAE levels 4\nand 5 is a multi-faceted challenge for which classical statistical\nconsiderations become infeasible. For this, contemporary approaches suggest a\ndecomposition into scenario classes combined with statistical analysis thereof\nregarding the emergence of criticality. Unfortunately, these associational\napproaches may yield spurious inferences, or worse, fail to recognize the\ncausalities leading to critical scenarios, which are, in turn, prerequisite for\nthe development and safeguarding of automated driving systems. As to\nincorporate causal knowledge within these processes, this work introduces a\nformalization of causal queries whose answers facilitate a causal understanding\nof safety-relevant influencing factors for automated driving. This formalized\ncausal knowledge can be used to specify and implement abstract safety\nprinciples that provably reduce the criticality associated with these\ninfluencing factors. Based on Judea Pearl's causal theory, we define a causal\nrelation as a causal structure together with a context, both related to a\ndomain ontology, where the focus lies on modeling the effect of such\ninfluencing factors on criticality as measured by a suitable metric. As to\nassess modeling quality, we suggest various quantities and evaluate them on a\nsmall example. As availability and quality of data are imperative for validly\nestimating answers to the causal queries, we also discuss requirements on\nreal-world and synthetic data acquisition. We thereby contribute to\nestablishing causal considerations at the heart of the safety processes that\nare urgently needed as to ensure the safe operation of automated driving\nsystems.\n", "date": "2022-10-27", "categories": []}
{"id": "2210.14573", "title": "Learning Causal Graphs in Manufacturing Domains using Structural\n  Equation Models", "abstract": "  Many production processes are characterized by numerous and complex\ncause-and-effect relationships. Since they are only partially known they pose a\nchallenge to effective process control. In this work we present how Structural\nEquation Models can be used for deriving cause-and-effect relationships from\nthe combination of prior knowledge and process data in the manufacturing\ndomain. Compared to existing applications, we do not assume linear\nrelationships leading to more informative results.\n", "date": "2022-10-26", "categories": []}
{"id": "2210.14706", "title": "Rhino: Deep Causal Temporal Relationship Learning With History-dependent\n  Noise", "abstract": "  Discovering causal relationships between different variables from time series\ndata has been a long-standing challenge for many domains such as climate\nscience, finance, and healthcare. Given the complexity of real-world\nrelationships and the nature of observations in discrete time, causal discovery\nmethods need to consider non-linear relations between variables, instantaneous\neffects and history-dependent noise (the change of noise distribution due to\npast actions). However, previous works do not offer a solution addressing all\nthese problems together. In this paper, we propose a novel causal relationship\nlearning framework for time-series data, called Rhino, which combines vector\nauto-regression, deep learning and variational inference to model non-linear\nrelationships with instantaneous effects while allowing the noise distribution\nto be modulated by historical observations. Theoretically, we prove the\nstructural identifiability of Rhino. Our empirical results from extensive\nsynthetic experiments and two real-world benchmarks demonstrate better\ndiscovery performance compared to relevant baselines, with ablation studies\nrevealing its robustness under model misspecification.\n", "date": "2022-10-26", "categories": []}
{"id": "2210.13583", "title": "Learning Latent Structural Causal Models", "abstract": "  Causal learning has long concerned itself with the accurate recovery of\nunderlying causal mechanisms. Such causal modelling enables better explanations\nof out-of-distribution data. Prior works on causal learning assume that the\nhigh-level causal variables are given. However, in machine learning tasks, one\noften operates on low-level data like image pixels or high-dimensional vectors.\nIn such settings, the entire Structural Causal Model (SCM) -- structure,\nparameters, \\textit{and} high-level causal variables -- is unobserved and needs\nto be learnt from low-level data. We treat this problem as Bayesian inference\nof the latent SCM, given low-level data. For linear Gaussian additive noise\nSCMs, we present a tractable approximate inference method which performs joint\ninference over the causal variables, structure and parameters of the latent SCM\nfrom random, known interventions. Experiments are performed on synthetic\ndatasets and a causally generated image dataset to demonstrate the efficacy of\nour approach. We also perform image generation from unseen interventions,\nthereby verifying out of distribution generalization for the proposed causal\nmodel.\n", "date": "2022-10-24", "categories": []}
{"id": "2210.13647", "title": "Temporally Disentangled Representation Learning", "abstract": "  Recently in the field of unsupervised representation learning, strong\nidentifiability results for disentanglement of causally-related latent\nvariables have been established by exploiting certain side information, such as\nclass labels, in addition to independence. However, most existing work is\nconstrained by functional form assumptions such as independent sources or\nfurther with linear transitions, and distribution assumptions such as\nstationary, exponential family distribution. It is unknown whether the\nunderlying latent variables and their causal relations are identifiable if they\nhave arbitrary, nonparametric causal influences in between. In this work, we\nestablish the identifiability theories of nonparametric latent causal processes\nfrom their nonlinear mixtures under fixed temporal causal influences and\nanalyze how distribution changes can further benefit the disentanglement. We\npropose \\textbf{\\texttt{TDRL}}, a principled framework to recover time-delayed\nlatent causal variables and identify their relations from measured sequential\ndata under stationary environments and under different distribution shifts.\nSpecifically, the framework can factorize unknown distribution shifts into\ntransition distribution changes under fixed and time-varying latent causal\nrelations, and under observation changes in observation. Through experiments,\nwe show that time-delayed latent causal influences are reliably identified and\nthat our approach considerably outperforms existing baselines that do not\ncorrectly exploit this modular representation of changes. Our code is available\nat: \\url{https://github.com/weirayao/tdrl}.\n", "date": "2022-10-24", "categories": []}
{"id": "2210.14011", "title": "Are All Spurious Features in Natural Language Alike? An Analysis through\n  a Causal Lens", "abstract": "  The term `spurious correlations' has been used in NLP to informally denote\nany undesirable feature-label correlations. However, a correlation can be\nundesirable because (i) the feature is irrelevant to the label (e.g.\npunctuation in a review), or (ii) the feature's effect on the label depends on\nthe context (e.g. negation words in a review), which is ubiquitous in language\ntasks. In case (i), we want the model to be invariant to the feature, which is\nneither necessary nor sufficient for prediction. But in case (ii), even an\nideal model (e.g. humans) must rely on the feature, since it is necessary (but\nnot sufficient) for prediction. Therefore, a more fine-grained treatment of\nspurious features is needed to specify the desired model behavior. We formalize\nthis distinction using a causal model and probabilities of necessity and\nsufficiency, which delineates the causal relations between a feature and a\nlabel. We then show that this distinction helps explain results of existing\ndebiasing methods on different spurious features, and demystifies surprising\nresults such as the encoding of spurious features in model representations\nafter debiasing.\n", "date": "2022-10-25", "categories": []}
{"id": "2209.08885", "title": "Causal Effect Estimation with Global Probabilistic Forecasting: A Case\n  Study of the Impact of Covid-19 Lockdowns on Energy Demand", "abstract": "  The electricity industry is heavily implementing smart grid technologies to\nimprove reliability, availability, security, and efficiency. This\nimplementation needs technological advancements, the development of standards\nand regulations, as well as testing and planning. Smart grid load forecasting\nand management are critical for reducing demand volatility and improving the\nmarket mechanism that connects generators, distributors, and retailers. During\npolicy implementations or external interventions, it is necessary to analyse\nthe uncertainty of their impact on the electricity demand to enable a more\naccurate response of the system to fluctuating demand. This paper analyses the\nuncertainties of external intervention impacts on electricity demand. It\nimplements a framework that combines probabilistic and global forecasting\nmodels using a deep learning approach to estimate the causal impact\ndistribution of an intervention. The causal effect is assessed by predicting\nthe counterfactual distribution outcome for the affected instances and then\ncontrasting it to the real outcomes. We consider the impact of Covid-19\nlockdowns on energy usage as a case study to evaluate the non-uniform effect of\nthis intervention on the electricity demand distribution. We could show that\nduring the initial lockdowns in Australia and some European countries, there\nwas often a more significant decrease in the troughs than in the peaks, while\nthe mean remained almost unaffected.\n", "date": "2022-09-19", "categories": []}
{"id": "2210.11021", "title": "Independence Testing-Based Approach to Causal Discovery under\n  Measurement Error and Linear Non-Gaussian Models", "abstract": "  Causal discovery aims to recover causal structures generating the\nobservational data. Despite its success in certain problems, in many real-world\nscenarios the observed variables are not the target variables of interest, but\nthe imperfect measures of the target variables. Causal discovery under\nmeasurement error aims to recover the causal graph among unobserved target\nvariables from observations made with measurement error. We consider a specific\nformulation of the problem, where the unobserved target variables follow a\nlinear non-Gaussian acyclic model, and the measurement process follows the\nrandom measurement error model. Existing methods on this formulation rely on\nnon-scalable over-complete independent component analysis (OICA). In this work,\nwe propose the Transformed Independent Noise (TIN) condition, which checks for\nindependence between a specific linear transformation of some measured\nvariables and certain other measured variables. By leveraging the\nnon-Gaussianity and higher-order statistics of data, TIN is informative about\nthe graph structure among the unobserved target variables. By utilizing TIN,\nthe ordered group decomposition of the causal model is identifiable. In other\nwords, we could achieve what once required OICA to achieve by only conducting\nindependence tests. Experimental results on both synthetic and real-world data\ndemonstrate the effectiveness and reliability of our method.\n", "date": "2022-10-20", "categories": []}
{"id": "2210.10168", "title": "Granger causal inference on DAGs identifies genomic loci regulating\n  transcription", "abstract": "  When a dynamical system can be modeled as a sequence of observations, Granger\ncausality is a powerful approach for detecting predictive interactions between\nits variables. However, traditional Granger causal inference has limited\nutility in domains where the dynamics need to be represented as directed\nacyclic graphs (DAGs) rather than as a linear sequence, such as with cell\ndifferentiation trajectories. Here, we present GrID-Net, a framework based on\ngraph neural networks with lagged message passing for Granger causal inference\non DAG-structured systems. Our motivating application is the analysis of\nsingle-cell multimodal data to identify genomic loci that mediate the\nregulation of specific genes. To our knowledge, GrID-Net is the first\nsingle-cell analysis tool that accounts for the temporal lag between a genomic\nlocus becoming accessible and its downstream effect on a target gene's\nexpression. We applied GrID-Net on multimodal single-cell assays that profile\nchromatin accessibility (ATAC-seq) and gene expression (RNA-seq) in the same\ncell and show that it dramatically outperforms existing methods for inferring\nregulatory locus-gene links, achieving up to 71% greater agreement with\nindependent population genetics-based estimates. By extending Granger causality\nto DAG-structured dynamical systems, our work unlocks new domains for causal\nanalyses and, more specifically, opens a path towards elucidating gene\nregulatory interactions relevant to cellular differentiation and complex human\ndiseases at unprecedented scale and resolution.\n", "date": "2022-10-18", "categories": []}
{"id": "2210.10256", "title": "Causal Structure Learning with Recommendation System", "abstract": "  A fundamental challenge of recommendation systems (RS) is understanding the\ncausal dynamics underlying users' decision making. Most existing literature\naddresses this problem by using causal structures inferred from domain\nknowledge. However, there are numerous phenomenons where domain knowledge is\ninsufficient, and the causal mechanisms must be learnt from the feedback data.\nDiscovering the causal mechanism from RS feedback data is both novel and\nchallenging, since RS itself is a source of intervention that can influence\nboth the users' exposure and their willingness to interact. Also for this\nreason, most existing solutions become inappropriate since they require data\ncollected free from any RS. In this paper, we first formulate the underlying\ncausal mechanism as a causal structural model and describe a general causal\nstructure learning framework grounded in the real-world working mechanism of\nRS. The essence of our approach is to acknowledge the unknown nature of RS\nintervention. We then derive the learning objective from our framework and\npropose an augmented Lagrangian solver for efficient optimization. We conduct\nboth simulation and real-world experiments to demonstrate how our approach\ncompares favorably to existing solutions, together with the empirical analysis\nfrom sensitivity and ablation studies.\n", "date": "2022-10-19", "categories": []}
{"id": "2208.07331", "title": "Anticipating Performativity by Predicting from Predictions", "abstract": "  Predictions about people, such as their expected educational achievement or\ntheir credit risk, can be performative and shape the outcome that they aim to\npredict. Understanding the causal effect of these predictions on the eventual\noutcomes is crucial for foreseeing the implications of future predictive models\nand selecting which models to deploy. However, this causal estimation task\nposes unique challenges: model predictions are usually deterministic functions\nof input features and highly correlated with outcomes. This can make the causal\neffects of predictions on outcomes impossible to disentangle from the direct\neffect of the covariates. We study this problem through the lens of causal\nidentifiability, and despite the hardness of this problem in full generality,\nwe highlight three natural scenarios where the causal relationship between\ncovariates, predictions and outcomes can be identified from observational data:\nrandomization in predictions, overparameterization of the predictive model\ndeployed during data collection, and discrete prediction outputs. Empirically\nwe show that given our identifiability conditions hold, standard variants of\nsupervised learning that predict from predictions by treating the prediction as\nan input feature can indeed find transferable functional relationships that\nallow for conclusions about newly deployed predictive models. These positive\nresults fundamentally rely on model predictions being recorded during data\ncollection, bringing forward the importance of rethinking standard data\ncollection practices to enable progress towards a better understanding of\nsocial outcomes and performative feedback loops.\n", "date": "2022-08-15", "categories": []}
{"id": "2210.10010", "title": "Vision Paper: Causal Inference for Interpretable and Robust Machine\n  Learning in Mobility Analysis", "abstract": "  Artificial intelligence (AI) is revolutionizing many areas of our lives,\nleading a new era of technological advancement. Particularly, the\ntransportation sector would benefit from the progress in AI and advance the\ndevelopment of intelligent transportation systems. Building intelligent\ntransportation systems requires an intricate combination of artificial\nintelligence and mobility analysis. The past few years have seen rapid\ndevelopment in transportation applications using advanced deep neural networks.\nHowever, such deep neural networks are difficult to interpret and lack\nrobustness, which slows the deployment of these AI-powered algorithms in\npractice. To improve their usability, increasing research efforts have been\ndevoted to developing interpretable and robust machine learning methods, among\nwhich the causal inference approach recently gained traction as it provides\ninterpretable and actionable information. Moreover, most of these methods are\ndeveloped for image or sequential data which do not satisfy specific\nrequirements of mobility data analysis. This vision paper emphasizes research\nchallenges in deep learning-based mobility analysis that require\ninterpretability and robustness, summarizes recent developments in using causal\ninference for improving the interpretability and robustness of machine learning\nmethods, and highlights opportunities in developing causally-enabled machine\nlearning models tailored for mobility analysis. This research direction will\nmake AI in the transportation sector more interpretable and reliable, thus\ncontributing to safer, more efficient, and more sustainable future\ntransportation systems.\n", "date": "2022-10-18", "categories": []}
{"id": "2206.02013", "title": "Causal Discovery in Heterogeneous Environments Under the Sparse\n  Mechanism Shift Hypothesis", "abstract": "  Machine learning approaches commonly rely on the assumption of independent\nand identically distributed (i.i.d.) data. In reality, however, this assumption\nis almost always violated due to distribution shifts between environments.\nAlthough valuable learning signals can be provided by heterogeneous data from\nchanging distributions, it is also known that learning under arbitrary\n(adversarial) changes is impossible. Causality provides a useful framework for\nmodeling distribution shifts, since causal models encode both observational and\ninterventional distributions. In this work, we explore the sparse mechanism\nshift hypothesis, which posits that distribution shifts occur due to a small\nnumber of changing causal conditionals. Motivated by this idea, we apply it to\nlearning causal structure from heterogeneous environments, where i.i.d. data\nonly allows for learning an equivalence class of graphs without restrictive\nassumptions. We propose the Mechanism Shift Score (MSS), a score-based approach\namenable to various empirical estimators, which provably identifies the entire\ncausal structure with high probability if the sparse mechanism shift hypothesis\nholds. Empirically, we verify behavior predicted by the theory and compare\nmultiple estimators and score functions to identify the best approaches in\npractice. Compared to other methods, we show how MSS bridges a gap by both\nbeing nonparametric as well as explicitly leveraging sparse changes.\n", "date": "2022-06-04", "categories": []}
{"id": "2206.02063", "title": "Active Bayesian Causal Inference", "abstract": "  Causal discovery and causal reasoning are classically treated as separate and\nconsecutive tasks: one first infers the causal graph, and then uses it to\nestimate causal effects of interventions. However, such a two-stage approach is\nuneconomical, especially in terms of actively collected interventional data,\nsince the causal query of interest may not require a fully-specified causal\nmodel. From a Bayesian perspective, it is also unnatural, since a causal query\n(e.g., the causal graph or some causal effect) can be viewed as a latent\nquantity subject to posterior inference -- other unobserved quantities that are\nnot of direct interest (e.g., the full causal model) ought to be marginalized\nout in this process and contribute to our epistemic uncertainty. In this work,\nwe propose Active Bayesian Causal Inference (ABCI), a fully-Bayesian active\nlearning framework for integrated causal discovery and reasoning, which jointly\ninfers a posterior over causal models and queries of interest. In our approach\nto ABCI, we focus on the class of causally-sufficient, nonlinear additive noise\nmodels, which we model using Gaussian processes. We sequentially design\nexperiments that are maximally informative about our target causal query,\ncollect the corresponding interventional data, and update our beliefs to choose\nthe next experiment. Through simulations, we demonstrate that our approach is\nmore data-efficient than several baselines that only focus on learning the full\ncausal graph. This allows us to accurately learn downstream causal queries from\nfewer samples while providing well-calibrated uncertainty estimates for the\nquantities of interest.\n", "date": "2022-06-04", "categories": []}
{"id": "2210.08139", "title": "Partial Identification of Treatment Effects with Implicit Generative\n  Models", "abstract": "  We consider the problem of partial identification, the estimation of bounds\non the treatment effects from observational data. Although studied using\ndiscrete treatment variables or in specific causal graphs (e.g., instrumental\nvariables), partial identification has been recently explored using tools from\ndeep generative modeling. We propose a new method for partial identification of\naverage treatment effects(ATEs) in general causal graphs using implicit\ngenerative models comprising continuous and discrete random variables. Since\nATE with continuous treatment is generally non-regular, we leverage the partial\nderivatives of response functions to define a regular approximation of ATE, a\nquantity we call uniform average treatment derivative (UATD). We prove that our\nalgorithm converges to tight bounds on ATE in linear structural causal models\n(SCMs). For nonlinear SCMs, we empirically show that using UATD leads to\ntighter and more stable bounds than methods that directly optimize the ATE.\n", "date": "2022-10-14", "categories": []}
{"id": "2210.08203", "title": "Unit Selection: Learning Benefit Function from Finite Population Data", "abstract": "  The unit selection problem is to identify a group of individuals who are most\nlikely to exhibit a desired mode of behavior, for example, selecting\nindividuals who would respond one way if incentivized and a different way if\nnot. The unit selection problem consists of evaluation and search subproblems.\nLi and Pearl defined the \"benefit function\" to evaluate the average payoff of\nselecting a certain individual with given characteristics. The search\nsubproblem is then to design an algorithm to identify the characteristics that\nmaximize the above benefit function. The hardness of the search subproblem\narises due to the large number of characteristics available for each individual\nand the sparsity of the data available in each cell of characteristics. In this\npaper, we present a machine learning framework that uses the bounds of the\nbenefit function that are estimable from the finite population data to learn\nthe bounds of the benefit function for each cell of characteristics. Therefore,\nwe could easily obtain the characteristics that maximize the benefit function.\n", "date": "2022-10-15", "categories": []}
{"id": "2210.08453", "title": "Learning Probabilities of Causation from Finite Population Data", "abstract": "  This paper deals with the problem of learning the probabilities of causation\nof subpopulations given finite population data. The tight bounds of three basic\nprobabilities of causation, the probability of necessity and sufficiency (PNS),\nthe probability of sufficiency (PS), and the probability of necessity (PN),\nwere derived by Tian and Pearl. However, obtaining the bounds for each\nsubpopulation requires experimental and observational distributions of each\nsubpopulation, which is usually impractical to estimate given finite population\ndata. We propose a machine learning model that helps to learn the bounds of the\nprobabilities of causation for subpopulations given finite population data. We\nfurther show by a simulated study that the machine learning model is able to\nlearn the bounds of PNS for 32768 subpopulations with only knowing roughly 500\nof them from the finite population data.\n", "date": "2022-10-16", "categories": []}
