Logic and Games for Ethical Agents in Normative Multi-agent Systems

摘要

In this paper we study how to characterize ethical agents in normative multi-agent systems. We adopt a proposition control game together with input/output logic. Norms create the normative status of strategies. Agents' preference in proposition control games are changed by the normative status of strategies. We distinguish four ethical types of agents: moral, amoral, negatively impartial and positively impartial. Agents of different ethical types use different input/output systems and different procedures to change their preference. Preference changes induce normative proposition control games and notions like normative Nash equilibrium are then introduced. We study some complexity issues related to normative reasoning/status and normative Nash equilibrium.