The military conflict between Armenia and Azerbaijan over Nagorno-Karabakh has generated headlines across the world. Yet as Elise Thomas explains, a parallel battle has also been fought between the two sides online. She argues that better research is required to understand the spread of disinformation during armed conflicts, particularly as material is often removed by social media sites before it can be documented by researchers.
In 2020, armed conflicts are frequently accompanied by a ferocious battle being played out online for control over the narrative. Social media has emerged as a key vector for influencing the perceptions of ordinary people around the world, but also and perhaps more importantly for shaping international media coverage. Cash-strapped media organisations compensate for journalists on the ground with images, video and information from social media. Their coverage, in turn, is likely to play a role in informing policy decisions and in shaping the broader geopolitical response to the conflict. The shadow wars playing out online are therefore an increasingly significant element of the conflict itself.
However, these information battles are often poorly documented, and as a result are poorly understood. There are a couple of factors driving this, which a recent report that my colleague Albert Zhang and I wrote over the course of a few days at the beginning of the latest outbreak of violence in the ongoing conflict between Armenia and Azerbaijan helps to illustrate.
The online dimension to the Nagorno-Karabakh conflict
Just hours after hostilities broke out between the two countries over the disputed region of Nagorno-Karabakh on 27 September, hashtags related to the conflict began to trend on Twitter. This was how I, and undoubtedly many others, first became aware of the conflict. Even as I was looking at these hashtags, however, I could see tweets and entire accounts disappearing before my eyes as Twitter’s content moderators swung into action and began removing the duelling inauthentic, potentially bot accounts which were retweeting and boosting hashtags on both sides.
To be clear, this is what Twitter should do. Removing coordinated inauthentic activity is a key responsibility of the social media platforms, especially when it comes to contentious issues such as an extremely complex international conflict. Once Twitter had determined that the content was likely to be a part of an information operation connected to the conflict, removing that content was undoubtedly the right and responsible course of action.
However, for disinformation researchers, conflict analysts and ultimately for historians, this presents a dilemma. These digital artefacts – the accounts, the tweets, the images and videos and hashtags – are a part of the conflict. When they’re removed, often that data is simply lost to researchers, and our ability to fully understand what happened goes with it. The deletions may make it impossible to track back exactly where a particular piece of disinformation originated or, on the other side, frustrate the ability to gather evidence of potential war crimes and human rights abuses.
This is why my colleague and I felt it was necessary to set aside other projects and scramble to collect and analyse at least some of the social media activity linked to the conflict in the initial days after clashes broke out. In a week, or three months, or a year, the data will have had many systematic holes punched in it by content moderation and other factors, and these holes will skew the results of any later research or analysis. Capturing data in near real-time is crucial in order to support the accuracy of any future research.
Obstacles to disinformation research
The speed at which evidence degrades is one of the factors behind why these shadow conflicts are poorly understood. The other is simple, but fundamental: it’s not clear who should pay for it.
Disinformation research is a growing, but still small field. Outside of academia, most organisations doing this kind of research work on a project-based funding model. The upshot of this is a lot of research on the regions and issues which funders are interested in (for example, Russian or Chinese disinformation or US election interference) and severely limited research on a vast swathe of the rest of the world.
We were lucky to be able to undertake our research into the Nagorno-Karabakh conflict, in that we are permitted to do a certain amount of unfunded work and the outbreak of hostilities happened to occur during a window of time when we had no immediately looming deadlines for funded work. If it had come a week or two earlier, we would probably not have been able to do this report at all and much of the data we collected would simply never have been recorded or analysed.
This continues to be the case for many other issues and conflicts around the world. The massive, months-long information war between India and Pakistan over Kashmir; the sometimes joyous, sometimes riotous, always fascinating dynamics of social media in African countries, from the #EndSARS movement in Nigeria to #Shutitalldown in Namibia or #Congoisbleeding in Congo; the steadily rising temperature of the social media conversation around Kosovo; even the ongoing Turkish-linked campaign against the PKK and YPG in northern Syria. These are fraught, highly contested issues with historical significance, and yet key pieces of those digital histories slip away day by day.
In some cases, journalists do their best to cover this social media battle, but this in itself can be a complicated task. As mentioned above, journalists are often themselves the target of such activity, and can unwittingly convey exactly the message that inauthentic influence campaigns are intended to spread. Even where journalists are alert to the risks of disinformation or inauthentic efforts to influence their coverage, their reporting rarely includes large-scale efforts to capture or analyse data (although the growing use of data journalism holds out some hope for this in the future).
This is a difficult needle to thread. The social media companies have a responsibility to remove efforts to use their platforms to artificially distort or manipulate perceptions and narratives, in particular where it comes to issues such as armed conflicts. At the same time, those efforts at information warfare are a part of modern conflict, and evidence about when and how they were conducted has both analytical and historic value. It is not impossible to capture that data at the time, but currently in practice it’s difficult and ad hoc, as the funding model which underpins most non-academic disinformation research simply does not support this work for many issues and conflicts.
One option might be for the social media platforms to adapt the approach which Facebook and Twitter have taken to foreign influence operations. When Facebook or Twitter dismantle a network linked to a foreign influence effort, they may preserve that dataset and share it with researchers for analysis (and sometimes later make all or part of that dataset publicly available).
It is possible to imagine something similar for analysing social media activity related to specific conflicts or geopolitical incidents. Researchers could mix the dataset of removed activity back in with the activity which was not removed in order to have a relatively complete picture of what actually occurred in the crucial minutes and hours after an airstrike, for example, or after tanks roll into a disputed area.
Almost everything is easier in theory than in practice, and there will be flaws and complexities to work out. It is clear, however, that this is a conversation which needs to start happening amongst researchers both inside and outside the academic community, amongst the social media platforms and amongst potential funders who would support this work in the interests of accurate conflict analysis, and in the interests of history.