In this paper we extend the analysis of optimal monetary policy rules in terms of stability of an economy, started by Evans and Honkapohja (2003b), to the case of heterogeneous private agents learning. Following Giannitsarou (2003), we pose the question about the applicability of the representative agent hypothesis to learning. This hypothesis was widely used in learning literature at early stages to demonstrate convergence of an economic system under adaptive learning of agents to one of the rational expectations equilibria in the economy. We test these monetary policy rules in the general setup of the New Keynesian model that is a work horse of monetary policy models today. It is of interest to see that the results obtained by Evans and Honkapohja (2003b) for the homogeneous learning case are replicated for the case when the representative agent hypothesis is lifted. Zdroj anotace: OKCZ - ANOTACE Z WEBU