Listen to this article:
The consensus in the Western media seems to be that Ukraine is winning the information war against Russia. Ukrainian President Volodymyr Zelenskyy, a former comedian and actor, has taken to addressing Western audiences like a duck to water. His tours de force included addressing the US Congress, appeals to German lawmakers and invoking Churchill to British parliamentarians to seek support and protection for his country.
Russia, on the other hand, has banned Facebook and Instagram by labelling them ‘extremist’, shut down independent media and criminalised factual reporting about its invasion.
Game, set and match to Ukraine, at least in information warfare?
Carl Miller doesn’t think so. The director of the Centre for Analysis of Social Media (CASM), a part of London-based think tank Demos, thinks Russia is waging war in information spaces that fall below the radar of Western researchers and journalists.
In early March CASM unearthed an influence campaign, possibly Russian in origin, targeting the BRICS bloc, of which Russia is a member.
The campaign amplifies scepticism of the West and NATO and invokes themes like anti-colonialism and BRICS solidarity to support the Russian invasion. Since Ukraine is appealing to Europe and the US, Russia is apparently tapping into a genuine non-Western scepticism of NATO intervention.
Miller speaks about his findings in an interview with Tushar Dhara for The Wire. The interview has been edited lightly for style and clarity.
Can you walk me through your findings?
The starting point for the research was two hashtags, #IStandWithRussia and #IStandWithPutin, that we saw trending on Twitter on March 2 and 3. These hashtags trended globally, in India, in the UK and elsewhere, and early research suggested that activity around those hashtags was suspicious. Rather than trying to tell who was a bot and who was not, we were interested in the nature of the accounts.
We collected every tweet (23 million) contained in the hashtags and winnowed it down to 10,000 accounts that sent five or more tweets containing the hashtag. We then looked at the kind of language the accounts were using, English or Hindi, and what were the main themes, words and phrases used. That gives each account a vector space. That led us to see that there were distinct clusters of languages across the Twitter universe of accounts that had been heavily engaged in making the hashtags trend.
What were your findings with the Indian language clusters?
There were three communities that we mapped which related to India. The most striking was a very tight and linguistically distinct grouping of primarily Hindi language, pro-Bharatiya Janata Party (BJP) accounts. There were around 566 accounts and four million messages. They have the most retweets to tweet ratio, so they are very spammy. They just retweet other messages. There are lots of pro-Narendra Modi, pro-BJP messages which they keep re-sharing: local political rallies and events, basically campaigning material.
And on March 2 and 3, many of these accounts just flipped and started sharing English language pro-invasion memes. They did this on March 2 and 3, and now, as far as we can tell, they are back to Indian politics and don’t have much to say about the Russian invasion.
How many days were the pro-Russia hashtags tweeted?
Just March 2 and 3. There was a sharp spike, and then down, and now there is very little Russia-related activity. Interestingly, there was another cluster that is even more linguistically distinct. It was barely connected to any other account in the whole network.
This was a Tamil cluster – 228 users sending 1.7 million tweets. And these were anti-BJP [accounts], with lots of oppositional messaging, quite a lot of engagement. I am sorry I am not an expert on Indian politics, but it looked like engagement with Tamil politicians who are anti-BJP. But again, over those two days, they started sharing exactly the same memes as the pro-BJP cluster.
So the Tamil cluster was also sharing pro-Putin hashtags?
Yes. Of course, there was some counter-messaging in there and people were calling out the hashtags. But primarily, both clusters were flipped to share the same English language pro-invasion memes.
I am surprised that the anti-BJP Tamil cluster flipped around and started tweeting pro-invasion memes.
That’s a sign of inauthenticity. Accounts classically changing behaviour and doing something that doesn’t fit within the context of what you had expected the person to do if they are who they say they are, that’s a sign you look for someone who is acting out of genuine resentment towards the West and a Russian-controlled influence operation.
It’s that unity of behaviour on those days pumping out the hashtags which otherwise pretend to be very different from each other politically and linguistically, that’s the kind of thing we would be looking for.
So you’re saying this is a Russian-controlled influence campaign?
I don’t have any proof that this is controlled by Russia, but it fits within the Russian playbook. This is inauthentic for sure, and coordinated. It may be a series of campaigns interlocking with this aim in mind. And this is not just the Indian clusters, there are Urdu, Sindhi and Farsi networks, Zulu and English language ones and much more diverse Javanese and Malaysian clusters.
So at this moment, it’s not clear whether the Hindi and English language accounts in the Indian cluster are controlled by the BJP or Russia?
It never will be clear. That’s not something data science can find out. What we can do is find patterns that are suspicious.
I am now reminded that on March 2 and 3, messages on Indian social media claimed that Modi had called Putin to halt the war for six hours to evacuate Indian citizens from Ukraine. This messaging, coming in the middle of elections in Uttar Pradesh, went viral. My friends who were covering the polls said that it penetrated down to the lowest level.
What was the third cluster?
It was an Indian English cluster. It’s much less Hindi language centric in its nature, almost exclusively accounts that identify as being from India. Indian names, actually, almost like too much Indian nationalism. It feels slightly authentic, the number of Indian flags and Taj Mahal boutiques on their profile pictures, but they only use English. And they are a loose cluster that connects the Hindi-language hub to the other more English language parts of the network.
