Towards a Model of Agent-Assisted Team Search

Abstract

This paper is the first step in our research plan towards addressing the fundamental question of how software agents can best aid distributed human teams performing time-stressed critical tasks in uncertain and dynamic environments. Based on prior work, we hypothesize that to improve the performance of human teams, agents must do some combination of the following: (1) reduce the cost of the humans?information processing; (2) decrease uncertainty in the task; (3) improve coordination between team members; (4) directly assist in task completion. In order to (a)establish an experimental baseline of the performance of human-only teams for some particular task domain, and (b) best understand where agents can provide best utility in supporting human teamwork, we designed scenarios and performed experiments with human teams performing a time-stressed, collaborative search task in a multi-player gaming environment. The collaborative search task recreates some of the challenges faced by human teams during search and rescue operations, such as the one described in the Holistan scenario. In our experiments, we analyze (1) verbal communication between team members and (2) the effects of presenting or omitting task progress information. By ascertaining the information processing and coordination requirements of this team task, we expect to identify ?nsertion points?for agent assistance to human teams. Agent assistance will be particularly critical to military teams as their operations become more agile and situation specific. As unfamiliar forces are brought together for different coalition missions, agent support of teamwork becomes crucial.