Why Johnny can’t think, AI and the unmaking of American education

By Massoud Amin

MinnPost guest commentary through its Community Voices posts
Johnny is a sophomore at a good American university. Last night he finished a problem set in 11 minutes. It should have taken three hours. He typed the problems into an AI, copied the answers, and went to bed. He will get an A on the problem set. He will get a B on the exam. And in three years, when an employer asks him to solve a problem no machine has seen before, he will not know how. Nobody taught him. Everybody assumed somebody else would.
I taught from 1982 to 2023 — at the University of Massachusetts, at Washington University in St. Louis, and for over two decades at the University of Minnesota. I worked on artificial intelligence when it was unfashionable, teaching neural networks in 1992 to students who asked why they should bother. To be clear, I am not afraid of this technology. I helped build its ancestors. I am afraid of what we are doing with it, which is nothing.
Here is what we know. Not what we suspect — what we know.
Researchers at the University of Pennsylvania’s Wharton School ran a randomized trial with nearly 1,000 high school students, published in PNAS. Students given unrestricted GPT-4 scored 48% better on practice problems. Then the researchers took the machine away and gave an exam. Those same students scored 17% worse than students who never touched it. They felt like they were learning. They were forgetting how.
And here is the part every school board should tattoo on its agenda: A second group used AI designed to coach with hints instead of handing over answers, and the harm vanished. The problem was never the machine. The issue was the design — and we are shipping the wrong design into every classroom in America....What he is losing is the capacity to judge what he produces — and judgment is built only one way: through supervised struggle, error and correction, one teacher and one student at a time.
The machine can make the answer. Only education makes the person who knows whether the answer is right. We still know how to make that person. The question is whether we will bother.
Massoud Amin is professor emeritus at the University of Minnesota, where he held the Honeywell/H.W. Sweatt Chair in Technological Leadership, and a frequent contributor to MinnPost Voices.

To read more, subscribe to the Swift County Monitor News.