Detecting momentary reward and affect with real-time passive digital sensor data.

Area of Science:

• Digital Phenotyping
• Computational Psychiatry
• Behavioral Data Science

Background:

• Mental health monitoring often relies on subjective self-reports, which can be burdensome.
• Ecological Momentary Assessments (EMA) provide real-time subjective data but require active participant engagement.
• Passive sensing offers a potential for continuous, unobtrusive data collection.

Purpose of the Study:

• To investigate the predictive power of passive digital sensor data for subjective mental states.
• To assess the feasibility of using smartphone and smartwatch data for ecological momentary assessments (EMA).
• To establish a baseline for scalable, non-invasive mental health monitoring.

Main Methods:

• Collected data from 245 participants with depression and varying anhedonia levels.
• Utilized machine learning models to predict subjective states from sensor data.
• Aggregated behavioral and physiological features over time windows (15 minutes to 3 hours).

Main Results:

• Machine learning models successfully predicted 12 out of 15 EMA questions above random chance.
• Optimal prediction required aggregation periods of at least two hours of sensor data.
• Model performance varied based on sensor type, demographics, depression, and anhedonia severity.

Conclusions:

• Passive digital sensing is feasible for detecting momentary subjective states.
• This approach provides a foundation for scalable and non-invasive mental health monitoring.
• Further research can refine models and explore diverse populations.