Infinitely Repeated Games with Private Monitoring: An Experimental Analysis

该奖项资助的研究将采用实验室实验来测试战略互动的博弈论模型。许多经济情况涉及到反复的互动，决策者不能直接观察到其他参与者所做的决定。在许多这样的情况下，决策者对其他人的决定都有部分信息。任何游戏中，代理人观察其他人的错误行为（被观察到的内容只有他们自己知道），都是带有私人监控的游戏。现实世界中有很多这样的例子，经济理论最近在开发分析这些博弈的新方法方面取得了实质性进展。然而，这些理论还没有经过实证检验。这类游戏在现实世界中有很多例子，但从这些例子中收集数据来检验这一理论是很困难的。关于特工私下观察的数据很少。此外，从理论的角度来看，其他关键参数往往是不可观察的或难以测量的。考虑到这一点，该实验室为探索这种环境中的行为提供了一个有价值的选择。不完全私人监督的经济学理论主要考虑了两个极端：几乎完全监督和几乎公共监督。顾名思义，几乎完美就是不完美，但非常接近完美的监控（玩家的信号与他们的行动不相关），几乎公开的监控有私人成分，但非常接近公开的监控（信号是相关的）。这项由该奖项资助的新研究在一系列三个项目中研究了近乎完美和近乎公开的监测。要解决的第一个问题很简单：在监督不完善且不公开的情况下，主体能否支持合作？如果实验对象确实可以支持一些合作，那么下一个问题是，当监控是私人的，而不是公开的或完美的时，他们的行为有什么不同？这个问题的一个更苛刻的版本是，最常用的策略是什么，这些策略如何随着信息结构的细节而变化？通过在实验室中实施随机终止的囚徒困境游戏的三个项目来解决这些问题。这些项目将对我们所知道的带有私人监控的无限重复游戏产生重大影响。它们还将进一步加深我们对不完善监测行为的理解。无限重复的游戏是许多人类互动类型的特征：政治、心理、社会等等；使这个项目本质上是跨学科的。因此，这些结果将与许多应用相关。了解促进或阻碍合作的因素是社会科学中的一个关键问题，本研究将提供部分答案。PI还将为研究生和本科教育做出贡献；各种各样的学生将参与这项研究。

This award funds research that will employ laboratory experiments to test game theory models of strategic interaction.Many economic situations involve repeated interactions where decision makers cannot directly observe the decisions made by other participants. In many of these situations, decision makers each have partial information about others' decisions. Any game where agents observe what the others do with error (and what is observed is only known to them) is a game with private monitoring. There are many real world examples, and economic theory has recently made substantial progress in developing new methods for analyzing these games. However, these theories have not yet been subjected to empirical testing. There are many real-world examples of these games, but gathering data from these examples to test the theory is difficult. Data about what agents privately observe is rare. Furthermore, other critical parameters from the point of view of the theory are often not observable or difficult to measure. Given this, the laboratory offers a valuable alternative to explore behavior in such environments.The economic theory of imperfect private monitoring has mainly considered two extremes: almost perfect and almost public monitoring. As the names imply, almost perfect is imperfect but very close to perfect monitoring (where the players' signals are uncorrelated conditional on their actions) and almost public is has a private component but is very close to public (where signals are correlated). The new research funded with this award studies both almost perfect and almost public monitoring, in a series of three projects. The first question to be addressed is simple: can subjects support cooperation when monitoring is imperfect and private? If indeed subjects can support some cooperation, then the next question is how is their behavior different when monitoring is private as opposed to public or perfect? A more demanding version of this question to be explored is what strategies are most often used, and how do these vary with the particulars of the information structure? These are addressed by three projects implementing randomly terminated prisoner's dilemma games in the laboratory. These projects will have a substantial impact on what we know about infinitely repeated games with private monitoring. They will also further our understanding of behavior with imperfect monitoring more generally.Infinitely repeated games are characteristic of many types of human interactions: political, psychological, sociological, et cetera; making this project intrinsically interdisciplinary. Hence, these results will be relevant to many applications. Understanding factors that facilitate or hinder cooperation is a key question in the social sciences, and this research will provide partial answers. The PI will also be contributing to both graduate and undergraduate education; a wide variety of students will participate in this research.