Dual-Mode Drift in Conversational AI: Scenario-Based Analysis with Full-Stack Psychological Mechanisms

This white paper documents a real interaction in which a conversational AI system provided accurate, neutral information in response to a factual sleep-timing question and later produced unsolicited grounding and behavior-oriented framing without a new request for guidance. The scenario is analyzed step-by-step and the full stack of psychological mechanisms activated by the tone shift is explicitly specified. The case demonstrates how multi-layer drift alters the decision environment and how immediate user detection restores informational boundaries and preserves autonomy.