In Robert Jordan’s Wheel of Time novels, a generic-Tolkien-ripoff fantasy series I read in my teens, the magic-stuff of the universe was tainted, “like water with a thin slick of rancid oil floating on top”; male practitioners who used it had great power, but were slowly driven mad. It is how I feel about Twitter. I rely on it for work; it brings me a reach I could not have elsewhere. But sometimes I feel it is steadily making me crazy.

We spend a lot of time worrying about the impacts of social media: on our mental health and our political culture especially. The effect I feel it has on me doesn’t tell us very much, but what can we learn from research?

A new paper in the Proceedings of the National Academy of Sciences takes a look at the effect of social media on “filter bubbles”. Unusually, it didn’t rely on users remembering and self-reporting their use, but used software to log it, meaning its data is more reliable; also, it makes that data openly available so others can check it (although it is not preregistered, so be wary).

It finds that — perhaps contrary to expectations — people who use social media more are exposed to a wider range of news sites. Which given my both my own overuse of it and my publicly saying we should scaremonger less about technology is the sort of thing I want to believe. Also, two psychologists I know tell me that the authors of this study are good, careful and not prone to overstating their results.

But I’m sceptical, for two reasons. First, as the authors say, exposure to different outlets is not the same as exposure to different viewpoints. If I used to just read The Daily Telegraph and now I’m exposed to The Spectator, or The Guardian and now The New Statesman, my number of outlets has doubled but the views I’m exposed to probably still overlap quite a lot. The authors say that previous research suggests more outlets typically means “more-diverse overall news exposure”, but in the context of social media bubbles it may be different.

Second, in general, all research about social media is messy, because you’re dealing in noisy data, complex behaviour and difficult conditions — it’s hard to randomly make 10,000 people use Twitter for five years and 10,000 people use a placebo, for a start, so good long-term randomised control trials are out. And the effects are usually small: the study seems to find that about 1% of site visits by non-Twitter-users are to news outlets; for Twitter users, it’s more like 1.4%, to a slightly wider set of outlets.

And you can find other research that says less heartening things, if you want to be less heartened: for instance, that exposure to opposing viewpoints hardens, rather than reduces, political polarisation.

For what it’s worth: the messiness of the data and the frequently contradictory findings push me towards thinking that, whatever the effects of social media on our brains are, it’s probably quite small. It’s hard to do research into the effects of smoking, too, but those effects were so huge that the signal came clearly through the noise. We like to think that the media, or social media, is controlling us like a secret puppet master, but I suspect it’s much less interesting than that.

---

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

TomChivers

---