What Happens When AI Takes The Bottom Rungs Of The Learning Ladder: A Teacher’s Observation
The coffee in my mug has been cold for 45 minutes, and the goldfish crumbs on my desk are stuck to a leftover copy of last week’s 7th grade reading quiz when Javi knocks on my classroom door after school. He’s got his hood up, one shoe untied, and he’s been failing my writing assignments for three weeks, so I motion him in. He pulls the crumpled AI-generated paragraph he turned in two weeks prior out of his binder—he got an A on it, but he’s here because he still can’t tell me what a topic sentence is, let alone write one.
He says he just didn’t get the assignment, so he asked the tool to do it so he wouldn’t fail. I pull out a blank sheet of paper, write the prompt at the top: “Should our school extend middle school recess?” I ask him to write a topic sentence for it. He stares at the paper for five minutes, chews the end of his pencil, and says, “Anything I write is gonna sound dumb. Why can’t we just use AI to do the bad parts anyway?”
That line stuck with me. The “bad parts” he’s talking about? Those are the bottom rungs of the learning ladder. I started teaching 11 years ago, and back then, kids turned in topic sentences like “Recess is cool and we need more.” It was terrible. We’d mark it up, talk about how it doesn’t tell the reader what the paragraph will prove, and try again. That messy, low-stakes, bad work was how they got good. No one starts out writing perfect topic sentences. You have to write 20 bad ones first to build the muscle to write a good one.
I leaned across the desk and told Javi, “Don’t worry about it being good. What’s the first thing you think when I say longer recess? Just say it out loud, and write it down exactly like that.” He mumbled for a second, then said, “I’m tired all day from sitting in math, so longer recess would help us focus more in class.” I told him to write that. He did, it was a little run-on, a little casual, but that was a topic sentence. We crossed out one extra word, moved a comma, and suddenly it was his. He blinked and said, “That’s it? I thought it had to be fancy.”
This isn’t just one kid. I’ve had three other students come in for extra help this month who can’t outline their own essay, can’t work through the first two easy practice problems for algebra, can’t even come up with a bad first idea for a history project. All of them used AI to do the starting work for them, because why slog through the messy hard part when you can get a perfect result immediately? But the starting part, the wobbly first steps that no one sees and that almost everyone gets wrong at first? Those are the bottom rungs. You don’t need them to get a good grade on one assignment, but you do need them to build the skill to climb higher later.
I’ve adjusted my classroom rules to push back on this slowly, and I share three small, low-effort changes with parents at conferences that don’t require extra time or fancy tools. First, if your kid is learning a brand new skill, no AI until they have a full first attempt done by hand. It doesn’t have to be good, it just has to exist. If your 12-year-old is learning to write a thesis, that first terrible thesis they scrawl themselves is more valuable than any perfect one AI can spit out. Second, the 10 minute messy rule: set a timer for 10 minutes before they reach for AI. They have to write down at least three bad ideas or work through two practice problems, no matter how wrong they are. That’s 10 minutes, it doesn’t add much to their homework load, and it gives them the practice they need to actually learn the skill. Third, ask for the bad idea first. When they say “I don’t get this, I need AI,” don’t let them open the app right away. Say “what’s the stupidest, most wrong answer you can think of?” Nine times out of 10, that wrong answer is the starting point they needed.
I get why kids and parents reach for AI for the first steps. It’s frustrating to watch your kid struggle through something that feels simple, and it’s easy to just let the tool get it done faster. I’ve been there with my own 10-year-old, when he was stuck on a Lego build and asked AI to tell him exactly where every brick went instead of figuring it out himself. The problem isn’t that AI is bad at doing work. The problem is that it’s too good at doing the work we need our kids to do themselves to get better. If you never learn to write a bad topic sentence, you never learn to write a good one. If you never mess up simple division, you never get comfortable with algebra. Those bottom rungs aren’t glamorous. They’re wobbly, they’re messy, they feel like a waste of time when you could just jump straight to the good part. But they’re the only way to get up.
Last week, Javi turned in a full paragraph he wrote on his own, no AI. It got a B minus, because it had two spelling errors and the conclusion was a little off. He taped it to the front of his binder, he told me. Today he turned in the next assignment, got a B plus, and stopped by my desk to show me. He didn’t even need to ask for help this time. I don’t pretend this solves everything. There are still days when kids sneak AI to do their first drafts, when parents tell me they don’t have time to make their kid do the messy work when AI can just get it done faster. I don’t know what the long term picture looks like. But so far, holding onto those bottom rungs by forcing one small, messy, human first step works. The top of the ladder doesn’t mean much if you never get to climb up to it yourself.
Please indicate: Thinking In Educating » What Happens When AI Takes The Bottom Rungs Of The Learning Ladder: A Teacher’s Observation