data/evals/raw/dcm-code_ec708b7/README.md

Digital Consciousness Model

A Bayesian model for evaluating consciousness in various systems using PyMC.

Overview

This repository contains the implementation of the Digital Consciousness Model (DCM), which uses a Bayesian approach to evaluate consciousness stances and their supporting evidence through nested Beta-Bernoulli hierarchies.

Features

• Bayesian Modeling: Uses PyMC for probabilistic programming
• Evidence Integration: Incorporates multiple levels of evidence through hierarchical models
• Flexible Configuration: Configurable evidential strength constants and sampling parameters
• Fully Offline: Model specifications are read from the local model_data.json file; no network access is required (this is a recent update)
• Results Management: Processes and prints model results locally

Installation

Requirements

• Python 3.8+
• PyMC
• NumPy

Setup

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

Or install manually:

pip install pymc numpy

Usage

Run the main model:

python dcm_model.py

Run sensitivity tests:

python sensitivity_test_model.py

Model Components

Configuration

The ModelConfig class contains all model parameters including:

• Evidential strength constants (BASE, OVERWHELMING, STRONG, MODERATE, WEAK)
• Sampling parameters (number of samples, tuning steps, chains)
• Prior parameters (default alpha and beta values)
• Path to the local data file

Key Classes

• DataFetcher: Loads model data from the local JSON file
• EvidenceProcessor: Processes evidence and observations, calculates Beta parameters
• BayesianModelBuilder: Builds and manages PyMC Bayesian models
• ResultsManager: Manages model results and output

Model Structure

The model evaluates consciousness stances using a hierarchical structure:

1. Stance: Top-level hypothesis with Beta prior
2. Features: Supporting evidence with support and demandingness levels
3. Indicators: Observable evidence from specific systems

Each level uses Beta-Bernoulli conjugate pairs, where:

• Beta distributions model probability beliefs
• Bernoulli variables model binary outcomes
• Support and demandingness levels determine parameter values

JSON File location (updated recently)

The model specifications are stored in a JSON file located at:

model_data.json

License

This project is licensed under the MIT License.

Citation

https://arxiv.org/abs/2601.17060