Normative Multi-Agent Systems

Abstract

This report documents the program and the outcomes of Dagstuhl Seminar 18171 “Normative Multi-Agent Systems”. Normative multi-agent systems combine models for multi-agent systems with normative concepts, like obligations, permissions, and prohibitions. As such, they promise to be a suitable model, for example for (regulated) multiagent societies, organizations, electronic institutions, autonomous agent cooperation (with humans-in-the-loop) and much more. The aim of this seminar was to bring together researchers from various scientific disciplines, such as computer science, artificial intelligence, philosophy, law, cognitive science and social sciences to discuss the emerging topic concerning the responsibility of autonomous systems. Autonomous software systems and multi-agent systems in open environments require methodologies, models and tools to analyse and develop flexible control and coordination mechanisms. Without them, it is not possible to steer the behaviour and interaction of such systems and to ensure important overall properties. Normative multi-agent systems is an established area focussing on how norms can be used to control and coordinate autonomous systems and multi-agents systems without restricting the autonomy of the involved systems. Such control and coordination systems allow autonomous systems to violate norms, but respond to norm violations by means of various sanctioning mechanisms. Therefore it is crucial to determine which agents or agent groups are accountable for norm violations. The focus of this seminar laid on how the responsibility of autonomous systems can be defined, modelled, analysed and computed.