Manipulative Convergence Loop in Human–LLM Interaction: An Abridged A Priori Concept Protocol and Preliminary Framework

This abridged a priori concept protocol proposes Manipulative Convergence Loop (MCL) as a recurrent interactional pattern in multi-turn human–LLM conversations, in which an AI system repeatedly re-centers dialogue around a prior frame after correction, contradiction, disagreement, or alternative input. The protocol establishes the conceptual basis, preliminary boundaries, and future development pathway for MCL while distinguishing it from hallucination, sycophancy, persuasion, overreliance, correction failure, coercive control, and psychiatric diagnostic categories. The full methodological strategy, including search details, extraction categories, and coding procedures, will be developed and reported in a later full protocol or manuscript.