And these have more #ISupportPutin content?
Yes, and that’s when we realised that functionally this network is pouring amplification into a small number of pro-invasion memes. When you start eyeballing these accounts, you realise there are a huge number of retweets, barely any original messages. And a small number of accounts that get inward amplification. An account with 70 followers that has been tweeting in Hindi and suddenly on March 2 and 3, the same account is tweeting English language pro-invasion memes and getting 10,000 retweets a piece. By no means does this suggest we have the full network. We deliberately didn’t get the full network in order to get a higher signal.
How many accounts around the world did you track?
Around 10,000. There is a South African cluster, a Pakistani cluster, an Iranian one and so on.
How many accounts are from India?
This is a broad figure, but the Hindi language cluster had 566 accounts, the Tamil network had 228 and the English language but Indian identifying accounts had 1,314. They sent a total of 11.7 million messages.
Do you know whether these accounts are bots or have people operating them?
A researcher like me cannot guarantee whether any account is run by a bot or not. Sometimes human beings can act in very robotic ways and at other times bots can behave like human beings. But overall, I would say the network has attributes which imply an overlapping series of bot accounts, compromised accounts and cyborg accounts, which are partly human and partly automated. There definitely are signs of automation in the network, snapping into English language pro-invasion meme and then snapping back into Hindi suggests that.
Is there any reason why this coordinated activity took place on those two days?
My interpretation is that this was a campaign that was angled towards the United Nations vote condemning the Russian invasion of Ukraine. There is quite a lot of messaging which is explicitly about the UN vote, for instance, making note of Russia’s diplomatic support of India in the past or its relationship with anti-Apartheid activists in South Africa.
There is a lot of anti-colonial, anti-Western, pro-BRICS messaging and I think that was perhaps intended in the run up to the UN vote to either influence the general public, but also possibly political and diplomatic actors who might be sitting on the fence on the vote to regard the Russian invasion as more popular in their countries than it really was.
What are the most prominent accounts in the Hindi and English clusters?
The strange thing is there aren’t many very highly followed accounts. This is attempting to be a grassroots organic network. In the South African cluster, we have Jacob Zuma’s daughter who is certainly not a bot and extremely influential. But within the three Indian clusters, there aren’t any highly followed influencer accounts.
What about language clusters in other countries?
The Urdu, Sindhi and Farsi cluster is as linguistically distinct as the Tamil cluster. There were pro-Imran Khan and some anti-Indian or anti-Indian nationalism posting from that cluster. But these were very similar to its Russian related messaging: Western hypocrisy, NATO expansionism, anti-colonialism. The South African cluster is primarily English with some Zulu. It is pro-Zuma, BRICS solidarity and anti-colonial in its rhetorical positioning.
What about the Farsi network?
It is linked to the Urdu and Sindhi networks. They are all clustered closely to each other. It is based in Pakistan and Iran, not a massive network. Apart from Pakistan politics, it has the same theme, which is that all these networks share an interest in anti-colonial, pro-Russia themes. The push factor is Western hypocrisy and NATO aggression. And the pull factor is BRICS and anti-colonial solidarity, Indian and Russian flags crossed together, that sort of thing.
Nothing from Brazil or China, which are part of BRICS?
Not that we can see. It’s just one small bit of research pulling apart what seems to us to be the first Russian influence operation – again, this is our judgment – in the early stages of the invasion. No Portuguese or Mandarin language accounts, but then Twitter isn’t used in China.
What are your research conclusions?
If indeed this is a Russian influence operation, it is targeting information spaces that are not the ones we in the West have been looking at or know much about. There has been this idea across the US and UK and EU, which I think is complacent, that Ukraine is winning the information war. And in many ways, I think that is making a mistake, thinking that our information environments are more general than they really are.
I haven’t seen a single piece of pro-Russia messaging in my timeline at all, ever. And we take that to mean that Kyiv is winning. But it may not be. What might be happening is this: Zelenskyy has very successfully couched this conflict as being one not between Ukraine and Russia, but Russia and the West. He has very much positioned Ukraine as standing as a bulwark of Europe and defending European values.
Russian influence operations may be exploiting that. They may be saying if this is about the West, then there are plenty of parts of the world that don’t have a huge amount of warmth for Western military adventurism. And there may be audiences, especially in countries like India, in order to puncture the picture of global condemnation. For Russia, it would much rather be a Western condemnation.
I have been surprised by the breadth of the US/NATO scepticism and pro-Russia sentiment in India cutting across the political spectrum.
Influence operations that are the most successful have truth at the core of them. I have interviewed people who do these kinds of things and what they tell me is this: “We don’t lie about the world to try and change your mind. We tell you things you already believe about the world and try and direct them in certain ways.”
When you say there is truth at the core, do you mean NATO and the West have acted in ways that have turned peoples’ minds against them?
The core truth is that there is genuine anti-colonial and anti-NATO feeling in India. A lot of Indians may not have sympathy with Western military adventurism. That’s exactly the kind of fertile environment in which an influence operation could amplify and exaggerate and generate sympathy for the Russian invasion.
Tushar Dhara is an independent writer and researcher working at the confluence of democracy, technology and rights.