Dataset for: Academic Social Media Migration Through a Psychology of Science Lens

The acquisition of Twitter by Elon Musk in 2022, and the changes to the platform that came with it, started the so-called #TwitterMigration: Many users called for leaving Twitter and using alternative platforms like Mastodon. Applying social impact theory, we hypothesize that academics on Twitter are more likely to migrate to Mastodon if they are under high social influence from #TwitterMigration influencers. We also hypothesize that researchers who endorse the open science movement are more likely to migrate to Mastodon than researchers who do not. We use an available dataset of around 500,000 researchers on Twitter, as well as lists of researcher accounts on Twitter and Mastodon. We identify additional researchers on Mastodon. For each researcher with posts related to #TwitterMigration, we compute an influence score. Open science advocates are identified using natural language processing. Based on account activity on both platforms, we identify users who have migrated to Mastodon.

Social media is an essential tool for scholarly communication, yet dissatisfaction with platforms can drive researchers to migrate elsewhere, as seen in the shift from Twitter to Mastodon or from X to Bluesky. Despite the growing importance of these migrations, the underlying social dynamics within academia remain poorly understood. Analyzing data from over 250,000 researchers with a computational psychology of science approach, we used a machine learning model informed by social impact theory and the concept of self-consistency to identify predictors of the 2022 academic Mastodon migration. Our findings reveal that the strongest predictor of platform migration was the number of followed peers discussing the move, whereas individual influential strength contributed relatively little to model predictions. Open-science endorsement was also strongly associated with migration (odds ratio ≈ 11). These findings suggest that behavior change interventions in science may be fostered by collective efforts tailored to group characteristics.