Closing the Equation: What We’re Learning About The Math Gender Gap

After years of progress, the gender gap in middle school math is opening again. New data show that we are now on track to undo the gains girls have made in STEM subjects, reversing a trend that suggested parity was within reach.

This backslide matters.

Math isn’t just another subject; it’s a gateway to opportunity. Falling behind early can derail a young person’s future in high-paying fields, like engineering, data science, and AI. That’s bad for our economy as a whole, as these skills are in growing demand, and risks closing off opportunities for millions of girls.

While researchers have been studying this gender gap for decades, new efforts are revealing a lot about how to make math instruction both more effective and more accessible. Some of these lessons are coming through the Learning Engineering Virtual Institute, or LEVI for short, which Renaissance Philanthropy operates.  

LEVI supports five teams of learning scientists and tech developers who are working toward an audacious goal: to double the rate of middle school math progress for students from low-income backgrounds. The LEVI teams are harnessing insights from cognitive science, adaptive learning, and AI to develop learning platforms that improve math achievement. 

One LEVI team, working at the University of Florida, has made some exciting discoveries about how to make math instruction more accessible for girls. Its tutoring program, called ALTER-Math, is being piloted by teachers in several Florida districts.

ALTER-Math flips the teacher and student roles. When using the platform, students are the teachers. They guide AI chatbots through math problems, teaching them with step-by-step explanations and instruction. The idea is that students build their math skills by teaching others. What makes ALTER-Math especially engaging is how lifelike the interactions can feel. The virtual partners are designed to act like real kids – sometimes lost, sometimes only in need of a hint. 

Teachers who have used ALTER-Math said their students find the platform engaging because it allows them to choose problems that match their interests and skill levels. The AI encourages students without judgment, making math feel approachable, which helps level the playing field.

“I think it’s the way the AI talks with the kids,” said Brad Belk, who uses the platform with his students at the P.K. Yonge Developmental Research School in Gainesville. “It comes across more like a friend than a teacher, and they love chatting, sharing ideas, and figuring things out together. I also think all the choices it gives them really make a difference, so they can work in the way that suits them best.”

Sixth-grade teacher Barbi Wigen recalled one exchange between a female student, Avery, and her AI bot that professed interest in interior design, an enthusiasm Avery shared.

“She was able to use her knowledge about interior design to explore math concepts like how to calculate the area of a room,” Wigen said. “That was really cool, and we all fed off of that excitement.”

When the University of Florida ran a randomized controlled trial of 619 girls and 562 boys who were using the ALTER-Math platform, they saw something surprising: There was no gender difference in learning gains among platform users, while the control group of “business as usual” instruction without ALTER-Math showed a significant gap in math achievement between boys and girls.

I asked Professor Wanli Xing, who is leading the LEVI work at UF, why he thought ALTER-Math had eliminated the gender gap in that study.

He said he thought the platform’s “learning-by-teaching mechanism” required students to deeply learn the material in order to “teach” it to their AI chatbot “peers.” This requires both boys and girls to engage in metacognition about their math knowledge, he said. Metacognition – the act of thinking about your thinking – helps students become more aware of how they approach math problems, which strategies they use, and where they get stuck. This self-awareness allows them to adjust their methods, avoid mistakes, and build stronger problem-solving skills.

Xing also said the ALTER-Math platform seems to reduce social pressure and “stereotype threat” through its low-stakes lessons that mirror a private tutoring session. This enables students to persist, ask questions, and revise their thinking without fear of judgment, something other research has found can hold girls back.

Finally, Xing posited that the platform’s design allows students to do what Avery did – connect math to the things they’re most interested in – which makes math more relevant to their lives, and increases their interest in learning.

It’s too soon to tell if these design principles will eliminate the gap long-term. But LEVI teams are seeing similar results elsewhere using similar strategies. One team, working in Ghana, has built a math tutoring chatbot called Rori. A 2023 study measuring yearly student growth found that girls using Rori improved 10 percentage points while boys improved 9 points. Owen Henkel helps oversee the project from the University of Oxford. He said the study’s results “suggest Rori successfully engages both boys and girls, with no significant barriers to equitable math learning across gender lines.”

Tools like Rori and ALTER-Math show that the gender gap isn’t a mystery; it’s a design problem. And their early results offer a hope that AI-powered learning tools can not only boost academic achievement but also foster greater accessibility for all learners. If every young girl could experience math as Avery did – as a gateway rather than a gatekeeper – imagine the talent we could unleash in our classrooms, in our labs, and in our country.

Kumar Garg is President of Renaissance Philanthropy and the Founder of the Learning Engineering Virtual Institute.