A Model of Affect and Learning for Intelligent Tutors

A model of affect and learning for intelligent tutoring systems is proposed. The model considers both how a student feels and what a student knows, and then customizes how instruction is presented and how learning and performance are reinforced. The model was designed based on teachers’ expertise, which was obtained through interviews and interaction with an educational game on number factorization learning. The core of the model is a dynamic decision network, which generates tutorial actions balancing affect and knowledge. The student’s affect representation relies on a Bayesian network and theoretical models of emotion and personality. A controlled user study to evaluate the impact of the model on learning was performed. Current results are encouraging since they show significant improvement in learning when the model of affect and learning is incorporated.