Charismatic machines: On the epistemic power of generative AI within platform convergence

Abstract

This article develops a sociological theory of the epistemic power of generative AI within AI-centered platform convergence, as these systems are increasingly embedded in knowledge infrastructures. We define epistemic power as the capacity to structure collective perceptions of credibility and confer legitimacy in knowledge production. Drawing on Bourdieu and Weber, we introduce “charismatic machines”: AI systems that acquire authority not through actual understanding, but by convincingly performing it and leveraging their non-human status. Their charisma rests on a dual misrecognition, with AI perceived as both human-like and superhuman. However, this symbolic power is structurally unstable, coexisting with epistemic blame when manipulation, bias, or deception is attributed. To explain this ambivalence, we propose a sociotechnical circuit of epistemic attribution that spans models, interfaces, infrastructures, users, and social contexts. By redrawing boundaries between media, institutions, and algorithmic infrastructures, generative AI raises fundamental questions of governance, democracy, and epistemic inequalities in digital societies.