Cognitive model supported team skill training

Complex tasks require coordinated performance by multiple team members. To perform the task effectively each team member must not only master the individual task component but also needs to function in the overall team. To increase team performance, each team member will need to acquire the relevant teamwork skills. Smith- Jentsch, Zeisig, Acton and McPherson (1998) identified four teamwork skill dimensions that can be trained to improve team performance. The effective exchange of information, assessment of team member situation awareness and the ability to take initiative are examples of such dimensions. These dimensions all relate to communication between team-members. We can identify two prerequisite components for the exercise of teamwork skills, namely team-members and communication between team-members. However, the logistics of arranging teamwork training has been noted as a bottleneck (Schaafstal, 2001).
In order to alleviate this, it has been proposed to use software agents to play key roles in training (Pew & Mavor, 1998). A software agent is realised by first constructing a cognitive model of a human role-player and subsequently building a cognitive agent based on that model. An advantage of using software agents over human role players is that the behavior of the former is fully controllable. Many such agents have been investigated for various roles (Conati & Zhao, 2004). Despite the number of models that have been investigated, cognitive agents and their application to team skill training have received little attention. In this paper we describe how a cognitive model can be exploited in the design of a software agent that supports training team skills in the absence of human team members.