Groshell, a former Seattle-area fourth-grade teacher, had to laugh: “Nobody knows me here in the U.S.,” he said in an interview.

But in Britain, lots of teachers know his name. An in-demand speaker at education conferences, he flies to London “as frequently as I can” to discuss , his 2024 book on explicit instruction. Over the past year, Groshell has appeared virtually about once a month and has made two personal appearances at events across England.

The reason? A discipline known as cognitive science. Born in the U.S., it relies on decades of research on how kids learn to guide teachers in the classroom, and is at the root of several effective reforms, including the Science of Reading.

In nearly a dozen interviews, educators and policymakers on both sides of the Atlantic said that while it’s caught fire in England, from the classroom to the halls of government, the idea has made little traction in its home country. Benjamin Riley, founder of , a Texas-based group that has pushed to make cognitive science more central to U.S. teacher training programs, jokingly refers to it as a “reverse Beatles” effect, with British educators pining for American insights.

“Cognitive science gives you a vocabulary and a language, a common framing, to talk about how minds work,” said Riley. “That is one of the hallmarks, typically, of professions: There’s an agreed-upon body of knowledge that constitutes the things that professionals need to know in order to be practitioners in that space. And education, at least in the United States, has never really done that.”

The result, observers say, is slow, steady academic progress for 9 million English students, even as U.S. results falter.

From 2011 to 2021, English students’ average scores in the International Benchmarks of Reading Achievement, a key global comparison, rose six points, placing them fourth worldwide, while U.S. students’ dropped eight points, ranking the U.S. just below England. Essentially, American fourth-graders in 2021 read nearly as well as English students did .

In the bargain, English schools cut students’ gender gap in reading by more than half.

Other commonwealth countries have taken notice, with policymakers in , and working to duplicate England’s progress.

Is U.S. system ‘too big for things to catch fire’?

Developed in the 1950s, cognitive science essentially explains how we learn, think, remember and process information. Applied to education, it allows teachers to maximize learning by incorporating key principles, among them:

• working memory and cognitive load: Students have limited capacity to remember several important things at a time, so teachers should break down complex information into smaller chunks to avoid overwhelming them. For instance, a teacher introducing a lesson on multiplying fractions should first ensure that students’ recall of multiplication facts is solid and that they can multiply numbers automatically in their heads.
• spaced practice and retrieval: Rather than cramming a lot of information into a single session, teachers should space out learning over time and regularly ask students to retrieve information from memory via review sessions and low-stakes quizzes.
• prior knowledge activation: Teachers should explicitly connect new concepts to students’ existing knowledge and experiences before introducing unfamiliar material. For instance, in a lesson about how seeds grow into plants, teachers should begin by asking students if they’ve ever planted seeds in a garden and what they noticed.
• metacognition: Teaching students to “think about their thinking” helps them become more effective learners. For instance, in a lesson that features a word problem, a teacher might say, “Let’s slow down and figure out what to do first, second and third.” When students make errors, a teacher can ask, “Walk me through your thinking. What steps did you take?” 

In England these days, said Groshell, the Seattle teacher, such jargon is now mainstream: “It’s impossible now to find a teacher who doesn’t know about retrieval practice, cognitive load theory or explicit instruction.” 

What began as a grassroots movement among teachers coalesced into national policy around 2010, when a series of structural reforms made it easier to embrace cognitive science.

That is when Michael Gove, education secretary under Prime Minister David Cameron, allowed virtually any public school to convert to “academy” status — British educator Dylan Wiliam calls them “charter schools on steroids.”

Freed from local authority, but funded centrally, these schools can pool resources to hire research advisors, directors of teaching and learning and the like. “These people have really engaged with the research,” Wiliam said.

In an interview, former Minister of State for Schools noted the irony that most of these ideas are American-made, developed by U.S. researchers. In 2006, Gibb recalled first encountering . Authored by E.D. Hirsch, a University of Virginia scholar, it argued for a content-rich curriculum, traditional math and phonics-based reading lessons.

“It just explained everything I was instinctively feeling about our school system,” said Gibb, who recalled that English schools at the time were steeped in more progressive methods. He made everyone he met read the book — including Gove, the education secretary.

“That really formed the basis of our reform programming from 2010 onwards,” said Gibb. It gave rise to universal phonics screening and adoption of the more traditional, step-by-step . 

The movement really bloomed in 2013, when Scottish educator Tom Bennett created the first in a series of affordable research conferences for teachers. Dubbed , the conferences, which continue 12 years later, have built an international appetite for scientifically proven classroom practices.

In 2019, the government introduced an for teachers, which standardized training on “very practice-focused” principles, said Wiliam, the British educator. Since then, every school that recruits a teacher out of a university training program must report how well they succeed in classrooms. If programs don’t get positive reports about trainees, they can lose accreditation.

“There’s a really strong alignment between the needs of the system and what is being provided in initial teacher preparation programs, in a way that doesn’t actually happen in the U.S. at scale,” he said

It’s a source of frustration for Wiliam, who now works as an independent consultant in northern Florida. Despite the movement’s success in England, he said, just 10% of his work is based in U.S. schools. “I find it quite difficult to get any American school districts to engage me,” he said. But he’s got three scheduled trips to Australia this year, among others. 

Riley, the Deans for Impact founder, noted that American public schools are governed by 50 different state agencies that rarely row in the same direction. The U.S. may just be “too big for things to catch fire” the way they can elsewhere, especially in centralized systems like the United Kingdom.

Beyond state control, he said, most U.S. teachers’ colleges “are not designed with learning science principles at their core — quite frankly there’s just a lot of stuff in schools of education that is not very good from a research standpoint, but that nonetheless has become ingrained. It’s a generational battle to try to change that.”

Like Groshell, Riley laughed at the contrast with the U.K. “I am beloved over in England, and increasingly [in] Australia, in a way that just is simply not true here in the United States,” he said. 

Sarah Oberle, a Delaware first-grade teacher who is active in research and training, said U.S. teacher prep doesn’t typically focus on cognitive science because many think it favors a kind of “authoritative and cold” approach. “But when you really understand science, you realize just this knowledge gives me the power to make changes within my practice that will actually protect and support my students.”

Oberle stumbled upon cognitive science about five years ago, when the Science of Reading movement started building momentum in the U.S., and wondered why she never learned about it during her training. She went back to school and earned a doctorate in education science.

“Our business is learning,” she said. “How do we facilitate learning when we don’t understand how learning happens?” 

‘Comrades in arms’ 

While much of England’s progress is traceable to shifts in national policies, several British teachers described moments early in their careers when, like Oberle, they got a taste of cognitive science and began questioning their training.

Daisy Christodoulou, a former London high school English teacher, began her career in 2007 as a member of , the international iteration of Teach For America. She had an inkling that much of her training wasn’t just unhelpful but wrong, with discredited ideas held up as best practices with little evidence they worked. “I was just looking at [them], going, ‘Really? Is this really best practice?’”

In 2010, she came across Daniel Willingham’s book Subtitled, “A Cognitive Scientist Answers Questions About How the Mind Works and What It Means for the Classroom,” it revolutionized how Christodoulou thought about her work. Over the past 15 years, Willingham’s book has been “enormously influential here,” she said, turning the genial scholar into another American celebrity.

In an interview, Willingham agreed that many U.S. teaching candidates are exposed to views about how children learn that aren’t all accurate. For instance, he said, “This phrase that you hear so often, ‘Every child learns differently,’ is, in one sense, true. But it’s kind of true in a trivial sense, and in a more important sense, it’s really not true.”

Peps Mccrea, a former teacher in Brighton, on the southern British coast said blogs written by colleagues have become another way for educators to share research, finding “comrades in arms” in a movement that continues to grow. More than 20 years after he first entered a classroom, Mccrea hosts a that unpacks research-based teaching methods. 

And Gibb has taken to touting England’s advances more widely. Last month, he met in Washington, D.C., with U.S. Education Secretary , raising hopes that the British reforms might find an audience here. A spokesperson for McMahon did not reply to a request for comment.

Actually, said Oberle, the Delaware teacher, the Trump administration is moving in the opposite direction from U.K.-style national policies, pushing to abolish the U.S. Education Department and creating the potential for “even more individuality between states.”

If we’re ever to see cognitive science advance here, Oberle said, it’ll take both a top-down and bottom-up approach: word-of-mouth influence among teachers, via events like researchED, as well as federal and state pressure on training programs to bring the research to teachers. 

“Once they have it clearly and don’t have misconceptions about it, the benefits they will see in their own practice very quickly will make them want more — will make them demand more. It’s just gaining that entry point.”

But she added, “It’s such a long process. There are so many minds to change.”

The problem is that people of various ages have different educational experiences, different jobs and different circumstances, affecting how they develop and retain their skills.

In , my colleagues and I find that skills typically rise until the 40s, after which reading skills gently fall and math skills more steeply. Even here, however, the story is not so simple. These averages mask the fact that any decline is closely tied to how much the skills are used. 

Simply put, people who read and do math on a regular basis hold on to those skills at least into their 60s.

Economists are interested in understanding this  because reading and math skills are closely related to economic outcomes. More highly skilled individuals tend to earn more, and countries with more skilled populations grow faster. Here is the big issue: Most developed countries of the world have aging populations. Does this then imply worse economic outcomes as we go forward?

The research challenge in answering this question has been a lack of appropriate data. For the most part, existing data on age and skills do not come from observing a representative sample of people as they age. Instead, they come from comparing the skills of different people of different ages, say one at 30 and one at 40, and assuming that after aging for 10 years, the 30-year-old will look like the 40-year-old.

But these two people grew up in different circumstances, with differing quality schooling and other factors that might affect their skills. Thus, any effects of aging are mixed up with other societal factors.

We overcome this problem by using unique German data that follow a representative sample of 3,263 adults over a three- to four-year period. At the initial survey and again at the later survey, the individuals are given the same reading and math test. Thus, it is possible to observe directly the impact of age on skills. 

What we found was that skills, on average, continue to increase into the 40s, and they never dip below the levels the individuals enjoyed in their 20s.

