From Predictive Models to Instructional Policies

At their core, Intelligent Tutoring Systems consist of a student model and a policy. The student model captures the state of the student and the policy uses the student model to individualize instruction. Policies require different properties from the student model. For example, a mastery threshold policy requires the student model to have a way to quantify whether the student has mastered a skill. A large amount of work has been done on building student models that can predict student performance on the next question. In this paper, we leverage this prior work with a new whento-stop policy that is compatible with any such predictive student model. Our results suggest that, when employed as part of our new predictive similarity policy, student models with similar predictive accuracies can suggest that substantially different amounts of practice are necessary. This suggests that predictive accuracy may not be a sufficient metric by itself when choosing which student model to use in intelligent tutoring systems.