The Epistemic Authorship Crisis in the Age of Generative AI: Overcoming the Responsibility Gap through Hyper Justification Obligations

Abstract

The use of Large Language Models (LLMs) in scientific research is becoming increasingly widespread, but presents epistemic risks that are not yet fully understood. This article discusses how the probabilistic mechanisms of LLM can produce outputs that appear correct and justified but are actually dependent on epistemic luck, thus resembling the Gettier case pattern. Through a conceptual study approach, this research clarifies concepts, analytically reconstructs the generative structure of LLM, and conducts a normative analysis of its implications for scientific accountability and authorship. The results of the analysis show that Algorithmic Gettier Cases (AGCs) occur when linguistic coherence deceives users and creates the impression of justification, even though the truth that emerges is statistical coincidence and is not supported by valid causal relationships. This condition poses a serious challenge to the attribution of knowledge and author responsibility in the production of academic texts. To address this issue, this article proposes the principle of Hyper-Justification Obligation, which is the ethical obligation for researchers to actively verify and causally reason every AI output before using it in scientific works. This research provides a theoretical contribution to understanding the epistemic risks of LLM and offers an ethical foundation for academic practice in the era of generative AI.