Crisis-induced differences in attention towards Ukraine in Twitter 2008-2023

A team of researchers has published a novel longitudinal analysis of global attention on Twitter (now X) surrounding the conflicts in Ukraine. The study, led by Mark Mets, Maximilian Schich, and Peter Sheridan Dodds, maps the frequency of the word "Ukraine" across 28 languages from 2008 to 2023. Their key innovation is a data visualization technique inspired by biological DNA microarrays, which charts the "log over-expression" of the term relative to each language's baseline. This macro-scale approach makes major geopolitical events starkly visible while uncovering subtler, long-term attention patterns that would be impossible to detect by examining any single language in isolation.

The most striking finding is the emergence of two nearly distinct language clusters with almost no overlap. One cluster shows a pronounced peak in attention around the 2014 Russian annexation of Crimea, while the other cluster surges dramatically around the full-scale invasion in 2022. Each cluster exhibits unique onset and decay profiles, which the researchers suggest mirror the varying levels of national readiness or reluctance to support Ukraine in different parts of the world. The method is presented as a versatile tool for comparing relative bias across languages, user subgroups, or even historical print corpora.

However, the study concludes with a critical and sobering observation about the limits of public data. While their cartographic approach can approximate global attention patterns from accessible data, the researchers note that a complete and unfiltered understanding of global information flows remains locked behind the proprietary algorithms of major social media platforms. This asymmetry grants platform owners a disproportionately comprehensive view of world events, raising significant questions about transparency and power in the digital public sphere.