Approximate solutions of moral dilemmas in multiple agent system

Moral dilemmas are one of the major issues of current research in ethical reasoning. In particular, it is well known that admitting moral dilemmas in Standard Deontic Logic generates a family of inconsistencies that are intrinsically unsolvable. Since managing dilemmas means performing preferential reasoning, we argue that one simple approach to both types of problems is by ordering actions. We notice that in general, more than local orderings between two actions, agents have intrinsic preferences based on classification issues, like the action type, and that, once we have discharged the dilemma as it is intrinsically, preferential reasoning is performed by using a second-level choice approach. Decision theory has dealt with the problem of making decisions in presence of conflicting decision criteria, and some researcher has pointed out that this is the case of moral dilemmas as well. In practice, the choice of preferences in presence of conflicting criteria can be seen as a form of preferential-ethical reasoning. Although this is certainly an important topic in multiple agent investigations, it is definitely neglected in the current investigations. It is well known that humans are quite clever in solving moral dilemmas, and the usage they make of preferential reasoning is very complex. In this paper we address the problems of preferential-ethical reasoning in a combinatorial fashion and provide an algorithm for making decisions on moral dilemmas in presence of conflicting decision criteria. We then evaluate the complexity of the algorithm and prove that this approach can be applied in practice.