The Russia-Ukraine war undermines the international order, centered at the United Nations (UN), and in the security realm at the UN Security Council (SC), like no other event in decades. A key element of the conflict is the reception of the war and of the role of the UN by the public and elites across the world. What are the central narratives of the war that prevail in the public sphere around the world? And what is the image of the UN in these narratives?
There are two main competing narratives of the War. The narrative promulgated by Western powers sees the war as an unprovoked illegal aggression, undermining the international order. Russia presents a competing narrative in which the war is a defensive response to NATO’s Eastward expansion and a part of the global re-ordering. The resonance of the narratives varies across different parts of the World. The UN holds a distinct position in these competing narratives. On the one hand, it represents the pinnacle of multilateralism and one of the centers of the liberal international order (LIO). It is the UN Security Council that is uniquely mandated with maintaining international peace and security. On the other hand, the UN is often heavily criticised for its inability to deliver precisely on its core mandate and effectively compel Russia to stop its aggression. It is seen as increasingly immobilised by undemocratic states. For the legitimacy of the UN, which of these pictures prevails is of utmost importance. News media is the essential vehicle that carries these narratives and studying their content enables us to map the prevalence of these narratives in the public spheres around the world.
In this text, we offer the first estimate of the prevalence of the competing narratives of the war and images of the UN in online news media across high number of states. We do so by deploying advanced natural language processing tools to analyse close to 7 million news articles from around 1,900 media outlets in 2022-2023. We provide a quasi-representative sample of news media content covering the analyzed topics in 136 states, representing more than 90% of world population.
First, we use two supervised machine learning model based on a fine-tuned deBERTa-v3 large. The models achieve high classification accuracy: 85% for the above-mentioned war narratives, and 80% for the different views on the UN. These models are then deployed on around 2.5 million news articles on the war and, among them, around 150 000 articles referring to the UN. Finally, we turn to generative AI (GPT-4, GPT-3.5-turbo, Llama2) models to detect with greater nuance the different ways in which the UN is referred to in the context of the war. Combining the results from these techniques, we identify major geographical variation in how the war and the UN are depicted in news media worldwide.
