Replication code and data for DeAngelo, McCannon & Wyatt (2025).

OSF Repository: https://osf.io/a6f3u/

This repository contains code to replicate experiments testing how responsibility affects LLM decision-making in a Stag-Hare coordination game. We compare "play-for-pair" (deciding for a team) vs "play-for-self" (deciding individually).

All data is in dist/. Files are named {treatment}_df_{timestamp}.csv.

Each experiment has a corresponding notes_{timestamp}.yaml with prompts and parameters.

pip install -r requirements.txt

Create .env with your OpenAI API key:

OPENAI_API_KEY=your_key_here

jupyter notebook replication.ipynb

jupyter notebook game_play.ipynb

python chi_square_tests.py

LLMs take significantly more risk when responsible for others.

• game_play.ipynb - Main experiment runner
• game_play_options.ipynb - Concealed game variant
• persona_analysis.ipynb - Persona experiment analysis
• replication.ipynb - Results replication
• chi_square_tests.py - Statistical tests for paper
• statistical_results.md - Summary of all test results

@article{deangelo2025responsibility,
  title={The Role of Responsibility in AI's Strategic Risk-Taking},
  author={DeAngelo, Gregory and McCannon, Bryan and Wyatt, Aaron},
  journal={Journal of Economic Psychology},
  year={2025}
}

Open an issue or contact the authors with questions.