An Explainable AI Framework for Identifying Transferable Sustainable Urban Mobility Policies: Evidence from Istanbul

Abstract

Cities increasingly look to the experiences of other urban areas when designing policies to promote sustainable mobility. However, the transfer of transport policies across cities is rarely straightforward: interventions that perform well in one context may produce substantially different outcomes elsewhere because of differences in urban form, population density, socioeconomic conditions, transport systems, travel behaviour and institutional characteristics. Conventional benchmarking approaches often identify high-performing cities but provide limited guidance on whether their policies are transferable to a specific target context. This study proposes an explainable artificial intelligence (XAI) framework for identifying sustainable urban mobility policies that are potentially transferable to Istanbul, Türkiye. The framework combines cross-city benchmarking, machine-learning-based identification of comparable urban contexts, and explainable modelling of the relationship between city characteristics, policy interventions and observed mobility outcomes. Istanbul is positioned as the target city, while a set of international cities provides the empirical basis for identifying potentially transferable policy interventions. Rather than ranking cities according to overall performance, the proposed approach evaluates the contextual similarity between cities and investigates whether policies associated with positive mobility outcomes under particular urban conditions are likely to be relevant to Istanbul. Explainable AI techniques are used to identify the characteristics that contribute most strongly to policy-transferability predictions, thereby providing transparency about why a particular intervention is considered more or less suitable for the Istanbul context. The resulting framework provides a systematic approach for moving from descriptive international benchmarking towards context-sensitive policy learning and decision support. The study contributes to the sustainable urban mobility literature by demonstrating how explainable AI can support the identification and assessment of policy transferability while recognising the contextual dependence of urban transport interventions. The findings are intended to provide evidence-based guidance for Istanbul's sustainable mobility transition and a methodological framework that can subsequently be applied to other cities seeking to learn from international policy experience.