Doom scrolling: How Elon Musk is boosting the British right

0:00Kaitlin Tosh:Coming up, Elon Musk says his social media platform X is the home of free speech on the internet. Sky News Research says something very different.

0:11Tom Cheshire:How does a banana trigger a CIA-backed coup? Do AirPods herald the arrival of a new global order? What do LED lights say about the future of humanity? I'm Ed Conway, and in each episode of my new podcast, Stuff Matters, I take an object, crack it open and reveal the world-shaping forces hidden inside. This is economics told through the things we think we understand. Search Stuff Matters on your podcast app to listen and follow.

0:48Kaitlin Tosh:If anyone ever tells you Elon Musk has no involvement in British politics, well, you might just want to remind them of this.

0:57Tom Cheshire:Whether you choose violence or not, a plant is coming to you. You either fight back or you die.

1:04Kaitlin Tosh:That's the billionaire boss of SpaceX and Tesla beaming in live to the right-wing Unite the Kingdom march, organised by one Tommy Robinson back in September. And that level of involvement in British politics would be noteworthy enough in itself, backed as it is by all Musk's money and influence. Yet, since 2022, he's also owned his own social media platform, Twitter, since rebranded as X. And as Musk has cut staff, reinstated accounts and gone on a few rants of his own, there's been a feeling, a perception, that the site has shifted significantly to the right, that some accounts are being pushed and others suppressed.

1:44Kaitlin Tosh:A Sky News investigation has revealed the X effect, that, for whatever reason, right-wing material is being amplified. Well, the investigation was put together by members of Sky's data and forensics unit. Two of them are here today, Tom Cheshire, our data and forensics correspondent, and Caitlin Tosh, our digital investigations journalist. Michelle Innes-Simon, another member of the data team, well, she isn't here today. She was, however, instrumental in putting this together. Tom, we're going to start with you. Just explain exactly why we have been taking this closer look at what I still call Twitter, but everyone else calls it.

2:20Tom Cheshire:Yeah, I mean, when Elon Musk bought it, it was Twitter. And the question is, why is he buying it? You know, this is a guy who's made his name with rockets, with cars, done amazingly successful things. And then he's getting in the messy world of social media for quite a lot of money,$46 billion. And he said he bought it to eradicate, as he put it, the woke mind virus from progressive politics. And it was, you know, this huge thing. I mean, Twitter was in a lot of danger before as well. It needed rescuing. He had these effects, obviously, very quickly. He put people like Tommy Robinson back on the platform who'd been banned for hateful content, other accounts like that.

2:52Tom Cheshire:But then we want to see if there's more going on than just that. He wants to be a free speech platform, which is a laudable goal. Any journalist would agree with that. But has there been more than that? People have felt that perhaps we're seeing a shift towards more right wing content. That's the feeling. It's not data. And that's what we want to go and get is that data. Can we show, especially the algorithm thing that powers is it that sorts everything you see, especially on the For You tab, whether that isn't just impartial, it's not just neutral, actually, it's pushing a point of view.

3:21Kaitlin Tosh:He is a billionaire that owns one of the world's biggest social media platforms. So, I mean, he does wield an awful lot of power, doesn't he?

3:26Tom Cheshire:Yeah, a huge, huge amount of power, and it's just about how does that show up in your feed? You'd expect it to happen. That's why he bought it. And he's been really sort of, that intervention in British politics we've seen, he's got obsessed with it, really. It started with attacking Keir Starmer a lot, and now he's alighted on Tommy Robinson really is his pet favourite. He says it's hard men like him who are going to save the UK and that extends to sort of appearing at the Unite the Kingdom rally that Tommy Robinson did, you know, 150 ,000 people were there in the street and then Elon Musk is beaming into one of those four giant screens down in Whitehall.

3:57Tom Cheshire:So that is direct interaction but also when I've been going to protest over the summer, whether it's in Epping outside the hotel or a Britannia hotel in Canary Wharf, just two examples, you speak to people and they'll say thank you Elon, it's something Tommy Robinson says a lot. They say, thank you for giving us a voice. They say without Elon, there wouldn't be free speed. And again, a laudable goal. And the question is, what does that mean for the average user? You've got these power users, these people who are politically active, but does that translate to a skew, or does it not? And that's what we try to find out.

4:25Kaitlin Tosh:Caitlin, at the heart of all this is the X algorithm. I mean, first things first, what's an algorithm? And secondly, why is it important when it comes to social media platforms? Yeah, you phrase that like it's an easy question. So yeah, an algorithm is basically the back end of a social media website that it's a code that decides what is going to be shown to the algorithmically curated feeds on your social media platforms. So for X, that's the For You tab on X. That's kind of what tweets are sent there are decided by the algorithm. So because the other tab is people that you have chosen to follow.

4:57Exactly. Yeah, there's kind of two at the top, one's following, one's for you, the one that the algorithm picks.

