Cognitive Stunting, Not Cognitive Decline

The right question is not whether AI lowers our children's intelligence, but whether it stops that intelligence from forming in the first place.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/cognitive-stunting-not-cognitive?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Ask whether AI is harming children’s minds and the conversation reaches almost at once for a familiar fear: that it is lowering their IQ. It is the wrong fear, and getting it wrong matters, because it sends parents and schools looking for the wrong thing. Decline is what happens to an adult who loses a capacity once held. A child has not yet built the capacity. The real question is whether AI is quietly preventing it from forming at all.

CONTEXT AND BACKGROUND

This distinction has a name. Researchers at the Brookings Institution have proposed measuring not cognitive decline but cognitive stunting, arguing that when children use AI to complete thinking tasks, they are not offloading skills they already possess; they are short-circuiting the process of building those skills in the first place. Their image is exact. A child cannot learn to ride a bicycle by having a machine outline the steps or pedal on their behalf, and cannot learn to think if AI routinely supplies the ideas, the text and the answers. The concern is no longer hypothetical in South Africa, where the Basic Education department has moved to limit young children’s screen exposure, citing links to delays in language and cognition.

INSIGHT AND ANALYSIS

What AI removes is the struggle, and the struggle is not a flaw in learning but its engine. Effort, error, reflection and slow progress are how a young mind encodes knowledge and builds the habit of reasoning. Researchers at the Harvard Graduate School of Education have found that when a chatbot is readily available, some students outsource their thinking rather than engage in the productive struggle known to be essential for learning. A polished essay produced in seconds looks like progress, but for a child, the essay was never the point. The point was the mind built in the writing of it.

There is suggestive evidence about the cost, though it must be read carefully. A widely-discussed MIT Media Lab study found that people using an AI assistant to write showed weaker neural connectivity and less sense of ownership over their work than those writing unaided. The finding is a signal, not a verdict: it studied adults, used a small sample, and other researchers have rightly flagged its limits and called for more cautious interpretation. A review of the adolescent evidence reaches a more measured conclusion, that overreliance may diminish executive functions during a crucial developmental window, while stressing the picture is genuinely mixed.

That mixed picture is the honest part of the story, and it should temper the alarm. The same technology can help. A 2026 kindergarten study found that a well-designed reading chatbot produced real gains in early decoding, with the largest improvements among the weakest pupils.

I have previously written about how easily we mistake a curated digital reality for the real thing, and how carefully we must watch the environments shaping young minds. AI is not the enemy of a child’s development. Used without thought, it can be the enemy of the effort that development requires.

IMPLICATIONS

The practical difference between stunting and decline is everything. If the worry were decline, the answer would be to use AI less and recover what was lost. Because the worry is stunting, the answer is to protect the conditions under which capacity forms in the first place. That means guarding the friction zones in a child’s day, the slow reading, the unaided problem, the boredom that precedes original thought, and treating these not as inefficiencies to be optimised away but as the work of growing up. It also means measuring the right thing. A school or a parent who grades only the finished product is rewarding the tool and missing the child. The harder task is to value the messy, invisible process of thinking, which is precisely the thing AI is most able to hide.

CLOSING TAKEAWAY

We should be neither dismissive nor hysterical. AI can support a child’s learning when an adult shapes how it is used, and can hollow it out when it becomes a machine for skipping the hard part. The danger is not that our children will become less intelligent than we are. It is that they may never build the intelligence they would otherwise have had, and that we will not notice, because the homework will look immaculate. The question for every parent, teacher and policymaker is not how to keep AI away from children, but how to keep the struggle in.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net