Cooperation in Repeated Interactions

This thesis systematically reviews empirical research on human cooperation in repeated reciprocal interactions, and reports 6 new studies that aim to verify, clarify and extend previous findings. As a research paradigm, Rosenthal’s (1981) Centipede game is chosen because it provides a formal model of alternating reciprocal cooperation between two individuals. Despite potentially large gains to both players through long-term cooperation in the game, a logical argument based on backward induction shows that instrumentally rational players will never cooperate in this way. However, previous research showed reliable deviations from this rational game-theoretic solution in human decision makers. All studies reported in this thesis confirm this robust finding. Study 1, drawing on a qualitative design, further evidences a large variety of other-regarding experiences underpinning decision making in the game including previously neglected, low-level motives such as activity bias. Study 2 addresses inconsistencies across research designs reported in the literature, and suggests large effects of even subtle changes to the games’ payoff functions. Studies 3 and 4 aim to increase the Centipede game’s applicability to real-life decision contexts by increasing game length and introducing varying termination rules. It is found that human decision makers cooperate more frequently during longer decision sequences. When additionally faced with the uncertainty of an unknown end or the risk of random game termination, however, defection may increase. Studies 5 and 6 examine cross-cultural differences in cooperation and find that Japanese participants cooperate more frequently than European/UK participants, which is attributed to different, culturally-dependent types of trust. Drawing on the research presented, explanations for cooperation in repeated interactions are discussed including the possible invalidity of backward induction reasoning, the cognitive burden of backward induction, a possible breakdown of common knowledge of rationality, virtual bargaining theory, fuzzy-trace theory, and other-regarding preferences. Ultimately, it is suggested that other-regarding preferences provide the most persuasive explanation for cooperation.