Has AI Dependency Changed How Kids Solve Hard School Problems? A Teacher’s Observation
Last Wednesday, I stayed after school for my weekly 7th grade math extra help, the room smelled like floor wax and the cheese puffs a kid left under a desk, and Mia, a quiet 12-year-old who usually aces her quizzes, was staring at a percent discount problem that was stumping her: A $145 jacket has a 30% store discount, then an extra 10% off for customers who bring their own bag. What’s the final price? Before she even picked up her pencil to write down a number, she reached for her phone in her hoodie pocket. “Can I just check this?” she asked.
I said, “Wait two minutes. Just tell me what you’d try first, before you look it up.” She sighed, dropped her phone back, and slouched in her chair. “I don’t even know where to start,” she said. That’s when I really noticed the shift, after four years of generative AI being common in my classroom, I’ve been tracking how kids approach hard problems now versus before this tool was everywhere.
Five years ago, when a kid got stuck on this exact same problem, they’d at least take a guess. They’d write 145 minus 40%, cross it out, mumble that it’s a trick question, argue that the discounts should add together. They’d fumble through it, get it wrong, and that wrong step would lead to a conversation about why sequential discounts don’t stack that way. Now, half the time, kids don’t get that wrong step. They don’t even give themselves two minutes to fumble.
A month ago, I graded an 8th grade history assignment asking kids to explain one reason the Industrial Revolution started in Britain. Three of the responses were perfectly structured, cited obscure sources I’d never even assigned, and hit every point on the rubric. But when I pulled each kid aside and asked them to explain that point in their own words, none of them could. They’d just plugged the question into a tool, copied the answer, and called it done. They never engaged with the mess of figuring it out themselves.
It’s not all bad, of course. Javi, one of my students who has ADHD, uses AI as a hint tool. When he’s stuck, he asks it to give him just the first step, then he works the rest out on his own. That works for him; it takes the edge off the initial panic of being stuck. The problem isn’t AI itself, it’s the default approach to difficulty. It used to be that kids tried first, asked for help second. Now, the default is outsource first, try never, if they can get away with it.
I see this at home too; I have a 10th grader taking AP Biology this year. Last month, he was working on a lab report about enzyme activity, and he announced he was just going to have AI write the whole thing, since it was “just busy work.” I didn’t ban AI, but I told him one non-negotiable rule: before he opens the tool, he has to write down three things he already knows about his lab, and one rough draft of his conclusion. It took him 12 minutes, and when he was done, he said, “Wait, I already have half the report written anyway.” He ended up using AI to fix his run-on sentences and organize his messy notes, which was a good use of it.
From what I’ve tried in class and at home, small shifts work better than big bans or lectures. The most effective thing I’ve landed on is the 10-minute rule, which I share with parents all the time. It’s low-friction, no fancy tools needed: any time a kid reaches for AI to solve a problem for them, they have to spend 10 minutes working on it first, and write down at least three things they know, plus one possible first step. That’s it. No big speech about cheating, no taking phones away for good. It’s just a small pause that shifts the default back to “I try first.”
For parents helping with homework at home, I add one simple trick: when your kid says they’re stuck, don’t jump in to help, and don’t let them jump straight to AI. Just ask, “What’s one thing you know for sure about this problem?” That question gets them thinking instead of reaching for a shortcut. It’s also helpful to be clear with kids that homework isn’t just about getting the right answer to turn in. It’s okay if they get it wrong, as long as they tried to work through it. You can even pull up AI after they’re done to check the answer together, instead of letting AI do the work first.
Back to Mia that Wednesday. After I asked her to just try, she sat there for 8 minutes, twisted her hair, crumpled a scrap of paper, and finally wrote down: 30% off means I pay 70% of 145. Then she paused, and said, “Wait, the extra 10% is off that new number, right? Not the original?” I just nodded. She did the math, got the right answer, and smiled. “That wasn’t that hard,” she said. I asked her if she would have figured that out if she’d just looked it up on AI first. She shrugged and said, “Probably not. But why would I do it myself when it’s so fast?”
I don’t have a perfect solution. We’re all figuring this out as we go, and AI isn’t going away. The biggest change I’ve seen isn’t that kids are getting more answers wrong, or that they’re cheating more. It’s that they’re less willing to sit with the uncomfortable feeling of being stuck. That uncomfortable feeling is where problem solving grows. The 10 minute rule isn’t going to fix everything, but it’s a small start. It gets kids to practice that uncomfortable feeling once a day, one problem at a time. Some days it works, some days kids still reach for their phones before I can say anything. We just keep trying.
Please indicate: Thinking In Educating » Has AI Dependency Changed How Kids Solve Hard School Problems