Collaborative Research: The Evolution of Conventions in Intergenerational Games

这是一个与Barry Sopher， 9709079合作的项目。这个项目包括对社会习俗和其中包含的信息是如何在社会代理人的几代人之间产生和传递的实验调查。博弈论的大量文献研究了由一组固定的和不变的参与者进行的无限重复的博弈。在现实世界中，虽然游戏可能无限期地进行下去，但参与游戏的代理人的生命却是有限的。当这些代理人死亡或退休时，他们在这些无限生存的游戏中的角色由其他代理人接替，这些代理人可能会与他们的前任协商，在他们之后继续玩这个游戏。我们考察了从一代人到另一代人的过渡，在这个过程中，传统和规范要么被传承下去，要么被打破，新的规范被建立起来。我们使用另一种进化模型，拉马克模型来解释社会进化的过程，与更常见的达尔文模型形成对比。在这个拉马克的过程中，一代获得的行为习惯可以有意识地传递给下一代，因此这里不再需要达尔文模型中相对机械的随机匹配和复制因子动力学特征。我们希望这里提出的工作能够为不断增长的社会制度文献提供补充，并为研究重复博弈中的进化问题提供一个现实的框架。所进行的所有实验都具有相同的总体结构。研究对象是从大学生群体中招募的，并按顺序分成几代，每代人在预先规定的时间内，与相同的对手重复进行预先规定的游戏。在他们参与游戏的过程中，在不同的时间，他们被要求预测对手采取的行动，以便深入了解对对手的看法是如何随着时间的推移而变化的，并寻找几代人之间的信念趋同，作为惯例出现的证据。在他们参与游戏之后，被下一代所取代，他们能够看到一些在他们之前发生的历史。这个历史可以是他们在实验中的直接前辈的历史，也可以是他们之前所有受试者的全部历史。每一代的研究对象都能够以提出正式策略的形式，或简单地写下要做什么并解释为什么要给出这样的建议的形式，给他们的继任者提供建议。任何主体的收益都等于这一代人在其一生中所获得的收益加上取决于其继任者所获得的收益的贴现收益。我们研究了使用这种非重叠世代设计的四种不同游戏中的行为。调查的四个游戏包括一个协调游戏（性别之战），一个公平考虑很重要的游戏（最后通牒讨价还价游戏），一个信任游戏和一个个人学习任务。

Project Summary This is a collaborative project with Barry Sopher, 9709079. This project consists of an experimental investigation of how social conventions and the information contained in them are created and passed on between generations of social agents. A large literature in game theory examines infinitely repeated games played by a set of fixed and unchanging players. In the real world, while games may go on indefinitely, the agents who play them have finite lives. When these agents die or retire, their roles in these infinitely-lived games are taken by other agents who, perhaps in consultation with their predecessors, continue to play the game after them. We examine the transition from one generation to another, where conventions and norms are either passed along, or are disrupted and new ones established. We do so using an alternative evolutionary model, the Lamarckian model, in contrast to the more common Darwinian model, to explain the process of social evolution. In this Lamarckian process conventions of behavior acquired by one generation can be consciously passed on to the next, so that the relatively mechanical random matching and replicator dynamics characteristic of Darwinian models are no longer needed here. We expect that the work proposed here to add to the growing body of literature on social institutions, as well as provide a realistic framework within which to study the problem of evolution in repeated games. All the experiments performed have the same general structure. Subjects are recruited from undergraduate populations and ordered into generations in which each generation plays a pre-specified game repeatedly, with the same opponent, for a pre-specified length of time. During their participation in the game, at various times, they are asked to predict the actions taken by their opponent in order to obtain insight into how beliefs about one's opponent change over time, and to look for convergence in beliefs across generations as proof of the emergence of conventions. After their participation in the game subjects are replaced by a next generation, who are able to view some history of what has transpired before them. This history might either be that of their immediate predecessors in the experiment or the entire history of all subjects before them. Subjects in each generation are able to give advice to their successors either in the form of suggesting a formal strategy or simply writing down a suggestion as to what to do and explaining why such advice is being given. The payoffs to any subject are equal to the payoffs earned by that generation during their lifetime plus a discounted payoff which depends on the payoffs achieved by their successors. Behavior in four different games using this non-overlapping-generations design are investigated. The four games investigated include a coordination game (the Battle of the Sexes), a game where equity considerations are important (the Ultimatum bargaining game), a game of trust, and an individual learning task.