Analyzing student learning trajectories in an introductory programming MOOC

Abstract

Understanding student learning and behavior in Massive Open Online Courses (MOOCs) can help us make online learning more beneficial for students. We investigate student learning trajectories on the individual problem level in an MITx MOOC teaching introductory programming in Python, considering simple features of the student and problem as well as more complex keyword occurrence trajectory features associated with student code submissions. Since code is so problem-specific, we develop gold standard solutions for comparison. Anecdotal observations on individual student trajectories reveal distinct behaviors which may correlate with prior experience level. We build models to correlate these trajectories with student characteristics and behaviors of interest, specifically prior experience level and video engagement. Generative modeling allows us to probe the space of submitted solutions and trajectories and explore these correlations.