Dynamic role-switching in hypergraphs: Enhancing cooperation via adaptive punishment and reinforcement learning

Evolutionary game theory, enhanced by reinforcement learning, provides deep insights into cooperation dynamics crucial for collective behaviors in complex systems. As complex network structures, hypergraphs present a robust framework for examining the emergence of cooperation. In this study, we combine evolutionary game theory with an adaptive Q-learning algorithm optimized for hypergraphs structures to explore the effects of a dynamic punishment transition mechanism on collective cooperative behavior. This algorithm allows agents to dynamically adjust roles and engage in introspective learning, moving beyond simple imitation. Extensive Monte Carlo simulations demonstrate that increasing the probability and intensity of punishment significantly promotes cooperation, while moderate punishment costs can catalyze cooperation even under low synergy factors. Moreover, higher discount factors, increased learning rates, and smaller group sizes within hypergraphs further enhance cooperation. This research highlights the critical role of self-adjusting Q-learning and dynamic punishment transition mechanisms in fostering cooperation, providing valuable insights into social dilemma scenarios within complex environments.