{"id": "2210.08874", "title": "Probabilities of Causation: Role of Observational Data", "abstract": "  Probabilities of causation play a crucial role in modern decision-making.\nPearl defined three binary probabilities of causation, the probability of\nnecessity and sufficiency (PNS), the probability of sufficiency (PS), and the\nprobability of necessity (PN). These probabilities were then bounded by Tian\nand Pearl using a combination of experimental and observational data. However,\nobservational data are not always available in practice; in such a case, Tian\nand Pearl's Theorem provided valid but less effective bounds using pure\nexperimental data. In this paper, we discuss the conditions that observational\ndata are worth considering to improve the quality of the bounds. More\nspecifically, we defined the expected improvement of the bounds by assuming the\nobservational distributions are uniformly distributed on their feasible\ninterval. We further applied the proposed theorems to the unit selection\nproblem defined by Li and Pearl.\n", "date": "2022-10-17", "categories": []}
{"id": "2210.05816", "title": "Finding and Listing Front-door Adjustment Sets", "abstract": "  Identifying the effects of new interventions from data is a significant\nchallenge found across a wide range of the empirical sciences. A well-known\nstrategy for identifying such effects is Pearl's front-door (FD) criterion\n(Pearl, 1995). The definition of the FD criterion is declarative, only allowing\none to decide whether a specific set satisfies the criterion. In this paper, we\npresent algorithms for finding and enumerating possible sets satisfying the FD\ncriterion in a given causal diagram. These results are useful in facilitating\nthe practical applications of the FD criterion for causal effects estimation\nand helping scientists to select estimands with desired properties, e.g., based\non cost, feasibility of measurement, or statistical power.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.07518", "title": "Counterfactual Neural Temporal Point Process for Estimating Causal\n  Influence of Misinformation on Social Media", "abstract": "  Recent years have witnessed the rise of misinformation campaigns that spread\nspecific narratives on social media to manipulate public opinions on different\nareas, such as politics and healthcare. Consequently, an effective and\nefficient automatic methodology to estimate the influence of the misinformation\non user beliefs and activities is needed. However, existing works on\nmisinformation impact estimation either rely on small-scale psychological\nexperiments or can only discover the correlation between user behaviour and\nmisinformation. To address these issues, in this paper, we build up a causal\nframework that model the causal effect of misinformation from the perspective\nof temporal point process. To adapt the large-scale data, we design an\nefficient yet precise way to estimate the Individual Treatment Effect(ITE) via\nneural temporal point process and gaussian mixture models. Extensive\nexperiments on synthetic dataset verify the effectiveness and efficiency of our\nmodel. We further apply our model on a real-world dataset of social media posts\nand engagements about COVID-19 vaccines. The experimental results indicate that\nour model recognized identifiable causal effect of misinformation that hurts\npeople's subjective emotions toward the vaccines.\n", "date": "2022-10-14", "categories": []}
{"id": "2210.07536", "title": "A Reinforcement Learning Approach to Estimating Long-term Treatment\n  Effects", "abstract": "  Randomized experiments (a.k.a. A/B tests) are a powerful tool for estimating\ntreatment effects, to inform decisions making in business, healthcare and other\napplications. In many problems, the treatment has a lasting effect that evolves\nover time. A limitation with randomized experiments is that they do not easily\nextend to measure long-term effects, since running long experiments is\ntime-consuming and expensive. In this paper, we take a reinforcement learning\n(RL) approach that estimates the average reward in a Markov process. Motivated\nby real-world scenarios where the observed state transition is nonstationary,\nwe develop a new algorithm for a class of nonstationary problems, and\ndemonstrate promising results in two synthetic datasets and one online store\ndataset.\n", "date": "2022-10-14", "categories": []}
{"id": "2210.04958", "title": "Mining Causality from Continuous-time Dynamics Models: An Application to\n  Tsunami Forecasting", "abstract": "  Continuous-time dynamics models, such as neural ordinary differential\nequations, have enabled the modeling of underlying dynamics in time-series data\nand accurate forecasting. However, parameterization of dynamics using a neural\nnetwork makes it difficult for humans to identify causal structures in the\ndata. In consequence, this opaqueness hinders the use of these models in the\ndomains where capturing causal relationships carries the same importance as\naccurate predictions, e.g., tsunami forecasting. In this paper, we address this\nchallenge by proposing a mechanism for mining causal structures from\ncontinuous-time models. We train models to capture the causal structure by\nenforcing sparsity in the weights of the input layers of the dynamics models.\nWe first verify the effectiveness of our method in the scenario where the exact\ncausal-structures of time-series are known as a priori. We next apply our\nmethod to a real-world problem, namely tsunami forecasting, where the exact\ncausal-structures are difficult to characterize. Experimental results show that\nthe proposed method is effective in learning physically-consistent causal\nrelationships while achieving high forecasting accuracy.\n", "date": "2022-10-10", "categories": []}
{"id": "2210.06594", "title": "Sample Constrained Treatment Effect Estimation", "abstract": "  Treatment effect estimation is a fundamental problem in causal inference. We\nfocus on designing efficient randomized controlled trials, to accurately\nestimate the effect of some treatment on a population of $n$ individuals. In\nparticular, we study sample-constrained treatment effect estimation, where we\nmust select a subset of $s \\ll n$ individuals from the population to experiment\non. This subset must be further partitioned into treatment and control groups.\nAlgorithms for partitioning the entire population into treatment and control\ngroups, or for choosing a single representative subset, have been well-studied.\nThe key challenge in our setting is jointly choosing a representative subset\nand a partition for that set.\n  We focus on both individual and average treatment effect estimation, under a\nlinear effects model. We give provably efficient experimental designs and\ncorresponding estimators, by identifying connections to discrepancy\nminimization and leverage-score-based sampling used in randomized numerical\nlinear algebra. Our theoretical results obtain a smooth transition to known\nguarantees when $s$ equals the population size. We also empirically demonstrate\nthe performance of our algorithms.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.06610", "title": "A Neural Mean Embedding Approach for Back-door and Front-door Adjustment", "abstract": "  We consider the estimation of average and counterfactual treatment effects,\nunder two settings: back-door adjustment and front-door adjustment. The goal in\nboth cases is to recover the treatment effect without having an access to a\nhidden confounder. This objective is attained by first estimating the\nconditional mean of the desired outcome variable given relevant covariates (the\n\"first stage\" regression), and then taking the (conditional) expectation of\nthis function as a \"second stage\" procedure. We propose to compute these\nconditional expectations directly using a regression function to the learned\ninput features of the first stage, thus avoiding the need for sampling or\ndensity estimation. All functions and features (and in particular, the output\nfeatures in the second stage) are neural networks learned adaptively from data,\nwith the sole requirement that the final layer of the first stage should be\nlinear. The proposed method is shown to converge to the true causal parameter,\nand outperforms the recent state-of-the-art methods on challenging causal\nbenchmarks, including settings involving high-dimensional image data.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.07138", "title": "Counterfactual Multihop QA: A Cause-Effect Approach for Reducing\n  Disconnected Reasoning", "abstract": "  Multi-hop QA requires reasoning over multiple supporting facts to answer the\nquestion. However, the existing QA models always rely on shortcuts, e.g.,\nproviding the true answer by only one fact, rather than multi-hop reasoning,\nwhich is referred as $\\textit{disconnected reasoning}$ problem. To alleviate\nthis issue, we propose a novel counterfactual multihop QA, a causal-effect\napproach that enables to reduce the disconnected reasoning. It builds upon\nexplicitly modeling of causality: 1) the direct causal effects of disconnected\nreasoning and 2) the causal effect of true multi-hop reasoning from the total\ncausal effect. With the causal graph, a counterfactual inference is proposed to\ndisentangle the disconnected reasoning from the total causal effect, which\nprovides us a new perspective and technology to learn a QA model that exploits\nthe true multi-hop reasoning instead of shortcuts. Extensive experiments have\nconducted on the benchmark HotpotQA dataset, which demonstrate that the\nproposed method can achieve notable improvement on reducing disconnected\nreasoning. For example, our method achieves 5.8% higher points of its Supp$_s$\nscore on HotpotQA through true multihop reasoning. The code is available at\nsupplementary material.\n", "date": "2022-10-13", "categories": []}
{"id": "2210.05993", "title": "Feasible and Desirable Counterfactual Generation by Preserving Human\n  Defined Constraints", "abstract": "  We present a human-in-the-loop approach to generate counterfactual (CF)\nexplanations that preserve global and local feasibility constraints. Global\nfeasibility constraints refer to the causal constraints that are necessary for\ngenerating actionable CF explanation. Assuming a domain expert with knowledge\non unary and binary causal constraints, our approach efficiently employs this\nknowledge to generate CF explanation by rejecting gradient steps that violate\nthese constraints. Local feasibility constraints encode end-user's constraints\nfor generating desirable CF explanation. We extract these constraints from the\nend-user of the model and exploit them during CF generation via user-defined\ndistance metric. Through user studies, we demonstrate that incorporating causal\nconstraints during CF generation results in significantly better explanations\nin terms of feasibility and desirability for participants. Adopting local and\nglobal feasibility constraints simultaneously, although improves user\nsatisfaction, does not significantly improve desirability of the participants\ncompared to only incorporating global constraints.\n", "date": "2022-10-12", "categories": []}
{"id": "2210.06183", "title": "Transfer Learning on Heterogeneous Feature Spaces for Treatment Effects\n  Estimation", "abstract": "  Consider the problem of improving the estimation of conditional average\ntreatment effects (CATE) for a target domain of interest by leveraging related\ninformation from a source domain with a different feature space. This\nheterogeneous transfer learning problem for CATE estimation is ubiquitous in\nareas such as healthcare where we may wish to evaluate the effectiveness of a\ntreatment for a new patient population for which different clinical covariates\nand limited data are available. In this paper, we address this problem by\nintroducing several building blocks that use representation learning to handle\nthe heterogeneous feature spaces and a flexible multi-task architecture with\nshared and private layers to transfer information between potential outcome\nfunctions across domains. Then, we show how these building blocks can be used\nto recover transfer learning equivalents of the standard CATE learners. On a\nnew semi-synthetic data simulation benchmark for heterogeneous transfer\nlearning we not only demonstrate performance improvements of our heterogeneous\ntransfer causal effect learners across datasets, but also provide insights into\nthe differences between these learners from a transfer perspective.\n", "date": "2022-10-08", "categories": []}
{"id": "2210.05027", "title": "Probabilities of Causation: Adequate Size of Experimental and\n  Observational Samples", "abstract": "  The probabilities of causation are commonly used to solve decision-making\nproblems. Tian and Pearl derived sharp bounds for the probability of necessity\nand sufficiency (PNS), the probability of sufficiency (PS), and the probability\nof necessity (PN) using experimental and observational data. The assumption is\nthat one is in possession of a large enough sample to permit an accurate\nestimation of the experimental and observational distributions. In this study,\nwe present a method for determining the sample size needed for such estimation,\nwhen a given confidence interval (CI) is specified. We further show by\nsimulation that the proposed sample size delivered stable estimations of the\nbounds of PNS.\n", "date": "2022-10-10", "categories": []}
{"id": "2210.05030", "title": "Unit Selection: Case Study and Comparison with A/B Test Heuristic", "abstract": "  The unit selection problem defined by Li and Pearl identifies individuals who\nhave desired counterfactual behavior patterns, for example, individuals who\nwould respond positively if encouraged and would not otherwise. Li and Pearl\nshowed by example that their unit selection model is beyond the A/B test\nheuristics. In this paper, we reveal the essence of the A/B test heuristics,\nwhich are exceptional cases of the benefit function defined by Li and Pearl.\nFurthermore, We provided more simulated use cases of Li-Pearl's unit selection\nmodel to help decision-makers apply their model correctly, explaining that A/B\ntest heuristics are generally problematic.\n", "date": "2022-10-10", "categories": []}
{"id": "2210.05446", "title": "Disentangling Causal Effects from Sets of Interventions in the Presence\n  of Unobserved Confounders", "abstract": "  The ability to answer causal questions is crucial in many domains, as causal\ninference allows one to understand the impact of interventions. In many\napplications, only a single intervention is possible at a given time. However,\nin some important areas, multiple interventions are concurrently applied.\nDisentangling the effects of single interventions from jointly applied\ninterventions is a challenging task -- especially as simultaneously applied\ninterventions can interact. This problem is made harder still by unobserved\nconfounders, which influence both treatments and outcome. We address this\nchallenge by aiming to learn the effect of a single-intervention from both\nobservational data and sets of interventions. We prove that this is not\ngenerally possible, but provide identification proofs demonstrating that it can\nbe achieved under non-linear continuous structural causal models with additive,\nmultivariate Gaussian noise -- even when unobserved confounders are present.\nImportantly, we show how to incorporate observed covariates and learn\nheterogeneous treatment effects. Based on the identifiability proofs, we\nprovide an algorithm that learns the causal model parameters by pooling data\nfrom different regimes and jointly maximizing the combined likelihood. The\neffectiveness of our method is empirically demonstrated on both synthetic and\nreal-world data.\n", "date": "2022-10-11", "categories": []}
{"id": "2206.00416", "title": "In the Eye of the Beholder: Robust Prediction with Causal User Modeling", "abstract": "  Accurately predicting the relevance of items to users is crucial to the\nsuccess of many social platforms. Conventional approaches train models on\nlogged historical data; but recommendation systems, media services, and online\nmarketplaces all exhibit a constant influx of new content -- making relevancy a\nmoving target, to which standard predictive models are not robust. In this\npaper, we propose a learning framework for relevance prediction that is robust\nto changes in the data distribution. Our key observation is that robustness can\nbe obtained by accounting for how users causally perceive the environment. We\nmodel users as boundedly-rational decision makers whose causal beliefs are\nencoded by a causal graph, and show how minimal information regarding the graph\ncan be used to contend with distributional changes. Experiments in multiple\nsettings demonstrate the effectiveness of our approach.\n", "date": "2022-06-01", "categories": []}
{"id": "2206.07824", "title": "Large-Scale Differentiable Causal Discovery of Factor Graphs", "abstract": "  A common theme in causal inference is learning causal relationships between\nobserved variables, also known as causal discovery. This is usually a daunting\ntask, given the large number of candidate causal graphs and the combinatorial\nnature of the search space. Perhaps for this reason, most research has so far\nfocused on relatively small causal graphs, with up to hundreds of nodes.\nHowever, recent advances in fields like biology enable generating experimental\ndata sets with thousands of interventions followed by rich profiling of\nthousands of variables, raising the opportunity and urgent need for large\ncausal graph models. Here, we introduce the notion of factor directed acyclic\ngraphs (f-DAGs) as a way to restrict the search space to non-linear low-rank\ncausal interaction models. Combining this novel structural assumption with\nrecent advances that bridge the gap between causal discovery and continuous\noptimization, we achieve causal discovery on thousands of variables.\nAdditionally, as a model for the impact of statistical noise on this estimation\nprocedure, we study a model of edge perturbations of the f-DAG skeleton based\non random graphs and quantify the effect of such perturbations on the f-DAG\nrank. This theoretical analysis suggests that the set of candidate f-DAGs is\nmuch smaller than the whole DAG space and thus may be more suitable as a search\nspace in the high-dimensional regime where the underlying skeleton is hard to\nassess. We propose Differentiable Causal Discovery of Factor Graphs (DCD-FG), a\nscalable implementation of -DAG constrained causal discovery for\nhigh-dimensional interventional data. DCD-FG uses a Gaussian non-linear\nlow-rank structural equation model and shows significant improvements compared\nto state-of-the-art methods in both simulations as well as a recent large-scale\nsingle-cell RNA sequencing data set with hundreds of genetic interventions.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.15374", "title": "Verification and search algorithms for causal DAGs", "abstract": "  We study two problems related to recovering causal graphs from interventional\ndata: (i) $\\textit{verification}$, where the task is to check if a purported\ncausal graph is correct, and (ii) $\\textit{search}$, where the task is to\nrecover the correct causal graph. For both, we wish to minimize the number of\ninterventions performed. For the first problem, we give a characterization of a\nminimal sized set of atomic interventions that is necessary and sufficient to\ncheck the correctness of a claimed causal graph. Our characterization uses the\nnotion of $\\textit{covered edges}$, which enables us to obtain simple proofs\nand also easily reason about earlier known results. We also generalize our\nresults to the settings of bounded size interventions and node-dependent\ninterventional costs. For all the above settings, we provide the first known\nprovable algorithms for efficiently computing (near)-optimal verifying sets on\ngeneral graphs. For the second problem, we give a simple adaptive algorithm\nbased on graph separators that produces an atomic intervention set which fully\norients any essential graph while using $\\mathcal{O}(\\log n)$ times the optimal\nnumber of interventions needed to $\\textit{verify}$ (verifying size) the\nunderlying DAG on $n$ vertices. This approximation is tight as $\\textit{any}$\nsearch algorithm on an essential line graph has worst case approximation ratio\nof $\\Omega(\\log n)$ with respect to the verifying size. With bounded size\ninterventions, each of size $\\leq k$, our algorithm gives an $\\mathcal{O}(\\log\nn \\cdot \\log k)$ factor approximation. Our result is the first known algorithm\nthat gives a non-trivial approximation guarantee to the verifying size on\ngeneral unweighted graphs and with bounded size interventions.\n", "date": "2022-06-30", "categories": []}
{"id": "2210.03242", "title": "Disentangling Mixtures of Unknown Causal Interventions", "abstract": "  In many real-world scenarios, such as gene knockout experiments, targeted\ninterventions are often accompanied by unknown interventions at off-target\nsites. Moreover, different units can get randomly exposed to different unknown\ninterventions, thereby creating a mixture of interventions. Identifying\ndifferent components of this mixture can be very valuable in some applications.\nMotivated by such situations, in this work, we study the problem of identifying\nall components present in a mixture of interventions on a given causal Bayesian\nNetwork. We construct an example to show that, in general, the components are\nnot identifiable from the mixture distribution. Next, assuming that the given\nnetwork satisfies a positivity condition, we show that, if the set of mixture\ncomponents satisfy a mild exclusion assumption, then they can be uniquely\nidentified. Our proof gives an efficient algorithm to recover these targets\nfrom the exponentially large search space of possible targets. In the more\nrealistic scenario, where distributions are given via finitely many samples, we\nconduct a simulation study to analyze the performance of an algorithm derived\nfrom our identifiability proof.\n", "date": "2022-10-01", "categories": []}
{"id": "2210.01798", "title": "Latent Hierarchical Causal Structure Discovery with Rank Constraints", "abstract": "  Most causal discovery procedures assume that there are no latent confounders\nin the system, which is often violated in real-world problems. In this paper,\nwe consider a challenging scenario for causal structure identification, where\nsome variables are latent and they form a hierarchical graph structure to\ngenerate the measured variables; the children of latent variables may still be\nlatent and only leaf nodes are measured, and moreover, there can be multiple\npaths between every pair of variables (i.e., it is beyond tree structure). We\npropose an estimation procedure that can efficiently locate latent variables,\ndetermine their cardinalities, and identify the latent hierarchical structure,\nby leveraging rank deficiency constraints over the measured variables. We show\nthat the proposed algorithm can find the correct Markov equivalence class of\nthe whole graph asymptotically under proper restrictions on the graph\nstructure.\n", "date": "2022-10-01", "categories": []}
{"id": "2210.00035", "title": "Neural Causal Models for Counterfactual Identification and Estimation", "abstract": "  Evaluating hypothetical statements about how the world would be had a\ndifferent course of action been taken is arguably one key capability expected\nfrom modern AI systems. Counterfactual reasoning underpins discussions in\nfairness, the determination of blame and responsibility, credit assignment, and\nregret. In this paper, we study the evaluation of counterfactual statements\nthrough neural models. Specifically, we tackle two causal problems required to\nmake such evaluations, i.e., counterfactual identification and estimation from\nan arbitrary combination of observational and experimental data. First, we show\nthat neural causal models (NCMs) are expressive enough and encode the\nstructural constraints necessary for performing counterfactual reasoning.\nSecond, we develop an algorithm for simultaneously identifying and estimating\ncounterfactual distributions. We show that this algorithm is sound and complete\nfor deciding counterfactual identification in general settings. Third,\nconsidering the practical implications of these results, we introduce a new\nstrategy for modeling NCMs using generative adversarial networks. Simulations\ncorroborate with the proposed methodology.\n", "date": "2022-09-30", "categories": []}
{"id": "2209.14975", "title": "Causal Inference via Nonlinear Variable Decorrelation for Healthcare\n  Applications", "abstract": "  Causal inference and model interpretability research are gaining increasing\nattention, especially in the domains of healthcare and bioinformatics. Despite\nrecent successes in this field, decorrelating features under nonlinear\nenvironments with human interpretable representations has not been adequately\ninvestigated. To address this issue, we introduce a novel method with a\nvariable decorrelation regularizer to handle both linear and nonlinear\nconfounding. Moreover, we employ association rules as new representations using\nassociation rule mining based on the original features to further proximate\nhuman decision patterns to increase model interpretability. Extensive\nexperiments are conducted on four healthcare datasets (one synthetically\ngenerated and three real-world collections on different diseases). Quantitative\nresults in comparison to baseline approaches on parameter estimation and\ncausality computation indicate the model's superior performance. Furthermore,\nexpert evaluation given by healthcare professionals validates the effectiveness\nand interpretability of the proposed model.\n", "date": "2022-09-29", "categories": []}
