E1: Observability and Calibration

Goal: determine whether a coarse in/out-of-R classifier is observable from wearables + context with calibrated uncertainty.

Design

N = 100-300, 4-8 weeks, wearables + sparse EMA + optional brief capacity probes, hierarchical model with per-user adaptation.

Outcomes

Detection performance (AUC, precision/recall), calibration (Brier score), personalization lift vs global model.

Falsifier triggers

No lift over trivial baselines (e.g., time-of-day only); chronic miscalibration; unacceptable false positive burden.

Tests

Premises F1-F3 and prediction P2, and kill criterion 6.1.

Secondary yield

If E1 demonstrates observability, the features that drive classification may be informative for IWMT's biophysical claims. If large-scale synchrony patterns predict intervention receivability, this would provide evidence relevant to IWMT. Not a design goal -- a potential secondary yield.