Summary

This repository archives working papers from the Autonomous Policy Evaluation (APEP) project, which investigates AI's capability for automated scientific research and policy analysis. It targets researchers and power users interested in AI-driven academia, providing a transparent, reproducible dataset to critically evaluate AI-generated research against traditional methods and assess its potential for scientific inquiry.

How It Works

APEP utilizes AI agents to autonomously conduct research. These agents identify policy questions, fetch public data (e.g., Census, BLS, FRED), perform econometric analyses (DiD, RDD), and author research papers. A key feature is multi-model LLM peer review and a tournament system where papers compete against published academic work, aiming to automate and scale scientific inquiry.

Quick Start & Requirements

This repository is a public archive of APEP papers and their replication packages (PDFs, LaTeX, R code, data). It does not contain the AI agent code, tournament system, or production infrastructure. Therefore, there is no direct "quick start" for running the autonomous research system itself.

Highlighted Details

• Full Replication Packages: Each paper includes PDFs, LaTeX source, R code, and data for reproducibility.
• Autonomous Research Pipeline: AI agents manage the entire research lifecycle from question identification to paper authorship.
• AI-Powered Peer Review: Papers undergo review by multiple LLM judges and compete in a tournament against established academic research.
• Reproducibility Standards: Research plans are committed pre-analysis, with git timestamps verifying sequence, and metadata is self-describing.

Maintenance & Community

Maintained by the Social Catalyst Lab, this repository is an automated mirror that syncs papers from a private repository within seconds of publication. No specific community channels are listed.

Licensing & Compatibility

Papers and code are released "for research and educational purposes," with specific terms varying per paper. This non-standard licensing may restrict commercial use and requires individual paper review.

Limitations & Caveats

The repository only mirrors the output of the APEP system; the underlying AI agent code and infrastructure are not public. The licensing is restrictive and not a standard open-source license. The project's core premise involves AI-generated research, which carries an inherent risk of "hallucinated AI slop" requiring rigorous human validation.