Did AI Cause the Phones-Cause-Depopulation Hype Cycle?

First some AI humor, then some critique, then some data stuff.

One thing chatbots are great at is cooking up theories that combine random ideas. Throw a dart three times at the dartboard of memetic ideas, and ask a chatbot to put them together into a new theory — works every time! Here’s an example:

First I asked Google for three popular contemporary social theories, and got critical theory, symbolic interactionism, and structuration theory. So:

Prompt: Write a 150-word abstract for a sociology paper using critical theory, symbolic interactionism, and structuration theory to offer a new explanation for the rise of AI.

ChatGPT: This paper offers a sociological explanation for the rise of artificial intelligence by synthesizing critical theory, symbolic interactionism, and structuration theory. Against accounts that treat AI as an autonomous technological breakthrough, it argues that AI’s ascent reflects the convergence of capitalist imperatives, everyday meaning-making, and recursively reproduced institutional practices. From a critical-theory perspective, AI advances as a mechanism for extracting value, disciplining labor, and legitimating inequality through claims of neutrality and efficiency. Symbolic interactionism shows how users, developers, managers, and publics collectively define AI as intelligent, inevitable, useful, or threatening through situated interactions and cultural narratives. Structuration theory links these levels by showing how AI systems both draw on and reorganize rules and resources embedded in workplaces, markets, states, and platforms. The paper proposes that AI rises because actors repeatedly enact it as necessary while institutions materialize those enactments into durable infrastructures of power, dependence, and social reproduction.

Look for it coming to a mid-level sociology journal near you. But more to the point for this newsletter, try this one:

Prompt: Write a 150-word summary of a magazine article arguing that rampant AI use and viral media behavior caused the rise of the theory that mobile phones are driving the global fertility decline.

ChatGPT: The article argues that the theory linking mobile phones to global fertility decline spread less because of strong evidence than because of today’s AI-amplified media environment. It suggests that declining birthrates created a demand for simple explanations, and the mobile-phone hypothesis offered an appealing, concrete culprit. Viral posts, podcasts, and algorithmic feeds then helped transform a speculative idea into a widely discussed “theory.” AI tools intensified this process by generating summaries, charts, headlines, and pseudo-scientific explanations that made the claim appear more established than it was. The article emphasizes that correlation was repeatedly mistaken for causation: mobile-phone adoption and fertility decline occurred in overlapping periods, but many social, economic, and cultural factors also shaped reproductive behavior. Ultimately, it portrays the phone-fertility panic as a case study in how weak evidence can gain authority when packaged by influencers, recycled by AI, and rewarded by attention-driven platforms.

I wrote a short thread about this (paywalled) viral Financial Times piece by John Burn-Murdoch. The article offered supporting comments from Melissa Kearney (who called the theory “plausible”) and Lyman Stone (doing whatever he does). But it didn’t pass the smell test for me.

First, the article revealed its own hype process at work, as the the key figure is titled, “Could digital media by affecting birth rates?” but the story is headlined, “Why birth rates are falling everywhere all at once.” Obviously, the speculative data noodling was not sufficient. This is how one-big-cause / one-little-trick data journalism works. (Here is one breakdown of this figure, alleging egregious cherry-picking, with a replication package.)

One common problem, which is hard for a lot of people to grasp — and for which Malcolm Gladwell is partly to blame — is thinking that by identifying the moment of a change in a trend you can identify its cause. In real life, it is likely that whatever happened when you took the first step on the stairs did not also cause all the subsequent steps. It’s just one clue.

Once upon a time I spent way too much time trying to stop Pulitzer-winning New York Times reporter Matt Richtel from selling books with the preposterous claim that texting-while-driving was killing 11 US teens per day. A number that, incidentally, was greater than the number of all the teens killed in car accidents. I eventually got them to remove this ridiculously false claim from the website for the book, but the damage was done. Even as car accident mortality was falling, many people in journalism and the public continued to believe that texting-while-driving was fueling a meteoric rise in auto fatalities. (Of course, our roads are very dangerous and we should stop using them so much.) It was a good preview for the current social media / phone hype cycle.

But back to the fertility trend, there is another common problem. Behind much of the discourse on birth rates runs the assumption that because falling birth rates are terrible, whatever is causing them is also bad. But sometimes good things have tricky consequences. Falling birth rates of course are mostly great, signaling — throughout modern history — healthy social progress, with profoundly positive causes and consequences. If some people are now concerned that the trend is going to far, that does not mean its causes are the bad things we’re worried about: unhappy people, housing costs, teenagers not having sex enough, and so on. There are real problems for some segments of our wealthy-nation populations, which are part of the story of falling birth rates, but the good causes are still operating: youth autonomy, improved opportunities for women, increasing education, the freedom to be single and childlessness.

One consequence of the current population panic is the disappearance of climate from the list of things a lot of people are even thinking about — a disappearance that happens to serve the interests of the AI companies. It’s one thing to exaggerate the importance of your favorite topic, but quite another to produce such absurdism as this: “nothing will shift the future of human history more than AI and babies.” This is not an accident, of course, as the AI promoters want us to argue back and forth about AI all day, all the while agreeing to agree that their product is the most important thing in human history — and it’s definitely not exacerbating climate change, which it will probably solve in its spare time if we just don’t overregulate it.

Whatever the importance of AI, however, if you are completely sure that 200 years from now we will mark the birth rates of 2026 as the key event in history rather than, say, today’s climate policy, you may be high on your own supply. (I have no idea what anyone will think in 200 years, if anyone thinks anything at that point, and neither do you.)

Black and White fertility

I did not care for the Derek Thompson podcast episode with guest economist Jesús Fernández-Villaverde (though I didn’t make it to the end), but I learned from him that the total fertility rate for Black women in the U.S. is now lower than that for White women (both non-Hispanic). We were heading that way for a while — consistent with fertility declining only among people who aren’t married — but didn’t know it happened already. It seems like an important historical milestone, presumably unprecedented.

CDC reports the general fertility rate (all births divided by female population age 15-44) through 2025. This shows the Black-White crossover occurring in 2024. The White GFR actually increased in 2025, from 50.8 to 51.8, something I haven’t seen reported yet.

The official total fertility rates are only released through 2023, but the rapid response quarterly provisional estimates are up through the end of 2025. If you sum the age-specific fertility rates there, and treat the average of the quarterly estimates as an annual rate, you can make annual total fertility rates. I’m not sure why these don’t quite line up in the year they overlap with the earlier official rates, but I graphed them separately so you can see the difference.

Here are the Black and White TFRs from 2016 through 2025. They were the same in 2024, and then Black rates fell in 2025, while the White rates actually rebounded by .01. The 2025 estimates are 1.91 for Hispanic, 1.54 for White, and 1.47 for Black women.

That’s remarkable. And yet: I don’t know if what Tressie McMillan Cottom said is literally true — “No one is encouraging Black women to have more babies” — but it’s close enough.

Remembering the catastrophic abortion ban in Romania in 1967, which briefly doubled birth rates (even as maternal mortality and child abandonment exploded). So I made a figure.

And thinking about pronatalist policy. This is not an analysis of pronatal policy effects, of which there are many, or a causal claim; it’s just an observation. For the years 2015-2019, the UN in 2021 reported that 52 countries had policies intended to raise fertility rates, and 69 were trying to lower fertility; 19 wanted to maintain current rates, and 52 specified no national policy. As it happens, in the subsequent decade total fertility rates fell fastest in the pronatal policy countries.