A cognitive model for the generation and explanation of behaviour in virtual training

services. In most of the current systems instructors play a major role; they enter the effects of trainee actions into the system, play the role of other characters in the scenario, and provide instruction, guidance and feedback to the trainee. Cognitive models can be used to support instructors by autonomously generating virtual character behavior. However, without knowing the reasons for an action, it is more difficult for a trainee to understand the situation and learn from it. Some virtual training systems with explanation components have been proposed, e.g. Debrief [3] and the XAI system [5, 1]. These components explain a characters actions by giving information about its state at the time it performed an action. However, the explanations do not provide the characters underlying motivations for its actions. We propose a cognitive model based on BDI theory (belief desire intention) that is able to generate character behavior and explanations. Because the character reasons and plans with BDI concepts, the model is also able to generate explanations in terms of the characterss beliefs and goals. Such explanations in high level concepts are easy to understand for humans. Moreover, they not only provide information about the characters state, but yield insight in the reasons why it performed a certain action