{"id": "2209.14279", "title": "Causal Proxy Models for Concept-Based Model Explanations", "abstract": "  Explainability methods for NLP systems encounter a version of the fundamental\nproblem of causal inference: for a given ground-truth input text, we never\ntruly observe the counterfactual texts necessary for isolating the causal\neffects of model representations on outputs. In response, many explainability\nmethods make no use of counterfactual texts, assuming they will be unavailable.\nIn this paper, we show that robust causal explainability methods can be created\nusing approximate counterfactuals, which can be written by humans to\napproximate a specific counterfactual or simply sampled using metadata-guided\nheuristics. The core of our proposal is the Causal Proxy Model (CPM). A CPM\nexplains a black-box model $\\mathcal{N}$ because it is trained to have the same\nactual input/output behavior as $\\mathcal{N}$ while creating neural\nrepresentations that can be intervened upon to simulate the counterfactual\ninput/output behavior of $\\mathcal{N}$. Furthermore, we show that the best CPM\nfor $\\mathcal{N}$ performs comparably to $\\mathcal{N}$ in making factual\npredictions, which means that the CPM can simply replace $\\mathcal{N}$, leading\nto more explainable deployed models. Our code is available at\nhttps://github.com/frankaging/Causal-Proxy-Model.\n", "date": "2022-09-28", "categories": []}
{"id": "2209.09018", "title": "A Causal Intervention Scheme for Semantic Segmentation of Quasi-periodic\n  Cardiovascular Signals", "abstract": "  Precise segmentation is a vital first step to analyze semantic information of\ncardiac cycle and capture anomaly with cardiovascular signals. However, in the\nfield of deep semantic segmentation, inference is often unilaterally confounded\nby the individual attribute of data. Towards cardiovascular signals,\nquasi-periodicity is the essential characteristic to be learned, regarded as\nthe synthesize of the attributes of morphology (Am) and rhythm (Ar). Our key\ninsight is to suppress the over-dependence on Am or Ar while the generation\nprocess of deep representations. To address this issue, we establish a\nstructural causal model as the foundation to customize the intervention\napproaches on Am and Ar, respectively. In this paper, we propose contrastive\ncausal intervention (CCI) to form a novel training paradigm under a frame-level\ncontrastive framework. The intervention can eliminate the implicit statistical\nbias brought by the single attribute and lead to more objective\nrepresentations. We conduct comprehensive experiments with the controlled\ncondition for QRS location and heart sound segmentation. The final results\nindicate that our approach can evidently improve the performance by up to 0.41%\nfor QRS location and 2.73% for heart sound segmentation. The efficiency of the\nproposed method is generalized to multiple databases and noisy signals.\n", "date": "2022-09-19", "categories": []}
{"id": "2209.06998", "title": "Stochastic Tree Ensembles for Estimating Heterogeneous Effects", "abstract": "  Determining subgroups that respond especially well (or poorly) to specific\ninterventions (medical or policy) requires new supervised learning methods\ntailored specifically for causal inference. Bayesian Causal Forest (BCF) is a\nrecent method that has been documented to perform well on data generating\nprocesses with strong confounding of the sort that is plausible in many\napplications. This paper develops a novel algorithm for fitting the BCF model,\nwhich is more efficient than the previously available Gibbs sampler. The new\nalgorithm can be used to initialize independent chains of the existing Gibbs\nsampler leading to better posterior exploration and coverage of the associated\ninterval estimates in simulation studies. The new algorithm is compared to\nrelated approaches via simulation studies as well as an empirical analysis.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.06262", "title": "Unifying Causal Inference and Reinforcement Learning using Higher-Order\n  Category Theory", "abstract": "  We present a unified formalism for structure discovery of causal models and\npredictive state representation (PSR) models in reinforcement learning (RL)\nusing higher-order category theory. Specifically, we model structure discovery\nin both settings using simplicial objects, contravariant functors from the\ncategory of ordinal numbers into any category. Fragments of causal models that\nare equivalent under conditional independence -- defined as causal horns -- as\nwell as subsequences of potential tests in a predictive state representation --\ndefined as predictive horns -- are both special cases of horns of a simplicial\nobject, subsets resulting from the removal of the interior and the face\nopposite a particular vertex. Latent structure discovery in both settings\ninvolve the same fundamental mathematical problem of finding extensions of\nhorns of simplicial objects through solving lifting problems in commutative\ndiagrams, and exploiting weak homotopies that define higher-order symmetries.\nSolutions to the problem of filling \"inner\" vs \"outer\" horns leads to various\nnotions of higher-order categories, including weak Kan complexes and\nquasicategories. We define the abstract problem of structure discovery in both\nsettings in terms of adjoint functors between the category of universal causal\nmodels or universal decision models and its simplicial object representation.\n", "date": "2022-09-13", "categories": []}
{"id": "2209.06367", "title": "A Review and Roadmap of Deep Learning Causal Discovery in Different\n  Variable Paradigms", "abstract": "  Understanding causality helps to structure interventions to achieve specific\ngoals and enables predictions under interventions. With the growing importance\nof learning causal relationships, causal discovery tasks have transitioned from\nusing traditional methods to infer potential causal structures from\nobservational data to the field of pattern recognition involved in deep\nlearning. The rapid accumulation of massive data promotes the emergence of\ncausal search methods with brilliant scalability. Existing summaries of causal\ndiscovery methods mainly focus on traditional methods based on constraints,\nscores and FCMs, there is a lack of perfect sorting and elaboration for deep\nlearning-based methods, also lacking some considers and exploration of causal\ndiscovery methods from the perspective of variable paradigms. Therefore, we\ndivide the possible causal discovery tasks into three types according to the\nvariable paradigm and give the definitions of the three tasks respectively,\ndefine and instantiate the relevant datasets for each task and the final causal\nmodel constructed at the same time, then reviews the main existing causal\ndiscovery methods for different tasks. Finally, we propose some roadmaps from\ndifferent perspectives for the current research gaps in the field of causal\ndiscovery and point out future research directions.\n", "date": "2022-09-14", "categories": []}
{"id": "2209.06790", "title": "Drawing Causal Inferences About Performance Effects in NLP", "abstract": "  This article emphasizes that NLP as a science seeks to make inferences about\nthe performance effects that result from applying one method (compared to\nanother method) in the processing of natural language. Yet NLP research in\npractice usually does not achieve this goal: In NLP research articles,\ntypically only a few models are compared. Each model results from a specific\nprocedural pipeline (here named processing system) that is composed of a\nspecific collection of methods that are used in preprocessing, pretraining,\nhyperparameter tuning, and training on the target task. To make generalizing\ninferences about the performance effect that is caused by applying some method\nA vs. another method B, it is not sufficient to compare a few specific models\nthat are produced by a few specific (probably incomparable) processing systems.\nRather, the following procedure would allow drawing inferences about methods'\nperformance effects: (1) A population of processing systems that researchers\nseek to infer to has to be defined. (2) A random sample of processing systems\nfrom this population is drawn. (The drawn processing systems in the sample will\nvary with regard to the methods they apply along their procedural pipelines and\nalso will vary regarding the compositions of their training and test data sets\nused for training and evaluation.) (3) Each processing system is applied once\nwith method A and once with method B. (4) Based on the sample of applied\nprocessing systems, the expected generalization errors of method A and method B\nare approximated. (5) The difference between the expected generalization errors\nof method A and method B is the estimated average treatment effect due to\napplying method A compared to method B in the population of processing systems.\n", "date": "2022-09-14", "categories": []}
{"id": "2209.03427", "title": "Causal discovery for time series with latent confounders", "abstract": "  Reconstructing the causal relationships behind the phenomena we observe is a\nfundamental challenge in all areas of science. Discovering causal relationships\nthrough experiments is often infeasible, unethical, or expensive in complex\nsystems. However, increases in computational power allow us to process the\never-growing amount of data that modern science generates, leading to an\nemerging interest in the causal discovery problem from observational data. This\nwork evaluates the LPCMCI algorithm, which aims to find generators compatible\nwith a multi-dimensional, highly autocorrelated time series while some\nvariables are unobserved. We find that LPCMCI performs much better than a\nrandom algorithm mimicking not knowing anything but is still far from optimal\ndetection. Furthermore, LPCMCI performs best on auto-dependencies, then\ncontemporaneous dependencies, and struggles most with lagged dependencies. The\nsource code of this project is available online.\n", "date": "2022-09-07", "categories": []}
{"id": "2209.01547", "title": "Conditional Independence Testing via Latent Representation Learning", "abstract": "  Detecting conditional independencies plays a key role in several statistical\nand machine learning tasks, especially in causal discovery algorithms. In this\nstudy, we introduce LCIT (Latent representation based Conditional Independence\nTest)-a novel non-parametric method for conditional independence testing based\non representation learning. Our main contribution involves proposing a\ngenerative framework in which to test for the independence between X and Y\ngiven Z, we first learn to infer the latent representations of target variables\nX and Y that contain no information about the conditioning variable Z. The\nlatent variables are then investigated for any significant remaining\ndependencies, which can be performed using the conventional partial correlation\ntest. The empirical evaluations show that LCIT outperforms several\nstate-of-the-art baselines consistently under different evaluation metrics, and\nis able to adapt really well to both non-linear and high-dimensional settings\non a diverse collection of synthetic and real data sets.\n", "date": "2022-09-04", "categories": []}
{"id": "2209.01956", "title": "Moderately-Balanced Representation Learning for Treatment Effects with\n  Orthogonality Information", "abstract": "  Estimating the average treatment effect (ATE) from observational data is\nchallenging due to selection bias. Existing works mainly tackle this challenge\nin two ways. Some researchers propose constructing a score function that\nsatisfies the orthogonal condition, which guarantees that the established ATE\nestimator is \"orthogonal\" to be more robust. The others explore representation\nlearning models to achieve a balanced representation between the treated and\nthe controlled groups. However, existing studies fail to 1) discriminate\ntreated units from controlled ones in the representation space to avoid the\nover-balanced issue; 2) fully utilize the \"orthogonality information\". In this\npaper, we propose a moderately-balanced representation learning (MBRL)\nframework based on recent covariates balanced representation learning methods\nand orthogonal machine learning theory. This framework protects the\nrepresentation from being over-balanced via multi-task learning.\nSimultaneously, MBRL incorporates the noise orthogonality information in the\ntraining and validation stages to achieve a better ATE estimation. The\ncomprehensive experiments on benchmark and simulated datasets show the\nsuperiority and robustness of our method on treatment effect estimations\ncompared with existing state-of-the-art methods.\n", "date": "2022-09-05", "categories": []}
{"id": "2208.12610", "title": "NeurIPS Competition Instructions and Guide: Causal Insights for Learning\n  Paths in Education", "abstract": "  In this competition, participants will address two fundamental causal\nchallenges in machine learning in the context of education using time-series\ndata. The first is to identify the causal relationships between different\nconstructs, where a construct is defined as the smallest element of learning.\nThe second challenge is to predict the impact of learning one construct on the\nability to answer questions on other constructs. Addressing these challenges\nwill enable optimisation of students' knowledge acquisition, which can be\ndeployed in a real edtech solution impacting millions of students. Participants\nwill run these tasks in an idealised environment with synthetic data and a\nreal-world scenario with evaluation data collected from a series of A/B tests.\n", "date": "2022-08-17", "categories": []}
{"id": "2209.00115", "title": "An evaluation framework for comparing causal inference models", "abstract": "  Estimation of causal effects is the core objective of many scientific\ndisciplines. However, it remains a challenging task, especially when the\neffects are estimated from observational data. Recently, several promising\nmachine learning models have been proposed for causal effect estimation. The\nevaluation of these models has been based on the mean values of the error of\nthe Average Treatment Effect (ATE) as well as of the Precision in Estimation of\nHeterogeneous Effect (PEHE). In this paper, we propose to complement the\nevaluation of causal inference models using concrete statistical evidence,\nincluding the performance profiles of Dolan and Mor{\\'e}, as well as\nnon-parametric and post-hoc statistical tests. The main motivation behind this\napproach is the elimination of the influence of a small number of instances or\nsimulation on the benchmarking process, which in some cases dominate the\nresults. We use the proposed evaluation methodology to compare several\nstate-of-the-art causal effect estimation models.\n", "date": "2022-08-31", "categories": []}
{"id": "2209.00137", "title": "Partial Counterfactual Identification for Infinite Horizon Partially\n  Observable Markov Decision Process", "abstract": "  This paper investigates the problem of bounding possible output from a\ncounterfactual query given a set of observational data. While various works of\nliterature have described methodologies to generate efficient algorithms that\nprovide an optimal bound for the counterfactual query, all of them assume a\nfinite-horizon causal diagram. This paper aims to extend the previous work by\nmodifying Q-learning algorithm to provide informative bounds of a causal query\ngiven an infinite-horizon causal diagram. Through simulations, our algorithms\nare proven to perform better compared to existing algorithm.\n", "date": "2022-08-31", "categories": []}
{"id": "2208.12809", "title": "Incrementality Bidding and Attribution", "abstract": "  The causal effect of showing an ad to a potential customer versus not,\ncommonly referred to as \"incrementality\", is the fundamental question of\nadvertising effectiveness. In digital advertising three major puzzle pieces are\ncentral to rigorously quantifying advertising incrementality: ad\nbuying/bidding/pricing, attribution, and experimentation. Building on the\nfoundations of machine learning and causal econometrics, we propose a\nmethodology that unifies these three concepts into a computationally viable\nmodel of both bidding and attribution which spans the randomization, training,\ncross validation, scoring, and conversion attribution of advertising's causal\neffects. Implementation of this approach is likely to secure a significant\nimprovement in the return on investment of advertising.\n", "date": "2022-08-25", "categories": []}
{"id": "2208.10981", "title": "Causal Entropy Optimization", "abstract": "  We study the problem of globally optimizing the causal effect on a target\nvariable of an unknown causal graph in which interventions can be performed.\nThis problem arises in many areas of science including biology, operations\nresearch and healthcare. We propose Causal Entropy Optimization (CEO), a\nframework that generalizes Causal Bayesian Optimization (CBO) to account for\nall sources of uncertainty, including the one arising from the causal graph\nstructure. CEO incorporates the causal structure uncertainty both in the\nsurrogate models for the causal effects and in the mechanism used to select\ninterventions via an information-theoretic acquisition function. The resulting\nalgorithm automatically trades-off structure learning and causal effect\noptimization, while naturally accounting for observation noise. For various\nsynthetic and real-world structural causal models, CEO achieves faster\nconvergence to the global optimum compared with CBO while also learning the\ngraph. Furthermore, our joint approach to structure learning and causal\noptimization improves upon sequential, structure-learning-first approaches.\n", "date": "2022-08-23", "categories": []}
{"id": "2208.08345", "title": "Discovering Agents", "abstract": "  Causal models of agents have been used to analyse the safety aspects of\nmachine learning systems. But identifying agents is non-trivial -- often the\ncausal model is just assumed by the modeler without much justification -- and\nmodelling failures can lead to mistakes in the safety analysis. This paper\nproposes the first formal causal definition of agents -- roughly that agents\nare systems that would adapt their policy if their actions influenced the world\nin a different way. From this we derive the first causal discovery algorithm\nfor discovering agents from empirical data, and give algorithms for translating\nbetween causal models and game-theoretic influence diagrams. We demonstrate our\napproach by resolving some previous confusions caused by incorrect causal\nmodelling of agents.\n", "date": "2022-08-17", "categories": []}
{"id": "2208.10916", "title": "Application of Causal Inference to Analytical Customer Relationship\n  Management in Banking and Insurance", "abstract": "  Of late, in order to have better acceptability among various domain,\nresearchers have argued that machine intelligence algorithms must be able to\nprovide explanations that humans can understand causally. This aspect, also\nknown as causability, achieves a specific level of human-level explainability.\nA specific class of algorithms known as counterfactuals may be able to provide\ncausability. In statistics, causality has been studied and applied for many\nyears, but not in great detail in artificial intelligence (AI). In a\nfirst-of-its-kind study, we employed the principles of causal inference to\nprovide explainability for solving the analytical customer relationship\nmanagement (ACRM) problems. In the context of banking and insurance, current\nresearch on interpretability tries to address causality-related questions like\nwhy did this model make such decisions, and was the model's choice influenced\nby a particular factor? We propose a solution in the form of an intervention,\nwherein the effect of changing the distribution of features of ACRM datasets is\nstudied on the target feature. Subsequently, a set of counterfactuals is also\nobtained that may be furnished to any customer who demands an explanation of\nthe decision taken by the bank/insurance company. Except for the credit card\nchurn prediction dataset, good quality counterfactuals were generated for the\nloan default, insurance fraud detection, and credit card fraud detection\ndatasets, where changes in no more than three features are observed.\n", "date": "2022-08-19", "categories": []}
{"id": "2208.10950", "title": "Deep Structural Causal Shape Models", "abstract": "  Causal reasoning provides a language to ask important interventional and\ncounterfactual questions beyond purely statistical association. In medical\nimaging, for example, we may want to study the causal effect of genetic,\nenvironmental, or lifestyle factors on the normal and pathological variation of\nanatomical phenotypes. However, while anatomical shape models of 3D surface\nmeshes, extracted from automated image segmentation, can be reliably\nconstructed, there is a lack of computational tooling to enable causal\nreasoning about morphological variations. To tackle this problem, we propose\ndeep structural causal shape models (CSMs), which utilise high-quality mesh\ngeneration techniques, from geometric deep learning, within the expressive\nframework of deep structural causal models. CSMs enable subject-specific\nprognoses through counterfactual mesh generation (\"How would this patient's\nbrain structure change if they were ten years older?\"), which is in contrast to\nmost current works on purely population-level statistical shape modelling. We\ndemonstrate the capabilities of CSMs at all levels of Pearl's causal hierarchy\nthrough a number of qualitative and quantitative experiments leveraging a large\ndataset of 3D brain structures.\n", "date": "2022-08-23", "categories": []}
{"id": "2208.11094", "title": "Dynamic Causal Collaborative Filtering", "abstract": "  Causal graph, as an effective and powerful tool for causal modeling, is\nusually assumed as a Directed Acyclic Graph (DAG). However, recommender systems\nusually involve feedback loops, defined as the cyclic process of recommending\nitems, incorporating user feedback in model updates, and repeating the\nprocedure. As a result, it is important to incorporate loops into the causal\ngraphs to accurately model the dynamic and iterative data generation process\nfor recommender systems. However, feedback loops are not always beneficial\nsince over time they may encourage more and more narrowed content exposure,\nwhich if left unattended, may results in echo chambers. As a result, it is\nimportant to understand when the recommendations will lead to echo chambers and\nhow to mitigate echo chambers without hurting the recommendation performance.\n  In this paper, we design a causal graph with loops to describe the dynamic\nprocess of recommendation. We then take Markov process to analyze the\nmathematical properties of echo chamber such as the conditions that lead to\necho chambers. Inspired by the theoretical analysis, we propose a Dynamic\nCausal Collaborative Filtering ($\\partial$CCF) model, which estimates users'\npost-intervention preference on items based on back-door adjustment and\nmitigates echo chamber with counterfactual reasoning. Multiple experiments are\nconducted on real-world datasets and results show that our framework can\nmitigate echo chambers better than other state-of-the-art frameworks while\nachieving comparable recommendation performance with the base recommendation\nmodels.\n", "date": "2022-08-23", "categories": []}
{"id": "2206.05490", "title": "Discovery and density estimation of latent confounders in Bayesian\n  networks with evidence lower bound", "abstract": "  Discovering and parameterising latent confounders represent important and\nchallenging problems in causal structure learning and density estimation\nrespectively. In this paper, we focus on both discovering and learning the\ndistribution of latent confounders. This task requires solutions that come from\ndifferent areas of statistics and machine learning. We combine elements of\nvariational Bayesian methods, expectation-maximisation, hill-climbing search,\nand structure learning under the assumption of causal insufficiency. We propose\ntwo learning strategies; one that maximises model selection accuracy, and\nanother that improves computational efficiency in exchange for minor reductions\nin accuracy. The former strategy is suitable for small networks and the latter\nfor moderate size networks. Both learning strategies perform well relative to\nexisting solutions.\n", "date": "2022-06-11", "categories": []}
{"id": "2208.09568", "title": "Probabilities of Causation with Nonbinary Treatment and Effect", "abstract": "  This paper deals with the problem of estimating the probabilities of\ncausation when treatment and effect are not binary. Tian and Pearl derived\nsharp bounds for the probability of necessity and sufficiency (PNS), the\nprobability of sufficiency (PS), and the probability of necessity (PN) using\nexperimental and observational data. In this paper, we provide theoretical\nbounds for all types of probabilities of causation to multivalued treatments\nand effects. We further discuss examples where our bounds guide practical\ndecisions and use simulation studies to evaluate how informative the bounds are\nfor various combinations of data.\n", "date": "2022-08-19", "categories": []}
