Norm-governed practical reasoning agents

This thesis describes a model of norm-governed practical reasoning agents and demonstrates the implementation of this model in the form of the NoA Normative Agent language and architecture. The introduction of normative concepts such as obligations, permissions and prohibitions into the practical reasoning of an agent is motivated by the need for effective coordination mechanisms in open systems. These are typically “multi-vendor” scenarios, where independently designed agents, as representatives of human organisations and individuals, form short to medium term coalitions and collaborate in the performance of specific tasks. Electronic commerce is one of the most prominent examples of such scenarios. The use of autonomous software agents provides the necessary flexibility in these scenarios, but the critical issues such as action coordination and trust remain. Agents joining a group, or society, of agents must undergo a process of socialisation – they are required to accept the “normative standard” of such a group that regulates the actions and interactions of members. Such a normative standard cannot be hard-coded into an agent, as this standard may not be known in advance to the designer of the agent. Agents need a form of “social awareness” – they must be able to adopt and, ideally, abide by societal norms. The importance of normative concepts as a regulatory mechanism can be seen by observing their analogue in human society. In fact, mature research exists in the understanding of how legal, or normative, systems are established within human societies and how they impact on the activities of social individuals. This thesis takes normative concepts from human societies as an inspiration for the development of norm-based artificial societies and norm-governed practical reasoning agents. The key contribution of the research presented in this thesis is a model of agents that are able to take normative positions into account during practical reasoning. With the ability to acquire new norms and the capacity to reconsider them as appropriate influences on its activities, the agent is enabled to not just apply norms, but to reason about whether to honour them – the agent becomes normautonomous. Salient Points of the Thesis • An abstract model of norm-governed agency that describes how agents choose their actions based on explicitly represented normative concepts such as obligations, permissions and prohibitions in a process of norm-governed practical reasoning • A language for the specification of normative concepts such as obligations, permissions and prohibitions and of realistic legally binding contracts as a set of such norms • A programming language for norm-governed agents that provides means for the declaration of plans as the agent’s pre-specified capabilities • The NoA Normative Agent Architecture, which is a reactive planning architecture based on this abstract model for the implementation of norm-governed practical reasoning agents • A detailed investigation of possible conflicts and inconsistencies in sets of norms adopted by a norm-governed agent and the presentation of detection and resolution strategies used during norm-governed practical reasoning • A specific form of interaction for norm-governed agents, called Supervised Interaction, to allow the execution of realistic legally binding contracts in a situation of trust