Factors Predicting Retention in Computer Science MOOC: A Survival Analysis of Pre-Computational Thinking and Auto-Feedback

• In Event: 31.060 - Perspectives on Educational Technologies

Abstract

we applied survival analysis on 20,134 students' characteristics, activities, and performance in the CS50x: Introduction-to-CS from the HarvardX. We have discovered a) predictors of dropout whose effects did not diminish over the course milestones. For example, students who made multiple trial-and-error were less likely to dropout – suggesting that the auto-feedback features to be useful for student retention. We also discovered b) predictors whose effects were important in the beginning, but diminished in importance over the course. One of which was pre-computational thinking skill, suggesting that a mismatch between prior intuition and the CS framework may pose an initial, yet temporary, hurdle to participation. Policy and pedagogical implications will be discussed.

Authors

• Gerhard Sonnert, Harvard University

• Chen Chen, Harvard University

• Philip M. Sadler, Harvard University