Theory Audit — Field Note 03: Baseline Drift

A systematic audit of 74 papers across six research domains to determine whether Baseline Drift can be distinguished from existing explanations. The audit examines whether repeated interaction with low-friction AI systems gradually recalibrates what ordinary human communication feels like.

Research domains reviewed

• Folder A — Adaptation, Contrast, and Contextual Judgment (Papers 1–11): Adaptation-level and range-frequency theories; sequential contrast; prevalence-induced concept change.
• Folder B — Processing Fluency and Effort (Papers 1–12): Schwarz et al. (1991) through Templeton et al. (2022). Processing ease as information, fluency effects on affect, judgment and effort prediction; conversational flow and timing.
• Folder C — Conversational Repair and Interactional Trouble (Papers 1–13): Sacks et al. (1974) through Raymond & Gill (2025). Repair as organised infrastructure; turn-taking; grounding; intersubjectivity.
• Folder D — Human–AI Interaction and Synthetic Fluency (Papers 1–12): Nass & Moon (2000) through current research. Social responses to machines; smart-reply mediation; parasocial and relational effects.
• Folder E — Repeated Exposure and Cross-Context Transfer (Papers 1–13): Learning from AI feedback; cross-context substitution and complementarity; longitudinal chatbot studies.
• Folder F — Longitudinal AI-Exposure and Psychosocial Outcomes (Papers 1–13): Extended AI interaction and its effects on social behaviour, loneliness, language, and relational judgments.

Audit conclusion

A narrower residual remains where: repeated synthetic fluency becomes normalised; the comparison standard transfers across agent classes; ordinary human delay and repair are evaluated more negatively; and the change persists beyond immediate priming. Baseline Drift provisionally names that cumulative, cross-context and partly unnoticed configuration. The label does not yet identify an independently demonstrated causal mechanism. Current evidential status: D — Integrative condition; no distinct mechanism established. Confidence: Medium.