MAILA: AI predicts mental health from cursor and touchscreen data
MAILA, a novel machine-learning framework, can accurately deduce mental health conditions from typical interactions between humans and computers, achieving biomarker-level precision. This system has been developed using 18,200 recordings of cursor movements and touchscreen interactions from 9,500 individuals, along with 1.3 million self-reported mental health assessments. MAILA monitors fluctuating mental states across 13 key clinical dimensions, effectively addressing circadian variations and changes in arousal and valence. It reaches near-maximum accuracy at the group level and captures insights not fully conveyed through verbal self-reports. Additionally, MAILA enhances the capacity of large language models to assess users' mental health, offering a scalable approach to evaluating mental well-being through digital behavior analysis.
Key facts
- MAILA is a machine-learning framework for Inferring Latent mental states from digital Activity
- Trained on 18,200 cursor and touchscreen recordings
- Data from 9,500 participants
- 1.3 million mental-health self-reports used for labeling
- Tracks mental states along 13 clinically relevant dimensions
- Resolves circadian fluctuations and experimental manipulations of arousal and valence
- Achieves near-ceiling accuracy at the group level
- Improves large language models' ability to infer user mental health
Entities
Institutions
- arXiv