TikTok and the algorithmic transformation of social media publics: From social networks to social interest clusters

Summary

Gerbaudo argues that TikTok inaugurates a “second generation” of social media that breaks with the networked logic of Facebook, Twitter, and Instagram. Where first-generation platforms organized “networked publics” around explicit interpersonal ties — friending, following, liking — TikTok produces “clustered publics”: algorithmically inferred neighborhoods of users grouped by behavioral similarity, primarily through implicit signals like watch time. The meaning of “social” itself shifts from interpersonal connection to statistical commonality of interest. Drawing on Simmel’s notion of social forms and Weberian ideal-types, the paper theorizes this as a morphological transformation of online publicity with three normative consequences: depersonalization, opacity, and subcultural fragmentation.

Key Contributions

• Introduces “clustered publics” as a conceptual counterpart to boyd’s “networked publics,” giving sociology a vocabulary for algorithmically curated platforms.
• Proposes a periodization of social media into a first generation (~2004–2014, network-centric) and a second generation (~2015–2024, cluster-centric).
• Builds an ideal-typical comparative framework contrasting networks and clusters as distinct social forms across platform logic, signals, focus, and visibility.
• Bridges technical literature on recommender systems (collaborative filtering, embeddings, neighborhood methods) with sociological theory of publics.
• Reframes filter-bubble debates by identifying depersonalization, opacity, and fragmentation as distinctive harms of cluster-based curation.
• Positions the concept relative to adjacent terms such as “refracted publics,” “imitation publics,” “algorithmic audiencing,” and “calculated publics.”

Methods

Conceptual and theoretical analysis. Gerbaudo constructs an ideal-typical typology (presented as a comparative table) contrasting networked and clustered publics, synthesizes literature on platform affordances and recommender systems, and analyzes TikTok’s published documentation, leaked materials, and technical descriptions of ByteDance’s “Monolith” recommendation algorithm. The argument is anchored in social theory (Simmel, Weber, Tarde, Habermas) and platform studies (boyd, Gillespie, Bucher, Van Dijck).

Findings

• Networked publics are people-centric, visible, and explicit; clustered publics are item-centric, opaque, and implicit.
• TikTok’s “For You” pipeline (signals → predictions → ranking) leans heavily on watch time as a behavioral proxy for interest.
• Interface design — default For You feed, full-screen autoplay, swipe navigation, endless stream — maximizes immersion and tightens the algorithmic feedback loop while minimizing deliberate choice.
• Follower count poorly predicts reach on TikTok, marking the declining role of interpersonal networks in visibility.
• Survey-style evidence suggests TikTok is used less for maintaining personal ties than other platforms, supporting the depersonalization claim.
• Clustering generates niche subcultures (BookTok, CottageCore, WitchTok) but as “silosociality”: users are assigned to communities rather than opting in, unlike subreddit-style affiliation.

Connections

This piece sits in productive tension with work that examines how creators and users actually navigate TikTok’s clustered architecture, such as Hollingshead2026-vx and Cabbuag2024-me, both of which can be read as empirical complements to Gerbaudo’s macro-theoretical claim about the shift from networked to algorithmically clustered publicity. Gerbaudo’s framework provides the structural backdrop against which such ethnographic and practice-oriented studies of TikTok culture can be situated.

Podcast

A research-radio episode discusses this paper: Listen