Learning about a population of agents and the evolution of trust and cooperation

Various reputation models consider how cooperation might emerge between rational adversaries in non-cooperative games. We suggest and analyze an additional determinant: uncertainty and learning about the population from which agents are selected. The analysis shows how the degrees of trust by principals and cooperation by agents can depend on the past behavior of other agents. A dramatic implication is that trust and cooperation can permanently and inefficiently break down due to revised beliefs about the population of agents. We consider lessons the theory might provide concerning the relationship between the U.S. Department of Defense and its contractors. The model suggests two explanations of events during the mid-1980s, which many interpret as a very socially costly breakdown of trust and cooperation.