We have developed an affective behavior model for intelligent tutoring systems that considers the affective and the pedagogical state of the student; and based on both selects the tutorial actions to be presented to the student. The model establishes the affective student state with base on the OCC model by means a dynamic Bayesian network. To select the tutorial action the model uses a dynamic decision network designed with base on the opinions of knowledgeable teachers. The results of a pilot study show that the actions selected by the tutor are in general in agreement with the students’ expectations.