The report covers issues pertinent to the design and evaluation of the communication between humans and intelligent software agents necessary to enable a collaborative relationship. For human–agent interaction to be robust in a dynamic real-world situation, both software agents and humans must be able to communicate their overall intent in terms of mission objectives. Because of differences in their reasoning processes, capabilities, and knowledge bases, humans and agents are not an analog for human teams. We discuss the technical issues involved in effective communication including models of mutual transparency, natural language processing (NLP), artificial intelligence (AI), and explainable AI. Lacking a theory of mind, which enables humans to gain insight into their teammates? mental processes, agents have a difficult time anticipating human information needs and future actions. Research in collaborative planning involving multiple agents and research into synthetic shared mental models are used as exemplars of attempts to integrate humans and agents into a synergistic unit. However, we conclude that progress in NLP, explainable AI, and human science will be necessary before humans and agents communicate like human teams during complex, uncertain missions.
展开▼
