Detecting Perceived Unfair Treatment Among US College Students Using Mobile Sensing: Pilot Machine Learning Study

Area of Science:

• Digital Phenotyping
• Machine Learning in Mental Health
• Mobile Sensing Applications

Background:

• Experiences of unfair treatment are linked to negative health outcomes in college students.
• Current detection methods rely on self-reports, limiting timely intervention.
• No prior research has explored passive mobile sensing for detecting perceived unfair treatment (PUT).

Purpose of the Study:

• To investigate the feasibility of using mobile sensing for passive detection of daily PUT experiences.
• To develop and evaluate machine learning models for PUT detection.
• To establish a benchmark for future research in this area.

Main Methods:

• Collected data from 201 undergraduate students over two 10-week terms.
• Utilized ecological momentary assessment (EMA) for daily self-reported PUT.
• Implemented user-independent supervised classification and user-dependent anomaly detection models.

Main Results:

• User-dependent anomaly detection models, particularly LSTM-AE, showed superior performance in detecting PUT.
• Key behavioral patterns identified include changes in mobility, sleep, and screen time.
• LightGBM and Random Forest models outperformed baselines in user-independent classification.

Conclusions:

• Mobile sensing demonstrates potential for passive detection of PUT in college students.
• Identified behavioral patterns can inform the development of targeted interventions.
• Mobile technology offers opportunities for timely support to improve student well-being.