Abstract
In this dissertation, we propose a computer-supported method, called insightful learning systems, for tracking student progress that yields human actionable insights. We cover our theories, experiments and applications that aim to pave the way for more effective understanding of student progress, in key skills for a course, by looking at students answers to questions. Our goal was to find a sweet spot balancing the complexity of a learning model needed for the analytic techniques to obtain accurate results, with the desire to keep everything simple enough, so a human could understand what’s happening and ensure the manual work needed to create the system is minimal. To demonstrate this, we chose the domain of Introductory Calculus. We worked with an expert to build a skill-based model for that domain. We developed several methods for using Analytic frameworks to predict student outcomes and mastery from that model. We implemented a system, the CalcTutor, to allow real-world testing of our theories. We then collected data, both within the CalcTutor and from regular Calculus classes, that were used to show our analytics were accurate and provided predictive insights. This dissertation provides evidence that Insightful Learning Systems could fundamentally transform STEM education over the coming decade.