A FIMI Outcomes Dataset and a Method for Structuring FIMI Linguistic and Visual Communication

The empirical foundations of FIMI research have long been fragmented: incident reports are produced by many different organisations using inconsistent taxonomies, making comparative analysis across campaigns or geographies structurally difficult. This deliverable from SWPS University and Debunk.org addresses that problem directly by presenting both a structured dataset and a methodological framework for codifying FIMI incidents. The dataset covers 501 reports from 110 organisations, cross-referenced against more than 12,600 URLs, 5,300 channels, and 3,000 domains — a scale that begins to support meaningful quantitative analysis of platform distribution and infrastructure patterns.

The methodological contribution is as significant as the data. The authors argue explicitly that “identifying FIMI requires more than analysing false content alone” — a corrective to content-centric approaches that miss the behavioural and infrastructural signatures that make influence operations distinct from ordinary misinformation. The integration of NLP, corpus linguistics, and AI-assisted detection methods positions this as a bridge between qualitative incident analysis and computational social science. Anti-Western messaging and pro-Russian narratives targeting Ukraine dominate the dataset, reflecting the current operational environment, but the methodological framework is designed to be generalised across other actor-context combinations.