Math achievement has been stagnating for over a decade. What could turn that around? As the “science of reading” makes big strides, one can’t help but ask if there’s an analogous “science of math.” Few people have thought as deeply about this as Larry Berger, the co-founder and CEO of Amplify, which has developed curricula and assessments used by nearly 20 million students. After years spent designing science of reading-aligned English curricula, Larry is now creating “next-gen” math curricula. Larry’s a fascinating guy. He’s been a Rhodes scholar, a White House Fellow working on educational technology for NASA, and he serves on the board of the Academy of American Poets. Here’s what he had to say about our math challenges.
—Rick

Rick: Larry, we’ve seen more than a decade of dismal performance in math on the National Assessment of Educational Progress. What’s going on?

Larry: NAEP isn’t designed to offer causal explanations, but it is important to try. Let’s first note that the top 20% of students have not experienced this decline. So, this may be one more place in American life where, for families outside that top quintile, the American Dream—work hard and you’ll succeed—doesn’t feel real. Focusing on policy: A decade and a half ago, we simultaneously introduced higher standards in math and retreated from the accountability systems that would reveal whether students were achieving those standards. Math learning needs to be measured and adjusted early and often. We can predict the 8th grade algebra gap from the 4th grade fractions gap; we can predict the 4th grade fractions gap from the 1st grade number-sense gap. But we retreated from the test-based accountability systems that helped us measure these learning deficits and pressure systems to act when students fell behind.

Rick: Any other factors you’d flag?

Larry: Another explanation for the decline is the smartphone’s assault on childhood—social media, anxiety, hyperstimulation, and lack of sleep. The attention of young people is being sliced, diced, and sold to the highest bidder. Math is not the highest bidder. We are also still living with a variety of pandemic effects we might call “educational long COVID.” Some schools have acute symptoms; others are merely winded. But all schools are struggling to get back in the game.

Rick: What’s the biggest problem in math instruction?

Larry: The biggest problem I see in math education today is an engagement crisis, which sits next to a looming “why bother if AI can do this for me?” crisis. Not enough kids are learning to love math or experience the feeling that math makes them powerful. But I’m optimistic we are on the verge of addressing this problem with a new generation of curriculum. There is nothing wrong with American math education that can’t be fixed by what’s right with it. The retreat from accountability has inspired new thinking about how we might measure math progress continuously, using an asset-based approach that builds on strengths rather than dissecting weaknesses. Parents and teachers are working together to ban phones, to help kids understand who is buying and selling their attention, and to offer ways to resist. There’s an emergent “attention activism” movement, and teachers are joining it in droves. But let’s not confuse addictive social media with quality educational software. Sometimes, communities get carried away and ban the good stuff, too.

Rick: You’ve spent a lot of years developing curricula aligned with the “science of reading.” Is there a “science of math” equivalent?

Larry: The science of reading is an overnight success that was 30 years in the making. It is tempting to hope there is something analogous in the wings for math. But, while there is a lot of learning science that can inform math teaching, there isn’t a math equivalent of the science of reading. The science of reading had a simple call to action—abandon disproven pedagogies and embrace a more precise, sequential system of instruction. You could learn the dance steps quickly: Stop doing A, B, and C and start doing scientifically validated X, Y, and Z. When it comes to math, the learning science says we need to learn a few dances, not just one. We need to encourage kids to develop their own ideas about how to solve compelling problems. And we need to embrace the pragmatic work of routine practice in order to develop automaticity.

Rick: Can you say more about the comparison between instruction in reading and math?

Larry: Kids need to learn math by doing the open, unstructured work of noticing, naming, and using patterns. They also need explicit instruction followed by enough practice to help them master math procedures. You can’t simplify it to one or the other. But I would say that the sequential precision for developing foundational reading skills that is the heart of the science of reading becomes less simple when we turn to developing reading comprehension and interpretation. Then we need to entwine a few pedagogical approaches and invite students to develop their own ideas. So, once the focus turns to reading comprehension, the learning science about both reading and math end up being similarly nuanced and multifaceted.

Rick: You’ve spent years developing a “next gen” math curriculum. Any thoughts as to what we’ve been getting wrong?

Larry: Math classrooms rarely achieve delight, and curriculum carries a lot of the blame. In our curriculum R&D for Amplify Desmos Math, we set the bar at “achieving delight routinely,” and the teachers using our program are experiencing it. Not every second should be a delight. Sometimes, you have to grind out the practice or memorize. But most things in math are beautiful and powerful, and they can be made compelling for kids. If you don’t do that, you’re swimming upstream against all the distractions vying for a kid’s attention. Delight alone doesn’t pass algebra tests or get you a good STEM job, but you are much more likely to do both if you have experienced delight routinely in math class. So, we’ve tried to work at the intersection of rigor and delight and build a system that ensures kids not only love math but also master the math they need to thrive in a society increasingly powered by it.

