The use of agent-based social simulation for policy-making involves ethical considerations of three different kinds: (i) in the agent-based policy model itself: the choice of values that are to be imbued in the simulated agents and in the policies; (ii) in the functionality of the system that is designed to support a policy-making process; and (iii) in the use of the system to design, negotiate, deploy and monitor an actual policy. In this paper we propose a value-driven framework to elucidate the corresponding ethical concerns of these kinds and then outline some guidelines to address them. We use the water policy domain to motivate and illustrate the proposal.