Interactional regions in cities: making sense of flows across networked systems

Do administrative boundaries correspond to the observable ways in which people interact in urban space? As cities grow in complexity, and people interact over long distances with greater ease, so partitioning of cities needs to depart from conventional gravity models. The current state-of-the-art for uncovering interactional regions, i.e. regions reflective of observable human mobility and interaction patterns, is to apply community detection to networks constructed from vast amounts of human interactions, such as phone calls or flights. This approach is well suited for origin–destination activities, but not for activities involving multiple locations, such as police patrols, and is blind to spatial anomalies. As a result of the latter, community detection generates geographically coherent regions, which may appear plausible but give no insights into forces other than gravity that shape our interaction patterns. This paper …