Rick: You’re emphasizing delight and engagement. That focus often seems to mean less attention is paid to computational mastery. What do you make of that tension?

Larry: There’s a funny assumption in the math wars—that if you want kids to love and understand math, you must hate computation. But can’t we have both? The research says kids need engagement and automaticity. One of the main reasons to care about engagement is that it motivates kids to practice until operations become automatic. And one of the main reasons we care about automaticity is that it makes extended mathematical work less exhausting—which helps kids stay engaged. It’s a virtuous cycle, not a tradeoff.

Rick: You’ve written about the challenges of evaluating curricula. What’s the right way to judge various offerings?

Larry: Try it before you buy it! In many places, curriculum is still chosen according to “the flip test”—teachers sit in a room and flip through textbooks. But today, all of the strong curricula are much more than pages of a book. They are hands-on activities, analytics that help teachers understand how their students are doing, tools that provide students with personalized support, interactive problem sets, simulations, and resources for parents. And there are huge variations in quality, usability, and learnability. Schools should try each contender for a month or two and visit schools that have used the program. Otherwise, they are judging a book by its cover (and it isn’t even a book).

Rick: What do teachers need to teach math effectively, especially in K-5?

Larry: Focus on giving teachers a curriculum that is rich, coherent, elicits thinking from students, and ensures that students receive responsive feedback. The curriculum should also help teachers develop their own understanding of the discipline. And then surround teachers with a community of other teachers and experts so they can discuss what’s working, what they are learning, what “great” looks like, and share tips about making tomorrow’s lesson sing.

Rick: Some tech enthusiasts have suggested that AI tutors will render many math teachers obsolete. You started in this work as a tech guy. What’s your take?

Larry: AI is developing so quickly that I’m reluctant to rule anything out. But anyone who claims they are replacing teachers with AI is almost certainly making things worse. Teachers do many things that AI can’t do yet and many other things that AI will never be able to do. Today, AI tutors in math can help students find the right answer to a problem given the limited data available to them, which is often just a number or some text. Teachers, on the other hand, can help kids care about that answer and also teach from a much larger set of data, including scribbles on the page, facial cues, and answers to probing questions. Teachers can give kids a kind of attention that kids know is valuable because it is scarce. Teachers can care. Kids are not fooled when AI pretends to care. Teachers can make kids feel pride, and they can cultivate community. Dan Meyer, the vice president of user growth at Amplify, has a lovely formulation about the unique role that humans—teachers and students—play. He writes, “Every day, kids come to school seeking answers to two big questions. Not just ‘How do I do this?’ but also, ‘Who am I?’ Chatbots can answer one but not the other, and for most kids, the two questions are inseparable. They are the same question.” It is only teachers, peers, and challenging learning processes that shed light on the “Who am I?” question.

Rick: Given your reservations, what do you think AI is more likely to mean for math education?

Larry: We’re excited about many of the things that AI can do for the math classroom, but the tools that have shown the most promise so far are those that extend the reach of teachers. For example, there’s a common fiction that math teachers scan—in real time—all the work their class is doing, interpret all the students’ thinking, and alter their lesson accordingly to feature their students’ novel ideas and to give individual kids a hint. Of course, that’s not reliably possible even if some great teachers approximate it. But AI can make that more of a reality. It can see in real time all the students’ work, give teachers an instant view of their class’ thinking, and instantly create teaching resources that enhance or fine-tune the original lesson.

Rick: Any big differences in what AI might mean for K-5 vs. high school math?

Larry: So much of the elementary grades is about learning how to learn. What does it mean to work on a problem, to have a possible solution and test it, to communicate your solution, to build on what you learned yesterday, to fold in the thoughts of others? AI wants to remove friction, but we learn from friction, and that starts in the earliest grades. By high school, we hope kids have learned a lot about how to learn. And then the AI can provide targeted help in areas where kids are stuck. It can provide multiple explanations that a teacher might not have time to give. I’m more bullish on student AI use in high school. And students are going to use it whether I’m bullish or not.

Rick: Is there one piece of research on math education that you find especially compelling?

Larry: I have long been devoted to the research on self-determination theory by Richard Ryan and Edward Deci. Curriculum designers ignore at their peril this research on what really motivates students.

Rick: What’s one piece of advice for leaders working to improve math instruction?

Larry: Education leaders should ask: Is there curiosity, delight, and rigor in our math classrooms? And if the answer is “no” or “rarely,” don’t shrug that off as the way it has always been. There are schools where something better is happening and new tools that make this available to all.

This conversation has been edited for length and clarity.