This mixed method study used social network analysis (SNA) and inductive qualitative analysis to compare social relationships and the underlying interactions they represent in discussions related and unrelated to the learning of course content in a statistics MOOC. It additionally examined the impact of how social relationships are conceptualized (via network tie definition) on resultant network structures and properties. Using a previously developed natural language classifier, 817 threads containing 3124 discussion posts from 567 forum participants were characterized as either related to the course content or not. Content, non-content, and overall interaction networks were constructed based on five different tie definitions: Direct Reply, Star, Direct Reply + Star, Limited Copresence, and Total Copresence. Results showed network properties were robust to differences in tie definition with the notable exception of Total Copresence. Comparison of content and non-content networks showed key differences at the network, community, and node (individual) levels. The two networks consisted of largely different people, and participants in the content network and communities had more repeated interactions with a larger number of peers. Analysis of the contributing threads helped to explain factors leading to some of these differences, showing the content discussions to be more diverse and complex in their communication purposes, conversation structures, and participants’ interaction techniques. Within content discussions, the network of learners surrounding each of the two instructors showed distinct characteristics that appeared related to the instructor’s facilitation approach. Finally, a group of learners tightly connected to each other through content discussions showed nascent learning community-like characteristics. This work contributes to the literature by (1) deepening understanding of MOOC discussion learning processes; (2) drawing connections between network structures and specific discussion practices; (3) providing evidence demonstrating the importance of separately examining content and non-content discussions; and (4) drawing attention to the empirical impact of the choice of tie definition in SNA studies of MOOC forums.
Massive open online courses; Social network analysis; Discussion forum; Network partitioning; Tie extraction