{"id": "2208.09569", "title": "Unit Selection with Nonbinary Treatment and Effect", "abstract": "  The unit selection problem aims to identify a set of individuals who are most\nlikely to exhibit a desired mode of behavior, for example, selecting\nindividuals who would respond one way if encouraged and a different way if not\nencouraged. Using a combination of experimental and observational data, Li and\nPearl derived tight bounds on the \"benefit function\", which is the payoff/cost\nassociated with selecting an individual with given characteristics. This paper\nextends the benefit function to the general form such that the treatment and\neffect are not restricted to binary. We propose an algorithm to test the\nidentifiability of the nonbinary benefit function and an algorithm to compute\nthe bounds of the nonbinary benefit function using experimental and\nobservational data.\n", "date": "2022-08-20", "categories": []}
{"id": "2208.09862", "title": "Twin Papers: A Simple Framework of Causal Inference for Citations via\n  Coupling", "abstract": "  The research process includes many decisions, e.g., how to entitle and where\nto publish the paper. In this paper, we introduce a general framework for\ninvestigating the effects of such decisions. The main difficulty in\ninvestigating the effects is that we need to know counterfactual results, which\nare not available in reality. The key insight of our framework is inspired by\nthe existing counterfactual analysis using twins, where the researchers regard\ntwins as counterfactual units. The proposed framework regards a pair of papers\nthat cite each other as twins. Such papers tend to be parallel works, on\nsimilar topics, and in similar communities. We investigate twin papers that\nadopted different decisions, observe the progress of the research impact\nbrought by these studies, and estimate the effect of decisions by the\ndifference in the impacts of these studies. We release our code and data, which\nwe believe are highly beneficial owing to the scarcity of the dataset on\ncounterfactual studies.\n", "date": "2022-08-21", "categories": []}
{"id": "2208.09329", "title": "Causal Intervention Improves Implicit Sentiment Analysis", "abstract": "  Despite having achieved great success for sentiment analysis, existing neural\nmodels struggle with implicit sentiment analysis. This may be due to the fact\nthat they may latch onto spurious correlations (\"shortcuts\", e.g., focusing\nonly on explicit sentiment words), resulting in undermining the effectiveness\nand robustness of the learned model. In this work, we propose a causal\nintervention model for Implicit Sentiment Analysis using Instrumental Variable\n(ISAIV). We first review sentiment analysis from a causal perspective and\nanalyze the confounders existing in this task. Then, we introduce an\ninstrumental variable to eliminate the confounding causal effects, thus\nextracting the pure causal effect between sentence and sentiment. We compare\nthe proposed ISAIV model with several strong baselines on both the general\nimplicit sentiment analysis and aspect-based implicit sentiment analysis tasks.\nThe results indicate the great advantages of our model and the efficacy of\nimplicit sentiment reasoning.\n", "date": "2022-08-19", "categories": []}
{"id": "2208.08247", "title": "Domain Knowledge in A*-Based Causal Discovery", "abstract": "  Causal discovery has become a vital tool for scientists and practitioners\nwanting to discover causal relationships from observational data. While most\nprevious approaches to causal discovery have implicitly assumed that no expert\ndomain knowledge is available, practitioners can often provide such domain\nknowledge from prior experience. Recent work has incorporated domain knowledge\ninto constraint-based causal discovery. The majority of such constraint-based\nmethods, however, assume causal faithfulness, which has been shown to be\nfrequently violated in practice. Consequently, there has been renewed attention\ntowards exact-search score-based causal discovery methods, which do not assume\ncausal faithfulness, such as A*-based methods. However, there has been no\nconsideration of these methods in the context of domain knowledge. In this\nwork, we focus on efficiently integrating several types of domain knowledge\ninto A*-based causal discovery. In doing so, we discuss and explain how domain\nknowledge can reduce the graph search space and then provide an analysis of the\npotential computational gains. We support these findings with experiments on\nsynthetic and real data, showing that even small amounts of domain knowledge\ncan dramatically speed up A*-based causal discovery and improve its performance\nand practicality.\n", "date": "2022-08-17", "categories": []}
{"id": "2208.08399", "title": "The Counterfactual-Shapley Value: Attributing Change in System Metrics", "abstract": "  Given an unexpected change in the output metric of a large-scale system, it\nis important to answer why the change occurred: which inputs caused the change\nin metric? A key component of such an attribution question is estimating the\ncounterfactual: the (hypothetical) change in the system metric due to a\nspecified change in a single input. However, due to inherent stochasticity and\ncomplex interactions between parts of the system, it is difficult to model an\noutput metric directly. We utilize the computational structure of a system to\nbreak up the modelling task into sub-parts, such that each sub-part corresponds\nto a more stable mechanism that can be modelled accurately over time. Using the\nsystem's structure also helps to view the metric as a computation over a\nstructural causal model (SCM), thus providing a principled way to estimate\ncounterfactuals. Specifically, we propose a method to estimate counterfactuals\nusing time-series predictive models and construct an attribution score,\nCF-Shapley, that is consistent with desirable axioms for attributing an\nobserved change in the output metric. Unlike past work on causal shapley\nvalues, our proposed method can attribute a single observed change in output\n(rather than a population-level effect) and thus provides more accurate\nattribution scores when evaluated on simulated datasets. As a real-world\napplication, we analyze a query-ad matching system with the goal of attributing\nobserved change in a metric for ad matching density. Attribution scores explain\nhow query volume and ad demand from different query categories affect the ad\nmatching density, leading to actionable insights and uncovering the role of\nexternal events (e.g., \"Cheetah Day\") in driving the matching density.\n", "date": "2022-08-17", "categories": []}
{"id": "2208.06701", "title": "Learning Linear Non-Gaussian Polytree Models", "abstract": "  In the context of graphical causal discovery, we adapt the versatile\nframework of linear non-Gaussian acyclic models (LiNGAMs) to propose new\nalgorithms to efficiently learn graphs that are polytrees. Our approach\ncombines the Chow--Liu algorithm, which first learns the undirected tree\nstructure, with novel schemes to orient the edges. The orientation schemes\nassess algebraic relations among moments of the data-generating distribution\nand are computationally inexpensive. We establish high-dimensional consistency\nresults for our approach and compare different algorithmic versions in\nnumerical experiments.\n", "date": "2022-08-13", "categories": []}
{"id": "2208.06748", "title": "Learning to Infer Counterfactuals: Meta-Learning for Estimating Multiple\n  Imbalanced Treatment Effects", "abstract": "  We regularly consider answering counterfactual questions in practice, such as\n\"Would people with diabetes take a turn for the better had they choose another\nmedication?\". Observational studies are growing in significance in answering\nsuch questions due to their widespread accumulation and comparatively easier\nacquisition than Randomized Control Trials (RCTs). Recently, some works have\nintroduced representation learning and domain adaptation into counterfactual\ninference. However, most current works focus on the setting of binary\ntreatments. None of them considers that different treatments' sample sizes are\nimbalanced, especially data examples in some treatment groups are relatively\nlimited due to inherent user preference. In this paper, we design a new\nalgorithmic framework for counterfactual inference, which brings an idea from\nMeta-learning for Estimating Individual Treatment Effects (MetaITE) to fill the\nabove research gaps, especially considering multiple imbalanced treatments.\nSpecifically, we regard data episodes among treatment groups in counterfactual\ninference as meta-learning tasks. We train a meta-learner from a set of source\ntreatment groups with sufficient samples and update the model by gradient\ndescent with limited samples in target treatment. Moreover, we introduce two\ncomplementary losses. One is the supervised loss on multiple source treatments.\nThe other loss which aligns latent distributions among various treatment groups\nis proposed to reduce the discrepancy. We perform experiments on two real-world\ndatasets to evaluate inference accuracy and generalization ability.\nExperimental results demonstrate that the model MetaITE matches/outperforms\nstate-of-the-art methods.\n", "date": "2022-08-13", "categories": []}
{"id": "2208.06935", "title": "Novel Ordering-based Approaches for Causal Structure Learning in the\n  Presence of Unobserved Variables", "abstract": "  We propose ordering-based approaches for learning the maximal ancestral graph\n(MAG) of a structural equation model (SEM) up to its Markov equivalence class\n(MEC) in the presence of unobserved variables. Existing ordering-based methods\nin the literature recover a graph through learning a causal order (c-order). We\nadvocate for a novel order called removable order (r-order) as they are\nadvantageous over c-orders for structure learning. This is because r-orders are\nthe minimizers of an appropriately defined optimization problem that could be\neither solved exactly (using a reinforcement learning approach) or\napproximately (using a hill-climbing search). Moreover, the r-orders (unlike\nc-orders) are invariant among all the graphs in a MEC and include c-orders as a\nsubset. Given that set of r-orders is often significantly larger than the set\nof c-orders, it is easier for the optimization problem to find an r-order\ninstead of a c-order. We evaluate the performance and the scalability of our\nproposed approaches on both real-world and randomly generated networks.\n", "date": "2022-08-14", "categories": []}
{"id": "2206.13973", "title": "Towards a Grounded Theory of Causation for Embodied AI", "abstract": "  There exist well-developed frameworks for causal modelling, but these require\nrather a lot of human domain expertise to define causal variables and perform\ninterventions. In order to enable autonomous agents to learn abstract causal\nmodels through interactive experience, the existing theoretical foundations\nneed to be extended and clarified. Existing frameworks give no guidance\nregarding variable choice / representation, and more importantly, give no\nindication as to which behaviour policies or physical transformations of state\nspace shall count as interventions. The framework sketched in this paper\ndescribes actions as transformations of state space, for instance induced by an\nagent running a policy. This makes it possible to describe in a uniform way\nboth transformations of the micro-state space and abstract models thereof, and\nsay when the latter is veridical / grounded / natural. We then introduce\n(causal) variables, define a mechanism as an invariant predictor, and say when\nan action can be viewed as a ``surgical intervention'', thus bringing the\nobjective of causal representation \\& intervention skill learning into clearer\nfocus.\n", "date": "2022-06-28", "categories": []}
{"id": "2208.06267", "title": "Causal Imitation Learning with Unobserved Confounders", "abstract": "  One of the common ways children learn is by mimicking adults. Imitation\nlearning focuses on learning policies with suitable performance from\ndemonstrations generated by an expert, with an unspecified performance measure,\nand unobserved reward signal. Popular methods for imitation learning start by\neither directly mimicking the behavior policy of an expert (behavior cloning)\nor by learning a reward function that prioritizes observed expert trajectories\n(inverse reinforcement learning). However, these methods rely on the assumption\nthat covariates used by the expert to determine her/his actions are fully\nobserved. In this paper, we relax this assumption and study imitation learning\nwhen sensory inputs of the learner and the expert differ. First, we provide a\nnon-parametric, graphical criterion that is complete (both necessary and\nsufficient) for determining the feasibility of imitation from the combinations\nof demonstration data and qualitative assumptions about the underlying\nenvironment, represented in the form of a causal model. We then show that when\nsuch a criterion does not hold, imitation could still be feasible by exploiting\nquantitative knowledge of the expert trajectories. Finally, we develop an\nefficient procedure for learning the imitating policy from experts'\ntrajectories.\n", "date": "2022-08-12", "categories": []}
{"id": "2208.06276", "title": "Sequential Causal Imitation Learning with Unobserved Confounders", "abstract": "  \"Monkey see monkey do\" is an age-old adage, referring to na\\\"ive imitation\nwithout a deep understanding of a system's underlying mechanics. Indeed, if a\ndemonstrator has access to information unavailable to the imitator (monkey),\nsuch as a different set of sensors, then no matter how perfectly the imitator\nmodels its perceived environment (See), attempting to reproduce the\ndemonstrator's behavior (Do) can lead to poor outcomes. Imitation learning in\nthe presence of a mismatch between demonstrator and imitator has been studied\nin the literature under the rubric of causal imitation learning (Zhang et al.,\n2020), but existing solutions are limited to single-stage decision-making. This\npaper investigates the problem of causal imitation learning in sequential\nsettings, where the imitator must make multiple decisions per episode. We\ndevelop a graphical criterion that is necessary and sufficient for determining\nthe feasibility of causal imitation, providing conditions when an imitator can\nmatch a demonstrator's performance despite differing capabilities. Finally, we\nprovide an efficient algorithm for determining imitability and corroborate our\ntheory with simulations.\n", "date": "2022-08-12", "categories": []}
{"id": "2207.05723", "title": "Latent Variable Models for Bayesian Causal Discovery", "abstract": "  Learning predictors that do not rely on spurious correlations involves\nbuilding causal representations. However, learning such a representation is\nvery challenging. We, therefore, formulate the problem of learning a causal\nrepresentation from high dimensional data and study causal recovery with\nsynthetic data. This work introduces a latent variable decoder model, Decoder\nBCD, for Bayesian causal discovery and performs experiments in mildly\nsupervised and unsupervised settings. We present a series of synthetic\nexperiments to characterize important factors for causal discovery and show\nthat using known intervention targets as labels helps in unsupervised Bayesian\ninference over structure and parameters of linear Gaussian additive noise\nlatent structural causal models.\n", "date": "2022-07-12", "categories": []}
{"id": "2208.05126", "title": "D-BIAS: A Causality-Based Human-in-the-Loop System for Tackling\n  Algorithmic Bias", "abstract": "  With the rise of AI, algorithms have become better at learning underlying\npatterns from the training data including ingrained social biases based on\ngender, race, etc. Deployment of such algorithms to domains such as hiring,\nhealthcare, law enforcement, etc. has raised serious concerns about fairness,\naccountability, trust and interpretability in machine learning algorithms. To\nalleviate this problem, we propose D-BIAS, a visual interactive tool that\nembodies human-in-the-loop AI approach for auditing and mitigating social\nbiases from tabular datasets. It uses a graphical causal model to represent\ncausal relationships among different features in the dataset and as a medium to\ninject domain knowledge. A user can detect the presence of bias against a\ngroup, say females, or a subgroup, say black females, by identifying unfair\ncausal relationships in the causal network and using an array of fairness\nmetrics. Thereafter, the user can mitigate bias by acting on the unfair causal\nedges. For each interaction, say weakening/deleting a biased causal edge, the\nsystem uses a novel method to simulate a new (debiased) dataset based on the\ncurrent causal model. Users can visually assess the impact of their\ninteractions on different fairness metrics, utility metrics, data distortion,\nand the underlying data distribution. Once satisfied, they can download the\ndebiased dataset and use it for any downstream application for fairer\npredictions. We evaluate D-BIAS by conducting experiments on 3 datasets and\nalso a formal user study. We found that D-BIAS helps reduce bias significantly\ncompared to the baseline debiasing approach across different fairness metrics\nwhile incurring little data distortion and a small loss in utility. Moreover,\nour human-in-the-loop based approach significantly outperforms an automated\napproach on trust, interpretability and accountability.\n", "date": "2022-08-10", "categories": []}
{"id": "2208.04892", "title": "Intrinsically Motivated Learning of Causal World Models", "abstract": "  Despite the recent progress in deep learning and reinforcement learning,\ntransfer and generalization of skills learned on specific tasks is very limited\ncompared to human (or animal) intelligence. The lifelong, incremental building\nof common sense knowledge might be a necessary component on the way to achieve\nmore general intelligence. A promising direction is to build world models\ncapturing the true physical mechanisms hidden behind the sensorimotor\ninteraction with the environment. Here we explore the idea that inferring the\ncausal structure of the environment could benefit from well-chosen actions as\nmeans to collect relevant interventional data.\n", "date": "2022-08-09", "categories": []}
{"id": "2208.01417", "title": "Bounding Counterfactuals under Selection Bias", "abstract": "  Causal analysis may be affected by selection bias, which is defined as the\nsystematic exclusion of data from a certain subpopulation. Previous work in\nthis area focused on the derivation of identifiability conditions. We propose\ninstead a first algorithm to address both identifiable and unidentifiable\nqueries. We prove that, in spite of the missingness induced by the selection\nbias, the likelihood of the available data is unimodal. This enables us to use\nthe causal expectation-maximisation scheme to obtain the values of causal\nqueries in the identifiable case, and to compute bounds otherwise. Experiments\ndemonstrate the approach to be practically viable. Theoretical convergence\ncharacterisations are provided.\n", "date": "2022-07-26", "categories": []}
{"id": "2208.01529", "title": "CIPCaD-Bench: Continuous Industrial Process datasets for benchmarking\n  Causal Discovery methods", "abstract": "  Causal relationships are commonly examined in manufacturing processes to\nsupport faults investigations, perform interventions, and make strategic\ndecisions. Industry 4.0 has made available an increasing amount of data that\nenable data-driven Causal Discovery (CD). Considering the growing number of\nrecently proposed CD methods, it is necessary to introduce strict benchmarking\nprocedures on publicly available datasets since they represent the foundation\nfor a fair comparison and validation of different methods. This work introduces\ntwo novel public datasets for CD in continuous manufacturing processes. The\nfirst dataset employs the well-known Tennessee Eastman simulator for fault\ndetection and process control. The second dataset is extracted from an\nultra-processed food manufacturing plant, and it includes a description of the\nplant, as well as multiple ground truths. These datasets are used to propose a\nbenchmarking procedure based on different metrics and evaluated on a wide\nselection of CD algorithms. This work allows testing CD methods in realistic\nconditions enabling the selection of the most suitable method for specific\ntarget applications. The datasets are available at the following link:\nhttps://github.com/giovanniMen\n", "date": "2022-08-02", "categories": []}
{"id": "2208.00894", "title": "Towards Computing an Optimal Abstraction for Structural Causal Models", "abstract": "  Working with causal models at different levels of abstraction is an important\nfeature of science. Existing work has already considered the problem of\nexpressing formally the relation of abstraction between causal models. In this\npaper, we focus on the problem of learning abstractions. We start by defining\nthe learning problem formally in terms of the optimization of a standard\nmeasure of consistency. We then point out the limitation of this approach, and\nwe suggest extending the objective function with a term accounting for\ninformation loss. We suggest a concrete measure of information loss, and we\nillustrate its contribution to learning new abstractions.\n", "date": "2022-08-01", "categories": []}
{"id": "2208.00926", "title": "Graphical Representations for Algebraic Constraints of Linear Structural\n  Equations Models", "abstract": "  The observational characteristics of a linear structural equation model can\nbe effectively described by polynomial constraints on the observed covariance\nmatrix. However, these polynomials can be exponentially large, making them\nimpractical for many purposes. In this paper, we present a graphical notation\nfor many of these polynomial constraints. The expressive power of this notation\nis investigated both theoretically and empirically.\n", "date": "2022-08-01", "categories": []}
{"id": "2207.11251", "title": "Variational Temporal Deconfounder for Individualized Treatment Effect\n  Estimation from Longitudinal Observational Data", "abstract": "  Estimating treatment effects, especially individualized treatment effects\n(ITE), using observational data is challenging due to the complex situations of\nconfounding bias. Existing approaches for estimating treatment effects from\nlongitudinal observational data are usually built upon a strong assumption of\n\"unconfoundedness\", which is hard to fulfill in real-world practice. In this\npaper, we propose the Variational Temporal Deconfounder (VTD), an approach that\nleverages deep variational embeddings in the longitudinal setting using proxies\n(i.e., surrogate variables that serve for unobservable variables).\nSpecifically, VTD leverages observed proxies to learn a hidden embedding that\nreflects the true hidden confounders in the observational data. As such, our\nVTD method does not rely on the \"unconfoundedness\" assumption. We test our VTD\nmethod on both synthetic and real-world clinical data, and the results show\nthat our approach is effective when hidden confounding is the leading bias\ncompared to other existing models.\n", "date": "2022-07-23", "categories": []}
{"id": "2207.11385", "title": "Causal Fairness Analysis", "abstract": "  Decision-making systems based on AI and machine learning have been used\nthroughout a wide range of real-world scenarios, including healthcare, law\nenforcement, education, and finance. It is no longer far-fetched to envision a\nfuture where autonomous systems will be driving entire business decisions and,\nmore broadly, supporting large-scale decision-making infrastructure to solve\nsociety's most challenging problems. Issues of unfairness and discrimination\nare pervasive when decisions are being made by humans, and remain (or are\npotentially amplified) when decisions are made using machines with little\ntransparency, accountability, and fairness. In this paper, we introduce a\nframework for \\textit{causal fairness analysis} with the intent of filling in\nthis gap, i.e., understanding, modeling, and possibly solving issues of\nfairness in decision-making settings. The main insight of our approach will be\nto link the quantification of the disparities present on the observed data with\nthe underlying, and often unobserved, collection of causal mechanisms that\ngenerate the disparity in the first place, challenge we call the Fundamental\nProblem of Causal Fairness Analysis (FPCFA). In order to solve the FPCFA, we\nstudy the problem of decomposing variations and empirical measures of fairness\nthat attribute such variations to structural mechanisms and different units of\nthe population. Our effort culminates in the Fairness Map, which is the first\nsystematic attempt to organize and explain the relationship between different\ncriteria found in the literature. Finally, we study which causal assumptions\nare minimally needed for performing causal fairness analysis and propose a\nFairness Cookbook, which allows data scientists to assess the existence of\ndisparate impact and disparate treatment.\n", "date": "2022-07-23", "categories": []}