Perhaps the more important finding is that even this later decline is not inevitable. These average patterns hide the dramatic differences in aging between those who use literacy and numeracy skills consistently at home or work and those who do not. The survey data asked about the frequency of doing separate items such as “calculating prices, costs, or budgets” for math or “reading letters, memos, or e-mails” for reading. 

Those with above-average usage never showed declining skills at least until age 65, when our data ended. Those who weren’t much using math or reading skills peaked in their early 30s.

Interestingly, based on assumed high-skill usage, some previous analyses followed the skill patterns for white-collar and highly educated workers. When we look at these factors, we find the same answers: Among professionals or highly educated individuals, those who use the skills never show declines with age, but those who do not use the skills do, in fact, start to decline. Women show a sharper drop in numeracy skills as they grow older than men, perhaps based on educational background or career choices.

While our results, in principle, offer some consolation for countries with aging populations, they also highlight the importance of policy attention toward not only the accumulation of skills in schools, but also their retention through using those skills and pursuing lifelong learning.

Fostering expanded learning opportunities takes on increased importance with such societal changes as the broad introduction of various forms of artificial intelligence, which could force a large number of people to change what and how they are doing their work. Unfortunately, while the idea of lifelong learning is frequently discussed in policy contexts, little has been done to make it a reality. 

The Hoover Institution provides financial support to ����Ӱ��.

Techno-optimists have high hopes for how AI will improve learning. But what’s the merit of the “bull case”, and what are the technology’s risks? To think through those questions, Michael and Diane sit down with Ben Riley of Cognitive Resonance, a “think and do” tank dedicated to improving decisions using cognitive science. They evaluate the cases made for AI, unpack its potential hazards, and discuss how schools can prepare for it. 

Listen to the episode below. A full transcript follows.

Diane Tavenner: Hi there, I’m Diane, and what you’re about to hear is a conversation Michael and I recorded with our guests, Ben Riley. It’s part of our series exploring the potential impact of AI in education, where we’re interviewing optimists and skeptics.

Here are two things from the episode that I keep thinking about:

First, our conversations are starting to make me wonder if AI is going to disrupt the model of education we’ve had for so long, as I think Ben perhaps fears, or if it’s actually going to strengthen and reinforce our existing models of the schoolhouse with classrooms filled with a teacher and students.

The second thing that I was really thinking about and that struck me was that Ben’s sort of one case for what could be beneficial about AI is something that’s directly related to his work and interest in understanding the brain. And kind of how learning occurs. To be fair, there’s a theme emerging across all the conversations we’re having with people where they see value in the thing that they value themselves. And perhaps that’s an artifact of the early stages and who knows, but it’s making me curious.

And speaking of curious, a reflection I’m having after talking with Ben is about the process of change. Ben is a really well reasoned, thoughtful skeptic of AI’s utility in education. He comes to his views at least partially from using AI. I would consider myself much more of an optimist and yet I’m finding myself a little bit annoyed right now, that every time I want to write an email or join a meeting or send a text or make a phone call that I’ve got AI pretty intrusively jumping in to try to help me. And it’s really got me thinking about the very human process of change, which is one of the many reasons why I’m really looking forward to sense making conversations with Michael after all of these thought provoking interviews.

In the interim, we’d both love to hear your thoughts and reflections. So please do share. But for now, I hope you enjoy this conversation on Class Disrupted.

Michael Horn: Hey, Diane. It is good to see you again.

Diane Tavenner: You too. And I’m really excited to be back. Coming off of our last conversation around AI and education, it’s making me even more excited about what we’re going to be learning in this series. And I think today will be no exception in really stretching our minds and our thinkings about the possibilities, the limitations, the potential harms of AI and its intersection with education.

Michael Horn: Yeah, I think that’s right, Diane. And to help us think through these questions, today, we’re bringing someone on the show that I think both of us have known for quite a long time. His name is Ben Riley. He previously founded the Deans for Impact in I believe 2014. And Deans for Impact is a nonprofit that connects cognitive science to teacher training. And then Ben stepped aside a couple years ago, and has most recently founded Cognitive Resonance, which is a think and do tank, in its words, and a consultancy organization that’s really, its focus actually is on this topic of AI and learning, which is perfect and makes Ben the perfect guest for us today. So, Ben, welcome.

Ben Riley: Thanks so much for having me. We’ll see if you still think I’m the perfect guest by the end of it, but I appreciate being invited to speak to both of you.

Ben Riley’s Journey to the Work

Michael Horn: Absolutely. Well, before we get into a series of questions that we’ve been asking our guests, we’d love you to share with the audience about how you got into AI so deep, specifically because I will confess and I’ll give folks background, I’ve been reading. I’ve actually been an editor on a couple of the things that you’ve submitted into Education Next on AI, and I found them super intriguing. And then somehow I had no idea that you created this entire life for yourself around AI and education. And you have some language on this that I think is really interesting on the site where you say the purpose is to influence how people think about Gen AI systems by actually using the lens of cognitive science. And you believe that will help make AI more intelligible, less mysterious, which will actually help influence what people do with it in the years to come. And then you write that you see it as a useful tool, but one with strengths and limitations that are predictable. And so we really have to understand those if we want to harness them in essence. So how and why did you make this your focus?

Ben Riley: Yeah. Well. And thank you for clearly having read the website’s cognitiveresonance.net or the Substack Build Cognitive Resonance, in many ways, the organization reflects my own personal journey because several years ago I started to become aware that something was happening in the world of AI, and at the time it was called deep learning, and that was the phrase that was starting to emerge. And to be completely candid, my focus has always been, and in some ways still very much is on how human cognition works. And so AI, artificial intelligence, is considered kind of one of the disciplines within cognitive science, along with psychology and neuroscience and linguistics, philosophy. There’s like it’s an interdisciplinary field. And for me, quite honestly, AI was sort of like this thing happening somewhere over there that I had maybe a loose eye on. And I got in touch with someone named Gary Marcus at the time, and we’ll come back to Gary in a second, and then just said, hey, Gary, can you explain deep learning to me and what it is and what’s going on? And that, you know, sort of began that conversation. And then quite frankly, I just kind of squirreled away and didn’t think much about it. And then, like it did for all of us, ChatGPT came into our lives. And I was stunned. I was completely stunned when I first sat down with it and started using it. And what really irked me was that I didn’t understand it. You know, I was like, I don’t get how this is doing, what it’s doing. So I am now going to try to figure out how it’s doing, what it’s doing. And that is not easy. At least it wasn’t easy for me. I don’t think it’s even now. I don’t think it’s easy for those who might have spent their entire lives, much less those of us who are coming in late in the game or just trying to make sense of this new technology in our lives. And what I was able to draw upon was both sort of the things that I do know and have learned over the last decade plus around human cognition and frankly draw on a lot of relationships I have with people who are in cognitive science broadly, and just start having a bunch of conversations, doing a bunch of reading, and really trying to, you know, build a mental model of what’s taking place with these tools and with large language models specifically. And when I finished all that, I thought, well, geez, it seems like, you know, that took a lot of work. Maybe it would be helpful to sort of try to pass this along and bring others into the conversation. So that’s really the thesis of Cognitive Resonance.

AI’s Educational Upside

Diane Tavenner: Ben, everything you just described is just so consistent with my experience with you over the years and the conversations that we’ve had and what my perception is what you care about. And I’m so glad you brought it together in that way, because I’ll be honest, when I was like, wait, Ben is doing AI? Like, that didn’t totally land with me. And so what I’m hearing from you is like, well, I’m super curious for this conversation because I’m. I’m not getting the vibe that you’re a total AI skeptic. I’m not getting the vibe that you’re a total cheerleader. I’m guessing we’re gonna have a really nuanced conversation here about this right now. So let’s start there. Like, let’s start with kind of that polar, and then see where we go. Can you make the argument for us of how AI is going to positively impact education? And I’m not saying it has to be your argument, but can you just stand up an argument for us based on what you’ve learned about how it could. Like, what’s the best case to be made for AI positively impacting it?

Ben Riley: Yeah. So this is what people are now calling steel manning, right? Like, can you steel man the argument that you may not agree with. I had a law school professor who taught me that the best way to write a good legal brief is to take the other side’s best argument, make it even better than they can make it, and then defeat it. And you all gave me this question in advance, and I’ve been thinking about it since you did, and I don’t know if I can make one best case. What I want to do is make three cases which I think are the positive bull cases. So number one, one that I think should be familiar to both of you because we’ve been having this debate for nearly a decade, is sort of personalized learning, a dream deferred, but now it can be real. When we said we were going to use big data analytics and use that to figure out how to teach kids exactly what they want to know, when they need to know it. Like, what we meant was we needed large language models that could do that. And now, lo and behold, we have that tool. And as Dan Meyer likes to joke, it can harness the power of a thousand suns. It’s got all of the knowledge that’s ever been put into some sort of data form that can be scraped from the Internet or from other sources, not always disclose what those sources are, but nonetheless, there’s a lot of data going into them and using these somewhat mysterious processes that they have of autoregression and back propagation. And we can go as deep as you want in the weeds on some of those terms, but we doing that, we can actually finally give kids like an incredibly intelligent, incredibly patient, incredibly, some would even say loving, some have said that, tutor. And we can do that at scale, we can probably do it cheaply. And boom, Benjamin Bloom’s dream, two sigma gains. It’s happening finally. There we go. All right, so that’s argument number one. Call that personalized maximization argument. Argument number two, I think, is the sort of AI as a fundamental utility argument. And the argument here is something along the lines of, look, this is a big deal technologically in the same way the Internet or a computer is a big deal technologically, and it’s one of those technologies that’s going to become ubiquitous in our society, the same way the computer or the Internet has become ubiquitous in our society. And we don’t even know all the many ways in which it’s going to be woven into the fabric of our existence. But that includes our education system. And so some benefits will accrue as a result of its many powers. Okay, so that’s the utility argument. The third argument would say something like this. It would say the process of education fundamentally is the process of trying to change mental states in kids. And I mean, frankly, doesn’t have to be kids, but we’ll just talk about it from teachers to students.

Michael Horn: Sure.

