Approaches to Communication in Games

本研究包括四个项目，涉及代理之间的有效信息传递。一项研究将调查廉价谈话游戏的进化稳定结果。在进化的框架中，关于私人信息的交流是不可忽视的。它迫使玩家在不存在利益冲突的情况下获得有效回报。我们的目标是研究进化力量是否有利于消息灵通的一方，当谈话廉价但不免费时，信息是否得到有效利用，以及当玩家的兴趣部分一致时，是否必须进行重要的交流。该项目还将说明，当代理人有不同的利益时，必须进行多少沟通。一项相关研究将调查随着时间变化的交流游戏，以发现单词是否会在多个上下文中具有适用的含义，以及代理是否会在复杂的环境中诚实地使用单词，即使这违背了他们的短期战略利益。第三个项目将为研究基于偏差的均衡优化和游戏中的动态行为提供一个总体框架。这个话题与沟通有很强的联系，虽然它与沟通没有直接的关系。这项研究将更好地理解在进化稳定的策略集合不存在的情况下的结果。第四个项目将研究一种特定的沟通模式，以描述将决策权委托给知情专家（医患关系中的医生）的最佳条件。在这个模型中，病人可能无法获得完美的信息，要么是因为他是有限理性的，要么是因为完全沟通不符合医生的最佳利益。由于医生的偏好与患者不同，或者由于医生不知道患者的偏好，患者将完全的决策权交给医生可能是不合适的。初步研究表明，当医生和病人有不同的偏好时，充分的沟通可能是不可能的。

This research consists of four projects related to effective transmission of information between agents. One study will investigate evolutionarily stable outcomes in cheap-talk games. Communication about private information cannot be ignored in an evolutionary framework. It forces players to obtain efficient payoffs when no conflicts of interest exist. The goal is to study whether evolutionary forces favor the informed party, whether messages are used efficiently when talk is cheap but not free, and whether nontrivial communication must occur when players' interests partially coincide. The project will also characterize how much must be communicated when agents have different interests. A related study will investigate games of communication that change over time in order to discover whether words will take on meanings that apply in more than one context and whether agents will use words honestly in complicated settings even if it is against their short-term strategic interests. The third project will provide a general framework for studying deviation-based equilibrium refinements and dynamic behavior in games. This topic has strong connections with communication, although it is not directly related to communication. This study will better our understanding of results in situations where evolutionarily stable sets of strategies fail to exist. The fourth project will study a specific model of communication in order to describe conditions under which it is optimal to delegate decision-making authority to an informed expert (the doctor in a doctor-patient relationship). In the model the patient may be unable to acquire perfect information either because he is boundedly rational or because it is not in the best interests of the doctor to communicate completely. It may be inappropriate for the patient to give complete decision-making authority to the doctor either because the doctor's preference is different from the patient's or because the doctor does not know the patient's preference. Preliminary study suggests that full communication may be impossible whenever the doctor and the patient have different preferences.