Urban mobility network centrality predicts social resilience

A new study published on arXiv bridges urban science and network theory, using AI to predict which parts of a city remain socially resilient during major disruptions. The research team, led by Lin Chen and including Fengli Xu and Esteban Moro, analyzed anonymized human mobility data from over 103 million residents across 15 major US cities. They examined three distinct types of urban shocks—likely including events like the COVID-19 pandemic and natural disasters—to see how visitation and social mixing at various venues changed.

The key technical finding is that a venue's 'eigenvector centrality' within a constructed urban mobility network is a powerful predictor of resilience. This network metric, which measures how well-connected a location is to other well-connected locations, boosted the model's explanatory power by more than 80% for predicting both changes in racial/economic segregation and visitation patterns, compared to using more obvious features like venue type or location alone. The data revealed a split response: while overall visitation and mixing declined, 36-53% of venues saw reduced segregation, and 21-39% saw increased visitation.

The researchers propose a 'well-and-pool' analogy, linking the findings to economic theory. 'Core' venues with high centrality act like wells, providing essential services and facilitating frequent, brief interactions that maintain broad social ties. 'Peripheral' venues act like pools, fostering deeper engagement within specialized social circles. This framework helps urban planners and policymakers identify critical infrastructure—not just hospitals and grocery stores, but highly connected social hubs—that must be supported during crises to maintain a city's social fabric. The methodology offers a data-driven tool for building more adaptive cities.