Human Mobility Data Across Ten Cities Shows Pervasive Location-Specific Bias

In an analysis of GPS mobility data collected from anonymized smartphones across ten major cities in the United States, researchers found that data points can be more unequally distributed among individual users than wealth. The investigation revealed that data production is marked by pervasive demographic disparities that manifest differently depending on the specific urban area examined.

Large-scale human mobility records have come to play increasingly critical roles across modern algorithmic systems, commercial operations, and public policy decisions. Despite this widespread adoption, the researchers noted that little focus has been directed toward understanding bias and other fundamental shortcomings within these digital datasets, or determining how such deficiencies distort downstream computational analyses and predictive models.

To evaluate these systemic issues, the authors examined the concept of data production, measuring not only whether individual citizens are represented within big digital datasets, but also how heavily they are represented in terms of the total volume of data they generate. The research team evaluated this dynamic using an anonymized smartphone GPS tracking dataset gathered across ten major urban centers in the United States.

The investigators constructed statistical models designed to predict the expected volume of data points produced based on the demographic composition of residents living within individual census tracts. Through this predictive framework, the researchers identified strong effects of household wealth, ethnicity, and educational attainment on the volume of mobility data generated by tract populations.

The findings demonstrated that while data bias occurred ubiquitously across all ten examined metropolitan areas, each city exhibited its own distinct manifestation of the problem. Because the patterns of demographic skew differed from one location to another, the authors determined that uniform frameworks are insufficient and that location-specific models are necessary to capture bias accurately in each city.

The authors concluded that the localized nature of data production inequalities raises serious questions regarding the efficacy of generalized approaches intended to debias human mobility datasets. Consequently, the research team urged further research.

Subject of Research: Demographic bias and unequal data production in smartphone GPS human mobility datasets

Article Title: A tale of ten cities: data bias in human mobility is pervasive and highly location specific

Article References: den Nijs, K., Omodei, E., & Sekara, V. (2026). A tale of ten cities: data bias in human mobility is pervasive and highly location specific. EPJ Data Science, 15(1), Article 87. https://doi.org/10.1140/epjds/s13688-026-00685-8

Image Credits: AI Generated

DOI: 10.1140/epjds/s13688-026-00685-8

Keywords: human mobility, data bias, GPS mobility data, demographic disparities, data production, urban mobility, census tracts, algorithmic systems, tale, cities, data, bias