糖心原创

The evolution of taking roles

Abstract

In Hawk-Dove games or anti-coordination games there is a striking difference between the evolutionary outcomes in a single-population model vs those in multi-population models. In this paper we consider what happens if the role a player assumes evolves endogenously. More precisely, we consider a single population model in which players have payoff-irrelevant, but observable labels. Thus players can have strategies which are contingent on these labels. In any evolutionary stable state players with different labels manage to anti-coordinate. However, the emerging probability distribution over labels is not efficient and welfare gains from the emerging different roles are limited. Furthermore, we find a key distinction that influences the evolutionary outcome. In Hawk-Dove games where players would always prefer their opponent to play “dove” (independently of their own choice) only a single label has positive weight in any hierarchical structure, while in “true anti-coordination games” (in which you always prefer your opponent to mismatch your action) additional labels are played with positive probability, yet payoffs are bounded away from those attainable with the efficient probability weights.