Gerbaudo, P. (2026). TikTok and the algorithmic transformation of social media publics: From social networks to social interest clusters. New Media & Society, 28, 1019–1036. https://doi.org/10.1177/14614448241304106
Summary
Gerbaudo argues that TikTok inaugurates a “second generation” of social media (c. 2015–2024) that is qualitatively distinct from the first-generation social network sites like Facebook, Instagram, and Twitter (c. 2004–2014). Where earlier platforms organized what boyd termed networked publics around explicit interpersonal ties — friending, following, liking — TikTok produces what Gerbaudo calls clustered publics: statistically-constructed neighborhoods of users grouped by inferred similarity of interest and behavior, largely from implicit signals such as watch time. The essay theorizes this as a morphological transformation in the “social form” of online publics, drawing on Simmel and Weberian ideal-types, and warns of three consequences: depersonalization, opacity, and subcultural fragmentation of the public sphere.
Key Contributions
- Introduces clustered publics as a conceptual counterpart to networked publics, giving vocabulary to algorithmically-curated sociality.
- Proposes a periodization of social media into first- and second-generation forms tied to distinct logics of publicity.
- Offers a comparative ideal-typical framework contrasting networks and clusters as distinct social forms (platform logic, collective categories, focus, signals, visibility).
- Bridges technical recommender-system literature (collaborative filtering, embeddings, neighborhood methods, ByteDance’s “Monolith”) with sociological theory of the public sphere.
- Revives filter-bubble debates for the algorithmic-feed era, naming depersonalization, opacity, and fragmentation as normative concerns.
- Positions itself against adjacent concepts: refracted publics, imitation publics, algorithmic audiencing, and calculated publics.
Methods
Conceptual and theoretical analysis grounded in social theory (Simmel’s social forms, Weberian ideal-types, Habermas, Tarde) and platform studies. Gerbaudo builds an analytical typology contrasting networked and clustered publics along multiple dimensions (presented as a comparative table), synthesizes scholarship on affordances and algorithmic curation, and reads TikTok’s public documentation, leaked materials, and technical literature on its recommendation pipeline to unpack the mechanics of interest clustering.
Findings
- Networked publics are people-centric, visible, and driven by explicit signals; clustered publics are item-centric, opaque, and driven by implicit behavioral signals.
- TikTok’s “For You” feed operationalizes clustering through a signals→predictions→ranking pipeline in which watch time is a decisive implicit signal.
- Interface design (default For You feed, full-screen autoplay, swipe navigation, endless stream) maximizes immersion while minimizing explicit user choice, tightening the algorithmic feedback loop.
- Follower count is a weak predictor of reach, signaling the diminished role of interpersonal networks for visibility.
- Users report using TikTok less for staying in touch with friends, supporting the depersonalization thesis.
- Clustering favors niche subcultures (BookTok, CottageCore, WitchTok), producing a “silosociality” of automatic assignment rather than opt-in membership (contrast with Reddit’s subreddits).
Connections
This essay’s account of a shift from explicit networks to inferred, behavior-driven clusters resonates with work theorizing platform recommendation and algorithmic ordering of attention, such as Tornberg2026-lc and Brady2026-ln. Its concern with fragmentation and the health of the digital public sphere connects to broader platform-critique arguments in Bak-Coleman2026-mk and Lewandowsky2026-ob, while its periodization of platform generations and revisiting of foundational “networked publics” scholarship situates it alongside retrospective essays like Boyd2026-op and Baym2026-tr.
Podcast
A research-radio episode discusses this paper: 🎧 MP3 · Spotify · Apple Podcasts