Smartphone monitoring of smiling as a behavioral proxy of well-being in everyday life

2025-12-10

AI Summary

Subjective well-being (SWB) assessment has long suffered from recall bias and high participant burden inherent in self-report methods, resulting in poor ecological validity for daily affect research. To address this, we propose a novel objective measurement paradigm leveraging smartphone-based passive capture of naturalistic facial interaction videos. We develop a deep learning–driven model to quantify smile intensity and integrate circadian rhythm modeling with multivariate regression analysis. In a large-scale real-world study, we demonstrate—for the first time—that individuals’ mean daily smile intensity strongly correlates with national SWB survey data (r = 0.92), and its diurnal pattern aligns closely with conventional assessments (r = 0.80). Critically, this association remains significant after controlling for smartphone usage duration and is independently linked to physical activity and light exposure. The approach is unobtrusive, scalable, and ecologically valid, establishing a robust technical pathway and empirical foundation for objective, continuous measurement of positive affect.

Technology Category

• Cognitive Modeling & Cognitive Systems: Affective Computing
• Humans and AI: Interaction Techniques and Devices
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements

Abstract

Subjective well-being is a cornerstone of individual and societal health, yet its scientific measurement has traditionally relied on self-report methods prone to recall bias and high participant burden. This has left a gap in our understanding of well-being as it is expressed in everyday life. We hypothesized that candid smiles captured during natural smartphone interactions could serve as a scalable, objective behavioral correlate of positive affect. To test this, we analyzed 405,448 video clips passively recorded from 233 consented participants over one week. Using a deep learning model to quantify smile intensity, we identified distinct diurnal and daily patterns. Daily patterns of smile intensity across the week showed strong correlation with national survey data on happiness (r=0.92), and diurnal rhythms documented close correspondence with established results from the day reconstruction method (r=0.80). Higher daily mean smile intensity was significantly associated with more physical activity (Beta coefficient = 0.043, 95% CI [0.001, 0.085]) and greater light exposure (Beta coefficient = 0.038, [0.013, 0.063]), whereas no significant effects were found for smartphone use. These findings suggest that passive smartphone sensing could serve as a powerful, ecologically valid methodology for studying the dynamics of affective behavior and open the door to understanding this behavior at a population scale.

Problem

• Measuring well-being objectively through smartphone-captured smiles
• Correlating smile intensity with happiness and daily activity patterns
• Developing passive sensing as an ecological method for affective behavior

Innovation

• Passive smartphone video recording captures natural smiles
• Deep learning model quantifies smile intensity patterns
• Correlates smile data with well-being and activity metrics