Generative AI and the Redistribution of Epistemic Power - From Institutional Dependence to Individual Truth-Testing Capacity

Abstract

Generative artificial intelligence is commonly discussed as a technology for text production, information retrieval, automation, and decision support. This article proposes a broader interpretation: generative AI may contribute to a redistribution of epistemic power. The ability to examine complex claims, compare evidence, formulate counterarguments, detect inconsistencies, and articulate publicly defensible judgments has historically depended strongly on education, professional expertise, institutional resources, publishing access, and social networks. Generative AI can lower some of these barriers by making sophisticated analytical and linguistic assistance available to individual users. The article introduces the concept of AI-enabled epistemic empowerment: the expansion of an individual's capacity to examine, challenge, substantiate, and publicly articulate truth claims through structured interaction with generative AI. AI is not understood here as an authority on truth. The proposed model is based on AI-assisted truth-testing through a five-step procedure: Source → Counter-check → Level of evidence → Judgment → Separation of behavior from person The central thesis is that generative AI can reduce the epistemic transaction costs involved in transforming observations and questions into robust public arguments. In doing so, it may redistribute the capacity to question, verify, argue, and participate in public knowledge processes. The article also examines the corresponding risks, including confabulation, bias, epistemic dependency, homogenization, and new concentrations of infrastructural power. Its main proposition is: Generative AI may not redistribute truth itself. It may redistribute the power to test truth.