From Wearables to Classrooms: A Person-Centered Feasibility Study of HRV-Based Physiological Monitoring for Learning Analytics

Abstract

Heart-rate variability is a promising index for learning analytics, but most prior work uses clinical-grade ECG ill-suited to naturalistic classrooms. We assess a consumer-grade Garmin watch with a custom iOS companion app (UeduPAD) during 3-hour university programming sessions, finding that person-centered HRV is feasible at the individual level but group-level aggregation can mask meaningful patterns.

Problem & Motivation

HRV is a promising physiological index for learning analytics, reflecting cognitive load, stress, and engagement. However, most education-side HRV work depends on clinical-grade ECG that is unsuitable for naturalistic classroom settings. Two open questions remain: can consumer-grade wearables (e.g., Garmin) deliver valid HRV measurement in real classrooms? And given the large between-person differences in autonomic baselines, will group-level analysis mask meaningful person-level patterns?

Method

In an undergraduate Python programming course we used the Garmin vívoactive 5 watch with a custom iOS companion (UeduPAD), streaming beat-to-beat intervals live via the Garmin Health SDK from five university students across two 3-hour sessions. We computed standard HRV indices (RMSSD, pNN50, lnHF, LF/HF) over 5-minute sliding windows, and validated them against Garmin's proprietary stress index at the individual level. Class events (calm baseline, lecture, programming lab, breaks) were manually time-stamped by the instructor for alignment with the physiological stream.

Findings

• Wearable validation: HRV indices showed moderate-to-good person-level agreement with Garmin's stress index (median ICC: RMSSD = 0.68, pNN50 = 0.74).
• Individual differences dominate: HRV responses to class events were highly heterogeneous; group-level Friedman tests were all non-significant — underscoring the need for person-centered analysis.
• Autonomic balance and learning: students who sustained lower sympathetic activation (LF/HF reactivity) during programming labs showed larger grade improvements (Spearman ρ = −1.0, n = 4).

Implications

Consumer-grade wearables are a practical tool for person-level classroom physiological monitoring — but researchers and instructors must adopt a person-centered analytic stance, since group-level aggregation can be misleading. The finding that autonomic balance during programming labs tracks learning gains points toward closed-loop interventions: when learner stress is over-elevated, the system can timely scaffold support, turning physiological sensing into actionable pedagogy.

Citation

@inproceedings{chang2026hrv_wearable,
  author    = {Chia-Kai Chang and Kuei-Hao Li and Cheng-Lin Cheng and Ting-Chuan Lin},
  title     = {From Wearables to Classrooms: A Person-Centered Feasibility Study of {HRV}-Based Physiological Monitoring for Learning Analytics},
  booktitle = {Proc. IEEE Swiss Conf. on Data Science and AI (SDS)},
  pages     = {186--189},
  year      = {2026},
  month     = may,
  doi       = {10.1109/SDS70563.2026.00035},
}