In our new paper published in Communications Psychology “The Russian war in Ukraine increased Ukrainian language use on social media”, we investigate language changes in Ukraine-geo-located tweets before and during the Russian war in Ukraine. Our analysis shows that substantial behavioural shifts from Russian to Ukrainian take place with the outbreak of war.

Background

Many people globally, including Ukrainian citizens, are multilingual, and the languages they use reflect different facets of their identity in various contexts. Relating to our social roles—who we are around and the context we find ourselves in, we will adapt our behaviour according to the situation, which may include our choice of language. At the same time, language also has a political dimension, as it can be a source of conflict, is part of national and cultural identity, and has often been intentionally engineered with language policies in the process of nation-building.

In Ukraine, language has played a key role in post-Soviet identity. After independence, many citizens considered themselves Russians by nationality or Ukrainian with Russian as their main native language. The newly formed government implemented various language policies to achieve a return to the country’s linguistic roots. However, as the 2001 census and various small-scale surveys have demonstrated, only very recently there seems to be a gradual shift towards Ukrainian language and identification, following the Euromaidan protests and the Russian military intervention in Crimea and the Donbas. Hence, our study takes a closer look at language use on social media before and during the outbreak of the war, at scale.

Our Study

We collected over 4 million tweets from over 62,000 users with a Ukrainian geo-location from January 2020 to October 2022, which – as we show – includes almost all tweets in this period before and after the Russian invasion (24th February 2022). Due to the Ukrainian geo-location, we can – with some caveats – assume that most of these tweets stem from Ukrainian citizens and (to a smaller extent) non-citizens living there. We subsequently conducted an extensive cleaning routine, in which we try to identify and exclude any spam from e.g. bots, and focus on analyzing tweets in the three most common languages (Ukrainian, Russian, English).

The aggregate tweet numbers (Figure 1) reveal substantial changes over time, particularly with the outbreak of the war (second black vertical line). However, as we discuss in the paper, a descriptive analysis in itself does not allow us to distinguish whether the observed patterns are due to the in- and outflux of users, i.e. user turnover, or whether the actively tweeting users actually change their behaviour over time. In order to address this challenge and to separate effect sizes, we use advanced statistical models.

Our models show indeed an influx of English-speaking users, but even more so a substantial behavioural shift with the outbreak of the war, as already active users tweet substantially more in English (+135%), independent of the language they were normally tweeting in. Most likely, users wanted to let the world know what was happening and called for aid, which is supported by the fact that we observe a heavy spike in English tweets discussing the war, according to our multilingual topic modelling.

More importantly, we already find a steady long-term shift away from Russian towards Ukrainian before the war, as the average Ukraine-based Twitter user is more likely to tweet in Ukrainian than Russian over time (from 33% to 48%). This shift drastically speeds up with the start of Russian aggression in November 2021 and the subsequent outbreak of the war (from 48% to 76%) with a rise in Ukrainian (+56%) and a decrease in Russian tweeting activity (-20%), and is majorly driven – around three quarters of this shift – by behavioural changes.

As further analyses show, many of the previously Russian-tweeting users shift towards Ukrainian exactly with the outbreak of the war, as visualized in Figure 2. There we plot the language proportion of the switching users (y-axis) over time (x-axis), with red dots indicating 100% of a user’s tweets being phrased in Russian and blue dots 100% in Ukrainian. We can observe a clear shift with the outbreak of the war (second black vertical line).

Who switches & what do those behavioural shifts mean?

We theorize that this shift is a highly politicized response. Users want to distance themselves from any support of the war by no longer using Russian, and consciously change their self-expressed (online) identity. This may also explain why previously Russian users who switch to Ukrainian seem to be more active on Twitter and have a larger follower base, as audience pressure and interactions on social media may very likely contribute to this conscious decision. Overall, we argue that this language shift can be interpreted as citizens’ increasing opposition to Russia and a return to the country’s linguistic roots as well as a push towards a conscious self-definition of being Ukrainian.