Abstract:As mental health issues for young adults present a pressing public health concern, daily digital mood monitoring for early detection has become an important prospect. An active research area, digital phenotyping, involves collecting and analysing data from personal digital devices such as smartphones (usage and sensors) and wearables to infer behaviours and mental health. Whilst this data is standardly analysed using statistical and machine learning approaches, the emergence of large language models (LLMs) offers a new approach to make sense of smartphone sensing data. Despite their effectiveness across various domains, LLMs remain relatively unexplored in digital mental health, particularly in integrating mobile sensor data. Our study aims to bridge this gap by employing LLMs to predict affect outcomes based on smartphone sensing data from university students. We demonstrate the efficacy of zero-shot and few-shot embedding LLMs in inferring general wellbeing. Our findings reveal that LLMs can make promising predictions of affect measures using solely smartphone sensing data. This research sheds light on the potential of LLMs for affective state prediction, emphasizing the intricate link between smartphone behavioral patterns and affective states. To our knowledge, this is the first work to leverage LLMs for affective state prediction and digital phenotyping tasks.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict emotional states from smartphone sensor data using large - language models (LLMs). Specifically, the research aims to explore how to use zero - sample and few - sample learning methods to predict the emotional states of college students by analyzing their smartphone behavior data. This involves several key aspects: 1. **Emotional State Prediction**: The paper focuses on how to use LLMs to predict an individual's emotional state, especially that of young people (such as college students). Emotional states include positive emotions (such as active, determined, focused, inspired, alert) and negative emotions (such as depressed, hostile, ashamed, nervous, afraid). 2. **Application of Digital Phenotyping Technology**: The paper explores the application of digital phenotyping technology in mental health monitoring. Digital phenotyping technology infers an individual's behavior patterns and mental health status by collecting and analyzing data from personal digital devices (such as smartphones and wearable devices). 3. **The Potential of LLMs in the Mental Health Field**: Although LLMs perform well in many fields, their application in the mental health field, especially in integrating mobile sensor data, is relatively scarce. This research attempts to fill this gap and demonstrate the potential of LLMs in mental health monitoring. 4. **Data - Driven Method**: The paper proposes a new data - driven method to predict emotional states through smartphone sensor data without the need for a large amount of labeled data. This is especially important in resource - limited situations. 5. **Multimodal Data Analysis**: The research also explores how to use multimodal or multi - sensor activity data to more comprehensively understand an individual's mental health state, which is an important direction in current digital phenotyping technology. In summary, this paper aims to develop an effective method to predict the emotional states of young people by using LLMs and smartphone sensor data, thereby providing support for the early detection and intervention of potential mental health problems.

Leveraging LLMs to Predict Affective States via Smartphone Sensor Features

Predicting Affective States from Screen Text Sentiment

Predicting states of elevated negative affect in adolescents from smartphone sensors: a novel personalized machine learning approach

Inferring Mood Instability Via Smartphone Sensing

Learning Language and Multimodal Privacy-Preserving Markers of Mood from Mobile Data

From Classification to Clinical Insights: Towards Analyzing and Reasoning About Mobile and Behavioral Health Data With Large Language Models

A Survey on Big Data-Driven Digital Phenotyping of Mental Health.

Personalized Mood Prediction from Patterns of Behavior Collected with Smartphones

Multimodal Privacy-preserving Mood Prediction from Mobile Data: A Preliminary Study

Efficient and Personalized Mobile Health Event Prediction via Small Language Models

Mental-LLM: Leveraging Large Language Models for Mental Health Prediction via Online Text Data

Machine learning for passive mental health symptom prediction: Generalization across different longitudinal mobile sensing studies.

Towards Deep Learning Models for Psychological State Prediction using Smartphone Data: Challenges and Opportunities

Enabling On-Device LLMs Personalization with Smartphone Sensing

Exploring Large-Scale Language Models to Evaluate EEG-Based Multimodal Data for Mental Health

Objective Prediction of Tomorrow's Affect Using Multi-Modal Physiological Data and Personal Chronicles: A Study of Monitoring College Student Well-being in 2020

MindShift: Leveraging Large Language Models for Mental-States-Based Problematic Smartphone Use Intervention

Integrating Wearable Sensor Data and Self-reported Diaries for Personalized Affect Forecasting

DeepMood: Modeling Mobile Phone Typing Dynamics for Mood Detection

Understanding Student Sentiment on Mental Health Support in Colleges Using Large Language Models