I think one way to do so is to ask the student, "Tell me more about why you said this." Give them the chance to explain their thinking and you can celebrate their kindness and address the math.
You asked for people who are excited about AI tutoring to respond, and I'm the opposite, but here goes. I think a large amount of successful tutoring, like teaching, is building a relationship with a student, knowing their strengths, how and where to push them, etc. I just finished tutoring a girl I've worked with for 5 years (she's graduating 😢), and it was rarely the procedures she needed help with. What she needed was confidence, the time to talk through ideas, and reminders to check for the calculation errors she often makes when she rushes. I don't know how an AI tutor could provide those things (but I'm happy to hear about it if anyone thinks this is possible!).
Congrats on building rapport with your tutoring group! I love that "you can have mine" response. I love the number line idea for ordering integers where students have to discuss with each other about where to go.
I wonder what it would be like to have students go up into the front of the room to symbolize an equation. Students could then shout out, "hey, get with so and so, you're like terms! everyone who is holding an x term do a little dance" or, "distribute Silas with Sol and Kiyana." Solving the equation could be a social event that would probably be more memorably and funny than any worksheet. Hat tip to some colleagues who teach foreign language who just modeled this effectively with showing the difference in sentence construction with replacing direct objects (I think) with pronouns between Spanish and English." It was hilarious and memorable.
Great post. Thanks, Dan. A response to your invitation for perspectives on AI performing this kind of tutoring work:
Broadly, I think we all have seen that students are comfortable asking AI questions, but that most LLMs do so without much regard for the context (i.e. Is this a homework problem that perhaps I should not just answer for the student? Is the bell about to ring - how much time do I have to give them tips?). So it's incumbent on us as the designers of these AI tutors to make sure that the kind of conversation that ensues after a question is aligned with what a real tutor might do - not just give away the answer.
Students immediately start asking questions during Aristotle tutoring sessions and continue to do so at a meaningfully higher clip than they do in our human-to-human tutoring sessions. I would not say they yet build the kind of real, interpersonal relationship + emotional trust that you do with your students in person, or that many tutors do with their students, but we do see that they quickly start asking questions and voicing their confusions and frustrations.
Current is an interesting metaphor. It implies complexity, something that we have worked with a great deal and collaborated with the Cynefin team (see Dave Snowden's work) around for education. Here is a short piece in case you're interested in how we frame it: https://www.linkedin.com/pulse/layer-underneath-math-wars-dr-kevin-berkopes-ws5dc
Coherence of the story built at the local level couples nicely with the incredible tools, and in our opinion, is the practical pre-process that would accelerate teachers' use of Amplify's tools and curriculum.
What the teacher said: "That the pieces are equal because they are both 1/4."
https://x.com/latentjuice/status/2051373595968909469?s=61
Feels like there needs to be a way to honor the sentiment and math both here.
This question is why teachers should learn more about measurement. It's just a bad question if their goals was to elicit the mathematical thinking.
Someone on Twitter even pointed out the kid might know they were equal and be offering to switch just to avoid the argument.
I think one way to do so is to ask the student, "Tell me more about why you said this." Give them the chance to explain their thinking and you can celebrate their kindness and address the math.
You asked for people who are excited about AI tutoring to respond, and I'm the opposite, but here goes. I think a large amount of successful tutoring, like teaching, is building a relationship with a student, knowing their strengths, how and where to push them, etc. I just finished tutoring a girl I've worked with for 5 years (she's graduating 😢), and it was rarely the procedures she needed help with. What she needed was confidence, the time to talk through ideas, and reminders to check for the calculation errors she often makes when she rushes. I don't know how an AI tutor could provide those things (but I'm happy to hear about it if anyone thinks this is possible!).
Congrats on building rapport with your tutoring group! I love that "you can have mine" response. I love the number line idea for ordering integers where students have to discuss with each other about where to go.
I wonder what it would be like to have students go up into the front of the room to symbolize an equation. Students could then shout out, "hey, get with so and so, you're like terms! everyone who is holding an x term do a little dance" or, "distribute Silas with Sol and Kiyana." Solving the equation could be a social event that would probably be more memorably and funny than any worksheet. Hat tip to some colleagues who teach foreign language who just modeled this effectively with showing the difference in sentence construction with replacing direct objects (I think) with pronouns between Spanish and English." It was hilarious and memorable.
Great post. Thanks, Dan. A response to your invitation for perspectives on AI performing this kind of tutoring work:
Broadly, I think we all have seen that students are comfortable asking AI questions, but that most LLMs do so without much regard for the context (i.e. Is this a homework problem that perhaps I should not just answer for the student? Is the bell about to ring - how much time do I have to give them tips?). So it's incumbent on us as the designers of these AI tutors to make sure that the kind of conversation that ensues after a question is aligned with what a real tutor might do - not just give away the answer.
Students immediately start asking questions during Aristotle tutoring sessions and continue to do so at a meaningfully higher clip than they do in our human-to-human tutoring sessions. I would not say they yet build the kind of real, interpersonal relationship + emotional trust that you do with your students in person, or that many tutors do with their students, but we do see that they quickly start asking questions and voicing their confusions and frustrations.
That's a useful data point, Shan, though I note it conflicts somewhat with the experiences of Khanmigo developers at Khan Academy.
https://www.chalkbeat.org/2026/04/09/sal-khan-reflects-on-ai-in-schools-and-khanmigo/
I have a suspicion that the kids who are most inclined to ask questions of an AI tutor are the least likely to need a tutor at all.
Super excited to be supporting this curriculum implementation in my district next Fall!
Current is an interesting metaphor. It implies complexity, something that we have worked with a great deal and collaborated with the Cynefin team (see Dave Snowden's work) around for education. Here is a short piece in case you're interested in how we frame it: https://www.linkedin.com/pulse/layer-underneath-math-wars-dr-kevin-berkopes-ws5dc
Coherence of the story built at the local level couples nicely with the incredible tools, and in our opinion, is the practical pre-process that would accelerate teachers' use of Amplify's tools and curriculum.