Intent Inference for Users, Teams, and Adversaries:
Papers from the AAAI Fall Symposium
Eugene Santos Jr. and Benjamin Bell, Cochairs
Advances in AI have enabled decision support systems to assume substantive roles in supporting human operators of complex systems. As such systems become more capable of autonomous performance, they must engage more fully with human operators in negotiating task assignments, anticipating near-term needs, and proactively providing information, analysis, and alerts. As this need for such sophistication extends to systems with multiple operators, research into team and adversarial intent inference becomes critical. The notion of a team or crew is central to applications involving complex systems and organizations ranging from transportation systems to command and control centers. For adversarial intent inference, decision support for teams facing an intelligent opponent (hostile force) is limited without an understanding of the adversary's goals and actions. In response to this fast-emerging need, researchers are now focusing on team dynamics and workflow. From a socio- anthropological heritage come approaches to capturing workplace procedures and information flows; cognitive task analysis contributes tools for capturing process models; and principles of reasoning under uncertainty allow for such models to remain robust under the complex conditions typical of multiple operators of complex systems. More recently, collaborative agent research has opened new avenues in modeling and implementing teams of cooperative agents, and "keyhole" approaches to non-intrusive observation of users' promises to enable systems to reliably track users' progress. By bringing together researchers throughout the community, this symposium helped foster the emerging discipline of team intent inference and promote the development of intent-aware decision support for multi-operator complex systems.