Scaffolding Cooperation In Human Groups With Deep Reinforcement Learning

This article explores how deep learning can be used to structure networks of human participants playing a group cooperation game. Leveraging deep reinforcement learning and simulation methods, a ‘social planner’ was trained to make recommendations to create or break connections between group members. The strategy developed by the social planner was successful in encouraging pro-sociality in networks of human participants, with an average cooperation rate of 77.7%. In contrast to prior strategies, the social planner learns to take a conciliatory approach to defectors, encouraging them to act pro-socially.