Interest Networks (iNETs) for Cities: Cross-Platform Insights and Urban Behavior Explanations

Platform differences matter less at broader scales: The study found that while Google Places and Foursquare show different patterns when analyzing neighborhoods at fine granularity, these platform-specific behaviors converge when examining larger urban areas, suggesting that coarser-level urban analysis may be more reliable across data sources.

Geography and venue type drive user behavior more than socioeconomic factors: The research found that user interest in urban locations is "primarily shaped by geographic proximity and venue similarity," while socioeconomic and political contexts play a "lesser role"—an important finding for understanding what actually influences where people go in cities.

Different user types require different recommendation approaches: The system identifies distinct behavior profiles (such as "explorers" who prioritize proximity and "returners" who prefer familiar venues), demonstrating that one-size-fits-all urban recommendations may miss important individual differences in how people navigate cities.

Explainability is built into the recommendation system: Rather than using a "black box" approach, the researchers developed their model to provide "natural-language explanations using explainable AI (XAI) techniques," making it possible for users and planners to understand why specific urban areas are recommended—important for transparency in systems that influence urban mobility.

Open tools enable broader scrutiny and application: The release of h3-cities as a public tool and an interactive demo allows independent verification and real-world testing of the findings, supporting responsible development practices through transparency and accessibility.