Rational Institutions for Nonrational Agents

点击翻译按钮获取中文摘要

9514809 Sobel The studies described in the proposal are linked by the idea that intelligently designed institutions may perform well even if individual participants are boundedly rational, and that a measure of an institution's success is how well it performs when agents make mistakes. The proposal contains several related projects. One project studies auction mechanisms to extend the ideas, introduced by others, that these markets do not rely on optimizing behavior to operate efficiently. The goal is to prove that prices converge to the competitive price when participants are budget constrained but unsophisticated, and to identify a class of internally consistent, adaptive behaviors that generate efficient outcomes in complicated auction environments. Another project analyzes a mechanism design problem with two-dimensional output in order to determine how the ability to observe one component influences the product of the second. Two studies focus on the relationship between poorly informed agents and experts. One model describes the function of a market in which agents must select an expert, either from information supplied through advertisements or word of mouth. The other investigates when a boundedly rational agent should delegate decision making authority to an expert. For the most part, the techniques are standard tools from game theory and the theory of mechanism design; in particular, incomplete information will be the primary reason that agents cannot make perfect decisions.