Navigating the fog of war during the Russia’s invasion of Ukraine: An exploratory network analysis of tweets about an alleged chemical attack in Mariupol

by Felipe Bonow Soares, Alyssa Saiphoo, Anatoliy Gruzd, Philip Mai

Misinformation and disinformation about the Russia-Ukraine war are rampant on social media, making it difficult to discern between them. As part of our ongoing research on misinformation and disinformation of various types (COVID-19 related, state-sponsored, and those driven by conspiracies theories), we conducted an exploratory analysis of tweets discussing an unverified report that Russian forces engaged in a chemical attack in Mariupol, Ukraine. This claim was made on April 11 by Ukraine’s Azov regiment. At the time when this claim was first reported, Mariupol was surrounded by Russian troops, making it difficult, if not nearly impossible, for journalists to gain access to the city and to interview local sources. We were interested in examining how this claim was discussed on social media because if it was true, it had the potential to galvanize the world’s sentiments in support of Ukraine and against Russia. 


Using Twitter’s Academic Track API, we retroactively collected 246,189 public tweets posted between April 6 and 13, 2022 to analyze how Twitter users were discussing this claim. We collected tweets related to this case a few days before and after April 11 to capture speculation before the accusation, and the reaction to it. We used the search query “chemical (weapons OR weapon) (Mariupol OR Ukraine)” to collect data. (For data completeness, we kept 12,193 tweets referenced by one of the tweets in the search results.)

English keywords were used to analyze English posts as previous research that focused on online activism in similar contexts has found that English is more widely used for political mobilization and advocacy at the international level than local languages (Bastos & Mercea, 2015; Altahmazi, 2020). Further, Ukrainians likely want to get a reaction from the West when tweeting about the war, so we might expect them to post in English, similar to what we saw during the 2014 Euromaidan revolution in Ukraine (Gruzd et al., 2016). 

To examine how information surrounding the chemical weapons claim was shared and reshared on Twitter, we used Social Network Analysis (SNA) to analyze a retweet network. Through retweets, certain claims can be amplified and spread further. Thus, by analyzing a retweet network, we uncovered accounts behind the most popular content in the network.

The resulting network visualization is shown in Figure 1 below; each dot (node) is a Twitter account connected by a tie (edge) that denotes an interaction between two or more Twitter accounts in the dataset.

Figure 1. Retweet network on chemical weapons in Mariupol (105,902 nodes and 187,355 edges)

To guide our examination, we used a community detection algorithm to locate densely-connected nodes that tended to reshare the same content with one another. This allowed us to identify the two largest clusters of nodes (depicted in blue and orange in Figure 1). We then used a common SNA metric called in-degree centrality to identify influential accounts within the network. In our case, in-degree is equal to the number of times an account was retweeted by others in the network.

The following sections summarize our results based on a manual review of the most retweeted accounts (i.e., accounts with an in-degree higher than 500) and the content of their tweets in each cluster. In total, there were 48 accounts in the dataset that met this 500+ in-degree (retweet) threshold.

Analysis: Blue cluster

Figure 2. The most retweeted accounts (Ukrainian journalists and Western media) in the blue cluster

In this cluster (Figure 2), the majority of accounts had an in-degree of 4k or lower. However, there were a total of seven accounts that were particularly popular based on the number of times their tweets were retweeted – their in-degrees were between 6k and 21k. Six of these seven most retweeted accounts belonged to local journalists, media organizations, and a blogger/news aggregator in Ukraine – and one belonged to an academic who studies Russian economy and political system. These accounts, which tended to tweet with an air of urgency and certainty, treated the claim as if the claim is already a verified and established fact. Their tweets called for immediate action to stop Russian hostilities in Mariupol in particular and in Ukraine generally (see Table 1).

Table 1. Top 7 tweets in the Blue cluster

DateTweetPosterUser-defined profile description In-degree (e.g. number of retweets)
Apr 11, 2022 Azov regiment: Russia used poisonous substance against Ukrainian troops in Mariupol. The substance has been distributed by a drone, and victims have shortness of breath and vestibulocerebellar ataxia. It may be Russia's first known use of chemical weapons in Ukraine.@KyivIndependent“Independent English-language journalism in Ukraine”
(verified account)
Apr 11, 2022 Russia just used chemical arms in Mariupol. This happened just after Putin appointed General Dvornikov, the butcher of Syria, who used chemical weapons there, to the Ukraine command. @POTUS has promised tough NATO action. Yes, time to deliver! Time is short!@anders_aslund“Economist & author”
(verified account)
Apr 11, 2022 Mariupol siege and total annihilation of the city by Russians should have been a turning point. Bucha massacre should have definitely been a game changer. They were not. Will the use of chemical weapons by Russia be? How many more #RussianWarCrimes will the world tolerate?@olgatokariukJournalist
(verified account)
Apr 11, 2022 Defenders of #Mariupol just said Russia used chemical weapons on them around an hour ago – soldiers are having difficulty breathing and vestibulo-atactic syndrome. Earlier today Russian puppets in occupied Donbas openly called for the use of chemical weapons against them@lapatina_“Ukrainian journalist”
(verified account)
Apr 11, 2022 The Azov Regiment in Mariupol reports what could be the Russian use of chemical weapons against the entrenched Ukrainian garrison in the city. An unidentified agent was delivered with a UAV. Revealed effects include respiratory and vestibular failures.@IAPonomarenkoJournalist
(verified account)
Apr 11, 2022 Chemical weapon used against soldiers and civilians in Mariupol. Is this the red fucking line yet?@MelaniePodolyakBlogger9,554
Apr 11, 2022 @igorlachenMedia Manufacturing Company6,080

