Loading...
Please wait, while we are loading the content...
Similar Documents
Modelling Predicting How People Play Games: Reinforcement learning in experimental games with unique
| Content Provider | Semantic Scholar |
|---|---|
| Author | Ido, Erev Roth, Alvin E. |
| Copyright Year | 1998 |
| Abstract | We examine learning in all experiments we could locate involving 100 periods or more of games with a unique equilibn'um in mixed strategies, and in a new experiment. We study both the ex post (" best fit ") akscrtptive power of learning models, and their ex ante predictive power, by simulating each experiment using parameters estimated from the other experiments. Even a one-parameter reinforcement learning model robustly outpelfonns the equilibrium predictions. Predictive power is improved by adding ' yorgemng " and " experimentation, " or by allowing greater ra-tionality as in probabilistic jictitious play. Implications for developing a low-rationality, cognitive game theory are discussed (JEL C72. C92) i Game theory has traditionally been developed as a theory of strategic interaction among players who are perfectly rational, and who (consequently) exhibit equilibrium behavior. This approach has been complemented by evolutionary game theory, which, motivated by biological evolution, seeks to understand how equilibria could arise in the long term by selection among generations of players who need not be rational or even conscious decision makers. Somewhere in between are models of learning, which consider the adaptive behavior of goal-oriented players who may not be highly rational, both to provide foundations also contributed to the design and programming of the new experiment. We are indebted to Barry O'Neill, Jack Ochs, and Amnon Rapoport for access to unpublished parts of their data. The present version reflects numerous comments by three anonymous referees on several earlier drafts. for theories of equilibrium and to model em-: pirically observed behavior. The present paper considers how well ~im-.~'! ple learning models, motivated by the pay-' chology of learning, can model the interaction :. of players who must learn about the game and. C each other in the course of playing the game, over time spans that may not be long enough to lead to equilibrium. Our goal will be to model observed behavior, starting with behavior observed in experimental settings. (In the conclusion we will also consider the implications of this approach for applied economics in naturally occurring, nonexperimental settings .) We will show that a wide range of experimental data can be both well described ex post and robustly predicted ex ante by a very simple family of learning theories. Economists have traditionally avoided explaining behavior as less than rational for fear of developing many fragmented theories of mistakes. Part of the attraction of highly rational models … |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.economics.harvard.edu/~aroth/papers/AER884.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
