Networks identify productive forum discussions

Discussion forums provide a channel for students to engage with peers and course material outside of class, accessible even to commuter and nontraditional populations. Forums can build classroom community and aid learning, but students do not always take up these tools. We use network analysis to compare three semesters of forum logs from an introductory calculus-based physics course. The networks show dense structures of collaboration that differ significantly between semesters, even though aggregate participation statistics remain steady. After characterizing network structure for each semester, we correlate students’ centrality—a numeric measure of network position—with final course grade. Finally, we use a backbone extraction procedure to clean up “noise” in the network and clarify centrality-grade correlations. We find that more central network positions are positively linked with course success in the two semesters with denser forum networks. Centrality is a more reliable indicator of grade than non-network measures such as postcount. Backbone extraction destroys these correlations, suggesting that the noise is in fact signal and further analysis of the discussion transcripts is required.

Figure 1

Structure of forum transcript data. The CN data largely shows a post with a single reply layer (left), in contrast to studies where more nested structure is retained and informs the network construction (right).

Figure 2

Bipartite network model for transforming forum transcript into a network object. Students are “actor” nodes, who post to thread or “event” nodes. The actor network projection links together student nodes who posted to the same thread.

Figure 3

Density distribution of forum activity (combined threads and comments) for class members by semester. The instructor’s contribution totals are included and are 94, 182, and 141 by semester.

Figure 4

Forum networks from weeks 7–8 in semester 1 (left) and semester 2 (right). Line opacity is scaled by edge weight, so darker lines indicate more threads in common for a student pair. Nodes are sized by total contributions over the semester and colored by grade (red low, yellow medium, blue high). Nodes without grades (withdrawals and instructor or CN staff accounts) are white, and the instructor’s node is labeled “I.”

Figure 5

Box plots of PageRank centrality for the semester 1 network backbones. The bottom, middle, and top line of the boxes show first quartile, median, and third quartile. The upper and lower “whiskers” extend to the maximum and minimum values or 1.5 times the interquartile range, whichever is larger. As alpha decreases, more node centralities cluster at the minimum value.