Deep Knowledge Tracing for Free-Form Student Code Progression

Knowledge Tracing, and its recent deep learning variants, have made substantial progress in modeling student knowledge acquisition through interactions with coursework. In this paper, we present a modification to Deep Knowledge Tracing to model student progress on coding assignments in large-scale computer science courses. The model takes advantage of the computer science education context by encoding students’ iterative attempts on the same problem and allowing free-form code input. We implement a workflow for collecting data from Jupyter Notebooks and suggest future research possibilities for real-time intervention.