Review of "The Anxious Generation"

Extremely over-exaggerated claims. Boys are fine, girls less so

A few days ago, Jonathan Haidt’s The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness released. Its central thesis is that social media is causing teenage mental health to fall. Because of this, Haidt wants 4 policies:

These are strong claims and strong policies. It’s important that the claims be correct beyond a reasonable doubt, because otherwise the policies could backfire and make teen mental health worse. What if teens rely on phones to form better friendships over the internet or to enhance their local friendships? Then these policies would make teen mental health worse, unless there is robust causal evidence that phones are a main cause in declining teen mental health.

Teens should also have civil liberties, including the right to use the internet to communicate with who they please, starting some years before 18. As I discussed in my first book, the brain finishes developing between 14 and 16. I appreciate that Haidt’s proposals fade out in this age range, instead of at 18 or “after high school”, as some would propose unironically.

Considering these points, I was shocked when I found that Haidt’s book doesn’t even contain a correlation coefficient or scatterplot, much less a causal estimate of the effect of social media on teen mental health.

This alone gets the book an F from me.

But it gets worse. The book is a whole 400 pages. 150 pages would be a lot for a book with no relevant data (it only contains trend lines showing that mental health is getting worse). But 400 pages? What does he do in all of this time? His central case, which is F tier, is contained in chapter 1. The rest of the book is teen brain neuroscience. The first chapter is “The Anxious Generation.” The rest tries to substantiate “The Great Rewiring.”

In my first book, I show that teen brain neuroscience is generally not a credible field. It’s a field wrought with pseudo-experts whose claims in the media fail to match their data, replication issues, thanks especially to low statistical expertise and small sample sizes, and even fraud, whereby experts allude to studies that don’t exist or egregiously mischaracterize ones that do. Haidt fails by taking the exoteric claims of the field at face value. Instead of diving into the data himself and discovering that the field doesn’t even have the methods to prove something like “phones are rewiring teen brains”, he uncritically cites the opinions of “experts” covered in my book like Laurence Steinberg. Steinberg’s opinions are first and foremost political and hardly match the data he claims to cite. For example, in one study, Steinberg’s data shows that 13 year old white boys are more mature than 30 year old black men. Steinberg ignores the race differences in the study and speaks in the media about how dumb and immature and undeserving of liberties white 13 year olds are. To me, this man has no credibility.

Yet here he’s cited. The claim that teen brains are vulnerable is unsubstantiated and based on rodent studies involving extremely young rats which are not equivalent to teens in age or species. Hilariously, this citation is directly covered in my book — that’s how thorough it was. The citation is to Steinberg’s textbook, which in turn cites this opinion paper. The opinion paper has Sarah-Jayne Blakemore on it, giving the same argument I discuss here:

Elsewhere in the book I also cover claims of teen brains being sensitive to alcohol, which are completely based on rat studies. Human studies fail to find any effect of moderate drinking on teen brains. The rat studies were pushed by an activist group to prevent teen drinking. Basically, if you want to understand why teen brain neuroscience is fake, read my book in full, and then you will start to see the same old claims I debunked ages ago sprout up in pop sci like this, and you will be in-the-know that they are all false.

The Anxious Generation, then, is one part an inadequate case for there being a causal effect of social media on teen mental health, and one part pseudoscience, and this sums to 400 pages. The book is definitely deserving of an F.

The Sexist Truth behind The Anxious Generation

I didn’t want to stop at the book. I previously respected Haidt for his work in The Righteous Mind. This helped advance the important idea that political left and right are natural moral differences long before I came along. I thought that, surely, this book was F tier because it was dumbed down for the masses. Haidt must just not think that the masses will understand OLS tables, the kinds of tables he needs in order to actually substantiate his claim to other researchers and the statistically literate.

I found that Haidt has a Substack article reviewing the actual literature (not just trends and teen brain pseudoscience), and his conclusion in that article is pretty reasonable. I recommend to all of my readers to skip the book and to just read Haidt’s blog posts if you want to know the truth about this topic.

My takeaway from the literature review Haidt gives is that when sexes are merged, effects are hit or miss. Sometimes they show up, sometimes not. This is in part because there is a strong sex-media interaction effect.

Source

First, we have Haidt’s own cross-sectional analysis.

Source

Weirdly no p-values given, but boy results are basically nothing, even positive with controls, while girl results are strongly negative, and slightly negative with controls. With controls, this data is not impressive or enough to convince me of there being any real effect.

Haidt discusses how the experimental evidence is likewise mixed (about 66% positive, 33% null). There are a lot of null results, so it’s unclear how, after accounting for publication bias, the effect would actually look.

But there is some better roll-out data.

Source

No demonstrable effect for boys, but a small one for girls. Importantly, a 3 SD internet increase would predict a 81% increase in serious girl mental health incidents. Yet the observed increases are higher, 100% - 500% as discussed by the paper.

So, the evidence shows that the proliferation of social media may explain part of the decline of female mental health, but not male. Yet there is a smaller male decline, and an unexplained component in the female decline. What could cause this?

Our declining genome is also important

Source

Above shows estimates of the selection pressure on various traits, from polygenic scoring. Due to missing heritability, these are likely underestimates by a factor of 2 to 5.

There are also fixed-effects estimates of the paternal age effects on mental illnesses. These control for older parents having more or less mental illness, but not for parent→child environmental effects that change for individual parents with age. From twin study literature, those effects should be small for adult mental illness, and as such they only show up for variables like “education under 10 yrs.” In other words, these should be robust paternal age effects.

Source.

It’s hard to compute just how much of an increase in self harming, suicide, and diagnoses we should see due to these two dysgenic pressures (selection and mutational load), since some of these are binary variables and diagnostic criteria are always changing. But GPT says that the above hazard ratio increase for paternal age is equivalent to an r=.10 paternal age correlation with suicidalness, similar to the one measured for leftism and IQ. If this is true, it can potentially predict a 3500% increase depending on assumptions. So, I think dysgenics can explain all of the male increase in mental illness and a component of that size in females, at least.

Conclusion

The decline of teen mental health is probably not monocausal. For girls, it’s one part girl social media, one part dysgenics. For boys, it may be only dysgenics, or dysgenics and other factors, but the evidence it’s phones is basically nonexistent.

Instead of being a reasonable treatment of this topic, The Anxious Generation is an attention-hungry anti-male work that overpromises and fails to deliver empirically. There is no reason to restrict teen male phone usage after 13 or 14, and restricting girl usage would only reverse a fraction of the decline of their mental health.