{"id": "2207.11652", "title": "Counterfactual Reasoning for Out-of-distribution Multimodal Sentiment\n  Analysis", "abstract": "  Existing studies on multimodal sentiment analysis heavily rely on textual\nmodality and unavoidably induce the spurious correlations between textual words\nand sentiment labels. This greatly hinders the model generalization ability. To\naddress this problem, we define the task of out-of-distribution (OOD)\nmultimodal sentiment analysis. This task aims to estimate and mitigate the bad\neffect of textual modality for strong OOD generalization. To this end, we\nembrace causal inference, which inspects the causal relationships via a causal\ngraph. From the graph, we find that the spurious correlations are attributed to\nthe direct effect of textual modality on the model prediction while the\nindirect one is more reliable by considering multimodal semantics. Inspired by\nthis, we devise a model-agnostic counterfactual framework for multimodal\nsentiment analysis, which captures the direct effect of textual modality via an\nextra text model and estimates the indirect one by a multimodal model. During\nthe inference, we first estimate the direct effect by the counterfactual\ninference, and then subtract it from the total effect of all modalities to\nobtain the indirect effect for reliable prediction. Extensive experiments show\nthe superior effectiveness and generalization ability of our proposed\nframework.\n", "date": "2022-07-24", "categories": []}
{"id": "2207.08963", "title": "The m-connecting imset and factorization for ADMG models", "abstract": "  Directed acyclic graph (DAG) models have become widely studied and applied in\nstatistics and machine learning -- indeed, their simplicity facilitates\nefficient procedures for learning and inference. Unfortunately, these models\nare not closed under marginalization, making them poorly equipped to handle\nsystems with latent confounding. Acyclic directed mixed graph (ADMG) models\ncharacterize margins of DAG models, making them far better suited to handle\nsuch systems. However, ADMG models have not seen wide-spread use due to their\ncomplexity and a shortage of statistical tools for their analysis. In this\npaper, we introduce the m-connecting imset which provides an alternative\nrepresentation for the independence models induced by ADMGs. Furthermore, we\ndefine the m-connecting factorization criterion for ADMG models, characterized\nby a single equation, and prove its equivalence to the global Markov property.\nThe m-connecting imset and factorization criterion provide two new statistical\ntools for learning and inference with ADMG models. We demonstrate the\nusefulness of these tools by formulating and evaluating a consistent scoring\ncriterion with a closed form solution.\n", "date": "2022-07-18", "categories": []}
{"id": "2207.09139", "title": "Heterogeneous Treatment Effect with Trained Kernels of the\n  Nadaraya-Watson Regression", "abstract": "  A new method for estimating the conditional average treatment effect is\nproposed in the paper. It is called TNW-CATE (the Trainable Nadaraya-Watson\nregression for CATE) and based on the assumption that the number of controls is\nrather large whereas the number of treatments is small. TNW-CATE uses the\nNadaraya-Watson regression for predicting outcomes of patients from the control\nand treatment groups. The main idea behind TNW-CATE is to train kernels of the\nNadaraya-Watson regression by using a weight sharing neural network of a\nspecific form. The network is trained on controls, and it replaces standard\nkernels with a set of neural subnetworks with shared parameters such that every\nsubnetwork implements the trainable kernel, but the whole network implements\nthe Nadaraya-Watson estimator. The network memorizes how the feature vectors\nare located in the feature space. The proposed approach is similar to the\ntransfer learning when domains of source and target data are similar, but tasks\nare different. Various numerical simulation experiments illustrate TNW-CATE and\ncompare it with the well-known T-learner, S-learner and X-learner for several\ntypes of the control and treatment outcome functions. The code of proposed\nalgorithms implementing TNW-CATE is available in\nhttps://github.com/Stasychbr/TNW-CATE.\n", "date": "2022-07-19", "categories": []}
{"id": "2207.07732", "title": "Partial Disentanglement via Mechanism Sparsity", "abstract": "  Disentanglement via mechanism sparsity was introduced recently as a\nprincipled approach to extract latent factors without supervision when the\ncausal graph relating them in time is sparse, and/or when actions are observed\nand affect them sparsely. However, this theory applies only to ground-truth\ngraphs satisfying a specific criterion. In this work, we introduce a\ngeneralization of this theory which applies to any ground-truth graph and\nspecifies qualitatively how disentangled the learned representation is expected\nto be, via a new equivalence relation over models we call consistency. This\nequivalence captures which factors are expected to remain entangled and which\nare not based on the specific form of the ground-truth graph. We call this\nweaker form of identifiability partial disentanglement. The graphical criterion\nthat allows complete disentanglement, proposed in an earlier work, can be\nderived as a special case of our theory. Finally, we enforce graph sparsity\nwith constrained optimization and illustrate our theory and algorithm in\nsimulations.\n", "date": "2022-07-15", "categories": []}
{"id": "2207.07908", "title": "Multiscale Causal Structure Learning", "abstract": "  The inference of causal structures from observed data plays a key role in\nunveiling the underlying dynamics of the system. This paper exposes a novel\nmethod, named Multiscale-Causal Structure Learning (MS-CASTLE), to estimate the\nstructure of linear causal relationships occurring at different time scales.\nDifferently from existing approaches, MS-CASTLE takes explicitly into account\ninstantaneous and lagged inter-relations between multiple time series,\nrepresented at different scales, hinging on stationary wavelet transform and\nnon-convex optimization. MS-CASTLE incorporates, as a special case, a\nsingle-scale version named SS-CASTLE, which compares favorably in terms of\ncomputational efficiency, performance and robustness with respect to the state\nof the art onto synthetic data. We used MS-CASTLE to study the multiscale\ncausal structure of the risk of 15 global equity markets, during covid-19\npandemic, illustrating how MS-CASTLE can extract meaningful information thanks\nto its multiscale analysis, outperforming SS-CASTLE. We found that the most\npersistent and strongest interactions occur at mid-term time resolutions.\nMoreover, we identified the stock markets that drive the risk during the\nconsidered period: Brazil, Canada and Italy. The proposed approach can be\nexploited by financial investors who, depending to their investment horizon,\ncan manage the risk within equity portfolios from a causal perspective.\n", "date": "2022-07-16", "categories": []}
{"id": "2207.08603", "title": "Abstraction between Structural Causal Models: A Review of Definitions\n  and Properties", "abstract": "  Structural causal models (SCMs) are a widespread formalism to deal with\ncausal systems. A recent direction of research has considered the problem of\nrelating formally SCMs at different levels of abstraction, by defining maps\nbetween SCMs and imposing a requirement of interventional consistency. This\npaper offers a review of the solutions proposed so far, focusing on the formal\nproperties of a map between SCMs, and highlighting the different layers\n(structural, distributional) at which these properties may be enforced. This\nallows us to distinguish families of abstractions that may or may not be\npermitted by choosing to guarantee certain properties instead of others. Such\nan understanding not only allows to distinguish among proposal for causal\nabstraction with more awareness, but it also allows to tailor the definition of\nabstraction with respect to the forms of abstraction relevant to specific\napplications.\n", "date": "2022-07-18", "categories": []}
{"id": "2207.01616", "title": "Breaking Feedback Loops in Recommender Systems with Causal Inference", "abstract": "  Recommender systems play a key role in shaping modern web ecosystems. These\nsystems alternate between (1) making recommendations (2) collecting user\nresponses to these recommendations, and (3) retraining the recommendation\nalgorithm based on this feedback. During this process the recommender system\ninfluences the user behavioral data that is subsequently used to update it,\nthus creating a feedback loop. Recent work has shown that feedback loops may\ncompromise recommendation quality and homogenize user behavior, raising ethical\nand performance concerns when deploying recommender systems. To address these\nissues, we propose the Causal Adjustment for Feedback Loops (CAFL), an\nalgorithm that provably breaks feedback loops using causal inference and can be\napplied to any recommendation algorithm that optimizes a training loss. Our\nmain observation is that a recommender system does not suffer from feedback\nloops if it reasons about causal quantities, namely the intervention\ndistributions of recommendations on user ratings. Moreover, we can calculate\nthis intervention distribution from observational data by adjusting for the\nrecommender system's predictions of user preferences. Using simulated\nenvironments, we demonstrate that CAFL improves recommendation quality when\ncompared to prior correction methods.\n", "date": "2022-07-04", "categories": []}
{"id": "2207.07458", "title": "Joint Application of the Target Trial Causal Framework and Machine\n  Learning Modeling to Optimize Antibiotic Therapy: Use Case on Acute Bacterial\n  Skin and Skin Structure Infections due to Methicillin-resistant\n  Staphylococcus aureus", "abstract": "  Bacterial infections are responsible for high mortality worldwide.\nAntimicrobial resistance underlying the infection, and multifaceted patient's\nclinical status can hamper the correct choice of antibiotic treatment.\nRandomized clinical trials provide average treatment effect estimates but are\nnot ideal for risk stratification and optimization of therapeutic choice, i.e.,\nindividualized treatment effects (ITE). Here, we leverage large-scale\nelectronic health record data, collected from Southern US academic clinics, to\nemulate a clinical trial, i.e., 'target trial', and develop a machine learning\nmodel of mortality prediction and ITE estimation for patients diagnosed with\nacute bacterial skin and skin structure infection (ABSSSI) due to\nmethicillin-resistant Staphylococcus aureus (MRSA). ABSSSI-MRSA is a\nchallenging condition with reduced treatment options - vancomycin is the\npreferred choice, but it has non-negligible side effects. First, we use\npropensity score matching to emulate the trial and create a treatment\nrandomized (vancomycin vs. other antibiotics) dataset. Next, we use this data\nto train various machine learning methods (including boosted/LASSO logistic\nregression, support vector machines, and random forest) and choose the best\nmodel in terms of area under the receiver characteristic (AUC) through\nbootstrap validation. Lastly, we use the models to calculate ITE and identify\npossible averted deaths by therapy change. The out-of-bag tests indicate that\nSVM and RF are the most accurate, with AUC of 81% and 78%, respectively, but\nBLR/LASSO is not far behind (76%). By calculating the counterfactuals using the\nBLR/LASSO, vancomycin increases the risk of death, but it shows a large\nvariation (odds ratio 1.2, 95% range 0.4-3.8) and the contribution to outcome\nprobability is modest. Instead, the RF exhibits stronger changes in ITE,\nsuggesting more complex treatment heterogeneity.\n", "date": "2022-07-15", "categories": []}
{"id": "2207.01722", "title": "A Causal Approach for Business Optimization: Application on an Online\n  Marketplace", "abstract": "  A common sales strategy involves having account executives (AEs) actively\nreach out and contact potential customers. However, not all contact attempts\nhave a positive effect: some attempts do not change customer decisions, while\nothers might even interfere with the desired outcome. In this work we propose\nusing causal inference to estimate the effect of contacting each potential\ncustomer and setting the contact policy accordingly. We demonstrate this\napproach on data from Worthy, an online jewelry marketplace. We examined the\nWorthy business process to identify relevant decisions and outcomes, and\nformalized assumptions on how they were made. Using causal tools, we selected a\ndecision point where improving AE contact activity appeared to be promising. We\nthen generated a personalized policy and recommended reaching out only to\ncustomers for whom it would be beneficial. Finally, we validated the results in\nan A\\B test over a 3-month period, resulting in an increase in item delivery\nrate of the targeted population by 22% (p-value=0.026). This policy is now\nbeing used on an ongoing basis.\n", "date": "2022-07-04", "categories": []}
{"id": "2206.10261", "title": "Interpretable Deep Causal Learning for Moderation Effects", "abstract": "  In this extended abstract paper, we address the problem of interpretability\nand targeted regularization in causal machine learning models. In particular,\nwe focus on the problem of estimating individual causal/treatment effects under\nobserved confounders, which can be controlled for and moderate the effect of\nthe treatment on the outcome of interest. Black-box ML models adjusted for the\ncausal setting perform generally well in this task, but they lack interpretable\noutput identifying the main drivers of treatment heterogeneity and their\nfunctional relationship. We propose a novel deep counterfactual learning\narchitecture for estimating individual treatment effects that can\nsimultaneously: i) convey targeted regularization on, and produce quantify\nuncertainty around the quantity of interest (i.e., the Conditional Average\nTreatment Effect); ii) disentangle baseline prognostic and moderating effects\nof the covariates and output interpretable score functions describing their\nrelationship with the outcome. Finally, we demonstrate the use of the method\nvia a simple simulated experiment.\n", "date": "2022-06-21", "categories": []}
{"id": "2207.05250", "title": "Efficient Real-world Testing of Causal Decision Making via Bayesian\n  Experimental Design for Contextual Optimisation", "abstract": "  The real-world testing of decisions made using causal machine learning models\nis an essential prerequisite for their successful application. We focus on\nevaluating and improving contextual treatment assignment decisions: these are\npersonalised treatments applied to e.g. customers, each with their own\ncontextual information, with the aim of maximising a reward. In this paper we\nintroduce a model-agnostic framework for gathering data to evaluate and improve\ncontextual decision making through Bayesian Experimental Design. Specifically,\nour method is used for the data-efficient evaluation of the regret of past\ntreatment assignments. Unlike approaches such as A/B testing, our method avoids\nassigning treatments that are known to be highly sub-optimal, whilst engaging\nin some exploration to gather pertinent information. We achieve this by\nintroducing an information-based design objective, which we optimise\nend-to-end. Our method applies to discrete and continuous treatments. Comparing\nour information-theoretic approach to baselines in several simulation studies\ndemonstrates the superior performance of our proposed approach.\n", "date": "2022-07-12", "categories": []}
{"id": "2207.05259", "title": "Language-Based Causal Representation Learning", "abstract": "  Consider the finite state graph that results from a simple, discrete,\ndynamical system in which an agent moves in a rectangular grid picking up and\ndropping packages. Can the state variables of the problem, namely, the agent\nlocation and the package locations, be recovered from the structure of the\nstate graph alone without having access to information about the objects, the\nstructure of the states, or any background knowledge? We show that this is\npossible provided that the dynamics is learned over a suitable\ndomain-independent first-order causal language that makes room for objects and\nrelations that are not assumed to be known. The preference for the most compact\nrepresentation in the language that is compatible with the data provides a\nstrong and meaningful learning bias that makes this possible. The language of\nstructured causal models (SCMs) is the standard language for representing\n(static) causal models but in dynamic worlds populated by objects, first-order\ncausal languages such as those used in \"classical AI planning\" are required.\nWhile \"classical AI\" requires handcrafted representations, similar\nrepresentations can be learned from unstructured data over the same languages.\nIndeed, it is the languages and the preference for compact representations in\nthose languages that provide structure to the world, uncovering objects,\nrelations, and causes.\n", "date": "2022-07-12", "categories": []}
{"id": "2207.04049", "title": "Learning Causal Effects on Hypergraphs", "abstract": "  Hypergraphs provide an effective abstraction for modeling multi-way group\ninteractions among nodes, where each hyperedge can connect any number of nodes.\nDifferent from most existing studies which leverage statistical dependencies,\nwe study hypergraphs from the perspective of causality. Specifically, in this\npaper, we focus on the problem of individual treatment effect (ITE) estimation\non hypergraphs, aiming to estimate how much an intervention (e.g., wearing face\ncovering) would causally affect an outcome (e.g., COVID-19 infection) of each\nindividual node. Existing works on ITE estimation either assume that the\noutcome on one individual should not be influenced by the treatment assignments\non other individuals (i.e., no interference), or assume the interference only\nexists between pairs of connected individuals in an ordinary graph. We argue\nthat these assumptions can be unrealistic on real-world hypergraphs, where\nhigher-order interference can affect the ultimate ITE estimations due to the\npresence of group interactions. In this work, we investigate high-order\ninterference modeling, and propose a new causality learning framework powered\nby hypergraph neural networks. Extensive experiments on real-world hypergraphs\nverify the superiority of our framework over existing baselines.\n", "date": "2022-07-07", "categories": []}
{"id": "2207.04055", "title": "Causal Discovery using Model Invariance through Knockoff Interventions", "abstract": "  Cause-effect analysis is crucial to understand the underlying mechanism of a\nsystem. We propose to exploit model invariance through interventions on the\npredictors to infer causality in nonlinear multivariate systems of time series.\nWe model nonlinear interactions in time series using DeepAR and then expose the\nmodel to different environments using Knockoffs-based interventions to test\nmodel invariance. Knockoff samples are pairwise exchangeable, in-distribution\nand statistically null variables generated without knowing the response. We\ntest model invariance where we show that the distribution of the response\nresidual does not change significantly upon interventions on non-causal\npredictors. We evaluate our method on real and synthetically generated time\nseries. Overall our method outperforms other widely used causality methods,\ni.e, VAR Granger causality, VARLiNGAM and PCMCI+.\n", "date": "2022-07-08", "categories": []}
{"id": "2207.04607", "title": "A Penalty Approach for Normalizing Feature Distributions to Build\n  Confounder-Free Models", "abstract": "  Translating machine learning algorithms into clinical applications requires\naddressing challenges related to interpretability, such as accounting for the\neffect of confounding variables (or metadata). Confounding variables affect the\nrelationship between input training data and target outputs. When we train a\nmodel on such data, confounding variables will bias the distribution of the\nlearned features. A recent promising solution, MetaData Normalization (MDN),\nestimates the linear relationship between the metadata and each feature based\non a non-trainable closed-form solution. However, this estimation is confined\nby the sample size of a mini-batch and thereby may cause the approach to be\nunstable during training. In this paper, we extend the MDN method by applying a\nPenalty approach (referred to as PDMN). We cast the problem into a bi-level\nnested optimization problem. We then approximate this optimization problem\nusing a penalty method so that the linear parameters within the MDN layer are\ntrainable and learned on all samples. This enables PMDN to be plugged into any\narchitectures, even those unfit to run batch-level operations, such as\ntransformers and recurrent models. We show improvement in model accuracy and\ngreater independence from confounders using PMDN over MDN in a synthetic\nexperiment and a multi-label, multi-site dataset of magnetic resonance images\n(MRIs).\n", "date": "2022-07-11", "categories": []}
{"id": "2207.02917", "title": "On The Universality of Diagrams for Causal Inference and The Causal\n  Reproducing Property", "abstract": "  We propose Universal Causality, an overarching framework based on category\ntheory that defines the universal property that underlies causal inference\nindependent of the underlying representational formalism used. More formally,\nuniversal causal models are defined as categories consisting of objects and\nmorphisms between them representing causal influences, as well as structures\nfor carrying out interventions (experiments) and evaluating their outcomes\n(observations). Functors map between categories, and natural transformations\nmap between a pair of functors across the same two categories. Abstract causal\ndiagrams in our framework are built using universal constructions from category\ntheory, including the limit or co-limit of an abstract causal diagram, or more\ngenerally, the Kan extension. We present two foundational results in universal\ncausal inference. The first result, called the Universal Causality Theorem\n(UCT), pertains to the universality of diagrams, which are viewed as functors\nmapping both objects and relationships from an indexing category of abstract\ncausal diagrams to an actual causal model whose nodes are labeled by random\nvariables, and edges represent functional or probabilistic relationships. UCT\nstates that any causal inference can be represented in a canonical way as the\nco-limit of an abstract causal diagram of representable objects. UCT follows\nfrom a basic result in the theory of sheaves. The second result, the Causal\nReproducing Property (CRP), states that any causal influence of a object X on\nanother object Y is representable as a natural transformation between two\nabstract causal diagrams. CRP follows from the Yoneda Lemma, one of the deepest\nresults in category theory. The CRP property is analogous to the reproducing\nproperty in Reproducing Kernel Hilbert Spaces that served as the foundation for\nkernel methods in machine learning.\n", "date": "2022-07-06", "categories": []}
{"id": "2207.01603", "title": "Invariant and Transportable Representations for Anti-Causal Domain\n  Shifts", "abstract": "  Real-world classification problems must contend with domain shift, the\n(potential) mismatch between the domain where a model is deployed and the\ndomain(s) where the training data was gathered. Methods to handle such problems\nmust specify what structure is common between the domains and what varies. A\nnatural assumption is that causal (structural) relationships are invariant in\nall domains. Then, it is tempting to learn a predictor for label $Y$ that\ndepends only on its causal parents. However, many real-world problems are\n\"anti-causal\" in the sense that $Y$ is a cause of the covariates $X$ -- in this\ncase, $Y$ has no causal parents and the naive causal invariance is useless. In\nthis paper, we study representation learning under a particular notion of\ndomain shift that both respects causal invariance and that naturally handles\nthe \"anti-causal\" structure. We show how to leverage the shared causal\nstructure of the domains to learn a representation that both admits an\ninvariant predictor and that also allows fast adaptation in new domains. The\nkey is to translate causal assumptions into learning principles that\ndisentangle \"invariant\" and \"non-stable\" features. Experiments on both\nsynthetic and real-world data demonstrate the effectiveness of the proposed\nlearning algorithm. Code is available at https://github.com/ybjiaang/ACTIR.\n", "date": "2022-07-04", "categories": []}
{"id": "2206.12689", "title": "Improving Data-driven Heterogeneous Treatment Effect Estimation Under\n  Structure Uncertainty", "abstract": "  Estimating how a treatment affects units individually, known as heterogeneous\ntreatment effect (HTE) estimation, is an essential part of decision-making and\npolicy implementation. The accumulation of large amounts of data in many\ndomains, such as healthcare and e-commerce, has led to increased interest in\ndeveloping data-driven algorithms for estimating heterogeneous effects from\nobservational and experimental data. However, these methods often make strong\nassumptions about the observed features and ignore the underlying causal model\nstructure, which can lead to biased HTE estimation. At the same time,\naccounting for the causal structure of real-world data is rarely trivial since\nthe causal mechanisms that gave rise to the data are typically unknown. To\naddress this problem, we develop a feature selection method that considers each\nfeature's value for HTE estimation and learns the relevant parts of the causal\nstructure from data. We provide strong empirical evidence that our method\nimproves existing data-driven HTE estimation methods under arbitrary underlying\ncausal structures. Our results on synthetic, semi-synthetic, and real-world\ndatasets show that our feature selection algorithm leads to lower HTE\nestimation error.\n", "date": "2022-06-25", "categories": []}
