Almost half a century intelligent tutoring systems have been developed towards imitating the learning process of a student and a tutor interaction in a one-to-one tutoring situation. However, the gap for this kind of systems still exists in showing the adaptation skills possessed by human-tutors, particularly, the systems lack emotional intelligence. The paper presents conceptual model of agent-based affective tutoring system for the simulation of human-tutors’ and students’ interaction using multi-agent approach for representation of involved parties. Such simulation would allow assessing the effectiveness of selected teaching approach on student’s emotional state, behaviour, and learning progress. Since ontologies play an important role in the agent interaction, the design and usage of knowledge structures necessary for ITS functioning including emotion ontology are considered in this paper as well.