Digital phenotyping with large language models to detect depressive state changes in patients

Yunhao Yuan, Ya Gao, Hans Moen, Erkki Isometsä, Pekka Marttinen, Talayeh Aledavood · 2026

Read Digital phenotyping with large language models to detect depressive state changes in patients by Yunhao Yuan, Ya Gao, Hans Moen, Erkki Isometsä, Pekka Marttinen, Talayeh Aledavood (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13631235.txt.

Description

Yunhao Yuan and five colleagues test whether large language models can spot worsening depression from phone data. Phones leave a passive trail of movement, sleep, and usage patterns. Such digital phenotyping collects that trail nonstop. The stream is messy and mixed in format, though. That strains standard prediction models. The study appeared in 2026.

The team worked with patients in major depressive episodes. They compared two ways of using the models. One prompts them with a few examples. The other fine-tunes them on the data. Both beat classic baselines at spotting changes in depression severity. Two fine-tuning styles performed much alike overall. One was stronger on single features. The other was stronger on combined inputs. The authors back the approach with caution. They call for clinical checking and ethical safeguards.

Catalog record

CREATORYunhao Yuan, Ya Gao, Hans Moen, Erkki Isometsä, Pekka Marttinen, Talayeh Aledavood
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE82107
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13631235

Browse the full library · About the Tower of Babel