Supporting teamwork using software agents in human-agent teams
Michael Anthony Lewis, Terri L. Lenox · Medical Entomology and Zoology · 2000
For this research, intelligent aiding strategies were evaluated for a team planning task. Planning environments are often complex and dynamic and planners often lack the knowledge, skills and time to perform successfully. Team planning has additional burdens due to the distribution of the resources, knowledge and skills across the individual team members. Recently, software agents have emerged as a suitable metaphor for interacting with computer technology particularly in these kinds of complex situations and tasks. Such software agents can handle parts of the workload associated with interacting with computer technology and allow the humans to concentrate on other activities, such as assessing the situation, making decisions or reacting to changes in the system (Zachary, Le Mentec and Ryder, 1996). In addition, agents can retrieve information on request, and actively and intelligently anticipate, adapt and seek ways to support humans (Sycara et al., 1996; Bradshaw, 1997; Lenox, Hahn, Lewis & Roth, 1999). The research described investigated how to effectively incorporate software agents into human teams for complex team tasks such as joint military operations planning. The aim of this research was to determine if the teamwork dimensions of information exchange and supporting behaviors could be supported in human-agent teams by supporting the individual members in their tasks; reducing their cognitive workload so that they can concentrate on teamwork and team-related tasks. A computer-based simulation environment for route planning called MokSAF was developed. Teams of three humans along with software agents participated in a cooperative planning task. The task studied was deliberative, iterative and time-critical and was theorized to have a high workload for team members. Two different software agents (Autonomous Agent, Cooperative Agent) and an unaided condition (Control) were developed for this simulated military planning environment. The research found across the board advantages for the agent-based aiding conditions for individual performance measures but little evidence to support the hypotheses that aided teams would perform better on teamwork and team-related tasks.