The AI Disconnect

25% of the sessions at the biggest edtech conference in the US were about AI. How does that compare to the math, English, science, and administrator conferences?

I’m back from math teacher fest in Chicago, IL (NCTM) and the disconnect has never been more apparent to me. A large group of people has drastically misunderstood the assignment of helping kids learn. We are not all working on the same page or even working from the same playbook here.

Leading one group are technologists like Sam Altman, the CEO of OpenAI, who recently promised “virtual tutors who can provide personalized instruction in any subject, in any language, and at whatever pace [our children] need.”

In another group, you’ll find the thousands of math teachers gathered in Chicago, each one in their own way celebrating joy, struggle, creativity, and perseverance in learning math, attributes that any of them will tell you are central to learning, not peripheral, and almost always mediated by humans.

Whenever possible at NCTM, I’d ask teachers, “How are you using AI in your practice lately?” Roughly 50% of teachers said “not at all” with the other 50% describing point solutions to very specific problems of practice—helping construct parent emails, for example—the kind of non-transformational “quality-of-life improvements” I do think generative AI is capable of delivering.

“I have my students work with their tutor chatbot for twenty minutes a day and they’re miles ahead of where they would have been last year,” said absolutely nobody.

You see the disconnect in RAND’s recently released survey of American Math Educators. (They haven’t released their analysis but the raw data is available with a free Bento account.)

RAND has S-tier analysts. They do rigorous and dependable work. This survey is in its second year. We aren’t dealing with yet another edtech company surveying its Facebook group about AI for content marketing here. And in Spring 2024, RAND found that 82% of surveyed math teachers have never used AI tools for their mathematics teaching. 1% said they use it “Often.”

I can promise you very little but I will promise you that these teachers are not answering out of spite. Teachers are drowning, generally speaking, and disinclined to refuse a good life preserver.

You see the disconnect when you look within education—comparing the conferences for administrators, math teachers, English teachers, science teachers, and education technologists. I scraped and analyzed each of their 2024 conference programs, coding each session as “about AI” based on keyword analysis.

The people who teach core content to our nation’s kids and the people who coach teachers on the use of technology are not on the same path. They aren’t even on on parallel paths. That’s apparent in 2024 and it’s apparent when you look at the change in their conference AI content from 2023 to 2024.

The disconnect is growing wider. Teachers are drowning and they are emphatic about their needs: higher pay, support with student behavior, fewer administrative burdens, and smaller class sizes. With generative AI, technologists think they’re throwing life preservers right next to the teachers. But it looks more and more to me like they’re throwing boxes of loose jet ski parts right on their heads.

Featured Comment

Maria H Andersen, over on LinkedIn, had this to say about your hopes for chatbot tutors:

I confiscated a school iPad from a 12 year old yesterday and in the previous hour he had done 50+ searches to try to find a game to play getting around the filters. Tab switching is definitely high.

I tried using the Magic School "Quiz Me" feature with my students this week in two middle school classes this week with a list of vocabulary words that they actually need to study for a quiz. Only 3 students engaged with the tutor. 2 of them asked for headphones. Everyone else saw the initial body of text and quit to do something else.

The engagement level of middle schoolers to want to learn is way way over estimated in edtech.

Odds & Ends

¶ It’s hard for me to imagine anyone writing a more impactful essay on edtech this year than Laurence Holt’s The Five Percent Problem. It is uniting nominally progressive-types like me, traditionalists like Paul Kirschner (over on Twitter), and academics like Larry Cuban (at his blog). Khan Academy’s Chief Academic Officer Kristen Dicerbo even popped up to comment on it, saying, “Agree that 4.7% getting to the 18+ hours per year is unacceptable. We report these numbers when others do not.”

¶ Huge news from the world of standardized assessments and edtech:

Update for the 2025 AP Exams: For all exams that allow or require calculators, except AP Statistics, you can use the built-in Desmos graphing calculator through the Bluebook testing application. Note that for Calculus AB, Calculus BC, and Precalculus, Desmos will only be available in the calculator-required parts of the exam. These exams continue to have parts where no calculator is allowed.

Desmos, if you didn’t know, makes a web-based graphing calculator that’s 1000% more powerful and fun to use than the plastic jobs you might have used when you were a student. Plus it’s free, so students can practice at home with the same tools they can now use on the AP exams.

¶ I am watching with tons of interest how Chris Lehmann, a tech-forward administrator I have followed for years, and his school network are implementing their cell phone-free policy.

¶ As the school year started, my Google alerts were pinging wildly with local and national news segments on AI in schools. A fun way to watch them? Take note of which roles are interviewed. I’m noticing roughly 95% administrators and tech coaches. 5% teachers and students. The headline of this CNBC segment, for example, is “Teachers turning to AI for help as students head back to school” and yet zero teachers are interviewed. There’s the disconnect again.

¶ Google commits $25M to train teachers in AI use. OpenAI hires a former Coursera executive to work with districts and schools. Of course, you could interpret these moves in the spirit of altruism. A small amount of money tossed in the direction of everyone’s favorite charity: schools and kids. Another way to look at it: these companies are setting giant piles of cash on fire every day building their frontier models and are not, in the eyes of many, making sufficient revenue to justify the cost. There’s a lot of available revenue in education, either by acquiring customers while they’re kids or (none of them will admit this publicly) by peeling away a few of the billions of dollars we spend on human teachers every year.

¶ Credit where it’s due to Alex Zimmerman and Chalkbeat for including a rare and important caveat in an article on personalized learning.

The idea, Banks said, is to make it easier for teachers to reach students at a range of academic levels who are all in the same classroom. Still, some previous efforts to promote personalized learning, including by Facebook founder Mark Zuckerberg, have fallen short of their lofty ambitions.