brycewang-stanford/Auto-Empirical-Research-Skills

What it solves

Auto-Empirical Research Skills (AERS) provides a comprehensive, automated, and intervention-ready pipeline for empirical research in social sciences. It bridges the gap between a vague research idea and a publication-ready paper by automating tasks like literature review, data acquisition, causal inference, robustness auditing, and writing.

How it works

The project acts as a library of over 1,100 "skills"—specialized tools that AI agents (like Claude Code or Codex) can call to perform specific research tasks. Users can install these skills globally or per project. The core of the pipeline is the Paper-WorkFlow skill, which orchestrates these individual tools into a sequential, end-to-end process. At any point in the pipeline, researchers can manually intervene to adjust methods or variables, and the pipeline will automatically resume from that point.

Who it’s for

This tool is designed for empirical researchers, economists, and social scientists who want to leverage AI agents to accelerate their research workflows while maintaining academic rigor and the ability to audit every step of the analysis.

Highlights

• End-to-End Automation: Covers 9 stages of research, from initial idea refinement to final manuscript submission and AIGC detection.
• Multi-Language Support: Includes dedicated "Full Empirical" stacks for Python, Stata, and R, allowing for seamless integration with existing research tools.
• Rigorous Validation: Features 19 numeric benchmark tasks and 42 behavioral evaluation scenarios to ensure the reliability of the agent's outputs.
• Intervention-Ready: Every step of the automated pipeline produces outputs that allow for human intervention, ensuring the researcher remains in control of the methodology.
• StatsPAI Integration: Deeply integrated with StatsPAI, a Python-based causal inference engine designed specifically for AI agents to perform complex econometric tasks like DID, RD, and IV analysis.