Learning in Games and in Market

点击翻译按钮获取中文摘要

SBR-9424013 Drew Fudenberg One of the things that separates human beings from other animals is their great capacity to learn about their environment through observation and experience. Yet, surprisingly, learning is often ignored in economic modeling and analysis. This research consists of four interrelated projects about the way "rational" people learn. The ultimate goal of this research is to introduce a new interpretation of "out-of-equilibrium" actions by a player in the course of a game: namely, that such puzzling actions are "experiments" designed to help the player learn the strategies being used by its opponents. By rigorously inferring the implications of the assumption that economic agents are constantly learning about their opponents and their environment this research has the potential to greatly extend the range of application of economic theory as well as increase its predictive ability. The first project looks at the choices made by consumers who try to use the experiences of friends and neighbors to choose among goods of differing quality. The second project analyzes models in which players try to learn of their opponents' strategies through repeated observation without the benefit of any other communication with them. The goal is to find learning rules that will perform well in a wide variety of strategic environments. The third project concerns the effect of learning in so-called "signaling games". These games consist of situations in which players have private information about the environment that cannot be independently verified. As a result, "talk is cheap." The only way players can communicate their information to other players is by taking costly actions that will "reveal" the information they possess. In signaling games, there are many equilibria, some of which seem implausible. The research examines which equilibria are likely to arise if players conduct "experiments" during the course of play to learn about their opponent's strategies. The last projec t examines the pricing of by a durable good monopolist when the good changes over time. Examples are computer hardware and software and automobiles.