Automating directed scenario-based training

In scenario-based training trainees prepare, ex-ecute and evaluate short storylines (scenarios) that resemble relevant real-world situations. Scenarios are situated within a simulated en-vironment, with role-players acting as team-mates, adversaries, etc. Role-players have to act in a representative and believable fashion. Therefore, subject-matter experts (usually staff of the school) are often asked to play these roles. This usually works well because (a) their domain knowledge enables them to act appro-priately, and (b) they choose to act in such a way that the trainee can achieve his learning goals. The downside of this solution is that training opportunities are limited to those occa-sions when staff is available. In addition, it sub-stantially elevates costs of training. New forms of training are needed that are less dependent on personnel, logistics and organization (train-ing anytime, anywhere). Recent research shows that intelligent agents can be developed that-within limits- show representative behavior for the role they represent. However, how to make these role-playing agents act in the best interest of the trainee is still an unsolved question. 2) Envisioned solution To bring about effective training, we suggest to develop a "director agent " (DA) for agent-based scenarios. A DA monitors trainee performance and the flow of events in the simulation. If the DA concludes that effective learning requires adjustments in the scenario, it can manipulate the course of events by delivering behavioral instructions to the role-playing agents