Tom Chivers

Tom Chivers is a science writer. His first book, The AI Does Not Hate You is out now.


Does vaping lead kids to marijuana? Does marijuana lead kids to harder drugs? Do video games lead kids to real-world violence? Does YouTube lead kids to extremism? Does anti-feminism lead to far-right beliefs?

These are serious questions that are regularly posed in the media. It was reported recently that young people who “vape” are more likely than their peers to use marijuana, according to a meta-analysis in the journal JAMA Pediatrics. The author of the study thinks it’s a causal link – i.e. that vaping causes young people to take up smoking weed. “It’s hard for me to find a better explanation,” he says.

Similarly, the UK government still claims that cannabis use leads to other drugs; Donald Trump says that video games lead to violence; a recent enormously viral twitter thread said that vloggers are driving white supremacy; and a recent Atlantic article said that anti-feminist views are the ‘gateway’ to the far right.

That last report contains an important word: all of these are examples of ‘gateway’ hypotheses. Cannabis-as-gateway-to-harder-drugs is just the most famous version of it. It’s superficially appealing as a theory, because – empirically – young people who take cocaine and heroin usually smoked weed first and went on to the harder drugs later. In all the other cases, the same is true: people who smoke weed have often vaped previously; school shooters have usually played video games; people who end up on 8Chan spreading QAnon conspiracy theories have usually been on YouTube before.

But there is an important logical misstep here, which President Bartlet called “post hoc ergo propter hoc” reasoning: happened after it, therefore happened because of it. When two things tend to happen together, there are four possible explanations. One is that it’s just a coincidence; another is that X causes Y; a third is that Y causes X; and the fourth that both X and Y are caused by some third factor, Z.

The classic example of the ‘third factor’ is that murders tend to go up at the same time as ice cream sales. It’s not that murder is hungry work, or that ice cream drives people insane with rage, but that both murders and ice cream sales tend to go up in hot weather.

There’s an obvious parallel here with the vaping/cannabis/hard drugs versions of “gateway” theory. Teenagers are naturally more prone than the rest of us to seek out new experiences and engage in risky behaviour, and some teens are more prone than others. The sort of kid who is more likely to try vaping is also the sort of kid who is more likely to try cannabis; the sort of kid who is more likely to try cannabis is also the sort of kid who is more likely to try heroin. I’d imagine that kids who vape are also more likely to drink alcohol, and (a bit more of a guess here, but I’m still confident) to drive above the speed limit or otherwise break the law.

But, of course, e-cigarettes are easier to get hold of than weed, and weed is easier to get hold of than cocaine, so if you just look at the sequence of events, you see ‘kids who take coke have previously smoked weed’ and ‘kids who smoke weed have previously vaped’. And if you’re not thinking about the third-factor problem, then you will be tempted into thinking that the one is caused by the other. Post hoc, ergo propter hoc.

I’d bet that this is what’s going on a lot of the time with the ‘gateway to the far right’ stuff as well. The sort of person who goes on anonymous messageboards talking about Bill Clinton running a paedophile ring is probably the sort of person who also thinks feminism has gone too far. It may be that they hang around in anti-feminist circles first, before they end up in the nasty corners of the web, but that’s because – like cannabis and vaping – ‘anti-feminist views’ are actually very common, even among people who agree with the stated aims of feminism. If you look back through the timelines of people who end up nodding along to alt-right videos, you’ll probably find them saying disobliging things about feminists, just as you’ll tend to find that most heroin addicts probably smoked weed when they were younger. 

Of course, it may be that there is some causal effect. But working out causal relationships is really hard. Studies of the real-world population can only show a correlation: people who do X also tend to do Y. The only really solid way of showing causation – that X causes Y – is to do a randomised controlled trial. But it’s ethically questionable to get 1,000 randomly selected children to vape nicotine and another 1,000 to vape a placebo, in order to see if it makes them more likely to smoke weed. So those studies don’t get done. You can try to tease out causality through longitudinal studies, but it’s very difficult and extremely uncertain.

The trouble is that we are extremely prone to seeing these causal relationships, especially where they make intuitive sense to us. If you’re an instinctively anti-smoking person (which I am, for the record), then it must be psychologically tempting to stick vaping – which seems so similar to smoking – in the same mental bin. So you see correlations between vaping and Some Bad Thing and you automatically believe they’re linked, because of motivated reasoning: instead of asking ‘must I believe this’, you ask ‘can I believe it?’

Similarly with the other stuff. Lots of us don’t like the sort of person who grumbles about ‘the feminists’ online, so it is tempting and easy to blame them for the rise (if there is one) of the far right. Lots of us are instinctively wary about our kids spending their time on YouTube, so it becomes easy to see it as a gateway to the far Right. 

I can’t rule it out, of course. Lots of serious people take it seriously – the New York Times has a big piece out arguing that YouTube “radicalised Brazil”, and the Atlantic piece I mentioned is deeply reported. Maybe they’re right – I can’t prove the relationship isn’t causal any more than I can prove that it is.

But I think we need to be wary. Most of the time, these things are not true: Dungeons & Dragons did not turn out to be a gateway to Satanism. And in the case of vaping, and video games, getting our causal story wrong will have serious consequences: vaping could save millions of lives if it helps smokers to quit, so if we scare people off it, we’d be killing people; blaming video games for violence rather than guns or cultural factors won’t stop the violence if the games aren’t causing it. And, let’s be honest, a large part of the reason that the far Right is more visible now is nothing to do with social media: it’s the bog-standard traditional media. If the newspapers and TV can shift the blame onto YouTube, they will do; let’s be careful about letting them.