Algorithmic Gatekeeping and the Reconfiguration of the Public Sphere: Generative AI, News Visibility, and Power in Digital Media Ecosystems

Abstract

This article examines how algorithmic gatekeeping in generative AI and news visibility systems reconfigures the public sphere by shaping whose knowledge circulates, what narratives are amplified, and which epistemic frameworks are legitimized. The article begins by distinguishing Habermas's consensus-oriented model of the public sphere from Chantal Mouffe's agonistic pluralism, arguing that the latter provides a more adequate framework for evaluating algorithmic systems precisely because it does not presuppose horizontal power or treat all claims as equally valid, but rather asks whether disagreement itself is institutionally enabled and made productive. Drawing on this distinction and on critical platform studies, the article employs a comparative governance methodology examining four AI systems — ChatGPT (OpenAI), Claude (Anthropic), Llama (Meta), and selected community-governed models — across three analytical dimensions: governance transparency, epistemic openness, and participatory legitimacy. The analysis further draws on an earlier ethnographic study of participatory media organizations in Cyprus (Author, 2026), while critically examining the limits of scaling community media principles to planetary-scale AI infrastructure. Findings demonstrate that governance architecture — rather than system scale or technical sophistication — systematically correlates with epistemic range. The article proposes that agonistic AI design requires transparency, substantive participatory governance, and institutional reflexivity, understood as interacting governance practices and design features rather than competing explanations. Theoretical contributions include a clarification of agonistic pluralism as distinct from epistemic relativism, and a sustained engagement with the analogy between community media governance and AI governance at scale.

Ähnliche Arbeiten