Jonathan Foster, a former high school math teacher who is now a professor in educational theory and practice at University at Albany, is working on an AI tool to give feedback to math teachers.
"I wanted that feedback from a peer or an instructional coach but I didn't always have access to that," he said.
Teachers often say they get little feedback on their teaching methods. Although many new teachers are assigned a mentor, that mentor also has their own classroom, and can usually only observe the teacher briefly. Principals also observe the classroom occasionally. But it's not enough, Foster said.
"I didn't have the ability to collect all the data that I wanted in my classroom environment," he said.
Enter AI. The AI tool uses cameras and audio recordings to report on whether the teacher looked at or walked through each section of the classroom, how often they used group work, and many other techniques. Even the words the teacher and students use are tracked.
"Am I engaging with student reasoning, am I asking students to explain their reasoning, am I giving my students opportunities to use rich mathematical vocabulary, am I using rich mathematical vocabulary?" Foster said, listing items the AI reviews.
The AI uses a common measure — the mathematical quality of instruction, which was developed by the Center for Education Policy Research at Harvard University. The measure is typically used to score teachers when a supervisor observes them teaching.
But the AI can give them daily feedback, without it going on performance reports.
"I hope we're helping teachers think more critically about their practice or shift their practice to high-quality instruction," Foster said.
The project is led by Peter Youngs and Scott Acton of the University of Virginia. The Bill & Melinda Gates Foundation funded it with a $1.4 million grant.
A pilot group of early-career teachers has been testing the tool. Starting this fall, they will use it throughout the school year. Foster and his fellow researchers will get a baseline measurement from the AI at the beginning of the year and then watch for improvements.
"See if we notice any subtle changes in their practice," he said.
The AI might tell a teacher to "move around and monitor and check in on their progress more" during group work, he said.
"Is she making those small adjustments in her practice that could become big?" he said.
So far, the teachers have been receptive, though they've also complained that the AI isn't good at noticing everything, he said.
Most AI research for math education recently has focused on tutoring students. The idea is that students would learn better with a one-on-one tutor. Foster doesn't think that is the right approach.
Students bond with teachers, not AI, he said.
"I see the act of teaching as such a human endeavor," he said. "I see teaching as an inquiry and a science and an art."
©2025 the Times Union (Albany, N.Y.). Distributed by Tribune Content Agency, LLC.
