Session T2F 1-4244-0257-3/06/$20.00 © 2006 IEEE October 28 – 31, 2006, San Diego, CA 36 th ASEE/IEEE Frontiers in Education Conference T2F-22 Incorporating an Affective Model to an Intelligent Tutor for Mobile Robotics Yasmín Hernández 1 , Julieta Noguez 2 , Enrique Sucar 3 , Gustavo Arroyo-Figueroa 1 1 Instituto de Investigaciones Eléctricas, Gerencia de Sistemas Informáticos, Cuernavaca, Morelos, México, {myhp, garroyo@iie.org.mx} 2 Tecnológico de Monterrey, Campus Cd. de México, México, D. F., México, noguez@itesm.mx 3 Instituto Nacional de Astrofísica, Óptica y Electrónica, Tonantzintla, Puebla, México, esucar@inaoep.mx Abstract - Emotions have been identified as important players in motivation, and motivation is very important for learning. When a tutor recognizes the affective state of the student and responds accordingly, the tutor may be able to motivate students and improve the learning process. We propose a general affective behavior model which integrates information from the student's pedagogical state, affective state, and the tutorial situation, to decide the best tutorial action, considering the tutor preferences from a pedagogical and affective point of view. Our proposal is based on emotions models, personality theories and teachers’ expertise. The affective model is implemented as a dynamic decision network, with utility measures on both learning and motivation, and is being incorporated to an intelligent tutor within a virtual laboratory for learning mobile robotics. This paper presents preliminary results in the construction of the affective behavior model. Index Terms - Affective state, decision networks, intelligent tutoring systems, student model, virtual laboratories. INTRODUCTION We have developed an intelligent tutoring system coupled to a virtual laboratory. This environment provides the student with the opportunity to learn through exploration within simulated experiments. Preliminary results show that students who had help of tutor improved their knowledge of the target knowledge objects [1]. During our studies, we detected that motivation is a very important aspect when students use a virtual laboratory, and their learning could be improved if students are motivated by means of appropriate actions. This hypothesis is consistent with what is stated in the literature: motivation is important for learning [2]. Likewise, emotions have been identified as important players in motivation; hence several authors have proposed to use the affective state of the student to give him a more suitable response that fits with his affective and cognitive state [3-6]. Accordingly, we want to improve learning within our virtual laboratory by means of a more personalized environment through recognizing the students’ affective state with the aim of reacting appropriately from a pedagogical and affective point of view. We propose an affective behavior model for an intelligent tutoring system, which combines the affective and cognitive state of the student to establish affective and pedagogical actions. The affective behavior model integrates an affective student model based on the OCC cognitive model of emotions [7] and relies on a probabilistic network. In the construction of the affective student model we use personality questionnaires based on the five factor model [8]. To select the tutorial actions, we propose the use of a dynamic decision network with two utilities measure on both learning and affect. By using the decision network, the tutor selects the best pedagogical and affective response considering the current state of the student. We have refined our model by means of questionnaires presented to university teachers. In the questionnaires we presented several scenarios of tutoring and we asked the teachers to select the appropriate pedagogical action for each scenario. The rest of the paper is organized as follows. First, we summarize related work in affective tutors. Then we describe the proposed affective model. Next, we present preliminary results in the incorporation of the affective behavior model to a tutor for mobile robotics. Finally, we discuss our ongoing and future work. RELATED WORK The affective state has been recognized as an important component in learning [2]; consequently, several authors have proposed to incorporate the affective state in the user model, to give a response according to the affective and cognitive state. One of first steps towards this end, has been the OCC Cognitive Model of Emotions [7], proposed by Ortony, Clore and Collins, with the aim to give artificial intelligence programs the capability to reason about humans emotions. This model considers that emotions arise as a result of a cognitive appraisal between goal and the situation. The OCC model has been used in artificial programs for education. One of the most detailed affect models is presented in [3, 9]. This affect model has two ways to establish the affect; on one hand, in a predictive way, they considered that emotions arise as OCC stated; and on the other hand, in a diagnostic way, they consider that emotions have an impact in biological signs and facial expressions. This model is developed as a probabilistic model and has been applied to an educational game for learning number factorization. The OCC model has been also used to generate emotions. For example, in [10] a robotic character with the ability to display facial expressions denoting an affective state is described.