Integrating artificial intelligence with game theory: a study on reinforcement learning in cooperative settings

Over the past two decades, the application of Artificial Intelligence—particularly Reinforcement Learning—to studying cooperative behaviour in the Public Goods Game has become increasingly widespread. We present a systematic review of this literature, highlighting factors that shape learning dynamics, such as agents’ strategy adjustments over time and the role of social norms in cooperative behaviour. A pre-analytical clustering of 70 studies since 2002 identifies key research topics, including learning and conditional cooperation, punishment mechanisms, agent-based networks, moral learning, cooperation under uncertainty, and dynamic multi-objective learning in multiplayer settings. Trends show a growing interest in topics such as moral learning, cognitive aspects, cooperative artificial intelligence, cooperation under uncertainty, and collective risk, which have become increasingly relevant, especially after 2015.