Everyday Smartphone Behaviors Predict Life Outcomes

Smartphone sensing provides an objective lens into contextualized human behavior, enabling ecologically valid perspectives on human lives. In this study, we evaluate the predictive power of smartphone-tracked behaviors relative to self-reported Big Five personality traits, the current gold standard for life outcome prediction. Using Random Forests to model data from a quota-representative sample of the German population (n = 752), we predicted 42 life outcomes spanning individual, interpersonal, and societal domains. Smartphone-based models outperformed the Big Five traits for 10% of outcomes and achieved comparable performance for 50%, with median Spearman correlations of up to r = .34 between predicted and actual outcome values. Behavioral data were particularly effective for outcomes closely tied to observable cues, including compulsive smartphone usage, substance abuse, and socioeconomic status. In contrast, subjective outcomes such as well-being were better predicted by personality traits. Feature importance analyses show differences in the number and importance of features across behavioral categories. While the number of features per category was determined by study design, patterns emerged in their importance: app usage and notifications include many features with relatively high importance; keyboard usage includes many with lower average importance; and screen and Wi-Fi categories include fewer features with relatively high importance. Our findings highlight that smartphone-tracked behaviors can complement traditional psychological measures in predicting life outcomes. They also suggest promising applications for behavioral interventions, while underscoring the need to clarify causal relationships and ensure ethical, responsible use.