{"id": "2206.11529", "title": "Explanatory causal effects for model agnostic explanations", "abstract": "  This paper studies the problem of estimating the contributions of features to\nthe prediction of a specific instance by a machine learning model and the\noverall contribution of a feature to the model. The causal effect of a feature\n(variable) on the predicted outcome reflects the contribution of the feature to\na prediction very well. A challenge is that most existing causal effects cannot\nbe estimated from data without a known causal graph. In this paper, we define\nan explanatory causal effect based on a hypothetical ideal experiment. The\ndefinition brings several benefits to model agnostic explanations. First,\nexplanations are transparent and have causal meanings. Second, the explanatory\ncausal effect estimation can be data driven. Third, the causal effects provide\nboth a local explanation for a specific prediction and a global explanation\nshowing the overall importance of a feature in a predictive model. We further\npropose a method using individual and combined variables based on explanatory\ncausal effects for explanations. We show the definition and the method work\nwith experiments on some real-world data sets.\n", "date": "2022-06-23", "categories": []}
{"id": "2206.11646", "title": "Invariant Causal Mechanisms through Distribution Matching", "abstract": "  Learning representations that capture the underlying data generating process\nis a key problem for data efficient and robust use of neural networks. One key\nproperty for robustness which the learned representation should capture and\nwhich recently received a lot of attention is described by the notion of\ninvariance. In this work we provide a causal perspective and new algorithm for\nlearning invariant representations. Empirically we show that this algorithm\nworks well on a diverse set of tasks and in particular we observe\nstate-of-the-art performance on domain generalization, where we are able to\nsignificantly boost the score of existing models.\n", "date": "2022-06-23", "categories": []}
{"id": "2206.11131", "title": "Variational Causal Dynamics: Discovering Modular World Models from\n  Interventions", "abstract": "  Latent world models allow agents to reason about complex environments with\nhigh-dimensional observations. However, adapting to new environments and\neffectively leveraging previous knowledge remain significant challenges. We\npresent variational causal dynamics (VCD), a structured world model that\nexploits the invariance of causal mechanisms across environments to achieve\nfast and modular adaptation. By causally factorising a transition model, VCD is\nable to identify reusable components across different environments. This is\nachieved by combining causal discovery and variational inference to learn a\nlatent representation and transition model jointly in an unsupervised manner.\nSpecifically, we optimise the evidence lower bound jointly over a\nrepresentation model and a transition model structured as a causal graphical\nmodel. In evaluations on simulated environments with state and image\nobservations, we show that VCD is able to successfully identify causal\nvariables, and to discover consistent causal structures across different\nenvironments. Moreover, given a small number of observations in a previously\nunseen, intervened environment, VCD is able to identify the sparse changes in\nthe dynamics and to adapt efficiently. In doing so, VCD significantly extends\nthe capabilities of the current state-of-the-art in latent world models while\nalso comparing favourably in terms of prediction accuracy.\n", "date": "2022-06-22", "categories": []}
{"id": "2206.09186", "title": "Causal Inference with Treatment Measurement Error: A Nonparametric\n  Instrumental Variable Approach", "abstract": "  We propose a kernel-based nonparametric estimator for the causal effect when\nthe cause is corrupted by error. We do so by generalizing estimation in the\ninstrumental variable setting. Despite significant work on regression with\nmeasurement error, additionally handling unobserved confounding in the\ncontinuous setting is non-trivial: we have seen little prior work. As a\nby-product of our investigation, we clarify a connection between mean\nembeddings and characteristic functions, and how learning one simultaneously\nallows one to learn the other. This opens the way for kernel method research to\nleverage existing results in characteristic function estimation. Finally, we\nempirically show that our proposed method, MEKIV, improves over baselines and\nis robust under changes in the strength of measurement error and to the type of\nerror distributions.\n", "date": "2022-06-18", "categories": []}
{"id": "2206.08448", "title": "Empirical Bayesian Approaches for Robust Constraint-based Causal\n  Discovery under Insufficient Data", "abstract": "  Causal discovery is to learn cause-effect relationships among variables given\nobservational data and is important for many applications. Existing causal\ndiscovery methods assume data sufficiency, which may not be the case in many\nreal world datasets. As a result, many existing causal discovery methods can\nfail under limited data. In this work, we propose Bayesian-augmented\nfrequentist independence tests to improve the performance of constraint-based\ncausal discovery methods under insufficient data: 1) We firstly introduce a\nBayesian method to estimate mutual information (MI), based on which we propose\na robust MI based independence test; 2) Secondly, we consider the Bayesian\nestimation of hypothesis likelihood and incorporate it into a well-defined\nstatistical test, resulting in a robust statistical testing based independence\ntest. We apply proposed independence tests to constraint-based causal discovery\nmethods and evaluate the performance on benchmark datasets with insufficient\nsamples. Experiments show significant performance improvement in terms of both\naccuracy and efficiency over SOTA methods.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.08531", "title": "Reframed GES with a Neural Conditional Dependence Measure", "abstract": "  In a nonparametric setting, the causal structure is often identifiable only\nup to Markov equivalence, and for the purpose of causal inference, it is useful\nto learn a graphical representation of the Markov equivalence class (MEC). In\nthis paper, we revisit the Greedy Equivalence Search (GES) algorithm, which is\nwidely cited as a score-based algorithm for learning the MEC of the underlying\ncausal structure. We observe that in order to make the GES algorithm consistent\nin a nonparametric setting, it is not necessary to design a scoring metric that\nevaluates graphs. Instead, it suffices to plug in a consistent estimator of a\nmeasure of conditional dependence to guide the search. We therefore present a\nreframing of the GES algorithm, which is more flexible than the standard\nscore-based version and readily lends itself to the nonparametric setting with\na general measure of conditional dependence. In addition, we propose a neural\nconditional dependence (NCD) measure, which utilizes the expressive power of\ndeep neural networks to characterize conditional independence in a\nnonparametric manner. We establish the optimality of the reframed GES algorithm\nunder standard assumptions and the consistency of using our NCD estimator to\ndecide conditional independence. Together these results justify the proposed\napproach. Experimental results demonstrate the effectiveness of our method in\ncausal discovery, as well as the advantages of using our NCD measure over\nkernel-based measures.\n", "date": "2022-06-17", "categories": []}
{"id": "2206.08311", "title": "Continuous-Time Modeling of Counterfactual Outcomes Using Neural\n  Controlled Differential Equations", "abstract": "  Estimating counterfactual outcomes over time has the potential to unlock\npersonalized healthcare by assisting decision-makers to answer ''what-iF''\nquestions. Existing causal inference approaches typically consider regular,\ndiscrete-time intervals between observations and treatment decisions and hence\nare unable to naturally model irregularly sampled data, which is the common\nsetting in practice. To handle arbitrary observation patterns, we interpret the\ndata as samples from an underlying continuous-time process and propose to model\nits latent trajectory explicitly using the mathematics of controlled\ndifferential equations. This leads to a new approach, the Treatment Effect\nNeural Controlled Differential Equation (TE-CDE), that allows the potential\noutcomes to be evaluated at any time point. In addition, adversarial training\nis used to adjust for time-dependent confounding which is critical in\nlongitudinal settings and is an added challenge not encountered in conventional\ntime-series. To assess solutions to this problem, we propose a controllable\nsimulation environment based on a model of tumor growth for a range of\nscenarios with irregular sampling reflective of a variety of clinical\nscenarios. TE-CDE consistently outperforms existing approaches in all simulated\nscenarios with irregular sampling.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.08353", "title": "Towards Understanding How Machines Can Learn Causal Overhypotheses", "abstract": "  Recent work in machine learning and cognitive science has suggested that\nunderstanding causal information is essential to the development of\nintelligence. The extensive literature in cognitive science using the ``blicket\ndetector'' environment shows that children are adept at many kinds of causal\ninference and learning. We propose to adapt that environment for machine\nlearning agents. One of the key challenges for current machine learning\nalgorithms is modeling and understanding causal overhypotheses: transferable\nabstract hypotheses about sets of causal relationships. In contrast, even young\nchildren spontaneously learn and use causal overhypotheses. In this work, we\npresent a new benchmark -- a flexible environment which allows for the\nevaluation of existing techniques under variable causal overhypotheses -- and\ndemonstrate that many existing state-of-the-art methods have trouble\ngeneralizing in this environment. The code and resources for this benchmark are\navailable at https://github.com/CannyLab/casual_overhypotheses.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.08363", "title": "Benchmarking Heterogeneous Treatment Effect Models through the Lens of\n  Interpretability", "abstract": "  Estimating personalized effects of treatments is a complex, yet pervasive\nproblem. To tackle it, recent developments in the machine learning (ML)\nliterature on heterogeneous treatment effect estimation gave rise to many\nsophisticated, but opaque, tools: due to their flexibility, modularity and\nability to learn constrained representations, neural networks in particular\nhave become central to this literature. Unfortunately, the assets of such black\nboxes come at a cost: models typically involve countless nontrivial operations,\nmaking it difficult to understand what they have learned. Yet, understanding\nthese models can be crucial -- in a medical context, for example, discovered\nknowledge on treatment effect heterogeneity could inform treatment prescription\nin clinical practice. In this work, we therefore use post-hoc feature\nimportance methods to identify features that influence the model's predictions.\nThis allows us to evaluate treatment effect estimators along a new and\nimportant dimension that has been overlooked in previous work: We construct a\nbenchmarking environment to empirically investigate the ability of personalized\ntreatment effect models to identify predictive covariates -- covariates that\ndetermine differential responses to treatment. Our benchmarking environment\nthen enables us to provide new insight into the strengths and weaknesses of\ndifferent types of treatment effects models as we modulate different challenges\nspecific to treatment effect estimation -- e.g. the ratio of prognostic to\npredictive information, the possible nonlinearity of potential outcomes and the\npresence and type of confounding.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.07195", "title": "Tearing Apart NOTEARS: Controlling the Graph Prediction via Variance\n  Manipulation", "abstract": "  Simulations are ubiquitous in machine learning. Especially in graph learning,\nsimulations of Directed Acyclic Graphs (DAG) are being deployed for evaluating\nnew algorithms. In the literature, it was recently argued that\ncontinuous-optimization approaches to structure discovery such as NOTEARS might\nbe exploiting the sortability of the variable's variances in the available data\ndue to their use of least square losses. Specifically, since structure\ndiscovery is a key problem in science and beyond, we want to be invariant to\nthe scale being used for measuring our data (e.g. meter versus centimeter\nshould not affect the causal direction inferred by the algorithm). In this\nwork, we further strengthen this initial, negative empirical suggestion by both\nproving key results in the multivariate case and corroborating with further\nempirical evidence. In particular, we show that we can control the resulting\ngraph with our targeted variance attacks, even in the case where we can only\npartially manipulate the variances of the data.\n", "date": "2022-06-14", "categories": []}
{"id": "2206.07386", "title": "Finite-Sample Guarantees for High-Dimensional DML", "abstract": "  Debiased machine learning (DML) offers an attractive way to estimate\ntreatment effects in observational settings, where identification of causal\nparameters requires a conditional independence or unconfoundedness assumption,\nsince it allows to control flexibly for a potentially very large number of\ncovariates. This paper gives novel finite-sample guarantees for joint inference\non high-dimensional DML, bounding how far the finite-sample distribution of the\nestimator is from its asymptotic Gaussian approximation. These guarantees are\nuseful to applied researchers, as they are informative about how far off the\ncoverage of joint confidence bands can be from the nominal level. There are\nmany settings where high-dimensional causal parameters may be of interest, such\nas the ATE of many treatment profiles, or the ATE of a treatment on many\noutcomes. We also cover infinite-dimensional parameters, such as impacts on the\nentire marginal distribution of potential outcomes. The finite-sample\nguarantees in this paper complement the existing results on consistency and\nasymptotic normality of DML estimators, which are either asymptotic or treat\nonly the one-dimensional case.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.06469", "title": "Invariant Structure Learning for Better Generalization and Causal\n  Explainability", "abstract": "  Learning the causal structure behind data is invaluable for improving\ngeneralization and obtaining high-quality explanations. We propose a novel\nframework, Invariant Structure Learning (ISL), that is designed to improve\ncausal structure discovery by utilizing generalization as an indication. ISL\nsplits the data into different environments, and learns a structure that is\ninvariant to the target across different environments by imposing a consistency\nconstraint. An aggregation mechanism then selects the optimal classifier based\non a graph structure that reflects the causal mechanisms in the data more\naccurately compared to the structures learnt from individual environments.\nFurthermore, we extend ISL to a self-supervised learning setting where accurate\ncausal structure discovery does not rely on any labels. This self-supervised\nISL utilizes invariant causality proposals by iteratively setting different\nnodes as targets. On synthetic and real-world datasets, we demonstrate that ISL\naccurately discovers the causal structure, outperforms alternative methods, and\nyields superior generalization for datasets with significant distribution\nshifts.\n", "date": "2022-06-13", "categories": []}
{"id": "2206.06685", "title": "Causal Discovery for Fairness", "abstract": "  It is crucial to consider the social and ethical consequences of AI and ML\nbased decisions for the safe and acceptable use of these emerging technologies.\nFairness, in particular, guarantees that the ML decisions do not result in\ndiscrimination against individuals or minorities. Identifying and measuring\nreliably fairness/discrimination is better achieved using causality which\nconsiders the causal relation, beyond mere association, between the sensitive\nattribute (e.g. gender, race, religion, etc.) and the decision (e.g. job\nhiring, loan granting, etc.). The big impediment to the use of causality to\naddress fairness, however, is the unavailability of the causal model (typically\nrepresented as a causal graph). Existing causal approaches to fairness in the\nliterature do not address this problem and assume that the causal model is\navailable. In this paper, we do not make such assumption and we review the\nmajor algorithms to discover causal relations from observable data. This study\nfocuses on causal discovery and its impact on fairness. In particular, we show\nhow different causal discovery approaches may result in different causal models\nand, most importantly, how even slight differences between causal models can\nhave significant impact on fairness/discrimination conclusions. These results\nare consolidated by empirical analysis using synthetic and standard fairness\nbenchmark datasets. The main goal of this study is to highlight the importance\nof the causal discovery step to appropriately address fairness using causality.\n", "date": "2022-06-14", "categories": []}
{"id": "2206.06699", "title": "Generalizing experimental findings: identification beyond adjustments", "abstract": "  We aim to generalize the results of a randomized controlled trial (RCT) to a\ntarget population with the help of some observational data. This is a problem\nof causal effect identification with multiple data sources. Challenges arise\nwhen the RCT is conducted in a context that differs from the target population.\nEarlier research has focused on cases where the estimates from the RCT can be\nadjusted by observational data in order to remove the selection bias and other\ndomain specific differences. We consider examples where the experimental\nfindings cannot be generalized by an adjustment and show that the\ngeneralization may still be possible by other identification strategies that\ncan be derived by applying do-calculus. The obtained identifying functionals\nfor these examples contain trapdoor variables of a new type. The value of a\ntrapdoor variable needs to be fixed in the estimation and the choice of the\nvalue may have a major effect on the bias and accuracy of estimates, which is\nalso seen in simulations. The presented results expand the scope of settings\nwhere the generalization of experimental findings is doable\n", "date": "2022-06-14", "categories": []}
{"id": "2206.00516", "title": "Feature Selection for Discovering Distributional Treatment Effect\n  Modifiers", "abstract": "  Finding the features relevant to the difference in treatment effects is\nessential to unveil the underlying causal mechanisms. Existing methods seek\nsuch features by measuring how greatly the feature attributes affect the degree\nof the {\\it conditional average treatment effect} (CATE). However, these\nmethods may overlook important features because CATE, a measure of the average\ntreatment effect, cannot detect differences in distribution parameters other\nthan the mean (e.g., variance). To resolve this weakness of existing methods,\nwe propose a feature selection framework for discovering {\\it distributional\ntreatment effect modifiers}. We first formulate a feature importance measure\nthat quantifies how strongly the feature attributes influence the discrepancy\nbetween potential outcome distributions. Then we derive its computationally\nefficient estimator and develop a feature selection algorithm that can control\nthe type I error rate to the desired level. Experimental results show that our\nframework successfully discovers important features and outperforms the\nexisting mean-based method.\n", "date": "2022-06-01", "categories": []}
{"id": "2206.05421", "title": "Greedy Relaxations of the Sparsest Permutation Algorithm", "abstract": "  There has been an increasing interest in methods that exploit permutation\nreasoning to search for directed acyclic causal models, including the \"Ordering\nSearch\" of Teyssier and Kohler and GSP of Solus, Wang and Uhler. We extend the\nmethods of the latter by a permutation-based operation, tuck, and develop a\nclass of algorithms, namely GRaSP, that are efficient and pointwise consistent\nunder increasingly weaker assumptions than faithfulness. The most relaxed form\nof GRaSP outperforms many state-of-the-art causal search algorithms in\nsimulation, allowing efficient and accurate search even for dense graphs and\ngraphs with more than 100 variables.\n", "date": "2022-06-11", "categories": []}
{"id": "2206.05871", "title": "Causal Inference-Based Root Cause Analysis for Online Service Systems\n  with Intervention Recognition", "abstract": "  Fault diagnosis is critical in many domains, as faults may lead to safety\nthreats or economic losses. In the field of online service systems, operators\nrely on enormous monitoring data to detect and mitigate failures. Quickly\nrecognizing a small set of root cause indicators for the underlying fault can\nsave much time for failure mitigation. In this paper, we formulate the root\ncause analysis problem as a new causal inference task named intervention\nrecognition. We proposed a novel unsupervised causal inference-based method\nnamed Causal Inference-based Root Cause Analysis (CIRCA). The core idea is a\nsufficient condition for a monitoring variable to be a root cause indicator,\ni.e., the change of probability distribution conditioned on the parents in the\nCausal Bayesian Network (CBN). Towards the application in online service\nsystems, CIRCA constructs a graph among monitoring metrics based on the\nknowledge of system architecture and a set of causal assumptions. The\nsimulation study illustrates the theoretical reliability of CIRCA. The\nperformance on a real-world dataset further shows that CIRCA can improve the\nrecall of the top-1 recommendation by 25% over the best baseline method.\n", "date": "2022-06-13", "categories": []}
{"id": "2206.06124", "title": "Causal Discovery in Hawkes Processes by Minimum Description Length", "abstract": "  Hawkes processes are a special class of temporal point processes which\nexhibit a natural notion of causality, as occurrence of events in the past may\nincrease the probability of events in the future. Discovery of the underlying\ninfluence network among the dimensions of multi-dimensional temporal processes\nis of high importance in disciplines where a high-frequency data is to model,\ne.g. in financial data or in seismological data. This paper approaches the\nproblem of learning Granger-causal network in multi-dimensional Hawkes\nprocesses. We formulate this problem as a model selection task in which we\nfollow the minimum description length (MDL) principle. Moreover, we propose a\ngeneral algorithm for MDL-based inference using a Monte-Carlo method and we use\nit for our causal discovery problem. We compare our algorithm with the\nstate-of-the-art baseline methods on synthetic and real-world financial data.\nThe synthetic experiments demonstrate superiority of our method incausal graph\ndiscovery compared to the baseline methods with respect to the size of the\ndata. The results of experiments with the G-7 bonds price data are consistent\nwith the experts knowledge.\n", "date": "2022-06-10", "categories": []}
{"id": "2206.04907", "title": "Efficient Heterogeneous Treatment Effect Estimation With Multiple\n  Experiments and Multiple Outcomes", "abstract": "  Learning heterogeneous treatment effects (HTEs) is an important problem\nacross many fields. Most existing methods consider the setting with a single\ntreatment arm and a single outcome metric. However, in many real world domains,\nexperiments are run consistently - for example, in internet companies, A/B\ntests are run every day to measure the impacts of potential changes across many\ndifferent metrics of interest. We show that even if an analyst cares only about\nthe HTEs in one experiment for one metric, precision can be improved greatly by\nanalyzing all of the data together to take advantage of cross-experiment and\ncross-outcome metric correlations. We formalize this idea in a tensor\nfactorization framework and propose a simple and scalable model which we refer\nto as the low rank or LR-learner. Experiments in both synthetic and real data\nsuggest that the LR-learner can be much more precise than independent HTE\nestimation.\n", "date": "2022-06-10", "categories": []}
