On the ‘overwhelming evidence’ for social media destroying kids’ brains

The year is 2017 2023. An eminent American psychology professor has written a lengthy think-piece about how modern technology is single-handedly destroying teens’ mental health. Soon, respected publications are reporting on this expert’s opinion, and why shouldn’t they? After all, the scientific community has been clamouring for journalists to listen to experts for years. 

Jonathan Haidt’s recent blog (as with Jean Twenge’s 2017 article in The Atlantic) paints social media as a compelling arch-nemesis in the fight against mental ill-health. In 2020, Haidt appeared in the Netflix documentary The Social Dilemma making similar arguments. The influence of statements like these has been pervasive, and the mind of the general public has long been made up. 

There’s just one problem: they’re wrong. 

I am not saying that social media cannot cause harm to teenagers. However, the claim that there is any single, straightforward consensus on the issue is simply untrue. A 2022 review of reviews demonstrated that most scholars in the field see the evidence as ‘weak’ and/or ‘inconsistent’, with a few outliers (including some familiar faces) concluding that there is a substantial deleterious relationship between social media use and mental health. 

The fact that has the views of most experts on social media have been overlooked to make these points has not stopped journalists from picking up on his post. I was disappointed to see that John Burn-Murdoch (of the Financial Times) has been uncritically parroting Haidt’s talking points. It speaks to the influence that certain senior academics have, and the seeming immunity their work has to critical appraisal. 

I have, for better or for worse, read Haidt’s post in full more than once. I don’t disagree with all of it. However, without taking it apart exhaustively, I would like to point out some of the reasons why I’m not convinced by his assertion that the scepticism that was “justified in 2019” is not also justified in 2023. 

The post begins with a consideration of correlational studies that assess the relationship between social media and mental health, drawing a conclusion that departs from most other reviews of the literature. One particular thorn in his side has been Orben & Przybylski’s (2019) paper suggesting a minimal relationship between screen media use and wellbeing. To refute it, Haidt relies on his (and colleagues’) 2022 replication of their specification curve analysis.

The argument goes something like this:

‘when we limited the analysis to social media use in girls, we got a median regression coefficient of –0.20, which is around the same as the value for binge drinking’

What he carefully avoids saying is this figure is obtained by ignoring some of the covariates, specifically ‘potential mediators’ of the social media–wellbeing relationship (e.g. such variables as closeness with parents, parental distress and negative attitudes towards school). The figure obtained from not doing this is –0.01, which is actually more positive than Orben & Przybylski’s headline figure of –0.032.

The authors’ questionable interpretation is that they showed a larger relationship “particularly when we removed possible mediators”, as opposed to only when they removed them. They also do not offer any empirical justification for removing these variables, such as evidence that closeness with parents or negative attitudes towards school actually mediate the relationship between social media use and wellbeing. 

The post considers longitudinal evidence only briefly. Because of this, I took it upon myself to delve into the 293-page(!) Google Doc which Haidt refers to frequently in his writing. Reading the comments that the authors offer under certain studies is particularly interesting. Several of the studies that show no significant relationship between social media use and mental health have disparaging comments added underneath, which would seek to discredit them.

These are generally despite

• similar flaws in those studies that show an effect and/or

• contradictory arguments that Haidt relies on in the blog post. For example, one study is criticised for having a 2-year spacing between waves of data collection: 

However, in the post, he asserts that those studies with longer time intervals between dose and response are more important: 

The longitudinal well being sufficiently poisoned, we move on to section 3: experiments. I note here that the ‘social media as poison’ hypothesis seems to be back in vogue: 

Also worth noting is that nearly all of the experiments are in adults, which I would say probably limits their utility in demonstrating the titular claim that social media are bad for teen girls. This does not appear to trouble him in the slightest. 

When we take publication bias into account (and even without doing so), the evidence hardly qualifies as ‘overwhelming’. And that probably explains why most academics in the field have long accepted that teens’ relationships with social media are far more complicated than can be explained in a single headline in the Financial Times. But why, then, are their views not heard in mainstream media publications? What allows a couple of individuals’ opinions to dominate the conversation, in spite of abundant evidence to the contrary? 

I have two possible explanations to offer.

Firstly, highly-cited academics are obvious targets for media attention, even if those citations were of publications not necessarily relevant to the topic at hand. But this means there are a number of individuals who have been working on social media for longer, but don’t receive the same attention in the press. And secondly, social media platforms are an easy scapegoat that distracts from much more difficult conversations about the social, economic and political factors behind apparent rises in mental health difficulties. As far as Haidt is concerned, there are only two important candidates: the internet and content warnings. 

It is easy to feel like the battle has been lost, that we have made no progress in the last six years and that society will be doomed to panic about new technologies forever.

But in recent years, a number of teams have been hard at work producing robust evidence around social media that goes on to inform policymakers and other stakeholders. Now, we need journalists and social media users to stop amplifying oversimplified accounts, so that parents and carers don’t resort to drastic measures such as banning their children from most or all social media use. After all, we don’t know whether doing so might do more harm than good, by isolating teens from their peers in a time where nearly all of them are online.  

In time, we will gain a better understanding of mental health in the digital sphere. With that understanding has come, and will come, a better-informed conversation about how to keep teenagers safe online. But in the meantime we must stop putting the cart before the horse. The hugely complex nature of social media interactions simply does not boil down to ‘social media use is dangerous’. 

 Declarations: 

Competing interests 

• I was employed as a research assistant in the lab of Dr Amy Orben between July and October 2021. Neither she nor any other current or former lab member had any influence over the content of this post. 

Funding & affiliation 

• I receive funding from the Wellcome Trust and am affiliated with University College London, but the views expressed here are my own and do not reflect those of either of those organisations. 

Acknowledgements 

Thank you to Humma Andleeb for her helpful comments on an earlier draft of this post.