SSE/EFI Working Paper Series in Economics and Finance 474
In this paper I define an evolutionary stability criterion for learning rules. Using Monte Carlo simulations, I then apply this criterion to a class of learning rules that can be represented by Camerer and Ho's (1999) model of learning. This class contains perturbed versions of reinforcement and belief learning as special cases. A large population of individuals with learning rules in this class are repeatedly rematched for a finite number of periods and play one out of four symmetric two-player games. Belief learning is the only learning rule which is evolutionarily stable in almost all cases, whereas reinforcement learning is unstable in almost all cases. I also find that in certain games, the stability of intermediate learning rules hinges critically on a parameter of the model and the relative payoffs.
Bounded rationality Evolutionary game theory Evolutionary Stability Learning in games Belief learning Reinforcement learning