How 'Curation Bubbles' Make Social Media Echoes More Partisan Than They Seem

What Are Curation Bubbles?

Jon Green, Stefan McCabe, Sarah Shugars, Hanyu Chwe, Luke Horgan, Shuyang Cao, and David Lazer introduce the idea of "curation bubbles": networked patterns on social media in which users select particular others to follow or pay attention to, and those chosen accounts in turn share stories that reflect their identities and interests. The result is an audience that repeatedly encounters content with consistent partisan appeal, even when that content originates from a range of different sources.

Why This Matters for Echo Chambers and Filter Bubbles

Current debates about political segregation online often measure partisanship at the source or domain level—for example, labeling an entire news site as left or right. The authors argue this source-level lens can misrepresent how audiences actually experience information, because a single source can publish stories with varying partisan appeal. This mismatch matters for claims about polarization, echo chambers, and who is exposed to cross-cutting information.

What the Authors Argue and How They Reframe Measurement

The paper critiques domain-level measures of partisan valence and proposes aligning theory and measurement around the networked curation process and the story level. Green et al. show that treating partisanship as a property of individual stories—rather than only of their domain—better captures the selective sharing and consumption dynamics that produce partisan audience clustering, and that these dynamics can make ostensibly moderate sources appear more polarized in practice.

Implications for Research and Practice

• Measurement: Researchers should consider story-level measures of partisan appeal and incorporate network selection when estimating audiences' political leanings.
• Interpretation: Findings about filter bubbles or exposure to cross-cutting views may change once analysis moves from domains to individual stories and accounts.

The authors' intervention is primarily conceptual and methodological: it reframes how scholars should think about and measure online political information flows so that empirical work on echo chambers, polarization, and platform effects better reflects how users curate and receive content on social media.