Collaborative Research: Individual and Team Behavior in Indefinitely Repeated Games

该奖项资助的研究使用实验室实验来检验经济学理论，即当相互作用的结束日期不确定时，决策者如何相互作用。通用汽车（General Motors）与福特（Ford）和克莱斯勒（Chrysler）的竞争就是一个无限重复博弈的例子。“博弈”是经济学文献中描述这些重复互动的专业术语。这项研究研究了个人和个人团队的行为。这一领域的大多数研究——理论的、实证的和实验的——都是研究个体的。然而，实验室之外的互动通常涉及团队中的个人群体。过去的研究表明，个人的经济行为可能与团队的经济行为大不相同。例如，个人通常会履行承诺，在短期经济互动中进行合作，从而提高经济福利。然而，团队经常以自利的方式行事，需要法律处罚来改善福利。这个项目将帮助我们理解在重复的经济互动中团队和个人之间的差异。研究方法将允许团队通过书面或口头沟通来协调他们的行动。研究小组将分析这些通信以及在实验中选择的行动。分析交流将提供新的见解，了解是什么激发了行为，并可以识别以前没有考虑到的行为因素。该项目的结果将帮助管理者、政策制定者和公民理解团队合作而不是个人合作的效果。这项新研究有几个特定的目标。博弈论理论家区分了完美和不完美的监控。在完全监控的情况下，代理人可以肯定地知道其他人做了什么，但在不完全监控的情况下，他们不确定其他人做了什么。研究人员将在监控不完善的情况下，将团队和个人在无限重复囚徒困境（IRPD）游戏中进行比较。他们特别感兴趣的是对比完美监控和不完美监控对团队玩法的影响，并利用团队讨论来帮助理解他们如何看待对手行动的变化，因为他们无法区分故意选择和随机噪音，这是许多经济互动中固有的。第二个项目将比较IRPD游戏中的玩法，其中延续概率在各个阶段游戏中变化，与具有固定预期交互结束的标准情况进行比较。很明显，在实验室外，贴现率会随着时间的推移而变化，从而改变了代理人面临的动态激励。受试者会解释动态激励的变化吗？这些变化会超出动态激励变化所能解释的范围吗？先前的结果表明，合作可能出现在通常很少合作的延续概率中。第三个项目将探索机器学习算法的使用，以从团队讨论中获得对行为的更深入了解。关于现代文本分析的文献越来越多，这可能会大大增加我们从团队讨论中学到的东西。第四个项目将比较个人和团队在公共利益游戏中的结果，包括金钱惩罚和其他选择的制裁（非金钱惩罚）。这里的目标是确定在类似环境中观察到的团队和个人之间的差异是否延伸到公共物品提供的货币和非货币惩罚。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This award funds research using lab experiments to test economic theories about how decision makers interact when the end date of the interactions is uncertain. General Motors competing with Ford and Chrysler is an example of an indefinitely repeated game. “Game” is the technical term in the economics literature for these repeated interactions. This research studies behavior of both individuals and teams of individuals. Most research in this area - theoretical, empirical, and experimental - studies individuals. However, interactions outside the laboratory often involve groups of individuals in a team. Past research shows that economic behavior of individuals can be quite different from the economic behavior of teams. For example, individuals generally follow through on promises to cooperate in short term economic interactions, improving economic welfare. However, teams often act in a self-beneficial manner, necessitating legal penalties to improve welfare. This project will help us understand the differences between teams and individuals in repeated economic interactions. The research method will allow teams to coordinate their actions through written or verbal communication. The research team will analyze these communications as well as actions chose during the experiment. Analyzing communications will provide new insights into what motivates behavior and can identify elements of behavior that had not previously been considered. The results of the project will help managers, policymakers, and citizens understand the effects of working in teams rather than as individuals.There are a number of particular goals for the new research. Game theorists distinguish between perfect and imperfect monitoring. With perfect monitoring, agents can tell what others have done with certainty, but with imperfect monitoring, they are not certain what others have done. The investigators will compare teams with individuals in indefinitely repeated prisoner dilemma (IRPD) games with imperfect monitoring. They are particularly interested in contrasting the effects of team play between perfect and imperfect monitoring, and using the team discussions to help understand how they think about changes in their opponent’s action given that they cannot distinguish between a deliberate choice and random noise, which is inherent in a number of economic interactions. A second project will compare play in IRPD games where continuation probabilities vary across stage games to the standard case with fixed expected end to the interactions. It is clear that discount rates vary over time outside the lab, changing the dynamic incentives agents face. Will subjects account for the change in dynamic incentives and will these changes go beyond what can be explained by the changing dynamic incentives? Previous results suggest that cooperation may emerge for continuation probabilities that normally produce little cooperation. A third project will explore the use of machine learning algorithms to gain greater insights into behavior from the team discussions. There is a growing literature on modern text analysis that may add substantially to what we learn from the team discussions. A fourth project will compare outcomes between individuals and teams in public good games with both monetary punishment and sanctioning of others choices (nonmonetary punishment). The goal here is to determine whether differences between teams and individuals observed in similar settings extend to public good provision for both monetary and non-monetary punishment.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.