| Peyton Young (University of Oxford). |
| Learning by trial and error |
| Abstract In classical game theory it is usually assumed that players are perfectly rational and possess common knowledge of the game. Yet there are many situations involving large numbers of players where the ‘game’ is so large and complex that for all practical purposes it is unknown to the players themselves. In such situations they may try to feel their way through a trial and error learning process. We show that in any normal form game with generic payoffs such trial and error learning eventually leads to Nash equilibrium behaviour. Moreover the basic features of such learning rules are validated in laboratory experiments where subjects are given no information about the game and cannot observe the actions of others. |