Optimal Policy with Heterogeneous Preferences