Two things in the news point the same way. A school ban doesn’t teach a child anything, and an AI tutor sitting one click away doesn’t either. My read: the missing ingredient is a human.
This fall, a lot of the school AI conversation has been about what to block. I understand why. But two pieces I read this week point somewhere else. Blocking a tool doesn’t teach a child how to use it well. And handing a child a well-built AI tutor doesn’t mean they’ll learn from it. My read: in both cases, what’s missing is an adult who sits with the child and makes the thinking happen. A parent or a teacher. The facilitator.
What’s in the news
- A ban buys time. It doesn’t teach. In EdSurge on October 26, Yvette Renteria, chief program officer at Common Sense Media, wrote about school AI bans. She cites a nationally representative study, completed before Los Angeles and New York City announced their AI bans: “of the 44% of teenagers who said they had had a tool banned at school, 59% of those same teens said they’d accessed the tool anyway on a personal device.” She isn’t against bans. “I support bans in that they can help schools buy time to figure out how to do AI literacy.” But her bottom line is plain: “Ban or no ban, we have to teach students how to safely navigate this new world.” Under a section she titles “A Block Isn’t a Lesson,” she says NYC’s requirement of two 45-minute AI lessons a year for high schools is “a good start, but not sufficient.” And she closes with this: “Teach kids what AI is. Teach them why AI flattens their critical thinking. Teach them why it acts like a friend, but isn’t.”
- An AI tutor, one click away, was rarely used. A new NBER working paper by Philip Oreopoulos and Nina Low, “One Click Away: AI Tutoring with Khanmigo in a Two-Year School Experiment,” ran a two-year randomized trial in 18 Tennessee middle schools. Students used Khan Academy with its AI tutor, Khanmigo, “configured to coach rather than give answers,” during daily remedial math sessions. Math scores went up, but the authors write that the gains “resemble those from Khan Academy practice without AI assistance.” Why? “96 percent of students tried Khanmigo at least once, but the median student messaged it on only a third of the days they practiced, and in only 17 percent of the exercise sessions in which they made a mistake.” Their conclusion: “The binding constraint appears to be engagement.”
- “Human attention” as the key ingredient. Chalkbeat’s Matt Barnum covered the study on August 25. He quotes the researchers: “Access was nearly universal but engagement was thin.” And: “It may be that human attention is the key ingredient for realizing the full benefits of personalized learning.” Oreopoulos put the problem simply: “making your life easier equals less effort, less learning.” Chalkbeat also reports that Khan Academy has since redesigned Khanmigo, so the study tested the earlier version, not today’s product.
What I’m taking from it
Side by side, these are two halves of one idea.
A block isn’t a lesson. If most kids who hit a school ban just open the tool on their own device, the ban hasn’t taught them anything. It has bought time, which Renteria says is useful. But time only helps if someone uses it to teach.
Access isn’t learning. The Khanmigo study gave students a tutor built to coach, not to hand over answers. That’s close to what many of us would ask for. Still, most students rarely reached for it, even after getting a problem wrong. The tool was ready. The engagement wasn’t.
What fills both gaps is a person. A parent at the kitchen table, a teacher in the room, someone who notices the wrong answer and says, “Let’s look at that together.” That’s the role I keep coming back to. My standing position: Augment, not Replace. Deepen, not Shortcut. AI is an assistant, never a replacement, and it never writes the child’s answers. The adult is the facilitator.
How much to trust it
- The 59% figure comes from a study Renteria cites. It was completed before the LA and NYC bans, so it tells us about earlier bans, not these ones. I’m using it as she reports it.
- Renteria’s piece is a first-person essay. Her views on bans and AI literacy are hers, and I’m quoting them as such.
- The Khanmigo study is one working paper, in one district, with middle schoolers in remedial math. It doesn’t show AI tutors can’t work. It shows this setup, at that time, didn’t add much beyond regular practice.
- The “human attention” line is the researchers saying “it may be.” It’s a hypothesis, not a finding. And in the study, the Khan Academy students were compared with students who got other digital programs plus small-group, teacher-led instruction. The Khan group made faster gains. So this isn’t a study that proves humans beat software. It’s a study that shows an AI tutor nobody talks to doesn’t do much.
- Khanmigo has been redesigned since. The current version wasn’t tested here.
- The facilitator approach below is mine. No source tested it, and I haven’t tested it either.
My own interpretation: be the one who makes it a lesson
This part is mine and it’s untested. It’s how I’d act on the two halves above, in a classroom or at the kitchen table.
- Sit with the tool, and with the child. Access alone didn’t produce engagement in the study, so don’t just hand it over. Use it together on one real problem, ideally one your child just got wrong. That’s the moment students in the study mostly skipped.
- Make it ask, not answer. If the tool hands over a finished answer, close it and ask your child what they’d try first. If it asks a question back, let your child answer it out loud. Deepen, not Shortcut.
- Teach the “why,” not just the rule. Renteria’s three lines make a good starting lesson at any age, in your own words: what AI is, why leaning on it can flatten your thinking, and why it can feel like a friend but isn’t.
A few things I’d keep in mind:
- For younger kids, keep it to one question: “Did it tell you, or did it ask you?”
- For homeschool parents, you’re already the facilitator. This just brings AI into the conversation you already have about any lesson.
- For teachers, if your school has a pause, that time is the window Renteria describes. A lesson on what AI is doesn’t need the tool turned on.
- Effort is the point. Oreopoulos’ line is worth sharing with kids: easier can mean less learning. The tool supports the work. It doesn’t do it.
Try this
This week, pick one moment when your child gets something wrong, on paper or on screen. Sit with them for five minutes and work through it together, with or without an AI tool. If you use one, notice whether it asked or told. What did you notice? Tell me what you tried and what came up. (Please don’t share names or identifying details about your kids or students.)
Continue this in the community: https://auxesislab.com/community/
Sources
- EdSurge, Yvette Renteria (Common Sense Media), “Pause AI in Schools. But Not Before You Do This First.” (opinion, Oct 26, 2026): https://www.edsurge.com/news/pause-ai-in-schools-but-not-before-you-do-this-first
- Philip Oreopoulos and Nina Low, “One Click Away: AI Tutoring with Khanmigo in a Two-Year School Experiment,” NBER Working Paper 35620 (2026): https://www.nber.org/papers/w35620
- Chalkbeat, Matt Barnum, Khanmigo engagement study (Aug 25, 2026): https://www.chalkbeat.org/2026/08/25/ai-tutoring-students-khanmigo-khan-academy-engagement-study/

Leave a Reply