How is the War in Ukraine Covered? Using Computer Vision to Map Media Information Bubbles across the European Union

Since its beginning in the spring of 2022, coverage of the war in Ukraine has constantly ranked high on the European countries’ media agenda. The influx of Ukrainian refugees as well as the need for military and economic support has been a staple of the political debate across the continent, with most, but not all countries supportive of the Ukraine military and humanitarian effort. Now as the war enters its third year, public opinion on the support for Ukraine is increasingly divided across Europe. In this context, the paper seeks to answer the following Research Questions: what information do European publics receive about the war in Ukraine, and what overlap, if at all, is there between these country-specific bubbles, when we consider the country-specific verbal and the nonverbal content in the coverage of this topic? We investigate this question using AMMICO, a tool we developed (https://doi.org/10.31235/osf.io/v8txj) that uses computer vision to automatically extract text and analyze images from media content. Our period of focus is the month of March 2024, corresponding to the beginning of spring and the expected new movements on the Ukrainian front. To determine the informational content of the bubbles different national publics are exposed to, we randomly sample stories containing the mention of Ukraine from the social media sites of the top three media news outlets in several European Union countries, specifically Hungary, Romania, Poland, Finland, Germany, Netherlands, France, and Italy. These countries are important to investigate due to them either being immediate neighbors and/or countries with significant influx of Ukrainian refugees. Using ammico, and in line with previous studies, we intend to analyze what political actors are being visually portrayed and in what manner (facial expressions, visual background elements), as well as measure the polarity of the sentiment in the text, and the topic construction of the statements. We expect to observe more negative statements and more instances of visual-verbal incongruity in countries with more negative public opinion. Our exploratory study aims to show (1) the benefits of using computer vision tools for more extensively mapping information bubbles, and (2) to provide a preliminary comparative perspective of the overlaps and differences in the European coverage of this salient topic.