Hermes: implementing goal-oriented agent interactions

Traditional approaches to designing agent interactions focus on defining agent interaction in terms of legal sequences of messages. These message-centric approaches are not a good match with autonomous proactive agents since they unnecessarily limit the agents' autonomy and flexibility. The Hermes methodology proposes an approach for designing agent interactions in terms of interaction goals. In this paper we focus on how Hermes designs can be implemented by mapping the design artefacts to collections of plans