Towards Detecting Wheel-Spinning: Future Failure in Mastery Learning

Wheel-spinning refers to a phenomenon in which a student has spent a considerable amount of time practicing a skill, yet displays little or no progress towards mastery. Wheel-spinning has been shown to be a common problem affecting a significant number of students in different tutoring systems and is negatively associated with learning. In this study, we construct a model of wheel-spinning, using generic features easily calculated from most tutoring systems. We show that for two different systems' data, the model generalizes to future students very well and can detect wheel-spinning in an early stage with high accuracy. We also refine the scope of the wheel-spinning problem in two systems using the model's predictions.