Abstract:

Most MOOC platforms collect learner activity and demographics, and record them in datasets for their analysis. Cross-analysis of learner activity against demographic features such as age, profession, and country of origin can provide some insights on the diversity of learners and their behaviours. These analyses are often performed on individual courses, which not always provide comparable or generalisable results. Moreover, despite the massive numbers of learners, demographic data is often difficult to obtain, and represents a small proportion of the learning community. The analysis of aggregate data can help to enhance the significance generalisability of the results when combining activity and demographic data. However, aggregating data entails analytical challenges that can sometimes lead to misleading results. In this paper we analyse aggregate data from a set of MOOCs in a British university that uses the FutureLearn platform. Some of the results suggest that certain demographic features such as education level do not have a significant influence in engagement measures such as completion of courses. However, other demographic features such as age do have an influence. The paper then considers some of the caveats of aggregating data from different courses, and proposes solutions to overcome them.

Keywords:

MOOCs, learning analytics, data aggregation.