fg-zettelkasten

When do parties lie? Misinformation and radical-right populism across 26 countries

3 min read

When do parties lie? Misinformation and radical-right populism across 26 countries

Summary

This conceptual paper argues that the “social media” paradigm — organized around user-generated content circulating through social networks — is unraveling, and proposes “post-social media studies” as a replacement analytical framework. Törnberg and Rogers identify three interlocking dynamics driving the transformation: a shift from social-graph to interest-based algorithmic recommendation, the rise of generative AI as a producer of synthetic content, and a retreat from public platforms into private, bounded spaces. From these they derive a typology of three emergent formations — algorithmic broadcasting platforms, semi-private spheres and micro-communities, and AI-mediated communication — each demanding new vocabularies, methods, and normative frameworks. (Note: the provided title does not match the paper’s actual content, which is “Towards a Post-Social Media Studies.“)

Key Contributions

  • Proposes post-social media studies as a programmatic research agenda for digital communication.
  • Develops a typology of three emergent post-social formations: algorithmic broadcasting platforms, semi-private spheres, and AI-mediated communication.
  • Introduces and synthesizes concepts such as algorithmic publics, imitation publics, and the user-spectator to displace networked publics.
  • Maps methodological reorientations to each formation: flow analysis and rhythmanalysis for broadcasting, ethnographic and design-analytic methods for semi-private spheres, AI auditing for chatbot-mediated communication.
  • Frames the democratic stakes of communicative infrastructures that no longer reliably produce shared publics or stable texts.

Methods

Conceptual and theoretical synthesis rather than empirical study. The authors integrate literatures across platform studies, communication theory, digital sociology, and human–machine communication, and draw on secondary evidence — including Meta’s antitrust court filings (7% of Instagram and 17% of Facebook time spent on followed content), industry developments (TikTok’s For You Page, Meta and YouTube reorientations), and survey data on chatbot adoption — to motivate a comparative framework that distinguishes the three emergent formations and their distinct research challenges.

Findings

  • Major platforms have shifted from social-graph distribution to algorithmic recommendation organized around inferred interests and passive attention signals (watch time, dwell time, completion rate).
  • User participation on legacy public platforms has declined sharply — roughly a 50% drop in posting/viewing on Facebook and Twitter/X between 2020 and 2024.
  • AI chatbot use (52% of U.S. adults in 2025) now exceeds posting activity on major social media (35% Facebook, 20% Instagram, 11% Twitter/X).
  • Generative AI lets platforms populate feeds with synthetic content (e.g., the “Shrimp Jesus” phenomenon), reducing dependence on user-generated content.
  • Communicative practice is migrating to semi-private spheres (messaging groups, Substack, invite-only communities) organized around trust and reciprocity rather than visibility and scale.
  • AI chatbots constitute a media form without publics: communication becomes individualized, generative, and ephemeral, leaving no shared text or public trace.
  • Each formation generates distinct research obstacles: algorithmic opacity, access/ethics for private spaces, and the absence of shared texts for chatbot communication.

Connections

This paper sits alongside other work in the register theorizing the reconfiguration of platform power and the analytic challenge of post-API, post-feed environments — most directly Tornberg2026-lc by one of the same authors, and broader infrastructural and platform-studies pieces like Helmond2026-ll and Rieder2026-pp. Its argument about AI-mediated communication as a new media form connects to research on chatbot use, persuasion, and synthetic content such as Hackenburg2025-dj, DeVerna2025-dl, and Le-Mens2025-qz, while its concern with the erosion of shared epistemic infrastructure resonates with Lewandowsky2026-ob and Bak-Coleman2025-pm.

Podcast

A research-radio episode discusses this paper: Listen

Graph View

Backlinks