Online Toxicity and Community Structure on Bluesky
SICSS-Saarbrücken group project — stance, toxicity, and author networks
A group project at the Summer Institute in Computational Social Science, Saarbrücken (2025). We studied about 156,000 Bluesky posts around a high-profile US political event, to see where toxic users sit in the conversation network.
I led the stance classification and the author-network analysis. We combined zero-shot stance classification, toxicity scores from the Perspective API, and community detection on a network of about 46,000 authors. The project is exploratory and was completed within the summer institute.
Methods: Python, zero-shot classification, Perspective API, network analysis.
Slides: link to Drive