In addition to the prominent Ukrainian accounts described above, we identified a few (less connected) accounts from Western media outlets such as Reuters, The New York Times and The Associated Press in the blue cluster (Figure 2). Most tweets from Western media outlets noted the report of a possible chemical attack, that the report was still an “unverified” claim and that verification was not yet possible. For example, Reuters posted the following tweet:

In our analysis of the blue cluster, we can see differences in how Western media and local Ukrainian journalists tweeted about the claim of a chemical attack in Mariupol. While Western media was more cautious and framed the case as unverified, local journalists were more likely to discuss the claim as if it was already a verified fact and tended to focus on the consequences of a chemical attack. This might be the reason why local sources were more influential and prevalent in the retweet network.

Analysis: Orange cluster

The orange cluster is relatively small in terms of the number of nodes (~9% of all nodes), with only 5 accounts meeting the 500+ in-degree (retweet) inclusion threshold. Their most popular tweets reinforced pro-Russia narratives or question Western and Ukrainian narratives on this incident (see sample tweets in Table 2). Some of the most retweeted users said the claim was disinformation spread by the U.S., while others discredited the accusations by highlighting the fact that the Azov regiment – which has been linked to far-right ideologies –  was the original source. These narratives are consistent with widespread Russian propaganda campaigns, such as the claims that Putin is trying to “denazify” Ukraine and that the U.S. is spreading disinformation to scale the Russia-Ukraine conflict

Table 2. Top 5 tweets in the orange cluster

DateTweetPosterUser-defined profile descriptionIn-degree
Apr 7, 2022 @ElectionWizMedia & News2,924
Apr 11, 2022 @aaronjmate Journalist
(verified account)
Apr 9, 2022 @JackPosobiec
Media & News Company
(verified account)
Apr 11, 2022 @MapsUkraine suspended account543
Apr 11, 2022 @stillgray
Social Media Influencer
(verified account)

Tweets in this cluster amplified narratives often associated with Russian propaganda. The continuation and repetition are central features of contemporary Russian propaganda to persuade others but also create confusion and disruption (Paul & Matthews, 2016). The repetition of some of the Kremlin’s propaganda claims might create an illusory truth effect, where repeated content is perceived as truthful (Hassan & Barber, 2021). In the context of the war, this seems to be a continuous strategy to decrease trust in Ukrainian sources, the U.S. government, and Western media outlets (Einsten & Glick, 2015; Valenzuela et al., 2021). 


In summary, our analysis shows that Twitter users participating in the conversation surrounding the claim of a chemical attack in Mariupol more frequently retweeted accounts that showed certainty in their messages. This finding is consistent with some previous research that found that messages containing certainty are associated with higher engagement on social media (Pezzuti et al., 2022) and conversely that the more certain people are of a belief, the more likely they are to act and express their opinions about it (Tormala & Rucker, 2015).

In contrast, Western media accounts that tweeted about the claim but cautioned it as being unverified, received fewer retweets, making them less central in the network. One previous study found that borderline fact-checking labels like “Lack of Evidence” can be perceived as false rather than neutral, which is likely motivated by human cognitive mechanisms to avoid risk and uncertainty (Park et al., 2021). It is therefore possible that Twitter users in this network looking for content about the accusation of a chemical attack might have perceived the claim as false in these tweets because of how they were framed, and decided not to retweeted them as a result.

The difference in how Western media and Ukrainian journalists tweeted about the claims of a chemical attack in Mariupol is likely influenced by several factors, including the “fog of war” –  an expression used to describe  informational uncertainties that often accompany wars. In the case of the Russia-Ukraine war, this is further fueled by the massive amount of information available in the media ecosystem. This creates challenges for journalists covering the war, for fact-checkers trying to verify various claims by each side, and for social media users in and outside Ukraine seeking the latest information about the war.

Journalists have to adapt their practices for verifying information to navigate the fog; with modern technology, doctored or repurposed images and videos, and the speed of information sharing online, identifying credible information and seeing through the fog is becoming more difficult. What works one day may not work the next, and sources may suddenly become invalid (Haughey et al. 2020). So, while verifying information to provide accurate reporting is a key practice in journalism, the fog of war makes this difficult.

For Western media, the fog of war is particularly thick: it is hard to verify claims, like the accusation of a chemical attack, as they have limited access to sources and to the location where the attack might have happened. As investigations of chemical attacks (in terms of war crimes) usually take a long time, the alleged chemical attack in Mariupol is an example of a case that cannot be easily fact-checked. As a result, Western journalists were more likely to be skeptical and report the use of chemical weapons as an “unverified” event. On the other hand, local journalists are closer to the events, and many of them are personally experiencing the war, which makes them directly exposed to the dangers of the war as potential targets of Russian attacks. They do not have the privilege of time and distance that Western media outlets have. This might make local journalists more likely to “cut through the fog”; to accept the claim as fact and focus on the consequences of a chemical attack.

*Note: An earlier version of this post included manually-assigned labels of top twitter accounts which were subsequently replaced with user-defined profile description to reflect user/account’s self-identification.