Ben Riley: And, there’s some really big challenges with that. When you just distill it down to the act of trying to make a kid think about something. One of the challenges is that we cannot see inside their head. So the process of what’s taking place, cognition or not, is opaque to us, number one. And number two, experiments are really, really hard. They’re not impossible. But you can’t really do the sort of experiments that you can do in other realms of life the same way. It’s just for ethical reasons, but also just frankly from like scientific, technical reasons. Because again, we can’t see what’s happening in the head. So even when you run an experiment, you’re getting approximations of what’s happening inside the head. Some would then say, well, now we have something that is kind of like a mind and we can kind of emphasis on kind of, see inside it. And we definitely can run experiments on it in a way that doesn’t implicate sort of the same ethical concerns and others. That argument, and I’ll call that the cognitive arguments, human and artificial, would say that can use this tool to better help us understand ourselves. In some ways it might help us by being similar to what’s happening with us, but in other ways it might help us by being different and showing those differences. So those are the three arguments that I see.

Evaluating the Case for AI

Diane Tavenner: Yeah. Super interesting. Thank you for making those cases. Which of any of them do you actually believe? Now you, I’m curious about your opinion and why?

Ben Riley: Yeah. So I have bad news for you. The first one, the personalized maximization dream, is going to fail for the same reason that I would like to say I predicted that personalization using big data analytics would fail. We could spend the entire podcast with me unpacking why that is. I’m not going to do that. So I’m going to limit it just to two arguments. Okay. The first would be that these tools fundamentally lack a theory of mind. Okay. So that’s a term that cognitive scientists will use for the capacity that we humans have to imagine the mental states of another. And these tools can’t do that. There’s some dispute in the literature and researchers will say, well, if you run these sort of tests, maybe they’re kind of capable of it. I’m not buying it. I don’t think it’s true. And there’s plenty of evidence on the other side as well saying that they just don’t have that capacity. Fundamentally, what they’re doing is making predictions about what text to produce. They’re not imagining a mental state of the user who’s inputting things into it. Number two, I would say, is that it obviously misses out on a huge part of the cultural aspect of why we do and why we have education institutions and the relationships that we form. And I think that the claim that students are going to want to engage and learn from digitized tutors the likes of which Khan Academy and others are putting out, I think is woefully misguided and runs counter to literally thousands, if not hundreds of thousands of years of human history. Okay, so number one, doomed. Number two is to me like a kind of like, so what? Right? So I use the example of computers and the Internet as ubiquitous technologies that AI might join. So, like, let’s say that’s true. Let’s say that comes to pass. So what? Like, we have the Internet now, we have computers now. We’ve had both of these things for decades. They have not, I would argue, radically transformed education outcomes. The ways in which technologies like this become sort of utilities in our lives, transforms our day to day existence. But just because a technology is useful or relevant in some way or form does not mean emphasis, does not mean that it is somehow useful for education purposes and for improving cognitive ability. So I have absent a theory as to in what ways these tools are going to do that. Whether or not they become, you know, ubiquitous background technologies is kind of a, so what for me. Number three, the argument, the cognitive argument that this tool could be a useful example and non example of human cognition, I have a great deal of sympathy for. I am very curious about. There’s a lot, a lot that has changed just within linguistics, I would say, in the last several years in terms of how we conceptualize what it is these tools are doing and what that says about how we think and deploy language for our own purposes. We may have just scratched the surface with that. The new models that are getting released that are now quote unquote reasoning models have a lot of similarities in their functionality to things in cognitive science like worked examples and why those are useful in helping people learn. A worked example being something that sort of lays the steps out for a student as to here, think about this, then think about this, then think about this. Well, it turns out if you tell a large language model, do this, then do this, then do this, do then this, or just sort of program it to do that, their capabilities improve. So you know, without sounding too much like I’m high on my own supply, this is the cognitive resonance enterprise. It’s sort of to say, okay, let’s put this in front of us and instead of focusing so much and using it as a means to an end, let’s study it as an end unto itself, as an artificial mind, quote unquote, and see what we can learn from that.

Michael Horn: Super interesting, Ben, on, on that one. And I’m just thinking about an article I read literally this morning about where it falls short of mimicking, you know, the true neural networks, if you will, in our brain. So I’m pondering on that one now. I guess I, before we go to the outright skeptic take if you will, I’m sort of curious on like other things that you think AI won’t help with in your view, beyond what you just listed in terms of, you know, this broad notion of personalizing learning or AI as utility, if you will, and, and the so what question, like are there other things that people are making claims around where they think AI is really going to advance the ball here. And you’re like, I just, I don’t see that as a useful application for it.

Ben Riley: Well, you know, we launched into this conversation and we didn’t define what we’re talking about when we talk about AI. Right, sure.

Michael Horn: There’s different streams of it. Yep.

Ben Riley: Yeah. And I think that, like, when I’m talking about AI, and least have been talking about it in this context thus far, I’m talking about generative AI, mostly large language models, but it includes any sort of version of generative AI that is in essence, sort of pulling a large amount of data together and then sort of trying to make predictions based on that, using sort of an autoregressive process or diffusion in the case of imagery, but sort of like trying to essentially aggregate what’s out there, and as a result of that, aggregation produce something that sort of relates to that. If you’re talking about beyond that, like, who knows? I mean, there’s just so many different varied use cases. There’s, I was mentioning off air, but I’ll say now on air, there’s a great book, AI Snake Oil, written by a couple of academics at Princeton, which talks about sort of the predictive AI, which they put in a sort of separate category from generative AI, and they’re very skeptical about any of those uses. My fundamental thing is that to the extent people think like the big claim, right? And unbelievably, Sam Altman, the CEO of OpenAI, just a few days ago declared that, like, we’ve already figured out how to create artificial general intelligence. In fact, that’s like a solved problem. Now we’re on to super intelligence. I think people should be very, very skeptical of that claim. And there’s a lot of reasons why I would say that, which again, could eat up the entire podcast. But I’ll just give you one. What we now know is true, I think from a scientific perspective about human thought, is that it exists, it does not depend on language. Language is a tool that we use to communicate our thoughts. So if that’s true, and I would argue in humans, it is almost unassailably true. And I can give you the evidence for why I think we think that or why we know that, then it would be very strange if we could recreate all of the intelligence that humans possess simply by creating something like a large language model and using all of the power of all the Nvidia chips to harness what’s in that knowledge. Now what people will say, and frankly, this is where all the billions and the leading thinkers on this are trying to do is okay, well now we can only go so far with language. How about we try to do it for other cognitive capacities? Can we do that? Can we create neuro symbolic, as it’s called, AI that is as powerful, powerful as generative AI with large language models and sort of start to piece this together in the same way that we may piece together various cognitive capacities in our own brain and then loop that together and call it intelligence. To which I say, well, good luck. I mean, honestly, good luck. But there’s no reason to think that just because we’ve done it with large language models that we’re going to have the same sort of breakthroughs in the other spaces. So don’t know if this fundamentally answers your question, Michael, but I would say that it’s sort of like, you can have progress in this one dimension. It can actually be quite fascinating and interesting. But I would urge people to sort of slow down in thinking that it just means that, you know, all of science and humanity and these huge questions around whether we will ever be able to fully emulate the human mind have suddenly been solved.

The Skeptical Take 

Diane Tavenner: Yeah. Wow. So fascinating. I have so many things coming to me right now, including my long journey and experience with people who make extraordinary com, you know, claims and then kind of make the work a little bit challenging for the rest of us who are actually doing it behind them. But let’s turn now, we’re kind of steering in that direction, but let’s go all the way in on the skeptical take. And so I feel confident you’ve got some good material here for us. Like what is AI going to hurt specifically in education? Let’s start there, and how’s it going to do harm?

Ben Riley: Yeah, well, I don’t think we should use the hypothetical or the future. Let’s talk about what it’s harming right now. So I mean, the big danger right now is that it’s a tool of cognitive automation. Right? So what it does is fundamentally offer you an off ramp to doing the sort of effortful thinking that we typically want students doing in order to build the knowledge that they will have in their head that they can then use in the rest of their life. And this is so fundamentally misunderstood. It was misunderstood when Google was starting to become a thing and the Internet was becoming a thing. You would hear in education, well meaning people say, well, why do we need to teach it? If you can Google it. Right? That was a thing that many people said, put up on slides. I used to stop and listen and look. It makes sense if you don’t spend any time with cognitive science and you don’t spend any time thinking about how we think. And so I don’t, I don’t want to throw those people too far under the bus, but just a little, because now we know. We know this. Like, this is a scientific, like, as established as anything else is established. It’s like our ability to understand new ideas in the world comes from the existing knowledge that we have in our head. That is the bedrock principle of cognitive science, as I like to describe it. So suddenly we have this tool that says, you know, to the extent you need to express whether or not you have done this thinking, let me do that for you. You know like, this exists in order to, to, to solve for that problem. And guess what? It is very much solving for that problem. Like, I think the most stunning fact that I have heard in the last year is that OpenAI says that the majority of its users are students. Okay, the majority. Now, I don’t know what the numerator and denominator is for that, and I’m talking to some folks trying to figure that out, but they have said that at the OpenAI education conference, Lea Crusey, who some of you may know who was over at Coursera, got up and said, and they said, and I think they meant this is like, they were happy about this, that their usage in The Philippines jumped 90% when the school year started. What are those kids using it for? Yeah, you know, what are those kids using it for? Like, I don’t think, like, we need to stop pretending that this isn’t a real issue. And for me, people sort of go, well, it’s plagiarism, you could always plagiarize. And it’s like, not exactly. Not exactly like. And I think it actually is sort of both overstates and understates the case to talk about it in the context of plagiarism. Because again, the real issue here is that we will lose sight of what the education process is really about. And we already have, I think, too many students and too much of the system sort of oriented around get the right answer, produce the output. And I think teachers make this mistake, unfortunately, too often, I think a lot of folks in the system make this mistake of we just want to see the outcome and we are not thinking about the process because that’s really what matters. And building that knowledge over time. And you’ve got now, I mean I literally sometimes lose sleep over this. You’ve got a generation of students whose first experience of school was profoundly messed up because of the pandemic. And then right on top of that, we have now introduced this tool that can be used as a way of offloading effortful thinking. And I don’t think we have any idea what the consequences are going to be for that cohort of students and the potentially, like, dramatic deficiencies in a quality education that they will have been provided. That’s one big harm. There’s another. I mean, there’s many others, but there’s another that I’ll highlight here, too. I don’t know if you, either of you watched, I imagine you did, the introduction of ChatGPT multimodal system last year, which included the family Khan, Sal Khan and his son Imran were on there. I thought it was fascinating and speaks again to the amount of users who are students that OpenAI chose Saul and his son to debut that major product. If you watch that video closely, and you should, you’ll see something, I think, that is worth paying attention to, which is at multiple points, they interrupt the multimodal tutor that they’re talking to. And why not, right? It’s not a life form. It doesn’t have feelings. And we know that, it’s a robot. You know, to a degree. I don’t think we’ve really grappled with the implications of introducing something like human like into an education system and then having students who are students who are still learning about how to interact with other humans, that’s another part of education and saying, you know what, it’s okay to behave basically however you want with this tool, right? Like the norms and the sort of, you know, ways in which schools inculcate values and inculcate, sort of how it is we relate to one another could be profoundly affected in ways that we haven’t even begun to imagine, except in the realm of science fiction. And I think it’s worth looking at science fiction and pointing to how we tell these stories. I don’t know if either of you watched HBO’s Westworld, particularly the first season before the show went off the rails. But if you watch the, if you watch.

