Learning, Memory, and Inertia

This paper explores the impact of memory in standard display imitation behavior, focusing on coordination games (as in Kandori et al (1993)) and N-player games where spiteful behavior allows to discard Nash equilibria. It is shown that the way interea is modeled in such examples actually entails a strong "no-memory" assumption. Once inertia is removed (or medeled otherwise), the addition of bounded memory changes the predictions dramatically. The analysis highlights the stability properties of Nash outcomes in purely evolutionary contexts with a finite population of agents.