aiqbench analysis
This study applies maximum-entropy inverse reinforcement learning to raw keystroke timing, successfully deriving an interpretable biomarker for Parkinson's disease severity without relying on clinical labels during training. By correcting feature collinearity issues and validating across multiple datasets including smartphone tapping data, the research demonstrates robust reliability that surpasses traditional statistical methods like SHAP and LASSO. The approach offers a promising model-free framework for passive motor function monitoring using standard input devices.
Inverse Reinforcement Learning for Interpretable Keystroke Biomarkers in Parkinson's Disease
arXiv:2606.25270v3 Announce Type: replace Abstract: Keystroke dynamics offer a passive window into motor function, but existing work extracts aggregate typing statistics and trains classifiers for PD/control discrimination, foregoing interpretability and rarely reporting reliability. We instead apply maximum-entropy inverse reinforcement learning (IRL) to raw keystroke timing, recovering a per-subject speed-preference weight (w_speed) reflecting the implicit cost assigned to fast movement, without any clinical label during fitting. On the neuroQWERTY MIT-CSXPD dataset (85 subjects, 42 PD), we diagnose and correct a feature collinearity failure in an initial four-parameter decomposition, yielding an identifiable three-parameter model. The recovered w_speed correlates with UPDRS-III motor severity at r=-0.607 (95% CI [-0.770,-0.364], p<0.001, n=42), replicates across two independent sub-cohorts (r=-0.720, r=-0.588), and retains significant partial correlation after controlling for mean and SD of flight time (r=-0.371, p=0.016). It outperforms SHAP and LASSO on the same proxy features (r=+0.362 and r=+0.410) while additionally providing per-subject, interpretable output. A model-free AUC of 0.605 and LOO-CV AUC of 0.750 (95% CI [0.644,0.847]) confirm discriminative value. Test-retest reliability across clinic sessions yields ICC(2,1)=0.903 (95% CI [0.842,0.971]); no prior keystroke-PD study has reported formal reliability. Two other recovered weights (consistency, hand-alternation) did not survive confound checks, strengthening credibility of the surviving signal. Cross-modality external validation on independent mPower smartphone tapping data (n=200) recovers the same signal (r=-0.639, p=2.52e-24, OR=14.19), confirming convergent validity across modality, device, and country.
Abstract:Keystroke dynamics offer a passive window into motor function, but existing work extracts aggregate typing statistics and trains classifiers for PD/control discrimination, foregoing interpretability and rarely reporting reliability. We instead apply maximum-entropy inverse reinforcement learning (IRL) to raw keystroke timing, recovering a per-subject speed-preference weight (w_speed) reflecting the implicit cost assigned to fast movement, without any clinical label during fitting. On the neuroQWERTY MIT-CSXPD dataset (85 subjects, 42 PD), we diagnose and correct a feature collinearity failure in an initial four-parameter decomposition, yielding an identifiable three-parameter model. The recovered w_speed correlates with UPDRS-III motor severity at r=-0.607 (95% CI [-0.770,-0.364], p<0.001, n=42), replicates across two independent sub-cohorts (r=-0.720, r=-0.588), and retains significant partial correlation after controlling for mean and SD of flight time (r=-0.371, p=0.016). It outperforms SHAP and LASSO on the same proxy features (r=+0.362 and r=+0.410) while additionally providing per-subject, interpretable output. A model-free AUC of 0.605 and LOO-CV AUC of 0.750 (95% CI [0.644,0.847]) confirm discriminative value. Test-retest reliability across clinic sessions yields ICC(2,1)=0.903 (95% CI [0.842,0.971]); no prior keystroke-PD study has reported formal reliability. Two other recovered weights (consistency, hand-alternation) did not survive confound checks, strengthening credibility of the surviving signal. Cross-modality external validation on independent mPower smartphone tapping data (n=200) recovers the same signal (r=-0.639, p=2.52e-24, OR=14.19), confirming convergent validity across modality, device, and country.
Machine Learning (cs.LG)
arXiv:2606.25270 [cs.LG]
(or arXiv:2606.25270v3 [cs.LG] for this version)
Full text under the publisher's open licence · 231 words
Related coverage
Q-BIOMED Secures £902K to Expand Healthcare Quantum Sensing
Quantum Zeitgeist · May 31, 2026
BenSParX: A Robust Explainable Machine Learning Framework for Parkinson's Disease Detection from Bengali Conversational Speech
arXiv · Sep 16, 2026
In open RLVR, “improvement” depends on the instrument — a small GRPO testbed separating what training optimizes, measures, and teaches
LessWrong · Jun 15, 2026
Enhancing Personalized Bladder Cancer Treatment Through Reinforcement Learning: A Recurrent Patient State Transition Decision Support Framework
arXiv · Jul 21, 2026