5:02Tom Cheshire:It's also the default one as well. I think that's really important. When you log on for the first time, that is quite easy not to get over to the following. So the four years, that's how people experience X. Investigating how these algorithms work is one, very tricky, but also really important. It's really important to understand what is being decided behind what you're seeing.

5:17Kaitlin Tosh:I thought X was one of the few companies that kind of open sourced this stuff. They put the algorithm out there. So do we not have an idea of how it works despite that? Yeah, it's a really good point. So one thing that Elon said when he took over Twitter and made it X was that he was going to be more transparent with the algorithm and was going to be one of the only social media platforms to put it publicly. And he did. It's true. It was published in 2023 for the first time, updated most recently in September of 25. And it does tell you some things about how the algorithm is making the decisions that it does.

5:47So, for example, we know that videos are more promoted than just text, for example, and things like that. But the experts that we've spoken to say that there's a lot that it doesn't tell you. And there's a lot that's still kind of what a lot of people term a black box where you just don't really know what's going on. It's not as clear as it might appear.

6:04Kaitlin Tosh:So how did we test it? One of the best ways to investigate an algorithm is to basically create new accounts and see what it sends them. It's kind of the way to decipher the black box that is unclear otherwise. So that's exactly what we did. We made nine new X accounts and each of them were British. We didn't hide the fact that we were creating them in the UK. So they were British users and they all had different political leanings. Three were created to be emulating left-leaning British people. Three right-leaning and three were neutral, so no interest in politics. Over the course of two weeks, we pulled the data from the For You pages from all of those accounts.

6:39And that gave us a database of about 90 ,000 tweets from about 22 ,000 users.

6:44Kaitlin Tosh:Okay, so a fairly substantial data set for you guys to analyze here. Yeah, it was painful.

6:50Tom Cheshire:Mainly for Caitlin, not for me. Yeah, it was a lot. So what we ended up doing was then using a large language model, an AI tool, to categorize all of that content. So probably the most rigorous part of this process in terms of the analysis was training that language model.

7:06Kaitlin Tosh:So it takes these tweets and does what with them? Analyzes text, basically. And we'll be able to categorize based on the text information from every single one of those tweets. And then also information like the location of where the account came from. So, you know, a tweet from Elon Musk, for example, it would read the tweet information. So what the tweet said, where Elon Musk was, so where his location is. and it will be able to kind of read it, learn about it, and then apply it to a political definition that we also fed the AI. Okay, Tom. So we've got the AI cracking on with this analysis. At its core, though, we were wanting to find out whether left-wing content was being suppressed and right-wing content was being pushed.

7:45Kaitlin Tosh:So what did we discover?

7:46Tom Cheshire:I think we found fairly robustly that right-wing content is being pushed by the algorithm. So across the board, and remember, these are brand new BoxFresh accounts. They got 62 % of the content they saw over the course of this experiment was right-wing content. But even when you break it down, so if you want the left-wing accounts, they still saw a lot of right-wing content, nearly half the content they saw. But then when you look at the right-wing accounts, they see overwhelmingly right-wing content. And the neutrals in the middle, they still see way more right-wing content. And the neutrals especially, these are people who are just following Sky Sports or football accounts or entertainment accounts.

8:19Tom Cheshire:They haven't expressed any interest in politics, but they're still ending up in that. And the other thing we found here is that the accounts that are posting are also posting a lot of extreme language. And there's some sensitivity around this because, you know, it's a definition which can be controversial. But we work with experts at the Oxford Internet Institute to understand the best way of labeling this. So there's people calling for violence, of dehumanizing language, conspiracy theories, or with links to established far left or far right groups. There's not a right wing sort of issue. This is across the spectrum.

8:47Tom Cheshire:But we found a lot of extreme content being published. I think that's some of how X has changed and the incentives behind it. I think it's pretty clear from the data there is, the algorithm is boosting that right-wing content. I think one thing that experts in the field of academia or in the field of social computer science, one thing that's been surprising and interesting to them has been the lack of equivalency, they would say. So left-wing users getting a lot of right-wing content, but right-wing users not getting a lot of left-wing content.

9:14Kaitlin Tosh:Just to be absolutely clear, these nine accounts that we set up, They were not out there on X interacting with other accounts. This was information that was coming to them. Yes, exactly. One thing we did do in order to kind of establish them as having a political leaning, that was based on who these new accounts followed. So they all followed accounts and interacted with who they followed content at the beginning. And again, not meaning to doubt the methodology. And actually, I should say, if people want to check out the way in which we put this together, they can go to the website and see it there.

9:42Kaitlin Tosh:But didn't a load of lefties leave X when Elon Musk came on board? I mean, wouldn't that in and of itself account for the fact less left-wing content is being shared amongst individuals? That's a question that we've had at the forefront of our mind the whole time that we've been doing this investigation. And it's really hard to control for. It's really hard to kind of account for people who don't exist on the platform. You know, scientifically, data-wise, it's really hard to do. But we did do a little bit of digging and we were able to do a slight control. So we looked at some of the top politicians, British politicians, that showed up in our data set and that was across the political spectrum.

