Extended Control-Value Emotional Agent Based on Fuzzy Logic Approach H.Moga, F.Sandu , G.M.Danciu, R.Boboc Faculty of Electrical Engineering and Computer Science “Transilvania” University, Brasov, Romania moga.horatiu@unitbv.ro , sandu@unitbv.ro , danciugaby@gmail.com , razvan.boboc@unitbv.ro , I.Constantinescu Faculty of Psychology “Spiru Haret” University Brasov, Romania ioanaconstantinescu71@gmail.com Abstract - The goal of our research was the implementation of a new emotional agent based on the control-value theory in psycho-pedagogy. This new type of emotional agent is included in the design of an affective tutoring system (ATS). The aim of the ATS platform is the adaption of an individualized teaching strategy considering the emotional state of the student user. One of the main problems in any kind of classroom refers to the inactive emotions. The inactive negative emotions categories are boredom, frustration and hopelessness. The majority of ATS platforms of the last two decades are based on the “OCC” (Ortony, Clore and Collins) emotional computational model. The paradigm of this generic theory brings some limitations if applied in the teaching field. In this paradigm, all the three inactive negative emotions have no approach. This is the rationale of our paper – to take into consideration a new approach of emotional agents that could be a good alternative not only in ATS but also in some other fields of human-machine interaction. Thus, we called our emotional agent “ECVEA” – Extended Control-Value Emotional Agen,t based on psychological control-value theory. Keywords - emotional agent; fuzzy logic; ITS; ATS; I. INTRODUCTION Our research aimed to the implementation of a new emotional agent based on the control-value theory in psycho-pedagogy. This new type of emotional agent is to be included in the design of an affective tutoring system (ATS) that should individualize the teaching strategy considering the emotional state of the student-user. One of the main problems in any kind of classroom are the inactive emotions – boredom, frustration and hopelessness. The ATS platform should be capable to identify this kind of emotions and to adapt the individualized learning path accordingly, aiming not only to increase educational efficiency but also to improve the emotional state of the student. The majority of ATS platforms of the last two decades are based on the Ortony, Clore and Collins (“OCC”) emotional computational model [1]. This generic theory has some limitations if applied in the teaching field, as in OCC paradigm the three inactive negative emotions have no approach. The goal of our paper is a new approach of emotional agents that could be a good alternative not only in ATS but also in some other fields of human-machine interaction like e-Commerce, emotional robots, virtual reality or computer games. We called this approach ”extended” because it comprises fourteen emotions. It is validated by a preliminary prototype that includes four emotions (joy, anger, boredom, frustration) [2]. Thus, we called our emotional agent “ECVEA” – Extended Control- Value Emotional Agent. II. MOTIVATION AND OBJECTIVES The objectives of our research were: - The general description of the ECVEA structure - The “action section” of ECVEA (as detailed below) based on fuzzy logic approach - The “retrospective outcome” section of ECVEA based on fuzzy logic approach - The “prospective outcome” section of ECVEA based on fuzzy logic approach The motivation of our research is that our ECVEA could be more competitive in the ATS field where the state-of-the-art includes implementations affected by the limits of emotions perception – e.g. for the Ekman’s platforms with video- camera based perceptions – and also by the limits of theoretical support for the OCC models and emotional sensors. This gives the opportunity of a new theoretical construction that might approach the teaching with new ways of emotions collection. III. SELECTIVE RELATED WORK In the intelligent emotional agents domain there were developed a couple of research branches. The most of the models are based on the Ortony, Clore and Collins paradigm inspired from psychology in 1988. This approach is the base for a lot of other theoretical concepts of emotional agents using artificial intelligence paradigms like neural networks (SHAME model [3]), or fuzzy cognitive maps [4], and fuzzy logic-like models of FLAME [5] and GEMA [6]. Other ways opened in this research field were frameworks like Cathexis [7, 8, 9], CLARION [10, 11], and CofAff [12, 13] with origins in psychology, ethnology and neurobiology. IV. THEORETICAL ASPECTS In this section there are presented introductions in psychological control-value theory [14] and item response theory elements, necessary in ECVEA definition and general structuring of ECVEA with all the three sections: action, retrospective outcome, prospective outcome.