Reputation-based cooperation: empirical evidence for behavioral strategies

Human cooperation in large groups can emerge when help is channeled towards individuals with a good reputation of helping others. Evolutionary models suggest that, for reputation-based cooperation to be stable, the recipient's reputation should be based not only on his past behavior (1st-order information) but also on the past behavior of the recipient's recipient (2nd-order information). Second-order information reflects the context of others' actions, and allows people to distinguish whether or not giving (or denying) help was justified. Little is known yet about how people actually condition their cooperation on 2nd-order information. With a behavioral experiment, we show that people actively seek 2nd-order information and take this into account in their own helping decisions. In an anonymous iterated helping game, donors learned if their recipients helped others in the past and could obtain 2nd-order information about these actions. Donors often requested this 2nd-order information and were especially interested to know why help was denied (i.e., defection). Justified defection was rewarded: help was generally directed towards those who defected against the selfish, and away from those who defected against helpful individuals. A detailed analysis of individual strategies reveals that many subjects based their decisions solely on 1st-order information about their recipients' past behavior. However, a substantial fraction of subjects consistently considered also the 2nd-order information about their recipients' behavior. Our results provide strong empirical support for the mechanisms that theoretically underpin reputation-based cooperation, and highlight pronounced individual variation in human cooperative strategies.