The Phenomenon of ‘Concept Drift’ in AI-Generated Content and Its Ontological Roots

Abstract

Against the background of the rapid development of large language models, the phenomenon of ``hallucination'' has become a central obstacle to AI credibility. This paper argues that AI hallucination is not merely a technical failure, but a systematic form of ``concept drift'' that occurs when AI processes human knowledge systems. From the perspective of the RID cognitive-dynamics model inKnowing and Saying, the stability of human concepts derives from anchoring in survival pressure (D) and embodied experience (I), whereas AI lacks an ontological ground. Its generated linguistic symbols (R), once detached from concrete contexts, are therefore highly prone to semantic slippage and alienation. The paper analyzes the internal mechanisms of concept drift and its destructive consequences in high-risk fields such as medicine and law. It argues that human-AI collaboration must uphold the regulatory principle that ``humans bear conceptual responsibility, while AI provides statistical reference,'' in order to reconstruct safe boundaries for human-machine cognition.