Emotion-coupled Q-learning with cognitive bias enhances cooperation in evolutionary prisoner’s dilemma games

This study investigates how emotion-coupled Q-learning strategy updates and history-dependent cognitive biases influence cooperation in evolutionary games. We propose a Q-learning framework where agents dynamically adjust their strategies based on emotional states shaped by social interactions and intrinsic social value orientations (competitive vs. non-competitive). Emotional states govern agents’ propensity to learn from neighbors or rely on biased Q-tables. Cognitive biases, determined by neighbors’ historical actions and quantified as deviations from rational Q-values, further reshape decision-making by amplifying or suppressing cooperation. Through Monte Carlo simulations, we find that emotion-coupled Q-learning can significantly boost cooperation, particularly when individuals are sensitive to emotional differences. We also find that, moderate cognitive biases optimize cooperation under adverse conditions. Furthermore, we demonstrate that emotion-coupled learning fosters cooperator clusters dominated by non-competitive individuals. These results underscore the critical role of emotional–cognitive interplay in resolving social dilemmas and advancing adaptive collective dynamics.