Dedemocratizing social media: the case of Russian Telegram channels

(edited lightly 28/12 for typos and a weirdly wrong sentence, nothing substantive.)

One of the benefits early optimists saw in social media and the blogosphere of roughly the first decade of the new millennium was the possibility of democratizing opinion, giving a voice to all and overcoming the asymmetry that arises from the fact that in previous models of mass media the powerful tended to monopolize media as well as politics.

While we’re much more pessimistic now and likely to view social media as mere implements of platform capitalism, the fact remains that one can to some extent speak out against the powerful — one can silence brands, one can ratio prime ministers.

A natural and interesting question is how this pseudo-democratization of opinion plays out in authoritarian regimes. I’ve had an independent interest in this question, as I’ve been exploring whether we could discern the sentiment of everyday Russians from how they reacted to posts on messaging apps. In doing that work, I often wondered why state-backed Russian channels bothered to allow their users to react.

The reason I wondered is because Russia has all the hallmarks of an authoritarian state, one of the features of which is control over the mass media, and indeed since Putin’s taking power we’ve seen dissenting voices gradually (NTV), then suddenly (Memorial, Dozhd, Meduza), be quieted.

Why, I asked myself, would Russia allow everyday Russians to express their dissatisfaction with the regime? That’s not how authoritarianism works! Why let citizens thumbs-down the leader when he does PR photoshoots as his army gets beaten back?

I think I have the beginnings of an answer, one which, as far as I can see, has gone unnoted elsewhere. Here it is.

Look at this first graph:

It shows the number of reactions to posts on the Russian news agency TASS’s Telegram channel for ten days before the end of February. Fairly steady, and then suddenly dropping off to zero on the day Putin announced his ‘special military operation’.

And look at this second graph, which also shows the number of reactions to posts on the same channel. Near zero, before suddenly jumping into action:

One could, if one didn’t want to incur the wraths of the Gods who manage the Telegram API, show that this isn’t cherry-picked. Before February 24th, there were a sizable number of reactions; then basically none until June 23rd; and then, and ongoingly, a sizable number again.

It doesn’t take much to work out what’s going on. On announcing the war, TASS turned off reactions (something one can do on Telegram, as can easily be verified if you set up a test channel), and at the end of June (when things were looking good for Russia?) they turned them back on.

And so, very simply, I have an answer to my question: if the suggestion here is right, then Russia has been controlling dissenting voices, or the possibility of them. To prevent such voices, when war broke out, they simply turned them off. When things were looking better, they felt comfortable enough to turn them back on, and so I reiterate an earlier prediction that, should things continue to look bad for the Russian side, we’ll see the reacts turned off again. And an important corollary is the generalization: by relying on media platforms with their own idiosyncratic designs, we thereby come to inhabit idiosyncratic and uncertain — to borrow a phrase — ‘public spheres’, places of information exchange that can be exploited at the click of a button to, pretty much literally, turn off voices that aren’t desirable.

(The funniest example of this, by the way, is Ramzan Kadryov. He has over 2 million Telegram followers, his posts receive about 50,000 reacts, and, as far as I can tell, he never receives any thumbs down or other negative emojis. The obvious explanation is that he turned off the capacity to do that, something both possible in Telegram and perhaps revelatory of the Kadyrovian personality.)

This is just suggestive. If one looks back to the period when emojis were turned on again, it certainly seems like it coincided with optimism on the Russian side, but I haven’t done enough work conclusively to make the point. Here’s an interesting and not rhetorical question: what happened on around 23–24th that make them make this move?

Because something definitely did happen. If you think the above circumstantial and non-probative, fair enough — but I have more. Look at this final graph, which shows something very curious:

Telegram, unlike Twitter (until recently), has had an edit function. The above graph shows long-range edits, a term I’ve just made up to refer to edits when the editing and the post it edits are separated in time by a decent amount (it’s thus a fraction of overall edits, the majority of which happen within 30 seconds of the original and presumably just correct typos).

We see the 24th again! And not only that. Here’s a very strange thing. I checked a bunch of the edits, and this pattern emerged: if a story was edited, then it was overwhelmingly likely to have an emoji, even if, per the hypothesis suggested above, emojis were disallowed at the original time of posting. What this suggests is that the editors, for some reason, went back, once reacts were allowed again, and reacted to old posts. For example, here are the edits that were performed on 7–23 and 7–24:

And here are screenshots of the second, third, and fourth, from the 23rd, in context:

In each case, the edited one is the one that has the emoji. This generalizes: scroll through the TASS telegram, and you’ll see pages on pages of unreacted to stories, then when one does have a reaction, inevitably it will have had a long-range edit.

(Admittedly, this doesn’t hold for the first case, which is, I think, the only example I’ve seen of it not holding. I link a pastebin of the edits performed on a given day if others want to check my work (the laborious procedure is: download the json archive of TASS from Telegram; search for the second:millisecond time-stamp (not the whole time-stamp because I’m in an unusual timezone), then open Telegram, find the post (you might need to know Cyrillic for this) and then see if it has an emoji.)

So we have this picture. The war starts; emojis are turned off; when things are looking good, they are turned back on; at the very same time, old posts are retroactively reacted to.

I admit, I’m baffled by why this is. If you look at the edited posts, they’ll typically only have one or two reactions, and some, which I can’t immediately find, have the ‘wrong’ reaction — a positive reaction to a Zelensky speech, for example. Moreover, they are months old posts that basically no one apart from weird obsessives like me are ever going to see. So why do it?

I don’t have an answer to this, right now. But we can, I think, conclude this. Something happened, and that something has to do with either silencing (banning reactions post-24 Feb) or allowing (enabling them in June, and seemingly retroactively reacting to stories). And when the community’s voice was allowed, a strange sort of stealth editing went on. I don’t know the how to fill in the story, but I think it’s worth continued attention, especially for anyone interested in how public opinion can be shaped and shoved in the digital era.