Drivers of cross-topic alignment in structurally inferred Bluesky communities

Determine whether the observed cross-topic issue alignment among user communities inferred via the stochastic block model on Bluesky repost networks is primarily driven by Bluesky’s algorithmic design (e.g., feed curation and visibility mechanisms) or by genuine differences in how subcommunities approach political topics.

Background

The paper compares content-based and structure-based stance groupings across multiple political topics on Bluesky and finds that normalized mutual information (issue alignment) is generally low for content-based groups but slightly higher for structurally inferred groups. The authors note that structural communities may capture clusters of users whose views are more consistently aligned across issues, reflecting broader subcommunities, but these may not necessarily be opposing camps.

This raises uncertainty about the underlying cause of the observed cross-topic alignment in structural groups: whether it arises from algorithmic factors that shape interaction patterns and exposure on Bluesky, or from organic differences in how subcommunities frame and engage with political issues. Resolving this would clarify the role of platform design versus user-driven dynamics in producing the observed structural patterns.