This is a guest post by Liz Caffrey, a former mentee and colleague, now a master middle school teacher near Boston. I’ll link to her previous blog contributions at the end.
— Henri
Teaching Math in the age of AI
by Liz Caffrey
I have been teaching since 2000, when I began my career as a completely inexperienced teacher at the Urban School of San Francisco. That is where I met Henri, and where he (along with three other talented educators) taught me how to teach math. There is no doubt that schools have changed dramatically over my quarter century in the classroom. That change seems to have accelerated post-COVID.
The most recent shock to the system has been access to AI. Until recently, ChatGPT felt clunky and barely accessible to all but the most tech-savvy of my middle school students. Now my students have entire “conversations” about homework with a Chat GPT bot that uses Gen-Alpha lingo (“Full send! Let’s get you ready to slay this assignment!”). There is no avoiding AI in education.
In 2025, at my school, the Atrium School in Watertown, MA, we noticed that we needed to adapt our media literacy curriculum to include AI topics. We started by having our students co-create our middle school AI guidelines. This process took several weeks. We began by having students research AI policies at other schools. Next, we looked at our own school’s mission and values, and thought about how our AI guidelines could align with those larger aspirations for students. The kids crafted a general statement and then drafted expectations for students and teachers. That last part took me aback at first. However, it soon became clear that the students are as distrustful of AI as we adults are, and realize that adults can misuse it too.
Our full Atrium School AI guidelines make up a three-page document, but there is one sentence that I keep coming back to when I think about AI use at school:
Teachers and students use AI with integrity to enhance learning instead of using it to bypass learning.
When we rolled out the final draft of the AI guidelines to the middle school student body, we discussed scenarios from each subject where AI use might or might not be appropriate. For the most part, the students knew exactly when this tool would be bypassing versus enhancing learning. For example, they knew that asking it to quiz you before a math test would align with our guidelines, especially if you asked it to coach you rather than giving you the answer outright. They knew that asking it to solve a homework problem would not be an approved use of AI. In general, the students were highly skeptical of any text or images generated by AI, and quick to discount their validity. They recognized that while in school, it is almost always necessary to ask a teacher before using AI, especially at their age, when they are still getting to know the power of technology.
As a teacher, I feel I still have a lot to learn about how to use AI for my own purposes (beyond having it figure out what I can make for dinner with random pantry ingredients). The AI guideline-drafting process with my students has felt collaborative and honest, and opened the door for ongoing conversations. Just today, I explained to my sixth graders the exact prompt I used to have Chat GPT generate a random set of 18 order of operations problems. Rather than have them simplify all of those, their task was to choose 5 problems that covered every skill needed to fully master the skill of order of operations, and then simplify that subset with a partner. Because Chat GPT randomized the order of the problems (rather than ordering them by difficulty (as you’d find in a typical textbook), I was able to ask my students to do a higher-order thinking task that allowed for student choice. My use of AI in my lesson-planning meant that my students were more engaged, more challenged, and had more buy-in than they would have had on a typical set of leveled problems. AI gave me a new way to raise the bar for my students by improving a math task.
When I think about AI in education, I feel that math teachers have a leg up on teachers of other subjects. We’ve been here before. We have adapted from the slide rule to the calculator to graphing technology. Upon the release of Desmos in 2011, I modified my curriculum to take advantage of that new graphing technology (especially the clever use of sliders). This allowed students to quickly discover concepts that normally took days to build an understanding of. We could go even deeper into certain concepts or have more time for projects and real-world connections. In a similar way, AI makes many basic math tasks easier, thus freeing us up to do more open-ended, innovative tasks that capitalize on student curiosity.
In order to determine what AI can allow us to let go of, we must also decide what we need to hold on to. In most cases, the core concepts that students are learning must remain a focus. For example, if students want to know what 44% of 55 is, they can simply ask AI (or even Google it). However, the concept of how to calculate percentages remains very important. Thus, we might need to shift the question from
What is 44% of 55?
to
Which of these expressions could be used to calculate 44% of 55 and why?
- 44×55
- 0.44×55
- 55÷.44
- 44÷55
- 44×0.55
This asks students to demonstrate an understanding of what a percentage means rather than simply produce “the answer.” Open-ended questions like this allow for greater creativity and student agency while making it harder for AI to bypass the critical thinking we want students to do.
Just as Desmos made me a better teacher, I am confident that, used appropriately, AI can do the same. And for my students, I know that AI is an evolving tool that they need help learning how to use appropriately. Properly utilized, AI can foster creative problem-solving, flexible thinking, and question-asking, skills that are essential for deep mathematical investigations. And so far, those are skills that machines cannot replace.
———————————————————————
Thanks, Liz!
Involving the students in designing the policy was a powerful educational move. It will be interesting to see how things play out in coming years.
Liz is a sophisticated user of learning tools, and not only electronic ones. Read her previous contributions to this blog:
— Henri
Henri's Math Education Blog 
I love your optimism, but I am wary. I guess Desmos could be used to bypass some thinking (formal symbol manipulation) but you were engaged in other forms of math reasoning to do it. I tried your percentage question in Claude…
0.44 × 55 is the correct expression.
Why: “Percent” means “per hundred,” so 44% is really 44/100, which equals 0.44 as a decimal. Finding a percent “of” a number always means multiply by that decimal form. So “44% of 55” translates directly into: 0.44×55=24.20.44 \times 55 = 24.20.44×55=24.2
Why the others don’t work:
44 × 55 — This treats 44 as a whole number instead of a percentage, so it’s off by a factor of 100. (It gives 2,420, way too big.)
55 ÷ 0.44 — Division would answer a different question, like “55 is 44% of what number?” Here we’re not dividing, we’re taking a portion of 55.
44 ÷ 55 — This actually calculates what percent one number is of another (in this case, what percent 44 is of 55 — about 80%). It’s the reverse of what we want.
44 × 0.55 — This mixes up the two numbers’ roles: it would calculate 55% of 44, not 44% of 55. (Interestingly, these give the same answer, 24.2 — a neat coincidence of percentages — but the setup describes a different problem.)
The key idea: to find “X% of Y,” convert the percent to a decimal (divide by 100) and multiply: (X/100) × Y.
Blech.
LikeLike