Authors

Discussion

This paper presents a comparative analysis of dehumanising language on popular Telegram channels in Ukraine and Russia during the 2022 Russian invasion, examining differences between the two countries and how this language evolved over time. The dataset comprises posts from 185 channels, spanning December 24, 2021, to April 24, 2022, covering both pre- and immediate post-invasion periods.

To explore dehumanising dynamics between aggressor and defender, we conducted classification to identify dehumanising posts. Group-specific labels, such as city names and prominent political figures, isolated relevant posts. A manually labelled test set of 225 posts informed the optimal prompt selection for classification using GPT-3.5 Turbo. We then conducted a two-iteration auto-classification, retaining consistently classified dehumanising posts and analysing borderline cases separately.

Our mixed-methods approach combines NLP tools with qualitative analysis to address two questions: who are the primary agents of dehumanisation in each country, and how did the use of dehumanising language develop both qualitatively and quantitatively following the invasion?

Preliminary findings show a significant rise in dehumanising language on both Russian and Ukrainian Telegram channels post-invasion. However, Russian dehumanisation predominantly stems from state-affiliated channels, indicating a top-down approach, while in Ukraine it arises more from independent voices, like bloggers and community leaders, suggesting a bottom-up response. Additionally, Russian dehumanisation relies on a narrow set of clichés amplifying historical grievances, often using disinformation. In contrast, Ukrainian dehumanisation appears more varied, emotionally expressive, showing a dynamic response to the immediate context. References to the dehumanised “other” are frequently tied to calls for national unity in both countries, though this rhetoric existed on Russian Telegram before the invasion but only emerged on Ukrainian channels afterward.

These findings suggest that dehumanisation in Russian Telegram may be strategically deployed to reinforce unity and historical grievances, whereas Ukrainian dehumanisation seems reactive, potentially serving as a protective response. This contrast highlights the importance of tailored strategies to address dehumanisation in social media from the perspectives of aggressor and defender in wartime.