Does AI Make Kids Worse At Sitting With Confusion? A Teacher’s Observation
The scratch paper in front of Javi was crumpled into a waffle of half-erased numbers, a half-eaten bag of cheddar Goldfish spilling crumbs onto his notebook. It was 3:15, we were 15 minutes into after-school help for 7th grade pre-algebra, and he’d just hit a wall on a multi-step word problem about shirt and pants prices. Before I could even lean in to ask what he was thinking, his hand was in his hoodie pocket, pulling out his phone.
“Can I just ask ChatGPT what the first step is?” he said, already unlocking the screen. I put my hand over the phone, not snatching it, just resting it there. “Give it two minutes first,” I said. “Just sit with not knowing, and write down what you do get. I’ll wait.” He huffed, dropped the phone back on the desk, and started twisting the drawstring of his gray hoodie so hard his knuckles turned white.
I’ve been teaching middle school math for 12 years, and I’ve noticed this shift get steeper every year since easy-access AI tools hit every pocket and laptop. Ten years ago, a kid stuck on a problem would slump, complain to their neighbor, stare out the window for three minutes, doodle a little, then circle back. They’d sit with the confusion, even if they hated it. Now, the second that uncomfortable itch of not knowing hits, they reach for a tool that will give them clarity before they have to feel that discomfort. That got me wondering, the same question a lot of parents have asked me at fall conferences lately: is AI making kids worse at sitting with confusion?
I don’t think AI is inherently bad, exactly. It’s that it removes the one part of learning that actually builds the problem-solving muscle we all need: the uncomfortable part where you don’t know what’s next. Last semester, I assigned an open-ended research project about local climate policy. Half the class ran their prompt through AI first to get an outline. When I asked them to adjust the outline to focus on a local issue that mattered to them, most of them just went back and asked AI to write a new outline. Only three groups actually sat and argued about what to cut, what to add, how to make an idea work that fit what they cared about. That back-and-forth, that confusion about how to shape a messy idea, that’s where learning sticks. If you skip that, you just have a paper someone else built for you, and you never learn how to work through a messy problem on your own.
I have a 10-year-old daughter at home, so I get to test small changes both in my classroom and our kitchen table. The first thing that’s worked for both my students and my kid is the 2-minute rule, which is so simple anyone can do it. Anytime a kid says they’re stuck and reaches for AI (or even just yells for you to fix it immediately), they have to wait two full minutes before they get an answer. In those two minutes, they just have to write down two things they do know about the problem, and one thing they don’t. That’s it. No pressure to solve it right then, just sit with the confusion and name what’s going on.
When I did that with Javi that day, he stared at the problem for a minute, then scribbled: 1. Total cost of both items is $120. 2. The shirt is twice as expensive as the pants. 1. I don’t know how to write that as an equation. After two minutes, I asked him what would happen if he called the price of the pants x. He thought for another 30 seconds, then wrote 2x + x = 120. He’d figured it out before we ever touched AI. After that, he said, “Wait, that was easier than I thought it would be.” He just needed to not panic about the confusion long enough to think through the next step.
The second concrete change that works is modeling your own confusion out loud. A lot of adults act like we should never be confused, especially in front of kids. If I’m helping my daughter bake cookies and the dough comes out way stiffer than the recipe says it should, I don’t immediately pull up AI to ask what’s wrong. I say out loud, “Hmm, this isn’t right. I don’t know what I messed up. Let me check what I did. Oh, I used butter instead of coconut oil, that’s probably why it’s stiffer. What if I add a teaspoon more milk?” That shows kids that confusion isn’t a failure, it’s just the step before you figure something out. It doesn’t need to be fixed right away.
I’m not saying kids should never use AI. It’s a great tool to check your work, or explain a concept after you’ve already tried to work through it. If Javi had sat for two minutes, written down what he knew, and still couldn’t get it, AI would have been a fine way to walk through the steps. The problem is when we use it to skip the confusion entirely. That muscle gets weaker the less you use it, just like any other.
This is still a work in progress, for me as much as for the kids. Last week I was stuck on a weird formatting issue with my online gradebook, and I found myself opening ChatGPT before I even checked if my laptop was connected to the wifi. I caught myself, closed it, spent two minutes checking the obvious things, and turns out that was all it was. Some days my daughter still grabs my phone before she’s even tried to figure out her homework. Some days Javi still asks for the answer before he’s written down what he knows. We’re all just practicing. The goal isn’t to ban AI. It’s just to make sure we don’t outsource the uncomfortable, useful part of learning that actually makes us better problem solvers. That part can’t be generated for you, anyway.
Please indicate: Thinking In Educating » Does AI Make Kids Worse At Sitting With Confusion