Explainable AI Analyzes Paris's 15-Minute City Concept

A study investigated the "15-minute city" concept in the Paris metropolitan area, which advocates for accessible everyday services within a short walk or bike ride. Researchers utilized mobility data from the NetMob 2025 Data Challenge, enriched with sociodemographic information and OpenStreetMap points of interest (POIs), to analyze approximately 70,000 trip segments. The analysis constructed indicators for local service availability based on walking and cycling distances, examining their correlation with trip duration, transport mode, and short-trip car use. Findings showed that greater POI availability was associated with reduced private motorized travel and increased active mobility, though this relationship was weaker in the outer agglomeration. Gradient-boosted tree models, interpreted with explainable machine learning methods, consistently identified trip purpose, home-work distance, local service availability, vehicle ownership, public transport subscriptions, and sociodemographic context as key predictors of mobility. For short trips, high POI density correlated with lower car use, while car ownership increased predicted car use. The study highlights how XAI can complement traditional accessibility indicators to inform urban mobility policy, revealing spatial and demographic heterogeneities consistent with the 15-minute city's core assumptions.