Motivated cooperation in autonomous agents

Multi-agent systems are underpinned by the notion of cooperation - the process by which independent agents act together to achieve particular goals. Cooperation between autonomous agents requires appropriate motivations on behalf of those agents, since an agent’s behaviour is guided by its motivations. Interaction with others involves an inherent risk and, to manage this risk, an agent must consider its trust of others in conjunction with its motivations in entering into, and continuing in, cooperative activity. The aim of this thesis is to develop a framework for motivated cooperation, focusing in particular on the motivational reasons an agent might have for cooperating, and how it can use the information it has about others (such as their capabilities and trustworthiness) to make informed judgements about the risk involved in cooperating. The main body of this thesis can be decomposed into four parts. First, we introduce the issues associated with motivated cooperation, identify the outstanding problems, and discuss the key related work that gives a context to the thesis. Second, we present the motivated agent architecture, SENARA, which forms the foundation of our framework. Third, we introduce the framework itself, drawing out the details related to motivation and risk, and describing how this framework can be instantiated in particular applications. Finally, we conclude the thesis by considering the contributions it has made, and identifying potential areas for future work