Diane Tavenner: Season one was a little intense, too.

Ben Riley: Season one was intense, but it was good. I thought it was good. And, and, but it was haunting. And one of the things that was haunting about it is it’s like for those who haven’t watched the show, it’s a It’s filled with cyborgs who are quasi sentient, but they, you know, people come and they’re at amusement parks and it’s like the old west and what can you do? You can kill them. You can kill them and people do that or worse.

Diane Tavenner: Right, yeah. Well, talk about the other bad thing.

Ben Riley: Right, right. I mean, but, you know, but it’s sort of like the fact that we now can imagine that sort of thing being a future where you could like humans, but not. The philosopher Daniel Dennett, who passed away, talked about the profound dangers of counterfeiting humanity. And I think that’s the sort of concern that is just almost not even being discussed at any real level as we start to see this tool infect the education system.

AI’s Impact on How We Think

Michael Horn: I suspect that’s going to be something we visit a few times in this series. But you’ve just, you’ve done a couple things there. One, you’ve, I think, more articulately answered, you know, a lot of the bad behavior we’ve seen on social media. How that actually could get exacerbated is not through deep fakes per se, but in terms of actually how we relate to one another. But you also answered another one of my questions that I’ve had, which is I can’t remember a consumer technology where education has been the featured use case in almost every single demo repeatedly. And you may have just answered that as well. I’m curious, a different question because I know you and Bror Saxberg have had sort of a back and forth about, you know, where is certain things that maybe it’s harming going to be less relevant in the future. And he loves to cite the Aristotle story. Right. About we’re not going to be memorizing Homeric length poems anymore. And maybe that’s okay because it freed up working memory for other things. I’m sort of curious to get your reflection on that conversation at the moment because I think Diane and I would strongly agree. Replacing effortful thinking, thinking that you can just, you know, have people not grapple with knowledge and build mental models and things like that, that’s going to have a clearly detrimental impact. Are there things where you say actually it’s going to hurt this, but that may be less relevant because of how we accomplish work or something like that in the future? I don’t know your take on that.

Ben Riley: Yeah, I don’t think you’ll like my answer, but I’m going to give you my honest answer.

Michael Horn: I don’t know that I have an opinion. Like, I’m just curious.

Ben Riley: Yeah, I mean, I’m not a futurist and I’ve made very few predictions ever in my life, at least professionally. One of the few that I did was that I thought personalized learning was a bad idea in education. And I’d be curious, I don’t know in this conversation another, whether you two reflecting back on that would go actually, you know, knowing what we know now, there were reasons to be skeptical of it and the, the I’m annoyed at the turn he seems to have taken because I used to like to quote Jeff Bezos. So with all the caveats around, you know, Jeff Bezos and anybody right now from big tech, he has said something that I think is relevant, which is he said, he’s asked all the time, you know, how the, what’s going to change in the future and how to prepare for that. And he says that’s the wrong question. He says, you know, the thing that you should plan around is what’s not going to change. He’s like, when I started Amazon, he was like, you know, I knew that people wanted stuff, they wanted variety, they wanted it cheap and they wanted it fast. And he’s like, that, as far as I could tell, wasn’t going to change. Like, people weren’t going to like, I want to spend more or take longer to get to me. And it’s like I said, once you have the things that won’t change, build around those. So I said it earlier, I’ll say it again. The thing that’s not going to change is fundamentally our cognitive architecture is the product of certainly hundreds of thousands, if not millions of years of biological evolutionary processes. It is further, I think, the product of thousands of years, tens of thousands of years of cultural evolution. We now have something, we have digital technologies that can affect that culture. So it does not mean, and I am not contending that our cognitive architecture is some sort of immutable thing, far from it. But on the other hand, it would suggest that what we should do is A, not plan around changes that we can’t possibly imagine, but B, maybe more importantly, and I would say this to both of you, not try to push for that future, you know, that we should fundamentally be small c, very small c, conservative about these things, because we don’t know, you know, I don’t know what the amount of time that took place back in Socrates and Aristotle’s time in terms of the cognitive transitions that took place, but they took place. My strong hunch not so much as the product of any deliberate choice, but to get a sort of social conversation about which ways in which should we talk to one another. And it was clearly the case that writing things down proved to be valuable in many dimensions. It may prove to be the case that having this tool proves very valuable in many dimensions. But let the time and experience sort that out rather than trying to predict it.

What Schools Can Do To Prepare

Diane Tavenner: Super helpful. I love where you’re taking us, which is into actual schools. So I appreciate that you’re like, let’s talk about what’s actually happening right now. And, you know, that is where my, like, heart and work always is, is in real schools. And so given what we are seeing, what you’re articulating about what’s actually happening right now in schools, and given that, well, I won’t say it as a given. What do schools need to do to mitigate the challenges you just said to, to recognize this as a reality that is coming our way that maybe can’t be put back in the box. Now, I’m going to say that with a caveat because I’m reading in the last day or two too, that it’s people declaring, you know, that they’ve won the cell phone war and cell phones are going to be out of schools here pretty soon. So maybe, maybe you actually believe it’s possible to kind of put it back in the box in schools. But, like, what’s the impact on schools and what do they do literally right now, given what you’re saying is actually happening already?

Ben Riley: Yeah. So great questions, all of them. So, I mean, thank you for bringing up the cell phone example, because I cite that often and even before there was this sort of wave now, both at the international level, national level, state by state, district by district, to suddenly go, these tools of distraction aren’t great for the experience of going to school and having you concentrate on hopefully what the teacher is trying to impart through the act of teaching. So we can, it’s not easy, but we can take control of this. Nothing is inevitable. So, you know, people always say, well, you can’t put it back in the box. You know, AI will exist, but how do we behave towards it? What ethics and norms do we try to impart around it? These are all choices we get to make. I like the phrase, and I’m borrowing this from someone named Josh Break, who’s a professor at Harvey Mudd. He has a wonderful Substack called I think It’s Just the Absent Minded Professor. But he writes a lot about AI in education. And his phrase is just you have to engage with it, but that doesn’t mean integrate. Right? So what I do think, you know, Diane, you kept saying schools. I just think it’s teachers, educators need to engage with it. That can still mean that the answer after you engage with it is no, not for me, and also no, not for my students. I think that’s a perfectly acceptable thing to say. And look, maybe the students won’t follow it, but that, you know, you’ve done what you can, right? And, and that is all you can do. There’s a teacher out there who I’m desperately trying to get in touch with, but she made waves. Her name is Chanea Bond. She teaches here in Texas. She made waves on Twitter a while back by saying, look, I’ve just banned it from my kids because it’s not good for their thinking. People are like, what? And it was like, she was like, yeah, no, it’s not good. Like it’s interfering with their thinking. So I’ve banned it. So that’s a perfectly reasonable answer. I also think that, you know, once you start to understand it at a basic level, I’m not talking about getting a PhD in back propagation and artificial neural networks, but just starting to understand it, you’ll start to understand why it’s actually quite untrustworthy and fallible and that you know, if you just think that everything it’s telling you is going to be accurate, you have another think coming, you know, and one of the things in the workshops that I’ve led that I’ve been very satisfied by is when people come out on the other side of them, they’re like, yeah, okay, so this thing isn’t reasoning and it’s not this all knowing oracle. And once you have that knowledge, once you’ve demystified it a bit, I think it gets a lot easier to sort of grapple with it and make your own choices and your own decisions about how you want to do it. I will say that right now, in the education discourse, it’s like, you know, things are way out of balance between sort of the hype and enthusiasm versus the sort of, hey, pump the brakes, or at least have you thought about this, if you’ll forgive me, but again, sort of, you know, it’s a, it’s a free resource. But if you go to cognitiveresidence.net we’ve put out a document called the Education Hazards of Generative AI, which literally just tries to, in very bite size and hopefully accessible form, sort of say, here are all the things you really need to think about and might be some cautionary notes across a number of dimensions, whether you’re using it for tutoring or material creation, for feedback on student work. Like, there’s a lot of things that you need to be thinking about and aware of. One of the things that frustrates me is that I see a lot of enthusiasts and this ranges from nonprofits to the companies that make these tools, sort of saying, well, teachers, fundamentally, it all falls to you. Like, if this thing is not factual or it hallucinates, like, it’s your job to fact check it. And it’s like, well, come on, like, A, that’s never going to happen, and B, like, not fair, you know, like not fair to put that on educators and just kind of wipe your hands clean. So I do think that’s something that, like, we’re still going to have to sort of sort through society on a, you know, social level as well as within schools and well as like individual teacher and ultimately students are going to have to bear some agency themselves about what choices they make around whether and how to use it at all.

