Epistemic Authority And Legitimacy In The Age Of Generative Ai: A Normative Framework For Evaluating Ai As A Knowledge-Formatter In Scientific Inquiry

Abstract

The growing integration of generative Artificial Intelligence (AI) into scientific inquiry raises foundational questions about epistemic authority and justified epistemic dependence. Classical accounts—ranging from Goldman’s reliabilism to Zagzebski’s preemption theory—treat epistemic authority as a normative relation between agents characterized by intentionality, reason-responsiveness, and accountability. Drawing on critiques by McMyler and Jäger and Shackel, this paper argues that such agential and relational conditions exclude AI systems from possessing epistemic authority in any classical sense. Yet AI increasingly shapes evidential practices, hypothesis formation, and interpretive processes within contemporary science. To evaluate this non-agential epistemic influence, the paper develops the Epistemic Authority Analysis Framework (EAAF), which distinguishes epistemic authority from epistemic legitimacy. The EAAF identifies three normative conditions—Epistemic Transparency, Delegative Trust, and Normative Reflexivity—under which reliance on AI can generate justified belief without conferring authority on the system itself. By framing AI as a “knowledge-formatter,” the paper demonstrates how epistemic responsibility can be preserved while integrating AI into scientific inquiry. The framework clarifies the epistemic status of AI and offers a principled foundation for human–machine epistemic collaboration in increasingly algorithmic epistemic environments.