Narrative Persistence Bias in LLM Dialogue: A Micro-Case Study of False Episodic Construction and Repair

This project introduces the concept of Narrative Persistence Bias (NPB) in large language model dialogue. It analyzes False Episodic Construction (FEC), where fabricated autobiographical narratives persist temporarily after explicit negation. We propose the Narrative Inertia Coefficient (NIC), defined as: NIC = P(N_active | Negation) A structured experimental protocol and statistical design are provided to operationalize episodic persistence and repair latency in conversational systems. Version 0.1 – Conceptual and protocol framework (pre-empirical stage).