Social media has created the ability for people to seamlessly share content online, with nearly 4.2 billion people on some form of Online Social Network (OSN) [2]. These new forms of connectivity offer several exciting use-cases (e.g. social activism, health, marketing) [16, 17, 32]. However, this brings a range of challenges related to information quality, control, and their effect on society [8, 20, 24].

This paper explores the information divide and information warfare on OSNs in relation to crisis situations, e.g. natural disasters, violent conflicts and pandemic control. We do not limit ourselves to a particular crisis and, instead argue that similar challenges often transcend many types of situations. For example, we have witnessed similar online challenges in several natural crises, such as the COVID-19 pandemic, as well as political unrest or military conflicts such as the current Russo-Ukrainian crisis. We first discuss anecdotes from recent years and build our case towards highlighting the challenges.

COVID-19 pandemic: Since the start of COVID-19, social media has become one of the primary sources of discourse for users and administrators. Different channels were set up to immediately communicate any relevant information to public. People started using several hashtags to raise awareness related to COVID-19. For instance, #WFH, #workfromhome, #flattenthecurve were used to raise awareness for social distancing and general protective measures. In parallel, some people were spreading conspiracies. In the US, #filmyourhospital was trending at the start of the pandemic to show that hospitals were working normally (and that COVID-19 is fake). Other hashtags, such #covidiot, emerged to refer to someone who in any way is helping to spread COVID-19. Then debate on COVID-19’ moved to vaxers (pro-vaccine) and anti-vaxers (anti-vaccine). Both groups have a large presence on social media with a clear information divide.Other cases of information division during COVID-19 include the use of masks and lockdowns. Each of these factions has been active in spreading content (including facts and conspiracies) that favours their opinion.

2022 Russo-Ukrainian Crises: Governments have realised the effectiveness of social media in building certain narratives. In February 2022, when Russia invaded Ukraine, the world was already split between pro-Russia and pro-Ukrainian stances. Hashtags like #IStandWithPutin emerged on social media [18]. Pro-Russian Twitter accounts started sharing posts about mass genocide by the Ukrainian government, justifying a military invention [5, 11]. Controversies and anti-NATO rhetoric also grew during this time [4]. Sharing false information (that appears reliable) is also evident during such events. For instance, during this crisis, several videos were shared either depicting battle zones and dead bodies, later found to be fake or altered to spread disinformation in support of each side [6]. To make it even more damaging, sometimes such videos are even shared by verified accounts, as was the Russian Embassy's case sharing a video originally filmed by climate change activists in Austria [3].

These anecdotes of social media discourse during crises raise concerns related to information dissemination. Whereas one could argue that stricter information controls are required, this raises the question of who should have these powers, how should they be executed, and in which circumstances. This is particularly problematic in polarised debates, where society can potentially divide into opposing factions. In the rest of the paper, we explore how this information divide develops and the associated risks it poses. Finally, we summarise future research opportunities and challenges.

The aforementioned crises show that social media can carry a bi-polar discourse e.g. pro-russia vs. pro-ukraine, masks vs. anti masks. This information divide is also referred to as polarisation, i.e. when a society splits into two major factions, typically pertaining to a specific issue [25]. The most common examples of polarisation are political divides due to partisan narratives (e.g. Republican vs. Democrat). Although we provide examples from particular crises, information divides could also relate to other controversial issues, such as gun control, school re-openings during COVID-19, immigration etc. It is often reported that polarisation on these issues has a negative impact on society [10]. This is particularly problematic during crises as it may undermine mitigation efforts and even result in exacerbation (e.g. as we have seen in the anti-vaxer campaigns).

This information divide is commonly linked with the enforcement of similar opinions because there is less exposure to the opposite views. Such a divide in the community is also harmful to democratic procedures as it can affect the outreach of balanced information and free speech. On the one hand, OSNs act as a mirror to society and provide a great platform for studying such an information divide. Yet, on the other hand, they let members of an OSN shape the mirror. An average social media user might think of herself as aware and in control of her feed, but the actual influence of OSN ranking algorithms remains unaccountable. However, we note that the onus of polarisation is not only on the algorithms. There is an intricate interplay between the machine and human traits, some implicit and others explicit. The first widely discussed stimulus comes from “personalisation”. This term is heavily used in marketing and is employed on almost every digital platform to make an experience more relatable for individual users. Algorithms are trained to know the user better and keep recommending content that enforces users’ interests and opinions [30].

Such enforcement can result in a user being trapped in a so-called “Echo Chamber”, where it is difficult for opposing views to penetrate. In fast-moving crisis situations, these impacts can be exacerbated where groups ossify towards different opinions via aggressive sharing of news, videos etc [26]. Alternatively, some researchers consider that algorithms might not be enforcing the bubble effect but rather are consolidating the information according to their objective, and their effect on polarisation might not be significant. Nonetheless, it is known that content rankings can contain a bias towards certain opinions [15, 21] and such ranking bias can also affect people's opinions [23]. For instance, Twitter is reported to prefer the amplification of right-wing content [19]. Consequently, in crisis situations, the content that aligns well with right-wing ideology will often reach a larger audience, e.g. ideologically right-wing people prefer fewer restrictions related to COVID-19 [12].

