What does this research mean for the field?

Wearable stress-detection models trained to distinguish rest from stress frequently misclassify exercise-induced physiological activation as psychological stress. Novelty: ClaimNovelty.CONTRADICTORY. Consensus alignment: ConsensusAlignment.CHALLENGES_CONSENSUS.

What question did this study set out to answer?

To determine whether wearable stress-detection models are genuinely specific to psychological stress or are confounded by broader physiological activation from physical exercise.

September 10, 2026Medical Engineering & PhysicsOpen Access

Stress or Arousal? Exercise Confounding in Wearable Stress Detection

View Full Paper

Ask AI

Bookmark

Authors

YAYiğit Aydoğan FPFederico Villagra Povina

Discussion

Member takes

Implication

Validation study demonstrates that wearable stress-detection models misclassify exercise as mental stress in humans, highlighting widespread vulnerability to general physiological arousal.

Key Points

To determine whether wearable stress-detection models are genuinely specific to psychological stress or are confounded by broader physiological activation from physical exercise.
Re-analyzed public wearable data (PhysioNet, N=29 participants with matched aerobic and anaerobic sessions) comprising 522 rest, 192 stress, and 3,431 exercise windows, alongside secondary WESAD transfer analysis.
Trained four feature-based classifiers (including XGBoost) with nested leave-one-participant-out evaluation to distinguish rest from stress, strictly excluding exercise data from model training and threshold calibration.
The top-performing model (calibrated XGBoost) achieved a balanced accuracy of 0.703 (stress detection 0.604, rest specificity 0.801), yet misclassified 82.6% of held-out exercise windows as stress.
Participant-level exercise false-stress rate was 0.840 [95% CI: 0.775, 0.897] compared to a rest false-stress rate of 0.221 [95% CI: 0.151, 0.301], producing a paired difference of 0.619 [95% CI: 0.537, 0.699].
Self-reported stress was significantly higher in stress blocks than rest blocks (mean difference 1.22 [95% CI: 0.89, 1.53], one-sided Wilcoxon p=4.64×10⁻⁶), but adding time-domain heart-rate-variability features failed to resolve the exercise confounding.

Cite This Study

Aydoğan et al. (2026) studied this question.

synapsesocial.com/papers/6aa27aca58559d80afc738ac https://doi.org/10.1088/1873-4030/aea41e

View Full Paper

Ask AI

Bookmark

September 10, 2026Medical Engineering & PhysicsOpen Access

Stress or Arousal? Exercise Confounding in Wearable Stress Detection

View Full Paper

Ask AI

Bookmark

Authors

YAYiğit Aydoğan FPFederico Villagra Povina

Discussion

Member takes

Implication

Validation study demonstrates that wearable stress-detection models misclassify exercise as mental stress in humans, highlighting widespread vulnerability to general physiological arousal.

Key Points

To determine whether wearable stress-detection models are genuinely specific to psychological stress or are confounded by broader physiological activation from physical exercise.
Re-analyzed public wearable data (PhysioNet, N=29 participants with matched aerobic and anaerobic sessions) comprising 522 rest, 192 stress, and 3,431 exercise windows, alongside secondary WESAD transfer analysis.
Trained four feature-based classifiers (including XGBoost) with nested leave-one-participant-out evaluation to distinguish rest from stress, strictly excluding exercise data from model training and threshold calibration.
The top-performing model (calibrated XGBoost) achieved a balanced accuracy of 0.703 (stress detection 0.604, rest specificity 0.801), yet misclassified 82.6% of held-out exercise windows as stress.
Participant-level exercise false-stress rate was 0.840 [95% CI: 0.775, 0.897] compared to a rest false-stress rate of 0.221 [95% CI: 0.151, 0.301], producing a paired difference of 0.619 [95% CI: 0.537, 0.699].
Self-reported stress was significantly higher in stress blocks than rest blocks (mean difference 1.22 [95% CI: 0.89, 1.53], one-sided Wilcoxon p=4.64×10⁻⁶), but adding time-domain heart-rate-variability features failed to resolve the exercise confounding.

Cite This Study

Aydoğan et al. (2026) studied this question.

synapsesocial.com/papers/6aa27aca58559d80afc738ac https://doi.org/10.1088/1873-4030/aea41e

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark