Learning in a Local Interaction Hawk-Dove Game

Abstract

We study how players in a local interaction hawk dove game will learn, if they can either imitate the most successful player in the neighborhood or play a best reply versus the opponent's previous action. From simulations it appears that each learning strategy will be used, because each performs better when it is less popular. Despite that, clustering may occur if players choose their learning strategy on the basis of largely similar information. Finally, on average players will play Hawk with a probability larger than in the mixed Nash equilibrium of the stage game.