Dartmouth Events

As we see rapid shifts in society brought on by advances in artificial intelligence and changes to the political (and literal) climate, it becomes increasingly important to understand how to foster cooperation within and between communities. Our work extends models from evolutionary game theory used to understand cooperation to more complex, intelligent agents. We characterize the stability of equilibria in the open problem of exploration vs exploitation under a range of models including multi-agent reinforcement learning, a popular machine learning technique. We also analyze an opinion dynamics model with the counterintuitive effect that introducing trivial topics can completely change whether a population will polarize or reach consensus. This work has applications to fields from biology, computer science, and economics, to psychology, ecology, and sociology, giving it broad potential impacts.