ICES: SMALL: Mechanism Design with Information-Sensitive Agents

机制设计研究的是当参与互动的主体都是自利的时候，如何设计出能够产生社会、政治和经济上所希望的结果的互动。虽然代理人的兴趣通常是通过对金钱和消费品等物质结果的偏好来捕捉的，但这往往不足以满足现实世界的互动。例如，今天的许多社会和经济互动都是在网上进行的，互动的记录被存档、维护并经常传播给其他人。在这种互动中，代理人不仅受到由此产生的物质结果的推动，而且可能还受到对他们的私人信息在互动过程中被泄露的内容和数量的担忧。例如，当他们的私人信息被披露时，代理人可能更喜欢一种结果而不是另一种结果，但如果没有信息披露，则可能更喜欢后一种结果。他们可能重视隐私，并希望防止他们的私人信息泄露。或者，他们可能希望可信地披露他们的私人信息，以此作为向其他人发出的信号。本项目开发了一个捕获这些类型的偏好的信息敏感代理的模型，并研究了为此类代理设计惰性操作的问题。本项目中解决的一些问题是：当代理具有信息敏感偏好时，当前理论文献中存在的机制和实践中的机制是否仍然有效？是否有可能设计出对不同级别的代理信息敏感的健壮机制？是否有机制具有额外的可取特性，例如保护隐私或可信地披露信息？该项目的更广泛影响是双重的。首先，一种普遍存在的信息敏感形式是对隐私的偏好，这是电子商务中最有争议的问题之一。由于这个项目考察了隐私的战略含义，它有可能为这场激烈的辩论提供有价值的见解。其次，当代理人参与许多机制时，信息敏感性尤其相关，在这种情况下，信息从一个机制向另一个机制的泄漏成为一个关键考虑因素。通过明确包括对信息泄露的关注，该项目促进了经济机制的设计和分析，这些机制既更强大，也更模块化。

Mechanism design studies the problem of designing interactions that yield socially, politically, and economically desirable outcomes when the agents participating in the interaction are self-interested. While agents' interests are typically captured by preferences over material outcomes such as money and consumption goods, this is often insufficient for real-world interactions. For example, many of today's social and economic interactions take place online, and records of the interactions are archived, maintained, and often spread to others. In such interactions agents are motivated not only by the resulting material outcomes, but possibly also by concerns about what and how much of their private information is revealed in the course of the interaction. Agents may, for example, prefer one outcome to another when their private information is revealed, but may prefer the latter outcome to the former if no information is revealed. They may value privacy, and wish to prevent the leakage of their private information. Alternatively, they may wish to credibly disclose their private information as a signal to others. This project develops a model of information-sensitive agents that captures these kinds of preferences, and examines the problem of designing ineractions for such agents.Some of the questions addressed in this project are: Do the mechanisms that exist in the current theoretical literature and in practice still "work" when agents have information-sensitive preferences? Is it possible to design mechanisms that are robust to various levels of information-sensitivity of the agents? Are there mechanisms that have additional desirable properties, such as the preservation of privacy or the credible disclosure of information?The broader impacts of this project are twofold. First, a prevalent form of information-sensitivity is a predilection for privacy, one of the most debated issues in electronic commerce. Since this project examines the strategic implications of privacy, it has the potential to contribute valuable insights to this heated debate. Second, information-sensitivity is particularly relevant when agents participate in numerous mechanisms, in which case the leakage of information from one mechanism to another becomes a critical consideration. By explicitly including concerns for information leakage, this project facilitates the design and analysis of economic mechanisms that are both more robust and more modular.