Adding a Demographic Lens to Cluster Analysis of Participants in Entry-level Massive Open Online Courses (MOOCs)

This paper provides insight into how underrepresented learners are engaging with entry-level MOOCs, leveraging cluster analysis [19] on previously unexamined data from entry-level MOOCs produced by a major research university in the United States (USA) [29]. From an initial sample of more than 260,000 learners, a subset of data from more than 29,000 participants who submitted an assignment for a grade in one of nine entry-level MOOCs is analysed. Manhattan Distance [37, 27] and Gower Distance [5, 11] measures are computed based on engagement and achievement data, and clusters are derived from application of the CLARA and PAM algorithms [35]. The clusters are enriched by demographic data, with a particular focus on education level, as well as by approximated socioeconomic status derived from the American Community Survey. Results indicate that, in the courses analysed, learners without a college degree are more likely to be high-performing compared to college-educated learners; additionally, learners from approximated lower socioeconomic backgrounds are no less likely to be successful than learners from approximated middle and higher socioeconomic backgrounds in the USA. These findings are noteworthy insofar as they indicate that, while MOOCs have struggled to improve access to learning broadly speaking, there are possibilities for more inclusive outcomes.