Altruists, Egoists and Hooligans in a Local Interaction Model.

This paper studies a population of agents, each of whom can be either an Altruist or an Egoist. Altruists confer benefits on others at a cost to themselves. Altruism is thus a strictly dominated strategy and cannot survive if agents are rational best-responders. We assume that agents choose their actions by imitating others who earn high payoffs. We also assume that interactions between agents are local, so that each agent affects (and is affected by) only his neighbors. Altruists can survive in such a world if they are grouped together, so that the benefits of altruism are enjoyed primarily by other Altruists, who then earn relatively high payoffs and are imitated.