Mining Large Open Online Learning Networks: Exploring Community Dynamics and Communities of Performance

Abstract

Online learning communities are becoming increasingly popular as they are known to support collaborative dialogue and knowledge building. Previous studies have typically focused on small, closed learning communities from an individual, static, and aggregated perspective. This research aims to advance our understanding of open and large online learning networks by exploring and characterizing community level dynamics and communities of performance. To achieve this goal, we mined a large open online learning network of over 30,000 students and approximately one million posts. First, we analyzed overall community network development by building temporal social networks. Subsequently, we studied sub-community dynamics using community detection algorithms, and, following that, investigated the interaction between community dynamics and communities of performance using best colleague correlation and Kruskal-Wallis test. Results found that large open online learning communities begin with a very large network having numerous small sub-communities. These communities consist of students who are similar in performance with strong links. The overall network size gradually shrinks, as does the number of sub-communities, and these communities evolve over time for their membership formulation with students who are more different in performance with weaker ties. Theoretical, practical, and methodological implications are then discussed. This study pushed the online learning community research to examine large and open networks by taking a more community-based and dynamic view of investigation.