Why we should fear technophobia

For all the claims that reducing screen use is an easy way to improve sleep, it’s not the case.

“[People] go on and on about how tech is the place we can easily intervene to bolster sleep,” Andrew Przybylski, the author of the study, told me. “But if the difference between an eight-hours-screen-time-a-day teen and a zero-hours-a-day teen is less than 30 minutes’ sleep, then we’re barking up the wrong tree.”

There have been other studies. But they’re usually looking at 15 people, or 25, or some other tiny number, and have almost no statistical validity. And even then, they are usually over-hyped.

One 2015 study looked at the use of e-books before bed, and declared ominously that e-book users “took longer to fall asleep and had reduced evening sleepiness” than non-users. But when you look at the details, the study found that people who read an e-book for four hours immediately before bed for five nights in a row fell asleep, on average, 10 minutes later than people who read a normal printed book.

This is not the stuff of a societal plague. Russell Foster, a professor of sleep science at Oxford, told me that while the findings were statistically significant, they were “biologically meaningless”.

It’s all part of a wider problem with our attitudes to tech, and to social psychology research in general. Small links are blown out of all proportion, and extrapolated from mildly interesting scientific finding to major lifestyle or societal intervention.

The key thing to note here is that when scientists say some finding is “statistically significant”, they don’t mean that it’s important. They just mean that it’s probably not a fluke. If you give 1,000 randomly selected people a copy of À La Recherche du Temps Perdu and 1,000 people a copy of Jaws, and find that every single person who read Proust fell asleep faster, then you’ve got a statistically significant result: it’s very unlikely that would happen by chance. But if they only fall asleep one minute earlier, then you have a tiny “effect size”. It’s real, just not very important.

Finding a statistically significant result is interesting from a scientific point of view: it means that one phenomenon may be linked to another. (Or it may not. Because the bar of “significance” is set very low, and because there are simple tricks you can do to make your results look more significant than they are.)

But from the point of view of someone trying to navigate the world – say, to decide whether or not their child should use phones before bed, or play violent video games – it’s no use at all. What you want to know is the “effect size”.

Too often, though, we (the media, and scientific journals) treat statistical significance as important in its own right. A study last year found that social media use was linked to a rise in teen suicide and depression, which doesn’t sound great, and was widely reported. The findings appeared statistically significant, so people started talking about how to reduce social media use.

But quite apart from the fact that it couldn’t show causation (perhaps depressed teens spend more time on their phones), the effect size was tiny. As Amy Orben, a colleague of Przybylski’s at Oxford, wrote in a blog post, social media use “only explains 0.36% of the covariance for girls’ depressive symptoms”, and only 0.01% of boys’. Even if the effect is real, the impact on depression of banning social media would be negligible.

Relatedly, research often finds that people who have played violent video games are more likely to behave more aggressively in laboratory situations: more willing to blast people with uncomfortable levels of noise, or to give them painfully hot chilli sauce, two common proxies for real-world aggression. But this (possibly) statistically significant finding, that people are more likely to behave somewhat unpleasantly for a brief period in the lab after playing Call of Duty, is extrapolated to video games being to blame for real-world school shootings.

Away from tech, the much-hyped educational theory of ‘growth mindset’, where children who are praised for hard work rather than praised for cleverness do better in life, because they believe their abilities can grow rather than being fixed, has been declared a “revolution” and introduced to hundreds of UK schools. But, again, the statistically significant effect is tiny.

Media over-hype isn’t solely to blame. Often the scientists and journals, keen to make an impact with their research, will leap onto a statistically significant finding and shout it from the rooftops, not bothering to explain that the effect is miniscule.

It’s not that there’s nothing to worry about. Foster says that there is evidence that some teens’ excessive phone use is delaying sleep – not through blue light or circadian rhythms, but just because they’re staying up chatting and think the social gain is worth the loss of sleep. And it may be that while the average effect is small, a few people suffer badly: an “average” delayed sleep onset of eight minutes could mean nine people falling asleep as usual and one person staying up for an extra hour and 20 minutes. If you want to stop using your phone before bed, go right ahead and do it.

What we shouldn’t do, though, is engage in a sort of moral crusade against digital technologies, like a modern-day Temperance Society campaigning against the demon drink. I understand the temptation: the world really is changing fast. It’s barely two decades since mobile phones became commonplace and now everyone’s live an ever-increasing fraction of their lives online through them. Wariness is understandable.

But technophobic scaremongering, on the other hand, is making things worse. It is filling the world with false information, and making everyone a little bit stupider.

---

Tom Chivers is a science writer. His second book, How to Read Numbers, is out now.

TomChivers