Robust agent teams via socially-attentive monitoring

Agents in dynamic multi-agent environments must monitor their peers to exec ute individual and group plans. A key open question is how much monitoring of other agents' states is required to be effective: The Monitoring Selecti vity Problem. We investigate this question in the context of detecting fail ures in teams of cooperating agents, via Socially-Attentive Monitoring, whi ch focuses on monitoring for failures in the social relationships between t he agents. We empirically and analytically explore a family of socially-att entive teamwork monitoring algorithms in two dynamic, complex, multi-agent domains, under varying conditions of task distribution and uncertainty. We show that a centralized scheme using a complex algorithm trades correctness for completeness and requires monitoring all teammates. In contrast, a sim ple distributed teamwork monitoring algorithm results in correct and comple te detection of teamwork failures, despite relying on limited, uncertain kn owledge, and monitoring only key agents in a team. In addition, we report o n the design of a socially-attentive monitoring system and demonstrate its generality in monitoring several coordination relationships, diagnosing det ected failures, and both on-line and off-line applications.