Moral sensitivity drives cooperative evolution in Q-learning based donation games

Abstract

This study introduces a novel donation game framework that integrates reinforcement learning with evolutionary game dynamics to explore how moral sensitivity influences cooperative behavior. By comparing Q-learning-based models with traditional evolutionary models on a two-dimensional lattice, the results show that Q-learning models significantly enhance adaptability and promote cooperation, especially under high moral sensitivity and optimized donation ratios. Through detailed spatial and parametric analyses, it is shown that Q-learning not only accelerates convergence but also fosters more uniform and stable strategy distributions. These findings provide valuable insights into the interaction between moral cognition and learning mechanisms in driving social cooperation and offer practical implications for designing systems that encourage collaboration.