Managing Conflict Resolution in Norm-Regulated Environments

Abstract

Norms, that is, obligations, permissions and prohibitions, are a useful abstraction to specify and regulate the actions of self-interested software agents in open, heterogeneous systems. Any realistic account of norms must address their dynamic nature: the norms associated to agents will change as agents act (and interact) - prohibitions can be lifted, obligations can be fulfilled and permissions can be revoked as a result of agents' behaviours. These norms may at times conflict with one another, that is, an action may be simultaneously prohibited and obliged (or prohibited and permitted). Norm conflicts prevent agents from rationally deciding on their behaviour. In this paper, we present mechanisms to detect and resolve normative conflicts. We achieve more expressiveness, precision and realism in our norms by using constraints over first-order variables. The mechanism to detect and resolve norm conflicts takes into account such constraints and is based on first-order unification and constraint satisfaction. We also explain how the mechanisms can be deployed in the management of an explicit account of all norms associated with a society of agents.