Learning What is Similar: Precedents and Equilibrium Selection