Together with Oleksandra Poquet and Liubov Tupikina, we asked a simple but often overlooked question: Are forum networks really social networks? In Learning Analytics, it’s common to analyze forum interactions as if they reflected genuine social ties. But how much of the observed structure is simply driven by posting behavior — who posts, how much, and when — rather than social dynamics?
In our paper, we introduced a null model approach, leveraging methods from network science to statistically test whether forum network patterns exceed what would be expected from posting activity alone. Analyzing data from 20 online courses, we showed that key metrics often used to infer social learning — such as degree or frequency of interaction — are largely artefacts of individual posting behavior. However, other network features, such as clustering, cannot be explained by posting activity and may better capture emergent social dynamics.
This work calls for more methodological rigor in how we interpret digital traces of learning interactions — and invites researchers to question when and how learning networks can be considered genuinely “social.”
📌 Read the paper
Published at LAK 2020
Marc Santolini 