Capturing the participation and social dimensions of computer-supported collaborative learning through social network analysis: which method and measures matter?
Mohammed Saqr, Olga Viberg, Henriikka Vartiainen
International Journal of Computer-Supported Collaborative Learning
Network design
Nodes
Five hundred ninety-eight students across 12 iterations of four medical higher-education courses.
Ties
A student's reply to another student in a problem-based learning discussion, represented differently in the three network configurations.
Methods
Multigraph, simplified, and text-weighted network construction; five centrality measures; robustness and reproducibility comparisons; and analyses against participation and performance.
Source: Springer Nature
Reviewed summary
The study addressed a reproducibility problem in learning analytics: researchers can turn the same discussion data into different networks, and those choices may change what centrality appears to measure. The dataset contained 13,428 interactions from 598 students in 12 medical higher-education course iterations.
The authors compared multigraph, weighted, and simplified networks. Multigraph degree centralities were the most reliable indicators of participatory effort and consistent predictors of performance, while eigenvector centrality was the most consistent representation of the social dimension across configurations.
SNA
How SNA was used
Moodle reply data were reconstructed three ways: a multigraph retaining repeated ties and loops, a simplified graph removing them, and a weighted graph using posted text volume. In-degree, out-degree, closeness, betweenness, and eigenvector centrality were then compared for robustness and relation to learning measures.
Source
International Journal of Computer-Supported Collaborative Learning, 15, 227-248
DOI: 10.1007/s11412-020-09322-6
