An affective student model to assess emotions in an educational game

Electronic educational games integrate the target domain knowledge in a game-like environment in order to help students learn. While, in general, these educational games are more engaging than the traditional computer-based educational software, they often do not necessarily trigger learning. One explanation is that many students play the games without actively reasoning about the underlying domain knowledge. To make learning more effective in educational games, we are designing intelligent pedagogical agents that can provide tailored interventions to students. However, in order not to compromise the high level of engagement that is the main advantage of educational games, it is important for these agents to consider students' emotional states in addition to their cognitive states (such as learning) to decide when and how to provide interventions. This thesis focuses on the creation of an affective student model that assesses the students' emotional states while they are playing an educational game Prime Climb. The affective student model explicitly represents the cognitive appraisal process of emotions by implementing the cognitive theory of emotions (OCC Model). It relies on Dynamic Decision Networks (DDNs) to deal with the high level of uncertainty involved in affective user modeling. The initial version of the model was built based on the observation of two preliminary Prime Climb user studies, our intuitions, and psychological findings. This model was then revised based on the results of a third Prime Climb study.Science, Faculty ofComputer Science, Department ofGraduat