{"id": "2206.04620", "title": "On the Generalization and Adaption Performance of Causal Models", "abstract": "  Learning models that offer robust out-of-distribution generalization and fast\nadaptation is a key challenge in modern machine learning. Modelling causal\nstructure into neural networks holds the promise to accomplish robust zero and\nfew-shot adaptation. Recent advances in differentiable causal discovery have\nproposed to factorize the data generating process into a set of modules, i.e.\none module for the conditional distribution of every variable where only causal\nparents are used as predictors. Such a modular decomposition of knowledge\nenables adaptation to distributions shifts by only updating a subset of\nparameters. In this work, we systematically study the generalization and\nadaption performance of such modular neural causal models by comparing it to\nmonolithic models and structured models where the set of predictors is not\nconstrained to causal parents. Our analysis shows that the modular neural\ncausal models outperform other models on both zero and few-shot adaptation in\nlow data regimes and offer robust generalization. We also found that the\neffects are more significant for sparser graphs as compared to denser graphs.\n", "date": "2022-06-09", "categories": []}
{"id": "2206.02962", "title": "Confounder Analysis in Measuring Representation in Product Funnels", "abstract": "  This paper discusses an application of Shapley values in the causal inference\nfield, specifically on how to select the top confounder variables for coarsened\nexact matching method in a scalable way. We use a dataset from an observational\nexperiment involving LinkedIn members as a use case to test its applicability,\nand show that Shapley values are highly informational and can be leveraged for\nits robust importance-ranking capability.\n", "date": "2022-06-07", "categories": []}
{"id": "2206.01931", "title": "Discovering Ancestral Instrumental Variables for Causal Inference from\n  Observational Data", "abstract": "  Instrumental variable (IV) is a powerful approach to inferring the causal\neffect of a treatment on an outcome of interest from observational data even\nwhen there exist latent confounders between the treatment and the outcome.\nHowever, existing IV methods require that an IV is selected and justified with\ndomain knowledge. An invalid IV may lead to biased estimates. Hence,\ndiscovering a valid IV is critical to the applications of IV methods. In this\npaper, we study and design a data-driven algorithm to discover valid IVs from\ndata under mild assumptions. We develop the theory based on partial ancestral\ngraphs (PAGs) to support the search for a set of candidate Ancestral IVs\n(AIVs), and for each possible AIV, the identification of its conditioning set.\nBased on the theory, we propose a data-driven algorithm to discover a pair of\nIVs from data. The experiments on synthetic and real-world datasets show that\nthe developed IV discovery algorithm estimates accurate estimates of causal\neffects in comparison with the state-of-the-art IV based causal effect\nestimators.\n", "date": "2022-06-04", "categories": []}
{"id": "2206.02216", "title": "Sequential Counterfactual Decision-Making Under Confounded Reward", "abstract": "  We investigate the limitations of random trials when the cause of interest is\nconfounded with the effect by formalizing a counterfactual policy-space where\nthe agent's natural predilection is input to a soft-intervention.\n", "date": "2022-06-05", "categories": []}
{"id": "2206.01496", "title": "Causality Learning With Wasserstein Generative Adversarial Networks", "abstract": "  Conventional methods for causal structure learning from data face significant\nchallenges due to combinatorial search space. Recently, the problem has been\nformulated into a continuous optimization framework with an acyclicity\nconstraint to learn Directed Acyclic Graphs (DAGs). Such a framework allows the\nutilization of deep generative models for causal structure learning to better\ncapture the relations between data sample distributions and DAGs. However, so\nfar no study has experimented with the use of Wasserstein distance in the\ncontext of causal structure learning. Our model named DAG-WGAN combines the\nWasserstein-based adversarial loss with an acyclicity constraint in an\nauto-encoder architecture. It simultaneously learns causal structures while\nimproving its data generation capability. We compare the performance of\nDAG-WGAN with other models that do not involve the Wasserstein metric in order\nto identify its contribution to causal structure learning. Our model performs\nbetter with high cardinality data according to our experiments.\n", "date": "2022-06-03", "categories": []}
{"id": "2206.01562", "title": "Prescriptive maintenance with causal machine learning", "abstract": "  Machine maintenance is a challenging operational problem, where the goal is\nto plan sufficient preventive maintenance to avoid machine failures and\noverhauls. Maintenance is often imperfect in reality and does not make the\nasset as good as new. Although a variety of imperfect maintenance policies have\nbeen proposed in the literature, these rely on strong assumptions regarding the\neffect of maintenance on the machine's condition, assuming the effect is (1)\ndeterministic or governed by a known probability distribution, and (2)\nmachine-independent. This work proposes to relax both assumptions by learning\nthe effect of maintenance conditional on a machine's characteristics from\nobservational data on similar machines using existing methodologies for causal\ninference. By predicting the maintenance effect, we can estimate the number of\noverhauls and failures for different levels of maintenance and, consequently,\noptimize the preventive maintenance frequency to minimize the total estimated\ncost. We validate our proposed approach using real-life data on more than 4,000\nmaintenance contracts from an industrial partner. Empirical results show that\nour novel, causal approach accurately predicts the maintenance effect and\nresults in individualized maintenance schedules that are more accurate and\ncost-effective than supervised or non-individualized approaches.\n", "date": "2022-06-03", "categories": []}
{"id": "2206.00885", "title": "Coordinated Double Machine Learning", "abstract": "  Double machine learning is a statistical method for leveraging complex\nblack-box models to construct approximately unbiased treatment effect estimates\ngiven observational data with high-dimensional covariates, under the assumption\nof a partially linear model. The idea is to first fit on a subset of the\nsamples two non-linear predictive models, one for the continuous outcome of\ninterest and one for the observed treatment, and then to estimate a linear\ncoefficient for the treatment using the remaining samples through a simple\northogonalized regression. While this methodology is flexible and can\naccommodate arbitrary predictive models, typically trained independently of one\nanother, this paper argues that a carefully coordinated learning algorithm for\ndeep neural networks may reduce the estimation bias. The improved empirical\nperformance of the proposed method is demonstrated through numerical\nexperiments on both simulated and real data.\n", "date": "2022-06-02", "categories": []}
{"id": "2206.01022", "title": "Learning Disentangled Representations for Counterfactual Regression via\n  Mutual Information Minimization", "abstract": "  Learning individual-level treatment effect is a fundamental problem in causal\ninference and has received increasing attention in many areas, especially in\nthe user growth area which concerns many internet companies. Recently,\ndisentangled representation learning methods that decompose covariates into\nthree latent factors, including instrumental, confounding and adjustment\nfactors, have witnessed great success in treatment effect estimation. However,\nit remains an open problem how to learn the underlying disentangled factors\nprecisely. Specifically, previous methods fail to obtain independent\ndisentangled factors, which is a necessary condition for identifying treatment\neffect. In this paper, we propose Disentangled Representations for\nCounterfactual Regression via Mutual Information Minimization (MIM-DRCFR),\nwhich uses a multi-task learning framework to share information when learning\nthe latent factors and incorporates MI minimization learning criteria to ensure\nthe independence of these factors. Extensive experiments including public\nbenchmarks and real-world industrial user growth datasets demonstrate that our\nmethod performs much better than state-of-the-art methods.\n", "date": "2022-06-02", "categories": []}
{"id": "2206.01081", "title": "Revisiting the General Identifiability Problem", "abstract": "  We revisit the problem of general identifiability originally introduced in\n[Lee et al., 2019] for causal inference and note that it is necessary to add\npositivity assumption of observational distribution to the original definition\nof the problem. We show that without such an assumption the rules of\ndo-calculus and consequently the proposed algorithm in [Lee et al., 2019] are\nnot sound. Moreover, adding the assumption will cause the completeness proof in\n[Lee et al., 2019] to fail. Under positivity assumption, we present a new\nalgorithm that is provably both sound and complete. A nice property of this new\nalgorithm is that it establishes a connection between general identifiability\nand classical identifiability by Pearl [1995] through decomposing the general\nidentifiability problem into a series of classical identifiability\nsub-problems.\n", "date": "2022-06-02", "categories": []}
{"id": "2606.21185", "title": "Two Layers of Instability in Causal Estimation", "abstract": "There is a precise sense in which drawing causal inferences from observational data is hard, even when identifiability is assumed. In particular, Robins and Ritov (1997) and Robins et al. (2003) showed that causal effects can be discontinuous as a function of the data distribution: two arbitrarily close data distributions might correspond to different causal effects. This is a fact independent of the choice of estimator; however, not all estimators are equally unstable. Our contribution is to surface a second layer of instability that depends on the choice of estimator. We show that many standard point estimates can be read as point summaries of multimodal distributions over the space of structural causal models. As such, estimators can jump discontinuously in the data distribution. This defines a taxonomy of estimators that admits a decision-theoretic reading: stability depends on whether the implicit loss function an estimator optimizes is aligned with the causal effect itself. Specifically, inverse propensity weighted estimators and regression estimators are examples of discontinuous summaries, while explicit posterior means and medians are shown to be continuous.", "date": "2026-06-19", "categories": []}
{"id": "2606.21212", "title": "DCD-PFN: A Decoupling-Aware Foundation Model for Causal Discovery", "abstract": "Causal discovery is critical for understanding complex data-generating mechanisms, yet traditional algorithms often struggle with highly non-linear and noisy systems, or suffer from severe computational bottlenecks. Recent tabular foundation models based on Prior-Data Fitted Networks (PFNs) have demonstrated remarkable zero-shot inference capabilities, but their potential for explicit structural causal discovery remains underexplored. To bridge this gap, we propose DCD-PFN, a decoupling-aware foundation model for causal discovery. Instead of directly amortizing global graph reconstruction, DCD-PFN focuses on local causal discovery through a decoupling-based paradigm. Through pre-training on diverse synthetic Structural Causal Models (SCMs), the model learns sample-wise decoupling weights that enable Markov boundary (MB) identification. Furthermore, by leveraging parallelized local discovery, DCD-PFN efficiently reconstructs global causal graphs while remaining grounded in the theoretical foundations of decoupling-based causal discovery. Experiments demonstrate that our foundation model achieves robust zero-shot generalization.", "date": "2026-06-19", "categories": []}
{"id": "2606.21668", "title": "A Framework for Directed Acyclic Hypergraph Learning", "abstract": "Continuous optimization methods for learning Directed Acyclic Graphs (DAGs) operate on weighted adjacency matrices and are therefore limited to pairwise causal relationships. We propose a framework for learning Directed Acyclic Hypergraphs (DAHGs) from observational data, capturing joint parental influences that pairwise models cannot represent. Our approach rests on three components: (i) a generalized linear structural equation model (SEM) with multiplicative interaction terms whose non-zero weights correspond one-to-one with directed hyperedges; (ii) a weighted adjacency tensor representation whose acyclicity is characterized via nilpotency under the tensor t-product; and (iii) a differentiable acyclicity constraint derived through the Fourier decomposition of the t-product, which reduces tensor nilpotency to slice-wise matrix nilpotency and enables least-squares learning via the augmented Lagrangian method.", "date": "2026-06-19", "categories": []}
{"id": "2606.21806", "title": "Causal Variational Deep Embedding: A Family of Interventional Generators for Confounded Images", "abstract": "Deep generative models reproduce the observational distribution of their training data, inheriting any spurious associations it contains. A common source is an unobserved confounder that shapes both an attribute the user wants to control at sampling time and an attribute expected to vary in response. Existing causal generative approaches resolve the resulting ambiguity by imposing structural assumptions strong enough to single out one interventional distribution; in image domains, such assumptions are rarely warranted, and the data is generally consistent with a set of distinct causal mechanisms -- a feasible region of interventional distributions. We propose CauVaDE (Causal Variational Deep Embedding), built on a canonical augmented SCM in which the unobserved confounder collapses, without loss of generality, into a discrete latent cluster of bounded support while continuous variation is absorbed into independent noises. We prove that this canonical class is dense, in both observational and interventional Wasserstein distance, in the class of augmented SCMs compatible with a given causal diagram, and instantiate it as a mixture variational autoencoder whose cluster variable plays the role of the canonical confounder. An entropy regularizer with weight $\\gamma$ on the cluster posterior then traces a family of candidate causal effects that fit the observational data to comparable likelihood while spanning the feasible region. Experiments on image data benchmarks show that CauVaDE produces diverse interventional samples and improves FID against an unconfounded reference.", "date": "2026-06-19", "categories": []}
{"id": "2606.21809", "title": "Causal Gaussian Processes for Robust Treatment Effect Evaluation with Unobserved Confounding", "abstract": "The presence of confounding bias poses a key challenge in policy evaluation, as the target causal effects of actions are not identifiable (i.e., underdetermined) from observational data. On the other hand, existing confounding-robust evaluation strategies require detailed prior knowledge about the environment or apply only to discrete treatments and outcomes. This paper investigates causal effect evaluation over the continuous domain from confounded observations, while requiring only basic temporal ordering between the treatment and the outcome. We introduce a universal discretization of the exogenous domains that approximates the observational and interventional distributions of any causal model with arbitrary accuracy using a finite number of latent states. Building on this newfound universal approximation property, we develop a novel family of Causal Gaussian process (CGP) models that effectively approximate the observational and interventional distributions of any causal model with confounded observations.", "date": "2026-06-20", "categories": []}
{"id": "2606.21821", "title": "Local Causal Attribution of Chain-of-Thought Reasoning", "abstract": "Understanding the causal structure of a language model's thought process is a problem of significant importance for both transparency and safety. In this work, we take a local approach toward this goal by analyzing the causal relationships among individual components, termed units, of a given, specific chain-of-thought trace. We construct a structural causal model on these units and relate each unit to the log probability of generating (subsequent) output units. Our algorithm, termed AttriCoT, is a black-box method that performs attribution by estimating importance parameters in the structural causal model using $O(U)$ forward passes through the model, where $U$ is the number of units. Evaluation of perturbation curves across 5 datasets and 4 reasoning models shows that AttriCoT produces attributions that are more faithful to the model's behavior than alternative methods. The attribution results also reveal notable differences in thought structure between models and domains.", "date": "2026-06-20", "categories": []}
{"id": "2606.22430", "title": "Words as Difference Makers: How Large Language Models Determine Causal Structure in Text", "abstract": "Because large language models (LLMs) are impressively successful in predicting text, it appears that they must have access to a 'world model' representing causal and definitional structure. However, the dominant formalisms of modern causal inference -- Judea Pearl's interventionist approach and the Neyman-Rubin potential outcomes framework -- struggle to illuminate how LLMs learn causal structure. I resolve this puzzle by arguing that LLMs employ a specific inductive approach based on a difference-making logic -- sometimes called variational induction. I demonstrate how central aspects of this logic are realized during training, where LLMs require enormous amounts of text data from a wide range of contexts to identify difference- and indifference-makers within word sequences. Furthermore, I analyze specific architectural features of LLMs -- such as token embeddings and self-attention -- to determine their roles in variational induction. The difference-making logic of LLMs fundamentally parallels the experimental method, where causal relations are derived by systematically varying individual circumstances to determine their influence on a phenomenon.", "date": "2026-06-21", "categories": []}
{"id": "2606.22449", "title": "Self-Evolving Cognitive Framework via Causal World Modeling for Embodied Scientific Intelligence", "abstract": "Current embodied world models are primarily optimized for predictive objectives, limiting their ability to generalize under distribution shifts and reason systematically about unseen situations and hypothetical interventions. We argue that embodied intelligence should move beyond predictive world modeling toward self-evolving cognitive systems that continually construct and refine internal causal representations through interaction with the environment. To this end, we propose a self-evolving cognitive framework via causal world modeling for embodied scientific intelligence, which integrates three complementary components: causal world modeling, intervention-driven causal reasoning, and continual cognitive refinement. The proposed framework continuously revises and expands its internal causal world model through causal discovery, intervention-driven feedback, and counterfactual reasoning, supporting continual cognitive refinement and enabling cognition itself to evolve over time. Furthermore, we reinterpret embodied interaction not merely as a means of trajectory optimization, but as an epistemic process for causal hypothesis generation, intervention-driven experimentation, and continual knowledge acquisition. This work provides a conceptual and theoretical foundation for a transition from predictive intelligence toward epistemic intelligence, in which intelligence emerges through the continual construction, revision, and refinement of causal world models via interaction with the environment. Accordingly, an intervention-driven causal-epistemic benchmarking paradigm is suggested for evaluating self-evolving embodied scientific intelligence.", "date": "2026-06-21", "categories": []}
{"id": "2606.22510", "title": "Fed-CausalDiff: Decoupled Synchronization for Federated Do-Simulation and Policy Evaluation", "abstract": "While federated learning enables collaborative modelling on decentralised data, standard methods merely fit historical observations. This purely observational approach is fundamentally insufficient for interventional inference and policy evaluation, as sequential actions dynamically alter future states. We propose \\textbf{Fed-CausalDiff}, a federated causal diffusion framework for do-simulation. The architecture decomposes the evolution of the latent state into a global causal score function and a local confounding score function. This design enables \\emph{decoupled synchronisation} (DSS), where clients aggregate only the shared causal mechanism while retaining site-specific confounders locally to handle heterogeneity. Experiments on four datasets demonstrate that Fed-CausalDiff achieves better ATE and policy-value estimation accuracy, offering a favorable trade-off between communication cost and inference fidelity.", "date": "2026-06-21", "categories": []}
{"id": "2606.23741", "title": "A Survey on Federated Causal Discovery and Inference", "abstract": "Causal reasoning, which encompasses the discovery of causal structures and the inference of causal effects, is fundamental to data-driven decision making. In practice, data for reliable causal analysis are often distributed across institutions and cannot be centralized due to privacy regulations or communication constraints. Federated learning (FL) addresses this by enabling collaborative analysis without raw data sharing, giving rise to the rapidly growing field of federated causal discovery (FCD) and inference (FCI). However, the interdisciplinary nature of this field and the absence of a comprehensive survey present barriers to entry for researchers. This paper bridges that gap by providing a systematic review through multi-dimensional taxonomies. Grounded in the three core design decisions underlying any FCD solution, namely how structures are learned, how data are partitioned, and what structural knowledge each party obtains, we organize FCD along three axes: methodological paradigm, federation topology, and structural scope. We further examine key practical dimensions, including temporal dynamics, data heterogeneity, missing data, and non-identical variable sets. For FCI, we categorize methods by target estimand (average versus individualized/conditional treatment effects) and by estimation strategy, from classical weighting methods to modern deep generative architectures. Unlike prior works that treat FCD and FCI separately, we formalize their connection as complementary stages of a unified federated causal reasoning pipeline, where FCD supplies the structural knowledge required for valid effect estimation in FCI. Finally, we highlight their shared concerns regarding privacy, communication efficiency, theoretical guarantees, and application domains, and conclude by identifying open challenges for future research.", "date": "2026-06-21", "categories": []}
{"id": "2606.23608", "title": "Causal Discovery in the Era of Agents", "abstract": "Recent attempts to combine large language models (LLMs) with causal discovery ask models to infer pairwise directions, propose graph structures, or inject language-model outputs as priors and constraints. These approaches promise faster analysis, but they also obscure whether a causal evidence is supported by data and assumptions or by textual associations, prompt artifacts and hallucinated mechanisms. We argue for a different role for agents in causal discovery. Agents should inspect data, retrieve context, explain method assumptions and clarify graph outputs, but they should not supply edges, orientations, priors, constraints or causal conclusions. We propose the principle that agents assist the workflow, while causal claims remain grounded in data, explicit assumptions, formal algorithms, diagnostics and user or domain-expert decisions. We instantiate this principle in causal-learn+, an online platform that coordinates data analysis, preprocessing, method recommendation, expert-knowledge incorporation, formal discovery and interpretation around the algorithmic ecosystem of causal-learn. A case study on Big Five personality data illustrates agent-assisted pipeline of causal discovery without turning language-model unreliability into causal evidence. The platform is available at causallearn.com.", "date": "2026-06-22", "categories": []}
