What I Teach 4th Graders to Prepare Them for an AI-Driven World (No Fancy Coding Required)
The first red flag was the word verdant. Javi sits in the back row, chews the eraser off at least one pencil a week, and spent last recess arguing that blue raspberry is objectively the best popsicle flavor. He does not use the word verdant, especially not to describe his grandma’s overgrown backyard peach tree.
I leaned against the edge of his desk after the rest of the class moved to silent reading, and nodded at his open Google Doc. “Tell me about that peach tree,” I said.
He shrugged, eyes on his scuffed sneakers. “It’s all there,” he mumbled.
“Verdant,” I said. “When did you start using words that sound like they came from a textbook?”
He chewed his lip for a minute, then admitted it. His older cousin showed him the free chatbot everyone uses for high school homework, and he’d typed in “personal narrative about grandma’s peach tree in Texas” because when he sat down to write, all he could think of was sour pie and thieving squirrels, and that felt too boring to turn in.
I didn’t give him detention or make him rewrite it as punishment. That’s the reaction a lot of people default to these days: AI is cheating, ban it from school, keep it away from kids as long as possible. Parents email me in a panic every other week, asking what skills their kid actually needs 15 years from now, when machines can write essays, solve math problems, and draw pictures in 10 seconds. Most mention coding tutors, pricey after-school STEM camps, online courses that promise to make kids “AI ready” by middle school. I get the panic; headlines make it sound like half the entry-level jobs will disappear next year. But after 11 years teaching 9- and 10-year-olds, the stuff that trips kids up already isn’t that they don’t know how to talk to a chatbot. It’s that they don’t know how to tell when the bot is wrong, when it’s giving them something generic that doesn’t fit their life, or when the thing it wrote will hurt someone.
We do 10-minute writing warm-ups every Monday, no devices allowed. I tell them to write down one tiny thing that happened in the last three days that no one else would care about but them. Recent entries include “my brother colored my homework blue with a crayon,” “I found a french fry under the park bench that looked exactly like Harry Styles,” and “my dog snored so loud he woke himself up.” No grades, no required sharing. The point isn’t to turn them into writers. It’s to get them used to noticing the weird, specific stuff only they carry around. AI can spit out a perfect generic peach tree story in two seconds. It can’t know your grandma lets you climb the branches even while she yells you’ll break your neck, that squirrels steal 80 percent of the fruit every year, or that last summer’s pie was so sour you both laughed until you cried after the first bite.
For parents who ask what to do at home, I tell them skip the formal drills. At dinner, instead of “how was school” (which always gets “fine”), ask “what’s one tiny silly thing that happened today that no one else would think is interesting.” It takes 30 seconds, and builds the same habit.
That habit of noticing what doesn’t fit comes up with math, too. Last month, we measured the classroom to price a new rug for the PTA. Half the class pulled up the AI calculator on their Chromebooks, typed in their wall measurements, and wrote down 120 square feet before I finished passing out graph paper. When I asked if that number looked right for a room that fits 22 desks and a reading nook, most just blinked. One girl, Mia, said “the computer gave it to me, so it’s right.” Turns out nearly everyone misread the tape measure and typed 10 feet instead of 18 for the room’s length.
Now, every time they use a digital tool for an answer, they do a gut check: write down one reason the answer might be wrong, and one way to test it with what they already know. That day, they stepped the length of the room heel to toe, counted 18 steps, and fixed the number before we sent the order. At home, this can look just as simple. If your kid asks a voice assistant how tall a giraffe is, ask “okay, how many of you stacked up would that take?” You don’t have to know the right answer yourself to make them practice checking.
We also spend time messing with problems that have no single right answer. Our class hamster Zorro kept escaping his cage earlier this semester. The kids first begged me to look up a fix; the chatbot told us to clip the locks tighter and set a book on top. He escaped two days later, chewed a hole in the homework bin, and napped on the extra crayon pile. So I told them to figure it out themselves. They watched him for three afternoons, and learned he shimmies up the water bottle to wedge himself in the door gap. They tried tape (he chewed through it in an hour) and a binder clip (he knocked it loose). Two weeks later, they settled on two small zip ties and a wooden ledge that blocks his climb. It’s not the fix any website would suggest, because no website knows Zorro is smaller and more stubborn than the average hamster. At home, resist the urge to Google a fix right away when your kid’s Lego tower keeps falling or their lemonade sign blows over. Give them 20 minutes to mess around, even if their first idea flops. AI excels at average answers to average problems. Most problems that matter aren’t average.
We’ve also had to work through what it means when AI writes something cruel. Earlier this year, a group of boys asked the bot to write roasts of classmates and passed them around at recess. They thought it was harmless—they didn’t write it, so it wasn’t their fault. One girl, Lila, cried after one called her light-up rainbow rain boots babyish. We sat down and typed the same prompt three times. The first take was the mean line. The second was about boots leaving rainbow footprints on the moon. The third was about the bright colors scaring away squirrels. AI pulls whatever floats around online, and it doesn’t care who gets stung. The person typing the prompt is the one who decides what’s okay to say out loud. Now, when kids show me a funny bot response, I always ask first: “How would you feel if that was about you?” It’s a tiny pause, but it sticks.
Javi rewrote that peach tree essay on his own. It had three run-on sentences and two spelling mistakes, earned a B+ instead of the A the bot’s polished version would have gotten, and when he read it aloud, three other kids raised their hands to say their grandparents have squirrel-plagued fruit trees too. I don’t ban AI in my room; I use it myself to make differentiated math worksheets and silly spelling sentences. It’s just a tool, like a pencil or a calculator. It doesn’t teach you what to pay attention to, or what to care about.
Last week, Javi turned in a new story about chasing a stray squirrel out of his garage. He told me unprompted he used the bot after he finished drafting, just to help split run-on paragraphs. The story still had typos, and the squirrel got away in the end, just like it did in real life. That felt like a win, but I have no master plan. Two months ago, two kids turned in book reports where the bot mixed up the plots of two same-titled novels, and we lost 20 minutes of class sorting it out. The tools shift every few months, and there’s no checklist that guarantees readiness for what comes next.
Yesterday I walked in early and found the kids clustered by Zorro’s cage. He’d chewed through one zip tie overnight, and they were passing around heavy-duty replacements they brought from home, arguing over whether a second latch was overkill. No one suggested looking up the answer. I leaned against the doorframe and watched for a minute. That’s the part I think matters, even if it never shows up on a standardized test.
Please indicate: Thinking In Educating » What I Teach 4th Graders to Prepare Them for an AI-Driven World (No Fancy Coding Required)