neurodeveloper's picture
Release Cognitive Agent Evaluator v2.0 (10-axis biases, Plotly radar, burnout & psychological safety telemetry)
c6c178c verified
|
Raw History Blame Contribute Delete
2.52 kB
---
title: Cognitive Agent Evaluator v2.0
emoji: 🧠
colorFrom: indigo
colorTo: purple
sdk: static
pinned: false
license: mit
short_description: Behavioral Telemetry & AI Alignment Engine
---
# 🧠 Cognitive Agent Evaluator v2.0: Behavioral Telemetry & AI Alignment Engine
[![Fabio Torres LinkedIn](https://img.shields.io/badge/LinkedIn-Fabio_Torres-0A66C2?style=flat-square&logo=linkedin)](https://www.linkedin.com/in/fabio-torres-39364b258/)
[![GitHub Repo](https://img.shields.io/badge/GitHub-cognitive--agent--evaluator-181717?style=flat-square&logo=github)](https://github.com/neurodeveloper11/cognitive-agent-evaluator)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg?style=flat-square)](LICENSE)
An autonomous evaluation agent and behavioral telemetry framework engineered to detect **10 cognitive distortions**, score **occupational burnout & psychological safety**, and synthesize **alignment & red-teaming directives** for Large Language Models (LLMs) and human decision-making.
Engineered by **[Fabio Torres](https://www.linkedin.com/in/fabio-torres-39364b258/)** (M.Sc. in Data Engineering & Cloud Infrastructure • 10+ Years Behavioral Science Leadership).
---
## 🎯 What's New in Version 2.0?
1. **Expanded 10-Axis Cognitive Distortion Engine**:
- Confirmation Bias, Anchoring Bias, Sunk Cost Fallacy, Availability Heuristic, Framing Effect.
- *NEW:* Catastrophizing, All-or-Nothing / Polarized Thinking, Overconfidence / Dunning-Kruger, Fundamental Attribution Bias, Outcome Bias.
2. **Occupational Telemetry & Burnout Risk Module**:
- Burnout Risk Index (0-100) & Psychological Safety Score (0-100) calibrated from real-world psychosocial risk methodologies (Resolución 2764/2022).
3. **Interactive 10-Axis Spider / Radar Chart**: Real-time visual distortion footprint powered by Plotly.
4. **Automated Red-Teaming Directives**: Ready-to-use prompt directives for debiasing reasoning chains in ChatGPT, Claude, or Llama.
5. **Production-Ready JSON Telemetry**: Exportable machine-readable payload for enterprise pipelines.
6. **100% Client-Side Privacy (Zero Data Leak)**: Real-time evaluation directly in the browser with sub-millisecond latency.
---
## 👨‍💻 Author & Professional Profile
- **Fabio Torres Benítez**
- **Email:** psicologofabiotorres@gmail.com
- **LinkedIn:** [linkedin.com/in/fabio-torres-39364b258](https://www.linkedin.com/in/fabio-torres-39364b258/)
- **GitHub:** [github.com/neurodeveloper11](https://github.com/neurodeveloper11)