Reasoning with agent preferences in normative multi-agent systems

Abstract

A fundamental feature of autonomous agents is the ability to make decisions in dynamic environments where alternative plans may be used to achieve their goals. From an individual perspective, agents can have personal preferences over some actions and they try to maximize their preference satisfaction in deciding the actions to achieve their goals. In normative multi-agent systems, however, agent actions are not only directed by their own personal preferences but also the normative constraints imposed by the system. Within this context, the agents decide on their actions based on the reasoning of (1) whether their actions violate the norms imposed by the system, and (2) to what extent their actions satisfy their individual preferences. As such, this necessitates mechanisms to provide the agents with information about both the normative consequence and the preference satisfaction of their actions in an integrated way. In this paper, we propose a unified framework to analyze agent interactions taking into consideration the agent preferences in the setting of normative multi-agent systems. To reason about the normative consequence, we use the normative language presented in [1]. To reason with agent preferences, we extend the work of Visser et al. [2] such that it