What We’re Reading and Watching 

Diane Tavenner: I’m so appreciative of this idea of agency here. And I do think that that’s like, certainly a place that I’ve always been and is core to my values and beliefs as an educator is the importance of agency, not only for educators, but for young people themselves. And so, I love that this is such a rich conversation. We go on and on and on. But I feel like maybe leave it there. Like really real people, real teachers, real students, real agency. So grateful for everything that you brought up, so much to think about. And we’re gonna pester you for one last thought, which is Michael and I have this ritual of, at the end of every episode, we share what we’ve been reading, watching, listening to. We try to push ourselves to do it outside of our day jobs. And sometimes we seep back into the work because it’s so compelling. And so we want to invite you, if you have thoughts for us and to share them.

Ben Riley: So I told you I had a weird one for you here. So I was just in New Orleans and when I was in high school, for reasons that I won’t go in detail here, my family got really into the Kennedy assassination and the movie JFK by Oliver Stone came out. And I don’t know whether either of you have watched that film in a long time. It’s an incredible movie. It’s also filled with lies and untruths, and it’s much like in large language.

Michael Horn: I think we watched it in high school, but keep talking.

Ben Riley: Yeah. Yeah. Well, the thing that, the reason I bring it up is because Lee Harvey Oswald lived in New Orleans in the summer of 1963. And that movie is based on the case that was brought by the New Orleans District Attorney, a guy named Jim Garrison. But there’s a bunch of real life people who are in that movie or portrayed in that movie. And I just started to think about accidents of history where all of a sudden you could be, you know, just a person of relative obscurity as far as, you know, anyone broadly paying attention to your life. And all of a sudden something happens and now you become sort of this focus of study. And trust me when I tell you that every single person who had any connection with Lee Harvey Oswald in his life has become this object of study to people and books have been written. And so I’m trying, this is very bizarre, I know, but what I’m trying to do is think about and understand what it is like for people in that situation. Like what it is like to suddenly have your story told that you don’t have control of it anymore, you know, and if you know where, this isn’t supposed to be work related but in a way I think it does connect backup because it goes back to the fact that these tools are taking a lot of human created knowledge and sort of reappropriating it for their own right. And we haven’t got touched on that. I don’t think we need to now. But it’s sort of like it’s, there are a lot of artists who feel a profound sense of loss because of what’s happening in a our society today. That’s another thing I think worth thinking about.

Diane Tavenner: Wow, you’re right. I didn’t see that one coming. But it’s fascinating. Thank you for sharing it. I am unfortunately not going to stray from work today. I can’t help myself. Three of my very good friends have recently released a book called Extraordinary Learning for All. And that’s Aylon, Jeff Wetzler, Janee Henry Wood. And it’s really about the story of how they work closely with communities on the design of their schools and in a really profound and inclusive way. And so I’m deep in that, been involved in that work for a long time and think it’s just a really powerful kind of inspiration slash how to guide of how communities can really take agency over their schools and own them and figure out what they want and what matters and what they need and how they design accordingly.

Michael Horn: So I was gonna say now, Jeff has appeared twice in a row in our book recs, I think, on episodes or something like that. So love that. Diane, I’ll wrap up with saying I’m gonna go completely outside of, I think, the conversation today. But, Ben, you may say it actually relates as well, because I’ve been binging on season two of Shrinking. I loved season one and season two, with the exception of a couple episodes in the middle has been no exception, I think. So I’m. I’m really, really enjoying that so far. And I suppose you could connect that back to.

Ben Riley: What is Shrinking? I don’t know. I have to. I don’t know what it is.

Michael Horn: Okay, it’s basically about three therapists in a practice and one who’s grappling with the deep personal tragedy. And Harrison Ford is outrageously hilarious. Yeah.

Diane Tavenner: So good. It’s so good. Okay, well, I’m gonna tag on to your, you know, out of work one and say yes, we love Shrinking as well.

Michael Horn: Perfect. Perfect. All right, well, we’ll leave it there. Ben, huge thanks for joining us. For all of you tuning in, huge thanks for listening. We look forward to your thoughts and comments off this conversation and continuing to learn together. Thank you so much as always, for joining us on Class Disrupted.

Sixteen tables, two seats to a table, all in rows, face front “because that’s where the instruction is coming from,” he said.

About the only technology in the room: small handheld whiteboards, dry-erase pens and small stacks of index cards. The walls are almost entirely bare. And phones are out of the question, stowed in backpacks before class.

It’s intentional, said Harvard, who teaches Advanced Placement Psychology at James Clemens High School in Madison, Ala., a suburb of Huntsville.

Over the past decade, he has become something of an expert in focus, memory, forgetting and distraction.

Harvard has put these principles into his first book, published last week, titled, appropriately, . 

Harvard hopes the book will offer practical advice to teachers on how to use the principles of cognitive science to create better learning environments.

The time is right for a new book about attention, said , a professor of English at the City University of New York and founding director of CUNY’s Futures Initiative. She said she’s excited to see Harvard’s work.

Davidson noted several indicators of rising inattention, from falling reading scores to the growth of media misinformation and the higher prevalence of young people who say they’re with traditional education. 

“I think people are really seeing that what it means to pay attention is important,” said Davidson, who wrote 2011’s . 

Harvard mostly focuses on more intentional teaching methods that reduce distractions and help students manage the vast amount of content they’re called upon to remember —  often called “.”

These ideas are decidedly not on tap in most teacher preparation programs, said Harvard, who earned his master’s degree in education in 2006. His coursework contained “nothing on cognition — there was nothing on the brain, nothing on how we learn.”

‘Why don’t I already know about this?’

It wasn’t until 2016, a decade after graduate school, that Harvard happened upon the now-defunct Twitter account “The Learning Scientists.” In plain language, educational psychologists from around the world laid out the basics of cognitive science for educators. 

Harvard was gobsmacked. Instead of just shooting in the dark, he finally saw research on the effectiveness of various learning strategies. 

He found himself instantly hooked and soon for the group. That led to his own website, which eventually became the popular blog .

Nearly a decade later, he’s traveling the world, speaking at conferences about strategies that affect students’ ability to channel ideas into long-term memory. He’s lost count of how many times he’s had to inform audiences that — humans can’t consciously focus on more than one thing at a time.

Harvard subscribes to something he calls the “SAR method,” an accessible way for students and teachers to think about memory. When they’re about to start a lesson, he tells students that memory follows a three-step process: Sense, Attend and Rehearse. 

“You can hear your teacher,” he said. “You can see your teacher. You can see the board. You can sense it. But are you attending to it? Are you paying attention to it, or are there things getting in your way? Are you trying to multitask? Is the person sitting next to you talking?”

Once a student attends to the material, the rehearsal happens. That’s perhaps the most important and tricky part. In the book, he likens it to an athlete’s ability to learn a new routine. If he or she doesn’t rehearse before the big game, he writes, “that would not be a good recipe for success on the playing field.”

Rehearsing in the classroom can take the form of a multiple-choice quiz, a discussion or a project. The key is to access the material from memory and use it appropriately.

Accordingly, he begins many classes by simply asking students to review what came the day, the week or even the month before. Retrieving those memories, he said, makes them more likely to be there the next time the brain goes looking for them.

Another principle he employs is “wait time.” When most teachers ask a question, they’ll settle for the first student with her hand up. But Harvard adds a step, ordering students to retrieve their handheld whiteboard. Before anyone can answer out loud, everyone must attempt an answer in writing.

“Now they’re committed to thinking,” he said. “They’re committed to writing something down. It seems like such a simple thing, but when you make the students do that, you give them time to think.”

As they’re studying, he’ll often give students a kind of slow-motion, three-stage assessment he calls “Brain-Book-Buddy” to offer a more honest take on what they actually know.

In the first assessment, they answer a series of questions from memory. Then they fill in the answers they couldn’t remember with the help of their notes. In the final test, they can talk to classmates.

“They end up getting all the right answers, but they’re also acutely aware of what they actually knew, what they knew with their notebook, and what they had to ask their buddies, their peers, about,” he said. “It’s an ongoing conversation of them thinking about their thinking.”

‘Attention Contagion’

Lately Harvard has been evangelizing most eagerly about an emerging topic in cognitive science known as “.” Only a handful of small-scale studies exist on the topic, but Harvard says the evidence is compelling.

In the research, students pose as attentive or non-attentive classmates, and researchers judge how well actual subjects attend to lessons in their presence — how many notes they take and their performance on post-lesson quizzes. The results suggest that seatmates’ behaviors have a profound effect: When a student is surrounded by inattentive peers, the behaviors are contagious. It works the other way as well: If a student is surrounded by peers who are visibly paying attention, they’re more attentive. 

had undergraduates watch a video lecture with a “classmate” posing as someone who either seemed attentive — leaning forward and taking notes — or slouched, shifting his gaze, glancing at the clock and taking infrequent notes. Researchers found that being seated behind these classmates had a profound effect: Subjects sitting near attentive students took significantly more notes and rated themselves as being on task. They also scored more than five points higher on a multiple-choice quiz.

Other studies have replayed the dynamic, with similar results. The findings even hold true for students observing one another in a Zoom-like virtual environment, where all that’s visible is a student’s face staring into a webcam.

In other words, Harvard notes, attention and inattention can actually pass through the Internet.

He considers the findings especially resonant because the “contagion” doesn’t come from obviously bad behavior like yelling, interrupting a teacher or staring at a phone. It’s stuff that he and most other teachers would typically let slide.

“They’re just slouching in their chair,” he said. “They’re just not taking notes. They’re gazing out the window.”

What the studies show is that attention operates by a kind of quiet osmosis, in some cases literally felt but not seen.

, the researcher who has pioneered this work, emphasized the “non-distracting” nature of the inattentiveness in his studies, noting that it’s “driven by more than just peer distraction.” Peers can detect these inattentiveness cues, he told ����Ӱ��, even via tiny changes in the case of the online environment, suggesting that students “pay attention to their peers on webcam — even when the video thumbnails are quite small.”

