The geography of interest in international conversations on Twitter

Current discourse on online social networks provides a valuable window into the dynamics, opinions, and behaviors of real-world social networks. This study leverages geolocated Twitter data and location references in tweets to examine patterns of international and interstate attention, evaluated through the number of mentions across three regions: Europe, South America and the United States. We construct directed networks where source nodes represent geographical areas (countries or states) tweets originate from, target nodes represent areas of locations mentioned in tweet text, and edges are weighted by mention frequency, a measure we define as “interest”. Our analysis reveals that these networks are remarkably dense, with nearly all areas mentioning most others, although with significant asymmetries in attention distribution. Within each region, there are some states that consistently receive or generate disproportionate amounts of mentions. To explain these patterns, we develop augmented gravity models that incorporate economic, geographic, linguistic, and demographic factors. These models demonstrate that the interest is primarily shaped by the GDP of both source and target areas, the geographic distance between them, and migration flows. The effect of distance varies notably across regions, with European discourse showing substantially weaker geographic constraints compared to the Americas, suggesting that regional integration may reduce spatial friction in international and interstate attention. Through topic modeling, we further identify both common discourse domains (tourism, sports, culture) present across all regions and distinctive regional preoccupations that reflect specific historical and political contexts. These findings illuminate how digital communication both reflects and potentially reshapes traditional patterns of international attention, offering insights into the evolving nature of global discourse in the digital age.