An Analysis of a Simple Reinforcement Dynamics: Learning to Play an "Egalitarian" Equilibrium

The paper analyses a simple reinforcing dynamics. The dynamics can be interpreted as a learning dynamics with fixed aspiration level. All payoffs are assumed to be above this aspiration level, therefore all strategies are reinforcing. Different versions of the dynamics exhibit different convergence properties. The analysis starts with one-agent decision problems and proceeds to games. Some results are available for decision problems and simple games. For complex games computer simulations are performed. The hypothesis is that the dynamics favors an "egalitarian" equilibrium even if it does not satisfy other refinements.