Opinion: Hold off on AI tutor for kids

Matthew Stublefield

In an October Springfield Business Journal Opinion piece, Andy Drennen asked us to imagine 2036. A household assistant gently wakes the kids for school, and “every student now has a personalized AI tutor that understands exactly what they know, where they struggle and how they learn best.” His warning was that we’ll underestimate how much artificial intelligence could reshape our everyday lives, and on that point I think he’s right.

I use AI all day. In this column back in May, I wrote that it lets me work at roughly eight times the pace I managed in 2024. Research that used to take me weeks now takes an afternoon. I’m also a geriatric millennial raising three boys – 8, 6, and 4 – and none of them will have a device of their own until they’re about 16. So consider this a counterpoint about the kids, from someone who isn’t an AI skeptic.

Last December, I argued here that young people need AI skills to get hired, and I still believe that about high schoolers. The City of Springfield has applied for an $8 million federal grant for an Ozarks AI Workforce Collaborative that would prepare 300 middle and high school educators to bring AI into their classrooms, and teenagers headed into the workforce need that. But a 6-year-old is in a different place developmentally, and so is a 10-year-old.

That 10-year-old is the subject of  an essay published this month by the University of California, Berkeley’s Greater Good Science Center. Tom Wooldridge, a psychoanalyst, describes his son coming to him with a question about black holes after talking it through with an AI model called Claude. The explanations were good, but the speed bothered him. “When the uncertainty is of a sort that once had to be lived with for a while,” he wrote, “my son reaches for an answer sooner than he used to.” He calls what’s being lost “productive friction.”

Let them wait
Productive friction is why the rules at our house are firm. Kids that age can benefit from sitting with uncertainty, going to the library and reading longer books, because the focus that takes, and the work of parsing and synthesizing what they’ve read, is developmentally important.

An AI’s synopsis skips all of it.

Our family homeschools using the Charlotte Mason approach, and by the time our boys graduate, they’ll have read more than 800 books.

Families can practice this in everyday ways: When your kid asks you something and you don’t know the answer, don’t immediately look it up on your phone. Let them spitball and guess wild things that probably aren’t true. Admit you don’t know, show them it’s OK not to know, and then go learn the answer together.

Leaders go first
As parents, we’re leaders. At work, we’re leaders, too. The kids growing up with these tools are the people we’ll be hiring, and they’re learning how to handle “not knowing” from us right now. We owe the people who come through our doors the chance to build foundational skills, and two of those skills deserve more attention than they get. One is admitting when you don’t know something. The other is understanding that experience and judgment don’t come from getting a quick answer from a large language model.

Drennen pictures department leaders in 2036 who “oversee teams of AI agents rather than employees.” Those leaders will need the judgment to know when an agent is wrong, and that kind of judgment gets built slowly, by being new, not knowing and working through the problem.

Springfield Public Schools Superintendent Grenita Lathan asked local business leaders here in August to provide internships and mentorships. That’s where we model this. It may be tempting to expect instant knowledge from an intern or a new hire, as if they were an AI that should already know everything. But that isn’t fair to a person. We can exercise patience instead, admit when we don’t know, and show them how we’d go about learning it.

It matters for our own thinking, too, because AI models can reflect biases in the data they’re trained on. When we were growing up, we were taught to think critically, and the strategist Alicia McKay argues that we now also need to be power literate. In  my May SBJ Opinion column, I applied her idea to AI: When AI gives you an answer, ask what it was trained on and whose worldview got baked in.

Like how best to raise our kids, that’s a question with no answer quickly forthcoming. But the friction of thinking about it is certainly productive for all of us.

Matthew Stublefield is founder and principal consultant at Fieldway LLC and a member of the Springfield Tech Council advocacy committee. He can be reached at matthew@fieldway.org.