Multiparty Dialogic Processes of Goal and Strategy Formation in Hybrid Teams

Andreas Wendemuth, Stefan Kopp

Abstract: A current trend which is already prevalent in highly structured (e.g. industrial) working environments is the cooperation of people, intelligent physical agents (robots) and, in parts, intelligent information agents (AIs, chatbots) in hybrid teams, mostly in paired settings. This position article discusses a major generalization and extension of this concept: (a) the goals, strategies, and actions are not fully prespecified, but develop in the course of a dialogic process; (b) the agents are not merely tools or assistants, but proactively intervene as peers; (c) the hybrid teams are multiparties with several humans and (situated or remote) intelligent agents, exhibiting and modelling pronounced group behavior. Cognitive, dialogic systems are the technical backbone of such team settings, bringing together techniques of multimodal processing, information retrieval, situated action planning and autonomous action generation, recognizing and anticipating task-related states of the actors.