More data needed

In an email, Forrin cautioned that attention contagion ”has not yet been studied in real classrooms,” only in laboratory settings with video lecturers. But he said he’s confident that attention and inattention “can spread between students during lectures,” and that this spread affects learning. Students “are attuned to their peers’ motivation to learn” and pay more attention when they infer that others have strong learning goals. They pay less attention when they sense weak or no goals. 

He suggested that teachers do their best to cultivate these goals in their students. They should also let students choose their own seats so they’re not consistently sitting near inattentive peers.

But he said more data are needed to determine whether these phenomena occur in real classrooms, especially with live teachers and different levels of student motivation.

Davidson, the CUNY scholar, said research on topics similar to attention contagion go back all the way to , who at the turn of the 20th century was studying the social aspects of “vivid” thoughts, distraction and focus. More recently, she noted, the psychologist Danie Kahneman, who helped establish what has become behavioral economics, studied .

And of course TV producers who pioneered the “canned laughter” of laugh tracks on early TV knew that suggestions of an engaged audience make viewers respond in kind. 

But perhaps the greatest experts in attention contagion, Davidson said, are stand-up comedians — she interviewed several for her 2011 book, and they told her that visibly bored audience members are “the kiss of death” in live performance. “People fall asleep in the front row, and pretty soon they’re falling asleep in the whole theater,” she said.

Harvard, for his part, is convinced that attention contagion in the classroom is real — and he tells students about the research.

“It’s powerful for students to hear that simply being inattentive can distract someone else from learning,” he said.

More broadly, he said, cognitive psychology has simplified his approach to teaching, allowing him to focus on proven strategies that are neither traditional nor progressive. 

The most cynical person, he said, would probably say his classroom is “too traditional. But I’m not thinking, ‘Do I want a traditional or a progressive classroom?’ When I designed it, I’m thinking, ‘How can I put my students in the best situation where they can pay attention to what they need to pay attention [to] and be distracted the least?’ That’s everything that I’m thinking about, and nothing else.”

But a new book argues that the current approach remains deeply flawed. Though phonics instruction has emerged as a key component of reading lessons, stagnant NAEP scores, among other measures, suggest that something is missing — a focus on substantive knowledge, including detail-rich lessons in science and history. 

Author Natalie Wexler, whose 2019 book advocated a greater emphasis on these topics paired with explicit instruction, has said these principles are supported by cognitive science. A content-rich curriculum, she maintains, allows students to go deeper, helping information stick and building an academic foundation that allows them to write more easily, creating a kind of virtuous circle of reinforcement: The more they know, the better they can write; the better they can write, the more they can learn.

Six years later, Wexler is back with a new warning. In her book, , out Feb. 3. (pre-orders open today, Jan. 21), she says the benefits of improved reading instruction will go to waste if we don’t offer students a more vibrant, content-rich set of lessons to go along with it. 

She spoke recently with ����Ӱ��’s Greg Toppo. Their conversation has been edited for length and clarity. 

����Ӱ��: Your book The Knowledge Gap came out in 2019. A lot has happened since then, including a pandemic and an explosion of interest in the science of reading, thanks in part, to the work of folks like . Would you say we’re in a better place in the knowledge discussion than we were in 2019?

Natalie Wexler: Yes, definitely. For one thing, there are now a number of knowledge-building curricula available that were not around when I was researching the book. There are more choices than there used to be. And although we don’t have really reliable data on what curricula are really being used, all indications are that more and more districts and schools are using those knowledge-building curricula. That’s been a very promising development. It’s still a minority, but certainly more than in 2019. Emily Hanford and other science of reading advocates have done a great service to the public and to the nation’s children by shining this spotlight on things that are problematic about typical phonics instruction. The risk is that it can lead, and has led in some places, to the assumption that if we just fix the phonics part of reading instruction, everything else is going to be fine. Unfortunately, that’s not the case. 

A lot of people see the science of reading as just “more phonics.” How do you describe this more comprehensive approach?

People outside the education world assume that schools are teaching social studies and science and all of those things. I have to do a lot of explaining when I talk about how we’re not building knowledge in school effectively. With The Knowledge Gap, the publishers expected that the audience would be primarily the general public and parents. But where it’s really taken off is among educators. And it’s because it’s a lot easier, certainly for elementary level and maybe some middle school level educators, to understand the argument, because they’re living what I’m describing: There isn’t much content in the elementary curriculum, and there is a lot of emphasis on teaching reading comprehension skills, making inferences as though they were abstract skills you can teach directly and apply generally. Many of them have seen that that doesn’t really work very well. 

As I was reading your book, it reminded me of some of the conversations I’ve had with Joy Hakim, who wrote the great series, A History of US and . Her books are favorites among people who are enlightened about this topic. One of the things she says is that we’re underestimating how much our kids can understand if they’re exposed to difficult material. Is that the right word, underestimating? 

“Underestimating” is the right word, and I use that a lot myself. But you have to be careful about what we’re underestimating. It is often assumed among educators that young children won’t be interested in history or can’t handle history because it’s just too abstract, too remote from their own experience. There’s no evidence to support that. And in fact, there’s anecdotal evidence that kids can get very interested in history.

I’ve seen this myself: second graders getting fascinated by the War of 1812. But at the same time, we’ve overestimated kids’ abilities sometimes to handle certain abstractions. I open The Knowledge Gap with a teacher who’s trying to teach kids the difference between a subtitle and a caption, which is abstract but not particularly interesting to them. They don’t get it. They want to know what’s going on in the picture. What is that shark eating? But the teacher feels that it’s more important. This is what her training in the curriculum has led her to do, to focus on the abstract difference between a caption and a subtitle.

You and I have both interviewed teacher in Baltimore, and I love what he tells you: He was initially skeptical that his students would like a Dust Bowl novel, , but as the drama unfolds, they’re hooked. I wonder what that tells you, not only about the topic, but about how he was able to approach it and make it come alive.

said that you can teach almost anything to a child of any age if you do it in a way that makes sense. Those weren’t his exact words, but if you engage kids, they will get interested in all sorts of things that have nothing to do with their own experience. If you basically tell them a good story, that’s the way you can teach history, science. This is what Joy Hakim does so beautifully in her work, both in history and science: telling stories that really hook kids, and then they learn a lot, almost effortlessly, along the way. 

There’s a lot of emphasis on having kids “see themselves” in what they’re reading, which is important. But it is at least as important to expand their horizons to other realms of experience. Fiction, novels especially, are a great way to do that. As Kyair said, when they learn that the main character’s little brothers died, they care. They care about this story and these characters. There’s also some evidence to show that this is the way empathy develops, through reading fiction about lives that are very different from our own.

In Chapter 3 of the new book, you talk about teachers colleges, and note that today’s teacher educators — that is, the people working in the colleges — have been shaped by “a system that devalues knowledge and prioritizes engagement.” In a way, you can’t blame teachers for this crisis.

Absolutely. It is no individual’s fault that we are where we are. It’s a systemic problem, so it’s not going to change overnight. It’s difficult, not just for teacher educators to step out of this system, but also the teachers themselves. If you’ve been teaching in a certain way for years in the sincere belief that you are doing a great job, and someone casts doubt on that, it’s a very difficult message to take in.

What really amazes me is how many teachers, despite the painfulness of the message, are nevertheless embracing it because they really care about the success of their kids. With teacher training, it’s going to be hard to change that overnight. We’re really trying to fix a broken system with the products of that system, which is very difficult. I don’t think we can rely on teacher training to change the system. Once teachers are on the job, we also need to continue communicating this message, doing training to undo some of the training they’ve gotten pre-service. 

Historically, teachers also haven’t learned much about cognitive science. Do you get a sense that’s improving?

As I say in the book, there are efforts. is an organization that is doing great work with some institutions of teacher training, but it’s going to be very slow. There are hundreds of programs that train teachers, and just a handful are signing up to bring their curricula in line with principles of cognitive science. Even within those programs, not every teacher is on the bandwagon. You can’t really, at the university level, control what goes on in the classroom. Professors are used to having a lot of autonomy. 

Let’s talk about writing. You’re the co-author of as well. Reading your new book, it seems that writing brings together a lot of your ideas. Can you talk a bit about the importance of writing?

Since I finished writing both of those books years ago, I have continued to think more about the relationship between reading and writing and learning in general. I’ve become more and more convinced that the combination of a content-rich curriculum and explicit, manageable writing instruction embedded in that curriculum can provide all the benefits of cognitive science-informed instruction, and possibly more. Without going into a lot of detail, we have evidence that when you write about what you’re reading or what you’re learning about, it enhances your learning. It enables you to retain the information better, it enables you to understand it better, and it enables you to think about it analytically.

The problem is that writing is really difficult. We have studies, like write-to-learn studies, where they have kids write about the content that they’re learning. Overall there’s a positive effect from that. But in one meta analysis, in 18% of these studies there was . In other words, kids writing about what they were learning actually retained less of it. It’s impossible to know why. But the reason is we sometimes ask kids to just write without giving them enough support, and that is cognitively overwhelming, so they don’t get the potential learning benefits. 

So what’s the key?

The key is to make writing manageable, not cognitively overwhelming, but still requiring some effort. The best way to do that is to start at the sentence level — because if writing is hard, then writing at length is only making it harder — and explicitly teach students how to construct sentences and eventually clear linear outlines for paragraphs and essays that are embedded in the content they’re learning about. If you do that, you’re having them engage in , which we know is a very powerful boost to retention of information. You’re also having them engage in elaboration, explaining what they’re learning about, giving examples, all of that. That has been shown to really help with comprehension. You’re familiarizing them in a powerful way with the complex syntax of written language, which can be a real barrier to reading comprehension.  

You say that content-rich curricula are under fire from both sides, the left and right. I love the anecdote where you visit a small town in Ohio where this group of parents objected to the use of the words “God” and “Goddess” in a second-grade unit on Greek mythology. You note, “It’s hard to imagine how children could truly understand Greek myths or ancient Greek culture without hearing those words.” I have two questions. Number one, how do we get out from underneath this? And number two, is there a way in which this is kind of a red herring? 

