Integrating social modelling and agent interaction through goal-oriented analysis

In the Agent-oriented software engineering (AOSE) community, it is now widely recognised that interaction is perhaps the most important single characteristic of complex software systems. Recently, the focus of AOSE research has shifted towards social (or organisational) modelling, which allows interaction to be modelled with higher-level constructs like organisational structures and social policies. However, a gap remains between the social models proposed in AOSE methodologies and the increasing number of sophisticated agent interaction frameworks outside AOSE, such as those based on auctions, commitments, and dialogue games. Hence, social models in AOSE methodologies need to have more 'hooks' to enable integration with rich domain-specific agent interaction frameworks. Towards closer integration between social modelling in AOSE and other agent interaction frameworks, we extend the ROADMAP methodology and provide a framework for enhancing the analysis of interaction requirements through goal-oriented analysis and social modelling. More precisely, goal models provide the purpose of interaction, while social models provide the social dependencies and policies that govern interaction. We show how the resulting requirements specification, complemented by protocol descriptions, can inform the design of the system's interaction model. The resulting approach bridges the gap described above by linking high-level social concepts to low-level, domain-specific design decisions about interaction.