social-dilemmas

A custom reinforcement learning environment for OpenAI Gym & PettingZoo that implements various Stag Hunt-like social dilemma games.

Multi-Agent Deep RL meets Darwinian evolution: predator/prey agents learn under sparse reproduction-only rewards (no shaping, no cooperation signal) while heritable traits mutate and are selected across generations — a live testbed for the Baldwin effect, and for whether cooperation, defection, and free-riding emerge unassisted.

An OpenAI gym multi-agent environment implementing the Commons Game proposed in "A multi-agent reinforcement learning model of common-pool resource appropriation"

Code and materials for the paper S. Phelps and Y. I. Russell, Investigating Emergent Goal-Like Behaviour in Large Language Models Using Experimental Economics, working paper, arXiv:2305.07970, May 2023

Learned cooperation. Updated reproduction of sequential social dilemmas in RLlib new API stack (2.54.0)

Cooperation and Exploitation in LLM Policy Synthesis for Sequential Social Dilemmas

Code for 'Quantifying the Self-Interest Level of Markov Social Dilemmas' (IJCAI 2025)

Multi-Agent Reinforcement Learning Environment

Zero-training entropy agent beats DeepMind's trained RL on Melting Pot social dilemmas. 350 lines of NumPy, no neural networks, no reward. DOI: 10.5281/zenodo.18717202

Participatory experiment during the open ceremony in FestivalTarrega with a set of games formed by Prisoner's Dilemma, Snowdrift Game and Dictator's Game

Code for 'Resolving social dilemmas with minimal reward transfer' (JAAMAS 38)

Participatory experiment, performed in the Biennal Ciutat i Ciència (Barcelona) and the citizen science festival Calidoscopi (Barcelona), that consist on a set of dilemmas to study how we behave in front of a situation of gender-based violence in public spaces. Natalia, Alex, Raquel and Sandra tell us situations lived in public spaces and partic…

Proximal Policy Optimization algorithm applied to the Harvest Game, a multi-agent environment

This fork adds multi-agent social dilemma environments in Gym: CoinGame, (Iterated) Prisonner Dilemma, Stag Hunt, Chicken, Matching Pennies. Gym: A toolkit for developing and comparing reinforcement learning algorithms.

A minimalist multi-agent implementation of the social dilemma problem with governance kernels