Date created: | Last Updated:

Category: Project

Description: Social interactions are essential for mental health. Therefore, researchers increasingly attempt to capture an individual's social environment to predict and explain changes in well-being. Digital phenotyping is an often used technology to assess a person's social behavior through passive sensing and without self-report. Additionally, the experience sampling method (ESM) can capture the subjective perception of specific interactions multiple times per day. Lastly, egocentric networks are often used to measure specific relationship characteristics. Although those different methods capture different aspects of the social environment that are related to well-being, they have rarely been combined in previous research. Combining those methods may be necessary to increase the predictive accuracy of well-being and thus its utility for clinical applications. In this study, we aim to investigate how accurately we can predict mood based on the social environment as measured through digital phenotyping, ESM, and ego-centric networks. We examine how much each of those three methods adds to the prediction, which allows us to identify the most important measurements for predicting health outcomes. We use data from a student sample collected over a 28-day period. We train individualized machine learning models and calculate feature importance scores. Overall, we investigate how feasible it is to predict mood, which might be useful for developing just-in-time interventions. Furthermore, identifying which parts of the social environment are most relevant, can help to deliver personalized interventions and to reduce the participant burden.