Mechanism Analysis — Field Note 03: Baseline Drift
A Field Note is an observation. A Mechanism Analysis is an investigation. The purpose of this page is not to defend the idea, but to determine whether it deserves to exist as a distinct concept.
Verdict: D — Integrative condition; no distinct mechanism established
Baseline Drift provisionally names a cumulative, cross-context and partly unnoticed recalibration pattern. The label does not yet identify an independently demonstrated causal mechanism.
The observed pattern
After repeated exposure to highly fluent machine interaction, a person may experience unchanged human conversation as more effortful because the comparison standard for normal responsiveness, clarity and repair has shifted.
The proposed sequence
repeated synthetic fluency → normalization of low-friction interaction → reference distribution shifts → expectations and fluency cues tighten → ordinary human delay or repair produces contrast → human conversation feels more effortful → comparison noticed only after evaluation has changed
10 competing explanations tested
- Adaptation level and range-frequency normalization — strongest basic explanation; unresolved step is domain transfer.
- Sequential contrast — explains immediate downgrade; residual requires cumulative exposure and persistence.
- Expectation–disconfirmation — covers much of the proposed movement; residual is cross-agent transfer before explicit expectation is formed.
- Processing fluency and metacognitive effort — explains why effortful feels worse; residual is whether exposure changes the threshold.
- Conversational flow, silence and response timing — establishes pre-existing timing sensitivity; does not establish AI exposure changes how much it matters.
- Repair, grounding and intersubjectivity — supports repair as functional infrastructure; residual is reclassification of necessary repair as avoidable failure.
- AI-mediated communication — supports synthetic fluency as socially consequential; does not test later unassisted conversation.
- Human–AI feedback loops and cross-context transfer — strongest adjacent evidence for learning outside awareness; outcome is judgment bias, not conversational standards.
- Chatbot-to-human-interaction transfer — shows context-dependent substitution and complementarity; measures trust and intention, not changed conversational standards.
- Longitudinal chatbot exposure studies — important boundary evidence; current studies do not measure Baseline Drift directly.