| dc.description.abstract | To facilitate autonomous training in virtual environments, recent studies have explored the integration of didactically founded methods, such as Scenario-Based Training (SBT), into serious games. The resulting Adaptive Educational Games (AEGs) monitor the performance of the trainee, and adapt the game to match this performance. To ensure effective training opportunities, an AEG should not only adapt the game during the scenario, but should also select appropriate scenarios based on the needs and abilities of the trainee. However, manual scenario creation is a time-consuming and often costly process. Writing personalised scenarios on a larger scale is simply not feasible. Consequently, the need arises for automated scenario generation techniques. In this thesis I present a framework for an automated scenario generation system. First, I consider the requirements of effective training scenarios. Next, the design, implementation, and evaluation of an automated scenario generator are discussed. The design of the framework is based on the idea that the content of the game, combined with the actions of the virtual agents, determines the training experience for the trainee. Therefore, the framework couples the planning of trainee actions with the selection of appropriate content for the virtual environment. To test the framework, a prototype has been implemented for the domain of burn-related incidents in First Aid training. The scenarios generated by this prototype have been compared to those written by human experts and laymen. The results followed the hypothesised trend with the expert scenario being judged best, followed by those generated by the sytem, and finally the laymen scenarios. However, all differences were not significant. It seems likely that there will be significant differences, at least between the expert and layman scenarios, in experiments with higher power. Further research will be required before conclusive answers can be given. For now, the combination of automated planning techniques with Smart Objects seems a viable approach to automated scenario generation. At the very least, the framework warrants further research, and several directions for additional studies have been identified. Adding an automated scenario generator to an AEG will provide the game with even more personalisation capabilities, and offer the trainee varied and appropriate training experiences. | |
