We describe preliminary research on intelligent pedagogical agents designed to maximize students' learning from educational games, while maintaining the high level of positive emotional engagement that games usually trigger. The agents' behavior is based on a decision theoretic model, dictating that agents act so as to maximize the expected utility of their actions. In oar framework, a pedagogical agent's utility function is defined to take into account both a user's learning and emotional state. In this paper we will first describe this general decision-theoretic model. We will then illustrate in detail how the user's emotional state is assessed within the model by taking into account the interaction 'with the game, the user's personality and the user's bodily expressions that are known to be directly influenced by emotional reactions.
展开▼
