Logics for Modelling and Verifying Normative Multi-Agent Systems

Abstract

A multi-agent system is a computerized system that is composed of multiple interacting agents within an environment. Such an agent can be a software program, but could equally well be a human actor. These systems are making their way into our everyday life, for example electronic marketplaces where agents can buy and sell goods, or smart grid energy systems where agents can produce or consume energy based on supply and demand. These systems are typically distributed and decentralized, and the participating agents may have diverging information, diverging interests, or both. The internals and architecture of these participating agents may be unknown to us, which disables us to make any strong assumption on the possible behaviour that these agents may exhibit. Moreover, these systems are generally composed or designed with a specific goal in mind. For example, an electronic marketplace is composed to facilitate safe transactions between currency and goods. Depending on the actual behaviour of the system these goals may, or may not, be achieved. Hence there is a need to regulate and control the behaviour of the system and the participating agents. In order to achieve this, norms have been proposed to regulate, coordinate, and control the behaviour of the agents and system, leading to the field of research called normative multi-agent systems. In this thesis we study enforcement norms, which are norms which can be violated, and which can lead to sanctions when a violation is detected. However, it can be the case that the norms are not well-designed, and that there exists ways for an agent to exploit the system. This may for example lead to dangerous situations within the system, or high costs for the designer. As these normative multi-agent systems can grow in size and scale, formal methods and tools are needed in order to prove whether such a system works as intended. This thesis aims to design and analyse formal logic-based frameworks in which normative multi-agent systems can be modelled and verified. Specifically, we design frameworks that aim to capture the behaviour of a normative multi-agent system, and we develop logics to reason about properties of such a system. Ultimately these logics can be used to verify the correctness of these systems. Apart from the many theoretical results, an important contribution this thesis makes is that it bridges the gap between the field of normative multi-agent systems and the modelling and verification of formal systems.