|
Can passive signals from phones, wearables, and the environment reliably model behavioral health over months and years? My work builds the methods, audits, and foundation models needed to answer that — with a focus on reliability, fairness, and clinical translation. I sit at the intersection of biomedical engineering, computational psychiatry, and climate–health systems. The field is moving fast on capability; my bet is that what makes these systems actually deployable is the unglamorous infrastructure underneath — signal forensics, drift monitoring, fairness audits, and honest evaluation. |
location: Hanover, NH
program: Biomedical Engineering
institution: Dartmouth College
labs:
- AIM HIGH Lab (Geisel)
- Empower Lab (Thayer)
status: pretraining v1 · auditing
open_to: collaborations, PhD talks |
Reliability of Passive Sensing Data: A Multi-Method Evaluation in Depression PsyArXiv, 2026 Langener, Lee, Castro-Alvarez, Lampe, Enbar-Salo, Ramrakhiani, Olcan, Dorris, Price, Heinz, et al. How reliably do wearable and smartphone passive sensing actually capture signal in depression? A multi-method evaluation.
Smartphone Missingness as a Depression Biomarker medRxiv, 2026 Olcan, C. Reframes smartphone data missingness — usually treated as a nuisance — as itself a depression biomarker, via a baseline-controlled re-analysis of the StudentLife cohort.
Moderate-to-Severe Olfactory Dysfunction Marks Accelerated Phenotypic Aging in U.S. Adults Research Square, 2026 Olcan, C. Population-scale evidence that moderate-to-severe olfactory dysfunction tracks with accelerated phenotypic aging.
Full list on ORCID and Google Scholar.
In passive-sensing studies, missing data is treated as noise — something to impute, weight, or drop. This paper shows the opposite: in a re-analysis of the StudentLife cohort, sparse days carry depression signal that dense days don't. The mechanism is intuitive once you see it. Depressed individuals withdraw — fewer screen unlocks, fewer SMS, fewer app launches. The missingness pattern itself is the symptom, not a measurement failure.
|
Self-supervised pretraining over months of wearable, smartphone, and environmental streams. Built for transfer to downstream behavioral-health forecasting. |
Auditing toolkit for wearable physiological signals — quality, drift, demographic fairness. Methodological companion to the Reliability of Passive Sensing preprint. |
|
NHANES pipeline linking olfactory dysfunction to circadian rhythm fragmentation and activity reduction. Companion to the Olfactory Dysfunction & Phenotypic Aging preprint. |
Risk analytics and strategic systems modeling engine. |
Behavioral health AI should be
Hanover, NH · Dartmouth College