Prosocial learning agents solve generalized Stag Hunts better than selfish ones

Alexander Peysakhovich, Adam Lerer

Deep reinforcement learning has become an important paradigm for constructing agents that can enter complex multi-agent situations and improve their policies through experience. On…