The Efficiency of Observability and Mutual Linkage

This paper studies the adaptive learning dynamics over multiple locations. Our focus is to show that, to observe the average performance of other locations is necessary or efficient. Through the analysis of information structure, the result about convergence and stochastic stability can be obtained. And, for a large class of information structures and games, only Pareto efficient symmetric outcome can emerge in the unique stochastically stable equilibrium. In the Prisoner's Dilemma, for example, observability and mutual linkage between different locations can guarantee that in each location, both players must ultimately cooperate most of the time