In addition to the algorithms, human psychology also plays a major part in enforcing the information divide. Humans are subjected to several biases that can be exacerbated during periods of high stress, such as crises. We would like to highlight two relevant ones: “Confirmation Bias” and “Demand Bias”. The first one deals with what we like and prefer to agree with. The second one exhibits bias towards our desire to achieve/produce something that we assume our peers would want from us. These biases can be affected by users’ online and offline social networks, and online content consumption can reinforce them. For instance, support for #IStandWithRussia has been different depending on geographic regions, as many south Asian and African regions have been supportive of Russia and vary of the West [27]. Users’ preferences for one group over another might also play an important role in picking a side during a crisis. Due to the global connectivity of OSNs, the uptick on a particular narrative can further keep the discourse alive, regardless of greater physical proximity. Similarly, the bias to hide one's true state can play a role in suppressing one's feelings — equivalent to letting the other side win the narrative. This is particularly risky with authoritarian regimes where publicly supporting certain narratives can be dangerous. We note that the information divide is not only limited to these factors though. There can be several other factors — both online and offline. However, the above-mentioned factors can significantly affect the information discourse and online narrative during the crisis.

Mal-actors have identified the above forms of polarisation and tried to exploit it. We next discuss the exploitation of the information divide: so-called information warfare. This is an umbrella term, and there could be many issues that lie beneath. In modern social media, mis(dis)information, fake news, and coordinated propaganda campaigns are evident. Similar to polarisation, there is a growing body of literature on identifying fake news and coordinated campaigns (e.g. astroturfing for election engineering [22]).

This is so powerful because social media users can mask their hidden motives, which an average user might not be able to detect easily. For instance, the use of Twitter bots or even a human workforce to generate a fake support base might impact other people's opinions [22], e.g. this was reported during the US Election in 2016 [1]. In information warfare, there is also the involvement of organisations and governments at a much more coordinated scale than average users [7]. Worryingly, such tactics can even be exploited to engineer artificial crises.

Another important factor in spreading such organised information campaigns is the role of “key opinion leaders” or “influencers” [33].Platforms like Instagram or TikTok have accounts that are followed by millions. TikTok has been influential in spreading pro-Russian content, as several western-based OSNs were banned in Russia. Such users with a large follower base can rapidly spread any information with a certain point of view to a global audience. In the case of social media, the likelihood of false information spreading faster is comparatively higher than that of truth [31]. Such information is often amplified by others due to the fast-moving nature of crises, making verification difficult.

In the Russia-Ukraine 2022 crisis, social media has been at the forefront of spreading content, particularly videos. With an apparently larger presence of pro-Ukrainian sentiments in the US and European countries, one might think that the west has successfully controlled the narrative. However, beyond those countries, the sentiment might not be the same. As reported in the media, users in other countries, such as the masses in India might have different views. It can be contested that such sentiments come due to larger geopolitical references. However, it is also how the narrative is built using those references in advance.

There is a vast literature that studies the information divide, polarisation, information warfare and its effect on society [28]. There is also extensive work looking at how social media is utilised during crises [9, 13, 14, 26, 29]. So far, the more significant focus of this research has been on detecting and analysing such phenomena in different societies. There have been fewer efforts towards the mitigation of information divides or polarisation. For example, warning labels to increase user awareness are commonly used. However, we argue that more innovative methods are required to increase users’ awareness about the content and themselves. Users should know if they are getting trapped in one side of a narrative and how the other side of the narrative differs. Efforts should be on promoting a healthy debate and cutting down the information divide. Future research should target the interfaces and methods of communication where both the users and algorithmic demands are balanced. Note, due to the scale of content generation and the fast-moving nature of crises [26], these actions may need to be fairly automated.

That said, social media companies and different regulatory bodies are already trying to mitigate information warfare activities. Fact-checking is one approach taken; however, finding a way to scale up these activities is difficult. It also creates numerous issues regarding defining what is “true” and what should undergo fact-checking. Cross-region and cultural considerations exacerbate these challenges on the world stage.

Inclusion of multiple stakeholders, and most importantly, the relevant ones, is crucial to keep the inclusivity of the platform. For example, OSNs could adopt jury-based methods alongside the standard guidelines that include local perspectives in content moderation. The actions taken by artificial intelligence-based methods should also be made more interpretable. This will not only help in achieving fairness but will also keep the online discourse closer to reality.

Thus, more research is also required on social media companies’ regulations (and ethics). This presents a question about who has the right to control information propagation and under what circumstances. Finally, we argue it is necessary for researchers from different fields to come together, particularly those within the law, computer science and social sciences.

This paper has explored the topic of information division and warfare on social media during crisis situations. Considering recent world events, we argue this is a particularly important topic to discuss. We cover several examples of information divides and manipulation and discuss their possible impact. Factors such as algorithmic raking, human bias, and government/OSN controlled mechanisms can affect this process. We also present future research directions that we believe should focus on mitigating such issues. Due to complex ethical, social and legal factors, this is a non-trivial problem, particularly considering the diversity of situations that crises cover. We conclude by stating that this will require a cross-disciplinary effort.

This research has been supported in part by project 16214817 from the Research Grants Council of Hong Kong, and the 5GEAR project (Grant No. 318927) and the FIT project (Grant No. 325570) funded by the Academy of Finland.

^⁎^Pan Hui has second affiliation with University of Helsinki, Finland

Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s).

HT '22, June 28–July 01, 2022, Barcelona, Spain

© 2022 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-9233-4/22/06. DOI: https://doi.org/10.1145/3511095.3536372