In a recent written testimony to the U.S. Senate, neuroscientist and educator Jared Cooney Horvath warned that children’s cognitive development in many domains has “stalled” and in some areas “reversed,” alongside declines in literacy, numeracy, attention, and higher-order reasoning (Horvath, 2026). In a widely shared clip from that conversation, he sharpens the warning: in schools, he argues, it doesn’t matter what kind of screen it is—screen-based technologies still hurt learning (YouTube, 2026).
That claim resonates because it fits what many educators feel: students’ attention is fragile, reading stamina is shrinking, writing is increasingly outsourced, and classrooms are turning into constant negotiations with devices. These are real physiological costs that we must manage. But we must distinguish between the ‘health of the body’ and the ‘capacity of the mind.’ A mind that no longer memorizes is not a broken mind; it is a mind being redirected.
I grew up in a Chinese village in the 1960s and 1970s. At that time, farming wasn’t “one job option.” It was life. Today, farming is far less central to most people’s livelihoods. Fewer people have deep farming skills not because humans became “less capable,” but because the economy changed, tools changed, and people adapted. They had to develop new skills.
And here’s the historical reminder: this isn’t the first time humans feared that a new technology would damage the mind. In Phaedrus, Plato has Socrates warn that writing would create forgetfulness and offer only the appearance of wisdom, because people would rely on external marks instead of exercising memory and understanding (Plato, trans. 1997). That is basically an ancient version of today’s complaint: “This tool is making us mentally weaker.”
So when we hear claims about declining intelligence or declining test scores, the right first question is not “Are humans getting dumber?” Instead, it is are we measuring tool-using humans with yesterday’s yardstick? We are judging a pilot’s intelligence based on their ability to ride a horse. They may have lost the ‘skill’ of the saddle, but they have gained the ‘intelligence’ of the cockpit. If our test only looks at the saddle, the pilot looks dumber.
There is credible evidence that some standardized indicators have fallen or stopped rising. For example, research on the “reverse Flynn effect” in Norway suggests that the rise and later decline in intelligence test scores across cohorts are largely explained by environmental causes rather than genetics (Bratsberg & Rogeberg, 2018). International achievement trends also show real setbacks. OECD reporting on PISA 2022 describes “unprecedented” drops in average mathematics and reading performance across OECD countries compared with 2018 (OECD, 2023).
Horvath and others interpret these patterns as evidence that modern technologies—especially screens—are contributing to cognitive decline (Horvath, 2026). Sometimes that might be true in specific contexts, especially when devices replace sleep, play, relationships, deep reading, and sustained practice.
But the leap from “some old measures are down” to “humans are declining” is not logically inevitable. It ignores something fundamental: humans have always used tools to extend their minds. Writing reduced the need to memorize everything—but it expanded what humans could preserve, share, and build across generations. Calculators reduced routine computation—but they enabled more complex modeling. Search engines reduced recall demands—but they expanded access to knowledge. Tools don’t simply “weaken thinking.” They shift cognition.
When powerful tools spread, two things happen at the same time:
- Some skills become less practiced because the tool makes them less necessary.
- New skills become essential because the tool changes what is possible—and what is valuable.
If we only measure #1, we can tell a convincing story of “decline.” But it may really be a story of adaptation—just like the fading of farming skills when the economy no longer demanded them from everyone.
The real problem is that we don’t measure #2 because it is not in the school’s definition of “learning.” Many of the things schools have traditionally treated as evidence of intelligence—producing essays, summarizing readings, doing routine problem sets, generating polished presentations—are exactly the kinds of tasks AI can now do quickly and well.
In order to measure #2, we must redefine learning in the new age so as to understand what #2 is. Horvath’s warning matters because it forces us to stop assuming that “digital” automatically means “better” (Horvath, 2026). But if our response is simply “less tech,” we risk fighting yesterday’s battle. The deeper question is curriculum-level: What is worth learning in an age when tools can do much of what school has traditionally required?
In Fix the Past or Invent the Future, I argue that we can’t keep trying to “fix” a one-size-fits-all model designed for another era—we need to invent schooling that fits the world students actually inhabit (Zhao, 2025). AI can solve problems, but it cannot decide which problems are worth solving.
When I was a child, farming skills were intelligence in context. Today, for many people, they are not. Humans didn’t become less intelligent. They became differently skilled because the world changed.
Socrates feared writing would weaken memory. He wasn’t crazy—he was noticing that tools change minds. But he missed the larger point: tools also expand what minds can do.
AI is changing the world again. If some traditional scores decline, we should not automatically conclude that humans are “getting dumber.” We should ask whether schooling is still preparing students for the world that exists—or for the world that used to exist.
The real danger isn’t that students will use tools. The real danger is that schools will keep teaching as if tools don’t exist.
References (APA 7)
Bratsberg, B., & Rogeberg, O. (2018). Flynn effect and its reversal are both environmentally caused. Proceedings of the National Academy of Sciences, 115(26), 6674–6678. https://doi.org/10.1073/pnas.1718793115
Horvath, J. C. (2026, January 15). Written testimony of Dr. Jared Cooney Horvath, PhD, MEd, before the U.S. Senate Committee on Commerce, Science, and Transportation [Testimony]. U.S. Senate Committee on Commerce, Science, and Transportation. https://www.commerce.senate.gov/services/files/A19DF2E8-3C69-4193-A676-430CF0C83DC2
Organisation for Economic Co-operation and Development. (2023). PISA 2022 results (Volume I): The state of learning and equity in education. OECD Publishing. https://doi.org/10.1787/53f23881-en
Plato. (1997). Phaedrus (A. Nehamas & P. Woodruff, Trans.). Hackett Publishing Company. (Original work published ca. 4th century BCE)
YouTube. (2026). Doctor on how screen time hurts kids’ cognitive development [Video]. https://www.youtube.com/watch?v=Fd-_VDYit3U
Learning in the Age of AI 