This is coming not just from the right and not just from the left. The same curriculum has been attacked, sometimes, from both sides for different reasons. What we need to fundamentally do is realize that compromise is essential, and it’s got to be compromise that doesn’t interfere with kids’ ability to learn. There’s been a lot of opposition from the right to teaching about Greek myths in a curriculum called . Sometimes it’s perceived as trying to proselytize about Greek myths or other religions, Buddhism, Hinduism. When school leaders have explained to the community, “No, this isn’t an attempt to convert kids to these other religions. It’s a part of teaching them about history and other cultures,” sometimes those controversies have been defused — not in every instance.

Another thing to bear in mind, though, is that sometimes the people who are protesting are not representative. They’re maybe a small but very vocal group of parents. You have to ask: Does it really make sense to deprive all students of exposure to some valuable information because a small group is protesting? Maybe there’s another way to handle it, some alternative texts or something for those kids. But fundamentally, everybody needs to realize that the curriculum is not going to align with your individual preferences about what you would like your kids to learn. And we have to find consensus. There’s more consensus than there appears to be, which kind of gets to your second question: The media have kind of elevated these conflicts. In many instances, there isn’t that much conflict.

Is there anything you see in the landscape that gives you hope that we are moving in the right direction?

For one thing, I have gotten many invitations to speak recently. The interest in this, at least from my limited personal perspective, is not dying down. It’s only growing. And that’s encouraging. There are other people taking up this message. I’m seeing the beginnings of a recognition that phonics instruction is important, but we may be overdoing it with all of the focus on it in some places — the one generalization you can make about American education is you can’t generalize, because who knows what’s really going on?

But in some places, schools are spending an hour a day on phonics and giving short shrift to some of these other important components of reading, like building knowledge. That really relates to reading comprehension. Even some of the people who have been in the forefront of the science of reading movement, like , have been saying this: Let’s not overdo it, because there’s an opportunity cost, and one of those opportunities that’s being lost is the chance to build a kind of knowledge that kids will need to read and understand the texts they’ll be expected to read in years to come.

On a recent afternoon, Caroline Able, a first-grade teacher at North Frederick Elementary School in Frederick County, Maryland, sat in a small office with her principal, Tracy Poquette, carefully practicing the next day’s math lesson.

Able, who is in her third year teaching, walked through each step, demonstrating how she was going to present comparisons between two numbers, then what students would do. She sometimes stopped to focus on granular details: Should she go over math vocabulary words like sum and difference beforehand, or will her students remember what they mean? Should students write down problems and answers in notebooks, or on mini-whiteboards?

Poquette recommended the whiteboards. “You’re going to ask them to hold them up,” Poquette coached Able, miming holding a whiteboard in the air. “Then you can see their answers, and how they got to that. Every student is responding.” 

Giving students multiple chances to “respond,” or provide answers, is a learning strategy , and part of why Able is here — to ensure that she’s incorporating evidence-based practices into her teaching. The sessions are meant to accelerate student learning and take some of the guesswork out of becoming an effective teacher, part of a larger district plan to incorporate research from the fields of neuroscience, educational psychology and cognitive science — often referred together broadly as the ‘science of learning.’

Frederick County, situated about 50 miles north of Washington, D.C., and 50 miles west of Baltimore, is a diverse district , and one of only a handful to use learning science research to try to improve schools at scale. Launched in 2015, it’s the centerpiece of a school improvement plan, and leaders say the goal is to raise academic achievement overall, as well as shrink stubborn gaps between more advantaged students and their less advantaged peers. 

“As a district, we’ve been talking about achievement gaps for a long time,” said Margaret Lee, Frederick County’s director of organizational development who has led the charge toward the science of learning. “I’ve seen it in every role that I’ve had, always looking at what could make the difference. Like every district in America, every silver bullet that people thought up had been peddled to us. It started to frustrate me that none of these things were making a difference, and that was a catalyst that led us here.” 

The district is seeing steady progress in a positive direction, even when accounting for pandemic-related learning loss. Third-grade , for example, on the Maryland Comprehensive Assessment Program test, rose from 49.5% proficient in 2018, to 60% proficient in 2023 — 12 points above the state average. In math, students from disadvantaged groups have also seen steady gains. African-American third graders were 38% proficient in 2018, but rose to 43.8% by 2023; over five years, low-income Title I third graders slowly grew from 32% to 37.6% proficient. 

Amid a of learning science and a spotlight on curriculum reform, experts are beginning to look to districts like Frederick County to gauge whether it can be a model for academic improvement. Unlike more common state plans reforming how , or increasing support for struggling students, the Frederick County plan is tackling learning as a whole — across subjects and grades — to systematically alter the paradigm of how teaching and learning happens throughout its schools.

Training adults on how the brain learns

Frederick County’s plan turns on a single premise: who work with kids don’t know how the brain learns, and haven’t been exposed to the body of research on which teaching practices are more likely to support it. 

that applying cognitive science principles and strategies to classrooms are “significant factors affecting rates of learning and its retention in many everyday classroom situations,” with certain caveats regarding the limitations of what scientists currently know about when and where to implement them. But within universities, scientific research on learning has historically been separate from teacher training, and misunderstandings about how learning happens are common in the field of education. They’ve led to such disproven ideas as children having “,” like being a “visual” or “kinetic” learner, or using the to teach reading, prompting students to try to guess at unfamiliar words using context clues like looking at pictures. 

The district has made educating faculty and staff on cognitive science a top priority. In 2017, Frederick County formed a partnership with and began training teachers, instructional coaches and leaders in the , including an understanding of how memory works and its pivotal role in academic learning; creating classroom environments that reduce obstacles and distractions while maximizing student memory; and creating effective ways to test whether students have learned the needed material.

The training homed in on how to translate findings from cognitive science and educational psychology into classroom practice, including in learning new material, meaning direct instruction heavily guided by the teacher, and why students need to understand what they read and form connections to new learning. Classroom changes also include specific learning strategies like retrieval practice and interleaving, in which teachers go back to learned material in multiple ways, spaced out over time, which has been students’ memory of what they learned. 

The training has changed the way the district is approaching content subjects like math. Stacy Sisler, a secondary math curriculum specialist for Frederick County, credits increased knowledge of learning research with steady gains middle schoolers have seen in math across the district. She first learned about the science of learning through the district training, and admits she was initially reluctant to adopt the changes. The more she learned, however, the more Sisler began to think the research made sense, and was applicable to every math classroom.

“As I started to learn more and gain a deeper understanding, then it became — how does instruction change because of this?” Sisler said. “We don’t just say it and it magically happens, so what does that actually look like?” 

Under her leadership, curriculum and instructional practices were re-designed to better reflect the research. Middle school math teachers have been trained in practices like teaching math more directly using example problems, checking student work multiple times during class time to gauge student understanding and incorporating more math practice both into each lesson and across lessons. 

Lee said even when considering how hard it often is to pinpoint what caused learning gains, the instructional changes coincided with significant improvements for students in Frederick County. Over five years since implementing the changes, middle school math students’ benchmark assessments have grown, in some schools by as much as 20%, especially among students of color and English learners. Over the same five-year period across the state of Maryland, students of color and English learners’ math proficiency has declined. In 2023 for example, only 8.2% of Black middle school students were proficient in math, down 8 percentage points from 2018. 

‘Using the time we do have differently’

New teachers across the district are onboarded in a three-year science-of-learning coaching program, which includes lesson coaching like first-grade teacher Caroline Able’s, but also group study. The aim is to give new teachers evidence-informed knowledge and tactics to decrease some of the trial-and-error that comes along with being a beginner. 

First-year eighth-grade math teacher Elizabeth Sypole’s monthly training is currently focused on evidence-based classroom routines that foster students’ attention.

Sypole has learned techniques like , a simple hand motion followed by a pause meant to help students get quiet quickly. Previously unaware of the technique, Sypole said it has been instrumental in her classroom management. “Literally within two days of doing it, everybody is quiet. It’s so much less stressful than trying to get everybody to quiet down. They know exactly what to do now and it’s just the routine.” 

Leaders get the training, too — principals, assistant principals and supervisors are focused on equity, and how schools can eliminate learning gaps between groups of students. Kent Wetzel, the district’s leadership development specialist, trains leaders in researcher , which include presenting new material in small, manageable steps and providing extra support for students if the task is especially difficult. The idea is to make learning as accessible as possible to everyone. 

The training, book studies and coaching sessions focused on the science of learning make up the heart of the district’s professional development, and therefore don’t require tons of extra funding or extra time for educators and leaders outside their contract hours, said Lee. In the past, professional learning brought in from outside vendors were “one-off” learning experiences not tied to any bigger picture or goal. Now all professional learning must meet a set of district standards for being “evidence-informed and equity-driven,” ensuring the entire district is swimming in the same direction. 

“We haven’t made extra time, we are just using the time we do have differently,” Lee said. 

While much of the district training is mandatory—like district-wide professional development and leadership training—other parts are optional or opt-in, like teacher book studies and principal coaching. The district is hoping that by making the science of learning training something gradual that takes hold naturally, it will win buy-in from the most experienced staffers over time because it was not a one-and-done push. 

Bernard Quesada, the veteran principal at Middletown High School, has embraced the science of learning approach to teaching. He said the organic approach and long-term picture has been key to its success at his school of mostly accomplished, veteran teachers. 

“When these things become mandates, and schools have to comply, you get a lukewarm reception,” Quesada said. “Schools get initiative fatigue.” 

Middletown teachers have adopted the new learning, Quesada said, because administrators have been intentional to connect the research to what teachers are already doing well. Quesada quoted learning researcher and retired University College London professor — a speaker he heard at a recent science of learning conference. 

“Wiliam said, ‘There’s no next new, big thing. It’s a lot of old, small things that work and are boring,’” Quesada laughed. “That’s about as true a statement as I’ve heard in my life.”

‘Guilty of chasing the next greatest thing’ 

On the other side of the country, in rural Delta County, Colorado, teachers are working on asking students better questions to get them thinking stronger and deeper — moving beyond basic factual answers to more “how” and “why” questions that require students to think not just about the answer, but how they got there.  