10:20So we had both right-wing and left-wing politicians in that comparison and looked at the amount that those politicians were tweeting at the time that we did the data pool and compared that to how much they were actually shown to our accounts.

10:32Kaitlin Tosh:I mean, you were looking very specifically at two politicians on different ends of the political spectrum. Kemi Badenoch, leader of the Conservatives and George Galloway, further to the left than practically anyone I can imagine. Yeah, exactly. I mean, so it's actually, Badenock was kind of average. Don't quote me on that. A little bit of politics, a little bit of politics there. By that, I mean, she was posting at about the same rate as she was sent to our accounts. Yes. So she's kind of representative of what you might expect to see. Galloway, on the other hand, was posting significantly more than the other politicians we were comparing to.

11:06Kaitlin Tosh:I follow him, I know he does, yeah. But I never see him and neither did our accounts. But on the opposite side of the scale in terms of somebody who has shown a lot to our accounts but not tweeting a lot is Rupert Lowe.

11:16Tom Cheshire:With Rupert Lowe as well. He's a former Reform MP who left after falling out with Farage and set up his own sort of right-wing party, further to the right of Nigel Farage. Elon Musk likes Rupert Lowe a lot and he's retweeted him a lot. And that's probably some of the – gets a boost from that. But again, that sort of engagement is sort of – it's beyond what you'd expect. And the reason we're talking about him is because he's such an outlier, isn't it? It wasn't let's go and look at what Rupert Lowe is doing. He's in the dating, like, who is that person getting a lot of engagement, although posting the same as anyone else.

11:44Tom Cheshire:And there he is. We got in touch with Rupert Lowe and he said, you know, mass deportations, which is one of his policy positions, they're popular and that's why it's doing well. But there's lots of people who talk about that on X and he goes well beyond that and ends up in a lot of people's timelines. Within all of this, there is the really, really hard, rigorous data. And then within that, you find these sort of illustrations, really, of how it's working as much as. so it wouldn't point to Rupert Lowe's account as the proof by itself but it shows how things do work within that larger data set.

12:12Kaitlin Tosh:This isn't just an outlier that these nine accounts that we have set up are representative of X as a whole.

12:19Tom Cheshire:I think it's in the volume of it. So it is nine accounts but you've got to remember they are seeing all these posts, all this content. They're running for weeks here. I think what's reassuring about this is the consistency and the checking that's been done after it. So we're using this AI tool, this LLM, to do the categorizations. But then we're going and checking those categorizations by hand on a proportional sample and ending up with, again, the academics that you've been speaking to saying, you know, this is extremely robust. It's data science, right? And I think maybe the word science, you know, it's closer to economics or something like that rather than physics.

12:55Tom Cheshire:But this is a huge, huge data set. We don't think anyone's ever done this sort of work this extensively. and we're pretty solid on those findings. Other academics have done very similar studies using very similar methodologies. And because it's so impossible to get the kind of data that we've got, it's hard to do on a large scale and the content is really significant. But most of the other studies that have been done academically don't look at the UK specifically. Our methodology is aligning with computer scientists and what they would do. But our focus is very much focused in the UK in a way that no other ones have in the past.

13:32Kaitlin Tosh:I suppose it boils down to this, though. I look at the evidence and look at what you guys have done. And I can see, yes, there is clear evidence that there is an imbalance in favour of right wing material. But I suppose, look, OK, take Donald Trump and Truth Social. Anyone who is on that platform knows what the content is likely to be like. Who is on X right now who doesn't know, A, who Elon Musk is, B, that Elon Musk owns the site and C, what Elon Musk's political ideology is? Well, now we know. But I think that's the issue is we didn't before. And I do think when Elon Musk tweets, it goes beyond Twitter.

14:02We write about it. Sky News is in headlines everywhere. So there is a world where not only what he tweets, but the conversations that are happening online are reflected in society.

14:12Tom Cheshire:And again, we're sort of describing a shift. I don't think there's a value judgment. Like Twitter is looking more right wing as a result. Is that the same as, you know, one example we've talked about is an oligarch, a billionaire buying a paper, putting their views out in the paper. that's something we recognise and people choose their paper. I think the difference with X is how Musk has presented it. He's talked about the word Mindvirus, but he's also said it wants to be a free speech platform. Some speech appears to be a bit more free than others. We might all decide, look, that's fine. Twitter can't, is a right-wing space, but I don't think it's advertised as such.

14:44Kaitlin Tosh:In a way in which Truth Social, frankly, is. Yeah, it's sort of... What's X been saying?

14:49Tom Cheshire:Well, Elon Musk has seen the story, which is nice of him taking interest. He slightly disagreed with the findings, saying he saw lots of annoyingly left-wing content in his timeline. But I'd argue our study is more rigorous than what he's seeing. We got in touch with X as well, and they replied, and they used to just send poor emojis to journalists asking questions. So I wonder if there's been a slight difference in approach. But they did engage with it, which is great. And, you know, we sent all our findings. We're not just sort of putting this out there. We went through in detail what we found, put these questions to them.