Student Modeling: Supporting Personalized Instruction, from Problem Solving to Exploratory Open Ended Activities

act intelligently is to endow them with the ability to understand the behaviors of other agents (human or artificial) in their environment. Intelligent tutoring systems (ITSs) are the ultimate example of this challenge: their goal is to provide instruction personalized to the specific needs of each learner, as good human tutors do. But understanding these needs can be extremely hard, because it entails modeling and capturing that complex ensemble of processes and states that constitutes human learning. These processes and states range from purely behavioral (for example, what actions a student takes to solve a problem), to cognitive (for example, domain knowledge, goals, cognitive load), to metacognitive (for example, domain-independent reasoning skills), to affective (for example, frustration, boredom, motivation). The more a tutor (human or artificial) understands about its learners at these different levels, the more personalized to the learners’ needs its instruction can be. Cognitive, metacognitive, and affective states, however, can be hard to assess accurately and unobtrusively from naturally occurring interaction events. Learner assessment is nontrivial even in its most basic incarnation, namely evaluating a learner’s understanding of a set of domain-dependent skills from ad hoc test items (for example, Desmarais [2011]). The assessment challenges increase with the complexity of the learner’s traits to be captured, because how a student behaves during an instructional activity generally provides partial and ambiguous information on the student’s underlying states, and the gap between what can be observed and what a learner actually thinks and feels increases as these states go from cognitive to metacognitive and affective. In ITSs, the research field concerned with addressing these challenges is known as student modeling, and a student model is the ITS component in charge of assessing student traits and states relevant to tailor the tutorial interaction to specific student needs. Articles