Abstract

Teaching methods must adapt to learners’ expectations. Computer game-based learning environments enable learning through experimentation and are inherently motivational. However, for identifying when learners achieve learning goals and providing suitable feedback, Intelligent Tutoring Systems must be used. Recognizing the learner’s affective state enables educational games to improve the learner’s experience or to distinguish relevant emotions. This chapter discusses the creation of an affective student model that infers the learner’s emotions from cognitive and motivational variables through observable behavior. The control-value theory of ‘achievement emotions’ provides a basis for this work. A Probabilistic Relational Models (PRMs) approach for affective student modeling, which is based on Dynamic Bayesian Networks, is discussed. The approach is tested through a prototyping study based on Wizard-of-Oz experiments and preliminary results are presented. The affective student model will be incorporated into PlayPhysics, an emotional game-based learning environment for teaching Physics.PRMs facilitate the design of student models with Bayesian Networks. The effectiveness of PlayPhysics will be evaluated by comparing the students’ learning gains and learning efficiencies.