Investigating the effect of Automated Feedback on learning behavior in MOOCs for programming

The challenge of learning programming in a MOOC is twofold: acquiring programming skills and learning online, independently. Automated testing and feedback systems, often offered in programming courses, may scaffold MOOC learners by providing immediate feedback and unlimited re-submissions of code assignments. However, research still lacks empirical evidence of their effect on learning behavior of MOOC learners, with diverse back-grounds and goals. Addressing this gap, we investigated the connections between the use of automated feedback system and learning behavior measures, relevant for MOOCs: engagement, persistence and performance. Further, two subjective measures of success are examined: sense of learning and intention fulfillment. In an experimental design, we analyzed data of active learners in a Python programming MOOC (N=4652), comparing an experimental group provided with automated feedback with a control group that did not. In examining the effect of automated feedback, prior knowledge of programming and Python was considered. Empirical evidence was found for the relation between automated feedback usage and a higher engagement and better performance, as well as higher attendance in "active watchers" and "high-performed completers" clusters, obtained by cluster analysis. Learners reports on their experience with the automated feedback system supported these findings. Regarding the subjective measures of success, however, no difference was found between groups. Our study and the offered future research may contribute to the considerations regarding the integration of automated feedback in MOOCs for programming.