Modeling affect by mining students’ interactions within learning environments

In the past decade, research on affect-sensitive learning environments has emerged as an important area in artificial intelligence in education (AIEd) and intelligent tutoring systems (ITS) [1-6]. These systems aspire to enhance the effectiveness of computer-mediated tutorial interactions by dynamically adapting to individual learners’ affective and cognitive states [7] thereby emulating accomplished human tutors [7,8]. Such dynamic adaptation requires the implementation of an affective loop [9], consisting of (1) detection of the learner’s affective states, (2) selection of systems actions that are sensitive to a learner’s affective and cognitive states, and sometimes (3) synthesis of emotional expressions by animated pedagogical agents that simulate human tutors or peer learning companions [9,10].