Group intentions

Groups of intelligent agents are studied in several subareas of Artificial Intelligence, notably, autonomous agents, multiagent planning and action, discourse understanding, and cooperative work. I motivate and present a formal theory of the intentions of a group of agents that analyzes the intentions of a group in terms of its internal organization, and the intentions of its members. This theory treats social structure directly in terms of the interactions among agents, and does not attempt to reduce it to psychological concepts. It makes few assumptions about the architecture of agents and about the manner in which they interact. Thus it is intuitively plausible, can be described in a simple formal model, and is applicable to a large variety of multiagent systems. This theory is applied to an extended example, and some interesting variations are outlined. ∗ I am indebted to Michael Huhns and to three anonymous referees for comments. 1