Longitudinal Autonomic Methodology for Single-Subject Sustained-Cognitive-Work Research: A Multi-Modal Biosensing Approach

Methods paper documenting a 26-day longitudinal, multi-modal, single-subject (N=1) approach to studying the somatic and phenomenological signatures of sustained engagement with conversational AI systems. Biosignal data from a consumer-grade wearable platform (HRV, sleep architecture, stress index, body battery, heart rate, respiration, activity recordings) are time-aligned with timestamped cognitive-interaction logs (475 segmented sessions, 278 bio-matched). Includes 13 ISO-704-formalized phenomenological terms (mean Aggregate Quality 0.842), a reproducible Python pipeline architecture, and anonymized day-level biosignal aggregates. Reproducible at consumer-grade hardware cost. Key findings: (a) within-hour pre-engagement physiological state strongly predicts during-engagement state (Pearson r = 0.63, N = 273); (b) sustained engagement produces a tonic stress plateau rather than acute spike; (c) post-decline recovery shows a phenotype-specific 4-day median time-constant; (d) physical activity has a chronically positive but acutely costly dual effect on autonomic recovery. Bundle V17. Access restricted; methodology is open via the Methods Paper. Raw biosignal data are not released (GDPR Article 9).