Abstract
Mapping individual differences in behavior is fundamental to personalized neuroscience. Here, we establish that statistical patterns of smartphone-based mobility features represent unique “footprints” that allow individual identification. Critically, mobility footprints exhibit varying levels of person-specific distinctiveness and are associated with individual differences in affective instability, circadian irregularity, and brain functional connectivity. Together, this work suggests that real-world mobility patterns may provide an individual-specific signature linking brain, behavior, and mood.
Introduction
In behavioral neuroscience, the "gold standard" for measuring behavior typically relies on controlled laboratory tasks or subjective self-reports. However, these methods often fail to capture the ecological complexity of daily life. "Digital phenotyping"—the use of smartphone sensors to passively track behavior—offers a promising alternative. A fundamental question remains: are these mobility patterns truly unique to the individual? Just as brain connectivity patterns (the "connectotype") can act as a neural fingerprint, this study investigates whether mobility patterns constitute a "mobile footprint." We explore whether these footprints are person-specific and if their "distinctiveness"—the degree to which a person's patterns are unique and predictable—relates to clinical outcomes like mood instability and sleep irregularity.
Methods
The researchers tracked 41 adolescents and young adults (mean age 23.4) over approximately three months using the "Beiwe" smartphone platform. Data collection focused on two primary streams:
- GPS Geolocation: Recorded in power-saving cycles (2 minutes on, 18 minutes off).
- Accelerometer Data: Measuring physical movement intensity.
To ensure privacy, raw coordinates were not used. Instead, 22 high-level summary statistics were extracted (15 GPS-based, 7 accelerometer-based), including metrics like time spent at home, radius of gyration, and "circadian routine" (the repeatability of daily paths).
The "Mobile Footprint" was operationalized as a covariance matrix of these features over time, capturing the unique relationships between different behaviors for each individual. To test for uniqueness, each participant's data was split into "reference" and "target" sets. Identification was considered successful if a target set was most similar to the corresponding reference set from the same individual.
Results
The study analyzed 3,317 person-days of geolocation data and 2,972 person-days of accelerometer data. The primary finding was that individuals could be identified with high accuracy based on their mobility patterns.
Identification Accuracy and Distinctiveness
The model achieved a mean identification accuracy of 63%, significantly exceeding the 3% chance level. While GPS data alone provided 55% accuracy and accelerometer data 28%, the combination of both modalities yielded the highest performance. Interestingly, "footprint distinctiveness" varied significantly across the cohort, with some individuals being nearly perfectly identifiable (99%) and others being much harder to distinguish (4%).
Clinical and Neural Correlates
Reduced footprint distinctiveness—representing more chaotic or less predictable mobility patterns—was a significant marker for instability. Specifically:
- Mood Instability: Lower distinctiveness correlated with higher levels of affective instability ($r = -0.37, p < 0.05$).
- Sleep Irregularity: Variability in sleep duration also correlated with reduced distinctiveness ($r = -0.36, p < 0.05$).
- Brain Connectivity: Multivariate patterns of resting-state fMRI connectivity, particularly within the somatomotor network, predicted an individual's footprint distinctiveness ($r = 0.29, p = 0.025$). This suggests that the neural circuits governing movement and sensory processing are linked to the "sharpness" of one's real-world mobility signature.
Discussion
These findings demonstrate that passive smartphone data can provide a stable, individual-specific signature of behavior. The link between "fuzzy" mobility footprints and mood/sleep instability suggests that digital biomarkers could serve as non-invasive tools for monitoring psychiatric health.
Limitations and Future Directions
The study's sample size (N=41) is relatively small and focused on a population with high affective instability, which may limit generalizability to broader populations. Furthermore, the reliability of the footprint is dependent on long-term data collection; shorter tracking periods significantly reduce identification accuracy. Future research should explore these patterns in larger, more diverse cohorts and investigate the potential for real-time clinical interventions based on changes in an individual's mobile footprint.
Summary updated