TL;DR
This paper explores the complex interplay between individual human values, technology use, and physical exercise. By analyzing a large dataset of US residents, the authors found that specific values, like prioritizing purity and de-emphasizing conformity, are linked to adopting health apps, and that these factors, along with demographics and online behavior, can predict exercise engagement.
What Problem Does This Paper Address?
In an era where sedentary lifestyles are a growing concern, there's a significant push to leverage technology, particularly mobile health and fitness apps, to encourage physical activity. However, understanding why people adopt these technologies and what factors truly drive exercise behavior remains a complex challenge. Previous research has touched upon psychological traits, demographics, and technology literacy, but the role of deeper human values – the fundamental guiding principles that shape our actions and beliefs – has been largely underexplored in this context. This paper aims to bridge this gap by investigating how a comprehensive set of human values, alongside demographic and digital usage data, influences both the adoption of health apps and actual physical exercise.
Methods
The study employs a multi-faceted approach combining survey data with digital trace data.
- Psychometric Measures: Participants completed questionnaires assessing:
- Moral Foundations Theory (MFQ): Measuring endorsement of five moral foundations (care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, purity/degradation).
- Schwartz Basic Human Values: Assessing ten basic values (self-direction, stimulation, hedonism, achievement, power, security, conformity, tradition, benevolence, universalism), clustered into higher-order values (Openness to change vs. Conservation, Self-enhancement vs. Self-transcendence).
- Emotional Contagion Scale (EC): Measuring susceptibility to "catching" emotions.
- Digital Trace Data:
- Desktop Browsing Data: Domain names, time spent online, and daily visits.
- Mobile App Usage Data: Application usage logs, including date/time stamps, time spent, and app categories (e.g., Google Play Store categories).
- Predictive Modeling:
- Logistic Regression: Used to identify the determinants of exercise, analyzing the contribution of various features.
- Random Forest Classifier: Employed to predict exercise engagement, iteratively adding different feature sets (demographics, values, health variables, web domain categories, app categories) to assess their predictive power.
Datasets
The study utilizes a unique, demographically representative dataset of 15,021 US residents. From this larger group:
- 5,008 participants provided desktop browsing data.
- 2,625 participants agreed to have their mobile app usage and browsing data captured.
This combined dataset, integrating psychometric, demographic, and digital trace data, appears to be a novel contribution for this type of analysis.
Key Insights and Ideas
- Values Drive App Adoption: It's not just about being tech-savvy; specific values predict the adoption of Health & Fitness apps. Users who prioritize "purity" and de-emphasize "conformity," "hedonism," and "security" are more likely to use these apps. This suggests that the why behind app use is deeply rooted in personal value systems.
- Wealth is a Strong Predictor of Exercise: Among demographic factors, wealth emerged as a particularly strong determinant of regular exercise, even more so than income. This highlights potential socioeconomic barriers to healthy lifestyles that go beyond mere financial resources.
- Belief in Choice Matters: An individual's belief that healthy habits are a conscious choice ("Health is a gift and there is not much I can do about it" - HQ 5) and that they have control over their lives ("We all have a choice about how to lead our lives, and healthy habits are just one example of that" - HQ 4) significantly correlates with exercise engagement. This points to the importance of self-efficacy and personal agency in behavior change.
- App Usage Outperforms Browsing for Prediction: When predicting exercise, mobile application usage proved to be a more powerful predictor than general internet browsing data. This underscores the specific relevance of dedicated health and fitness tools.
- Socioeconomic Status and Values are Key: The most robust models for predicting exercise combined demographic factors (specifically wealth), attitudinal beliefs (like the choice aspect of health), and technology use (Health & Fitness app usage).
Key Results
- Health App Adoption: Females are more likely to adopt Health & Fitness apps, with specific app types showing even stronger gender biases (e.g., weight loss apps are 168% more likely to be adopted by females).
- Exercise Prediction Accuracy:
- Basic demographics (gender, age) yielded an AUROC of .513.
- Adding richer demographics improved prediction to .616.
- Including moral values boosted it to .623.
- Health-related variables pushed it to .650.
- Web domain categories offered a slight improvement to .654.
- Application categories led to a notable increase to .671.
- The final model, incorporating app categories and average time spent on apps, achieved an AUROC of .673, statistically similar to models with just app categories.
- Determinants of Exercise (Logistic Regression): Significant positive predictors included education, wealth, and income, while age was a negative predictor. Among values, loyalty and stimulation were positive, while hedonism, power, security, and tradition were negative. Health variables like high blood pressure, chronic disease, and smoking were strong negative predictors.
Limitations and Future Work
- Self-Reported Exercise: The study acknowledges that reliance on self-reported exercise is a significant limitation, as it is prone to biases. Future work should aim to obtain direct user physical activity data.
- Cross-Sectional Data: The study captures only one month of technology use, making it non-longitudinal. This limits conclusions on application adoption and retention, as users might have used apps outside the observation window.
- Sample Size for Specific Apps: The analysis of specific app adoption might be limited by small eventual sample sizes per application.
- Future Directions: The authors suggest that future research should explore how to design persuasive technologies that consider individual values and beliefs, focus on specific activity tracking, and address socioeconomic barriers to healthy lifestyles. They also highlight the potential of using application usage data for predictive analytics when detailed demographic or value information is unavailable.
Summary updated