{"id": "2606.23767", "title": "One Ruler: A Same-Hands Re-Evaluation of Bivariate Causal Direction on Tuebingen, with a Parameter-Free Compression Baseline", "abstract": "Headline accuracies on the Tuebingen cause-effect pairs are routinely compared across papers even though each is measured under its authors' own protocol -- different pair subsets, weightings, model-selection, and decision rates. We argue this is the wrong comparison and run the right one: a same-hands re-evaluation in which every method is run by us on the identical 102 pairs, with one strict rule -- no tuning and a decision forced on every pair. As a clean reference point we introduce a deliberately minimal baseline: sorted-conditional compression, which feeds quantized, sorted, first-differenced data to an off-the-shelf compressor (bz2) and has zero fitted parameters. Under the common ruler the ranking differs sharply from the literature. Our baseline reaches 74.7% weighted accuracy (p = 3.7e-7); on the same 100 pairs that SLOPE is evaluated on it scores 76.0%, a 1.2-point gap below the authors' own forced-decision SLOPE (77.2%) that is well inside noise (McNemar p = 0.39). A faithful re-run of RECI lands at 70.7% -- inside the original authors' reported error bar, not the 77.5% often quoted (which we trace to a mis-copied cell). SLOPE's published 82.4% is a decided-subset figure: scoring the authors' own stored output only on the pairs its significance test chose to answer reproduces 81.7%. Under the common ruler the methods cluster in the low-to-mid 70s and the zero-parameter compressor ties the strongest of them. We document the mechanisms that inflate published figures (test-set model selection, significance-gated abstention) and contribute two further results: compression score magnitude is a model-free confounding flag (p = 2.8e-68), and a pre-registered falsification test fails in an instructive way that bounds the method's theoretical interpretation. Code, pre-registrations, and per-pair outputs are released.", "date": "2026-06-22", "categories": []}
{"id": "2606.23880", "title": "GRACE: Gated Refinement for Accurate Causal Edge Discovery in High-Dimensional Time Series", "abstract": "From climate teleconnections to gene regulation, modern time-series datasets encompass tens or hundreds of interacting variables, making causal discovery increasingly challenging. Constraint-based methods offer statistical rigor but their nonlinear CI tests are infeasible at scale, while score-based alternatives avoid CI testing but require arbitrary thresholds to binarize continuous edge scores. We propose GRACE ($\\textbf{G}$ated $\\textbf{R}$efinement for $\\textbf{A}$ccurate $\\textbf{C}$ausal $\\textbf{E}$dge discovery), which refines constraint-based discovery using Hard Concrete gates with $L_0$ regularization: each candidate edge has an independent gate whose values concentrate near 0 or 1, yielding a clean bimodal separation that makes the binary decision robust, unlike the narrow, overlapping score distributions produced by $L_1$ and attention-based methods. A fast linear CI skeleton provides high-recall candidates; a single gated model then prunes false positives by learning which edges genuinely improve prediction, with automatic regularization adapted to problem dimensions and skeleton density. Systematic experiments on synthetic benchmarks, spanning diverse graph topologies (scale-free, Erd\\H{o}s-R'enyi, small-world) and dimensionalities up to $d=100$, show that GRACE substantially improves F1 over its base CI method while maintaining high precision, and outperforms attention-based and score-based alternatives. GRACE matches or exceeds expensive nonlinear CI tests at a fraction of the cost ($75\\times$ faster). On a real-world river flow dataset, where rainfall confounders, variable propagation lags, and distributional shifts violate standard assumptions, a temporal bootstrap variant of GRACE recovers 9 of 11 causal edges along the Elbe River with only 1 false positive ($F_1 = 0.86$, AUROC${} = 0.99$), reducing the skeleton's 106 false positives by 99%.", "date": "2026-06-22", "categories": []}
{"id": "2606.24160", "title": "An Introduction to Causal Reinforcement Learning", "abstract": "Causal inference provides a set of principles and tools that allow one to combine data and knowledge about an environment to reason with questions of counterfactual nature, i.e., what would have happened had reality been different, even when no data of this unrealized reality is currently available. Reinforcement learning provides methods to learn a policy that optimizes a specific measure (e.g., reward, regret) when the agent is deployed in an environment and pursues an exploratory, trial-and-error approach. These two disciplines have evolved independently and with virtually no interaction between them. We note that they operate over different aspects of the same building block, counterfactual relations, which makes them umbilically connected. Based on these observations, novel learning opportunities arise when this connection is explicitly acknowledged and mathematized. To realize this potential, we note that any environment where the RL agent is deployed can be decomposed as a collection of autonomous mechanisms with different causal invariances, parsimoniously modeled as a structural causal model; any standard RL setting implicitly encodes such a model. This formalization allows us to put under a unifying treatment different modes of learning, including online, off-policy, and causal calculus learning, which appear unrelated in the literature. However, these modalities are not exhaustive: we introduce several natural and pervasive classes of learning settings that entail novel dimensions of analysis. Specifically, we introduce and discuss through causal lenses generalized policy learning, where to intervene, imitation learning, and counterfactual learning. These tasks lead to a broader view of counterfactual learning and suggest great potential for studying causal inference and reinforcement learning side by side, which we call causal reinforcement learning (CRL).", "date": "2026-06-23", "categories": []}
{"id": "2606.24488", "title": "RetiSEM: Generalising Causal Models for Fragmented Biomedical Data", "abstract": "Learning causal models from fragmented biomedical data is challenging because clinical, molecular, and imaging variables are often incomplete or not jointly observed. We propose RetiSEM, a domain-constrained structural equation modelling (SEM) framework for causal graph recovery and mediation analysis under limited multimodal resources. This proposed work organises variables into biologically informed blocks, applies forbidden-edge constraints, and decomposes pathway-level effects into TE, NDE, and NIE components. We evaluate RetiSEM across ten synthetic benchmark scenarios that vary in dimensionality, nonlinearity, causal depth, and pathway structure, together with a fragmented real-world setting that combines NHANES clinical variables with externally derived retinal representations. This approach achieves lower structural error and higher causal accuracy than unconstrained baselines across the synthetic benchmarks. In the real-data analysis, retinal variables behave mainly as downstream biomarker-like indicators, with smaller but detectable indirect effects. These findings support our strategy as an interpretable framework for testing structured causal hypotheses in limited-resource biomedical AI. The code and resources for this work are publicly available at: https://github.com/Inamullah-Colab/ReitSEM.", "date": "2026-06-23", "categories": []}
{"id": "2606.25762", "title": "OncoSynth: Synthetic data generation for treatment effect estimation in oncology", "abstract": "In oncology, access to patient-level data is often restricted. Synthetic data provides an alternative for analyzing treatment effectiveness, but existing methods for synthetic data generation fail to preserve the causal relationships between covariates, treatments, and outcomes, thereby leading to biased estimates of treatment effects. Here, we introduce OncoSynth, a generative, causally-aware machine learning framework designed to produce synthetic cohorts that enable accurate estimation of population- and patient-level treatment effects. OncoSynth uses a diffusion-based sequential approach to model how covariates influence treatment assignment and how treatment affects survival. We evaluate OncoSynth using large lung (N = 37,128) and breast cancer (N = 17,046) cohorts. Our results show that OncoSynth generates high-fidelity synthetic patient cohorts that preserve real-world patient, treatment, and outcome distributions. Notably, OncoSynth improves treatment effect estimation over existing approaches, by reducing population-level treatment effect error by up to 66%, and patient-level treatment effect error by up to 58%. Thereby, OncoSynth supports reliable evidence generation for precision oncology in settings where data sharing is restricted.", "date": "2026-06-24", "categories": []}
{"id": "2606.26467", "title": "A Causal Foundation Model for Structure and Outcome Prediction", "abstract": "We introduce TabPFN-CFM, a causal foundation model that can handle multiple causal problems. TabPFN-CFM predicts both causal structure and outcomes from observational data, supports queries on all three levels of Pearl's Causal Hierarchy and uses known graph structure when available to improve predictions. TabPFN-CFM is trained on synthetic datasets, and generalises to real datasets, demonstrating improved performance over both structural and outcome prediction baselines.", "date": "2026-06-25", "categories": []}
{"id": "2606.27114", "title": "Cross-Head Attention Uplift Network with Inverse Propensity Score under Unobserved Confounding", "abstract": "Uplift modeling, crucial for estimating individual treatment effects (ITE), faces dual challenges: flexibly leveraging inter-group similarity to enhance discriminative power and debiasing under unobserved confounding scenarios. In this paper, we propose the Cross-Head Attention Uplift Network (CHAUN) and Robust Adversarial Inverse Propensity Score (RA-IPS) method to address these limitations. CHAUN employs shared feature embeddings and cross-head attention mechanisms to dynamically integrate treatment-specific and control-specific representations, enhancing inter-group correlation modeling. Theoretically, we prove that access to the true propensity scores ensures ITE identifiability even with unobserved confounders. For practical scenarios lacking true propensity scores, RA-IPS adversarially optimizes propensity weights within constrained uncertainty sets to mitigate bias from unobserved variables. Experiments on public datasets (CRITEO-UPLIFT, LAZADA) and a production e-commerce dataset demonstrate CHAUN's superiority over state-of-the-art uplift models, achieving relative improvements of up to 25.6% in QINI scores. RA-IPS further enhances robustness, outperforming standard IPS by 5.4% under unobserved confounding. The results validate the effectiveness of our proposed methods in real-world causal inference tasks.", "date": "2026-06-25", "categories": []}
{"id": "2606.27510", "title": "The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching", "abstract": "Activation patching is the primary tool in mechanistic interpretability. It attributes causal responsibility for a model behavior to each of its individual components by estimating its natural indirect effect (NIE). Re-deriving the activation patching estimand from causal mediation analysis, we find that the NIE does not solely capture the causal effect through the specific component. It also contains interaction effects (INT) that measure how much the component's causal effect itself depends on the state of other components in the model. A natural response may be to try to eliminate INT by adjusting the estimator or unit of analysis, but each of these potential remedies has predictable failure modes. We demonstrate these failure modes in the GPT-2 IOI circuit; components whose causal importance is conditional on the state of other components are either invisible or artificially inflated, and INT variance explains the previously documented instability of faithfulness scores. We prove that INT scales with the distance between clean and patched component activations, is negligible when the model is locally affine, and decomposes combinatorially into pairwise and higher-order group interactions. Despite its inevitability, INT is not a nuisance to be eliminated, but rather a diagnostic for interpretability studies. Its individual and group-level magnitude and sign signal when causal conclusions are prompt-dependent, and when greedy NIE-based component ranking will miss mechanisms only discoverable through combinatorial search.", "date": "2026-06-25", "categories": []}
{"id": "2606.28024", "title": "Lifted Causal Inference", "abstract": "Lifted inference exploits indistinguishabilities in probabilistic graphical models by using a representative for indistinguishable objects, thereby speeding up query answering while maintaining exact answers. In this article, we show how lifting can be applied to efficiently compute causal effects in relational domains. More specifically, we introduce parametric causal factor graphs (PCFGs) to incorporate causal knowledge in lifted models and give a formal semantics of interventions therein. We further present the Lifted Causal Inference (LCI) algorithm to compute causal effects on a lifted level, thereby drastically speeding up causal inference compared to propositional inference, e.g., in causal Bayesian networks. In addition, we present partially directed parametric causal factor graphs (PD-PCFGs) as a generalisation of PCFGs to handle partial causal knowledge and extend LCI to perform lifted causal inference in a PD-PCFG, thereby extending the applicability of lifted causal inference to a broader range of models requiring less prior knowledge about causal relationships.", "date": "2026-06-26", "categories": []}
{"id": "2606.28225", "title": "Estimation--Prediction Tradeoff in Causal Probabilistic Temporal Graphs", "abstract": "Temporal link prediction is usually evaluated by predictive performance on unseen edges, but in probabilistic temporal graphs this criterion can conflate model error with irreducible uncertainty. We study this issue by characterising an inherent estimation--prediction tradeoff in binary logistic models where regimes that maximise Fisher information and improve parameter recoverability are also those with the highest entropy, making individual predictions intrinsically harder even under perfect parameter recovery. We propose a probabilistic causal framework for generating temporal graphs with transient edges and known ground-truth causal structure, allowing temporal link prediction to be evaluated jointly with causal parameter recovery. For the proposed binary logistic parametrisation, we derive the Cram\\'{e}r--Rao bound and validate the tradeoff between parameter estimation error and irreducible predictive loss. Our results show that predictive accuracy alone may not reflect whether a model has learned the underlying causal mechanism, motivating benchmarks that distinguish reducible model error from intrinsic process uncertainty.", "date": "2026-06-26", "categories": []}
{"id": "2606.28228", "title": "Disentangling Continuous-Time Latent Dynamics: Identifiability of Latent SDEs via Diffusion Shifts", "abstract": "Causal representation learning for time series has developed strong identifiability results in discrete-time latent causal models, but identifiability in continuous-time latent stochastic differential equation (SDE) models remains largely open. We address this gap using environment-induced shifts in diffusion covariance. We study additive-noise latent SDEs observed through an unknown nonlinear diffeomorphism, with shared drift but environment-specific diffusion covariance. We show that two diagonal diffusion regimes with pairwise distinct coordinate-wise variance ratios identify the latent coordinates up to permutation and scaling, without any sparsity assumption on the drift. We first prove this result for linear Ornstein--Uhlenbeck systems and then extend it to general additive-noise latent SDEs. Under mild smoothness, the instantaneous drift-Jacobian causal graph is identifiable up to the same permutation. We propose a two-stage estimator for latent disentanglement and optional graph recovery; experiments on synthetic systems confirm the predicted identifiability boundary, and an application to Hardanger Bridge monitoring data illustrates the approach on real sensor trajectories.", "date": "2026-06-26", "categories": []}
{"id": "2606.29911", "title": "A causal modeling perspective on decision theory", "abstract": "Decision theory provides a formal framework for how agents should make choices under uncertainty, drawing on ideas from philosophy, probability, and causality. Despite significant progress, the field still lacks a unified modeling language, and key concepts - such as the distinction between subjective and objective elements, or what it means for a decision theory to perform well - are often left implicit. This can make it difficult to evaluate and compare competing theories, particularly in controversial cases. In this paper, we address these issues by introducing a formal framework for decision theory based on nonparametric structural equation models (NPSEMs), a well-established tool in causal inference. NPSEMs provide a unified foundation for representing agents, counterfactuals, and causal relationships, allowing for unambiguous definitions of EDT and CDT. Building on this foundation, we propose a novel decision theory - personal decision theory - which instructs agents to maximize a subjective model of their own counterfactual utility. We introduce a formal performance metric based on hypothetical interventions that enforce a given decision theory across a population - such as might be achieved through education or policy -- and show that, under certain assumptions, personal decision theory is optimal with respect to this metric. Throughout, we use the smoking lesion problem as a running example and conclude with a formal analysis of Newcomb's problem. Our aim is to provide decision theory with a clearer modeling language and firmer evaluative ground, thereby enabling more rigorous comparisons and facilitating conceptual progress in the field.", "date": "2026-06-29", "categories": []}
{"id": "2606.30467", "title": "Non-parametric recovery of causal diffusion mechanisms from steady-state observations", "abstract": "We consider sparse multivariate stochastic systems that evolve in continuous time according to a causal mechanism and present methodology to recover the system's time-infinitesimal transition mechanism from mere cross-sectional data. This observational paradigm is motivated by applications such as gene expression analysis, where destructive experimental techniques may only allow recording data once over a cell's lifetime. Precisely, we assume the system follows a time-homogeneous diffusion process that has reached an equilibrium distribution at observation time. Further, we assume the causal mechanism is fully described by the diffusion drift, is acyclic, and its causal structure graph is known. In this setting, we prove that the full causal mechanism, i.e., the drift function, can be non-parametrically identified under a weak non-explosion criterion. We derive a non-parametric kernel estimator for this challenging inverse problem and prove its consistency. Moreover, we propose a cross-validation scheme for hyperparameter tuning, illustrate the behavior of our estimator in simulations, and we discuss connections with irreversible generative diffusion models and low-frequency sampled data.", "date": "2026-06-29", "categories": []}
{"id": "2606.30500", "title": "Doubly Robust Adaptive Conformal Inference for Causal Effects Under Temporal Dependence", "abstract": "We propose doubly robust adaptive conformal inference (DR-ACI), which constructs prediction intervals for doubly robust pseudo-outcomes under temporal dependence.", "date": "2026-06-29", "categories": []}
{"id": "2606.30992", "title": "Hierarchical Clustering As a Novel Solution to the Notorious Multicollinearity Problem in Observational Causal Inference", "abstract": "Multicollinearity is a long lasting challenge in observational causal inference, especially in regressions -- highly correlated independent variables make it hard to isolate their individual impacts on outcomes of interest. While common solutions such as shrinkage estimators and principal component regressions are helpful in prediction problems, a crucial limitation hinders their applicability to causal inference problems -- they cannot provide the original causal relationships. To fill the gap, we present an innovative and intuitive solution, by employing hierarchical clustering to aggregate data in a way that effectively alleviates collinearity. This method is generally applicable to causal problems featuring multicollinearity. We use a marketing application to demonstrate how and why it works.\n  Expenditures on different advertising channels often exhibit correlations, making it exceedingly difficult to separately measure their impact. Many previous studies proposed to leverage granular cross-sectional data for better identification but, to our knowledge, none explicitly addressed multicollinearity, which undermines causal identification even with granular data. We propose to hierarchically cluster geographic units based on marketing spend correlation to reduce collinearity, and to implement a Bayesian Marketing Mix Model with cluster-level data. Such clustering happens in two steps -- we first normalize and demean geo-level data to establish a common scale and to eliminate the common trends; we then calculate pairwise distance to summarize marketing spend correlation between geos and cluster the ones with moderate to strong correlation. Both descriptive evidence and regression analysis affirm that such hierarchical clustering effectively mitigates collinearity and facilitates the separate identification of the impact of different marketing channels.", "date": "2026-06-30", "categories": []}
{"id": "2606.30999", "title": "Estimating Supply Incrementality in Two-sided Marketplaces: A Causal Machine Learning Approach", "abstract": "In two-sided marketplaces with heterogeneous products, it is important to understand the causal relationship between additional supply and marketplace outcomes, such as the total quantity transacted or transaction value in the marketplace. This paper studies a causal machine learning approach to estimating this relationship across product segments. We use the Airbnb marketplace as an example, focusing on the impact of additional listing supply on total bookings, but the methodology applies to other two-sided marketplaces. Our approach combines double/debiased machine learning with a hierarchical Bayesian framework that leverages pre-existing knowledge as priors. We construct tractable and informative features for the model by leveraging measures of product segment similarity from the geospatial literature. We find that such a model provides plausible estimates of the marketplace returns to additional supply and strong out of sample performance.", "date": "2026-06-30", "categories": []}
{"id": "2606.31184", "title": "Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation", "abstract": "Adaptive experiments for average treatment effects (ATE) require randomized allocations balancing valid inference with statistical efficiency. The oracle design is a covariate-dependent Neyman rule governed by unknown arm-conditional outcome variances. We investigate whether this sequential variance-estimation and allocation process can be amortized via in-context learning. We introduce Bayesian in-context experimenters: transformer policies trained to imitate a Bayesian posterior Neyman teacher. The teacher updates nonparametric beliefs over potential outcomes using experimental history to assign posterior Neyman treatment probabilities. This design converges to the oracle rule, supporting efficient ATE inference. Transformers constructively implement this mapping through attention-based sufficient statistics and projected gradient descent, imitating Bayesian updating for Gaussian-series priors. To address unknown outcome smoothness, we combine smoothness-indexed experimenters using a mixture-of-experts transformer. The gate acts as a hierarchical posterior over smoothness classes, concentrating on near-oracle experts. By bounding the complexity of the transformer class, we prove this amortized policy can be learned via empirical risk minimization using supervised pretraining. Experiments confirm accurate teacher imitation, adaptive allocation, and improved ATE precision over baselines.", "date": "2026-06-30", "categories": []}
{"id": "2607.00267", "title": "Validating Causal Abstraction Metrics on Simulated Complex Systems", "abstract": "A central goal of science is to produce valid explanations of complex systems: high-level causal accounts that faithfully reflect the behavior of lower-level mechanisms. Yet no consensus exists on how to measure whether a proposed high-level explanation is actually valid. We introduce a benchmark of ten complex systems spanning both discrete and continuous state spaces, as well as static and dynamical regimes, each equipped with consensual ground-truth causal explanations and invalid contrastive conditions. Within a unified causal abstraction framework, we systematically evaluate over thirty candidate metrics drawn from observational, functional, information-theoretic, and causal families. Our results show that only the latter reliably discriminates valid from invalid abstractions, and only when incorporating faithfulness testing over unmapped variables. Building on these findings, we introduce the Causal Abstraction Error (CAE), a continuous validity metric with an explicit faithfulness test, which passes all discrimination tests across every system and can converge with as few as 30 sampled interventions. We offer it as a general-purpose metric for the discovery and validation of high-level explanations.", "date": "2026-06-30", "categories": []}