Learning Partner Selection and Opt-Out Strategies under Defection Bias in Social Dilemmas

Learning Partner Selection and Opt-Out Strategies under Defection Bias in Social Dilemmas

Computational social science

Mathematics and Statistics

Rachel Zhang & Hannah Wu

2025

Paolo Turrini and Chin-wing Leung

The project builds on Leung and Turrini (AAMAS 2024) and further studies learning-based partner selection and opt-out strategies under defection bias in social dilemmas. Agents learn whom to engage and when to exit in defector-heavy environments. Our findings demonstrate that cooperation breaks down nonlinearly once defection bias exceeds a critical threshold.

Reinforcement learning, Prisoner’s Dilemma, Partner selection,Q-learning, Boltzmann exploration, multi-agents