Like Frederick County, the small southwestern Colorado district with one-quarter English language learners and 65% low-income students has been training all their teachers and school leaders in the science of learning. Also like Frederick County, the district has taken a “no-silver-bullets” approach and has revamped professional learning, putting learning research at the center, with deep dives for teachers and leaders into cognitive science principles like “,” a technique where teachers design lessons that require students to evaluate, provide reasoning and detailed explanations for learned material. 

The district’s science of teaching and learning lead, Shawna Angelo, said she’s looking to help teachers “align how the brain learns with how we are delivering instruction.” 

The focus on effortful thinking was supported by , an organization that has worked for nearly a decade to improve teaching by getting the scientific principles of learning into more classrooms.  

Executive Director Valerie Sakimura sees districts like Frederick County and Delta County as models for improving academic achievement in more school systems across the country. “The priority for our work is helping teacher preparation programs and partnering districts trying to support teachers around the science of learning,” she said. “Our particular focus is aspiring and early-career teachers.” 

Deans for Impact is also brokering partnerships between school districts and local universities, offering coursework and training on cognitive science principles for student teachers. Teacher training facilities as varied as the and are breaking down the longstanding barrier between teacher training and research science, teaching future educators about how learning happens long before they step into a classroom. 

Frederick County has partnered with Hood College, where many local teachers get their degrees, to design coursework and provide instruction based on the science of learning for student teachers. District instructional coaches and mentor teachers work with teachers in training as well, giving them a chance to watch evidence-informed techniques in action and practice them in their student teaching.

Michael Markoe, deputy superintendent for Frederick County, said through all this work, the district is trying to create a throughline, where all teachers, coaches, principals — everyone is moving in the same direction, speaking the same language, all based on the research. When school leaders recently inquired about personalized learning, for example, where students progress and master subjects at their own pace, Markoe reminded them that the district is, for the time being, focused on only one thing: evidence of effectiveness. 

“I’ve been in education almost 30 years. I’ve been guilty of chasing the next greatest thing,” he said. “If we are going to advance personalized learning, we have to see the research behind it and ensure it’s the right thing for our children.” 

Getting the entire district on board is long, slow work. Because there are no mandates, some schools haven’t embraced the science of learning, or have chosen to focus on other priorities, despite leadership’s wholesale commitment to the methodology. 

But Lee, the district’s organizational developmental director, isn’t deterred.  

“I compare it to moving an aircraft carrier. To move the ship, you are making lots of tiny moves in the same direction. If you spin a wheel in a school system, you will throw people off the ship,” she said. “Public education isn’t patient. Everyone wants to fix it tomorrow, but those things don’t work.” 

And here’s a twist: The same core difficulties are bedeviling teachers too.

The findings, contained in a new working paper, are believed to be the first to identify brain changes as an explanation for why students have suffered, both inside and outside the classroom, since the pandemic drove millions out of the classroom. 

, a Harvard University psychologist who studies the effects of stress on executive functions and who is the study’s lead author, said the new findings offer the first evidence to help us “understand the ‘why’” of the pandemic downturn — “what is actually causing all these issues that we’re seeing and talking about in the news.”

The paper, from the educational assessment and services company“, examines the cognitive skills of students nationwide and finds that, simply put, over the past several years, kids’ famously ever-changing brains have changed for the worse. 

Since the pandemic’s onset, students across all ages and economic levels have begun to demonstrate weaker memory and “flexible thinking” skills — those represent the mental bandwidth needed for multitasking, shifting from one activity to another and juggling the day’s demands. But for a few groups, such as younger and lower-income children, the changes have been more profound.

They also show that their teachers’ brains are weaker in almost identical ways, which could help explain high rates of frustration and burnout. They suggest school districts have their work cut out for them if they want to keep their best employees on the payroll and returning to the classroom each fall. 

Understanding the ‘why’ of pandemic downturn

The data come from a large, widely-used assessment, the , developed in 2013 at the University of Pennsylvania. It consists of a series of cognitive tasks that measure subjects’ accuracy and speed in several major cognitive domains, including working memory, abstraction, sustained attention, episodic memory and processing speed. 

MindPrint has administered the assessment periodically to its clients over the past decade. The most recent rounds totaled 35,000 students and 4,000 teachers in 27 states.

By most measures, U.S. students are suffering. Last year, NAEP scores showed the average 13-year-old’s understanding of math dropping to levels last seen in the 1990s and reading levels dropping to 1971, when the test was first administered.

More recent research has shown that while older children are showing encouraging signs of academic recovery, younger kids aren’t making the same progress. Many students who weren’t even in a formal school setting when COVID hit are already falling behind — especially in math.

The Penn assessment found that children who attended elementary or pre-school during the pandemic and who are now 8 to13 years old showed the largest declines in memory. 

“Younger kids haven’t really developed a lot of these core cognitive skills,” Tsai said. “It hasn’t solidified for them, either through development or just through practice in the classroom. And so younger kids are more vulnerable to these pandemic shifts.”

But students across all age groups showed worse flexible thinking, which researchers now theorize contributes to lower academic performance — as well as challenging behaviors.

Tsai said kids from lower income backgrounds were more vulnerable to these changes, specifically in verbal reasoning and verbal memory, than their higher income peers, with bigger declines in verbal scores, which are highly correlated with academic achievement in all subjects.

Adults in the study had similar declines in both memory and flexible thinking, possibly explaining higher reported levels of and .

Nancy Weinstein, MindPrint’s CEO, said weaker flexible thinking isn’t necessarily a problem for experienced teachers who have developed strategies to cope with stressful situations and can modify plans on the fly. But those with less experience may be unable to change gears when lessons go astray or students act out in class. That may lead to higher teacher burnout.

Across the board, teachers’ skills suffered in areas such as verbal and abstract reasoning, spatial perception, attention and working memory, but they saw the greatest losses in verbal memory and flexible thinking.

“If we care about that, we need to know how to help them,” Weinstein said. “And there are some tried and true things you can do.”

She said schools should consider sharing data like this with teachers so they can understand that their frustration in class might not be due to students alone. That could make a big difference, she said, in “their willingness to put in the effort to change, as opposed to saying, ‘Why bother?’”

For students, Weinstein said, offering them more opportunities to practice skills with between study sessions could help. Schools should also consider “” techniques that break learning into chunks and address each individually.

Could such techniques help students — and teachers — regain a measure of pre-pandemic skills? Weinstein suggests the answer is “Yes.”

“The environment will matter, but certainly we can regain some of that if we do the right things,” she said. “And we know what the right things are to do.”

Crystal Green-Braswell, coordinator of staff wellness and culture for the Little Rock School District in Arkansas, said offering the Penn assessment to teachers and staff has helped many think more deeply about their work — and about their own thinking. 

“People who have had the assessment will say, ‘Now, you know my processing speed is slower — y’all are going to have to give me a moment,’” she said. 

That’s a huge change in a profession in which most workers have been asked “to take ourselves out of the equation and just get the work done,” Green-Braswell said. 

She sees offering such insights to educators as part of “rehumanizing” teaching. “When we provide this kind of assessment and we provide this kind of space for folks to actually get to know themselves, we are humanizing this profession and helping people to realize, ‘You play a role. You play an active role. You matter.’ ”

As the founder of in 2015, he enlisted university education school deans to incorporate findings from into teacher preparation. Before that, he spent five years as policy director of the , which underwrites new models of schooling. In his new endeavor, , which he calls “a think-and-do tank,” he’s pushing to help people think not only about how we learn, but how generative artificial intelligence (AI) works — and why they’re different.

His and regularly poke holes in high-flying claims about the power of AI-powered tutors — he recently offered choice words for Khan Academy founder Sal Khan’s of Open AI’s new GPT4o tool, saying it was “deployed in the most favorable educational environment we can possibly imagine,” leaving open the possibility that it might not perform so well in the real world.

In April, Riley ruffled feathers in the startup world in the journal Education Next that took and other AI-related companies to task for essentially using students as guinea pigs.

In the essay, he recounted asking to help him simplify an algebraic equation. Riley-as-student got close to solving it, but the AI actually questioned him about his steps, eventually asking him to rethink even basic math, such as the fact that 2 + 2.5 = 4.5.

Such an exchange isn’t just unhelpful to students, he wrote, it’s “counterproductive to learning,” with the potential to send students down an error-filled path of miscalculation, misunderstanding and wasted effort.

The interview has been edited for length and clarity.

����Ӱ��: We’re often so excited about the possibilities of ed tech in education that we just totally forget what science says about how we learn. I wonder if you have any thoughts on that.

Benjamin Riley: I have many. Part of my frustration is that we are seemingly living in a moment where we’re simultaneously recognizing in other dimensions where technology can be harmful, or at least not beneficial, to learning, while at the same time expressing unbridled enthusiasm for a new technology and believing that it finally will be the cure-all, the silver bullet that finally delivers on the vision of radically transforming our education system. And yeah, it’s frustrating. Ten years ago, for example, when everybody was excited about personalization, there were folks, myself included, raising their hand and saying, “Nope, this doesn’t align with what we know about how we think and learn. It also doesn’t align with the science of how we collectively learn, and the role of education institutions as a method of culturally transmitting knowledge.” All of those personalized learning dreams were dying out. And many of the prominent, incredibly well-funded personalized learning efforts either went completely belly-up, like , or have withered on the vine, like some of the public schools now named .

Now AI has revived all of those dreams again. And it’s frustrating, because even if it were true that personalization were the solution, no one 10 years ago, five years ago, was saying, “But what we need are intelligent chatbot tutors to make it real.” So what you’re seeing is sort of a commitment to a vision. Whatever technology comes along, we’re going to shove into that vision and say that this is going to deliver it. I think for the same reasons it failed before, it will fail again. 

You’re a big fan of the University of Virginia cognitive scientist , who has done a lot to popularize the science of how we learn.

He’s wonderful at creating pithy phrases that get to the heart of the matter. One of the counterintuitive phrases he has that is really powerful and important is that our minds in some sense “are not built to think,” which feels really wrong and weird, because isn’t that what minds do? It’s all they do, right? But what he means is that the process of effortful thinking is taxing in the same way that working out at the gym is taxing. One of the major challenges of education is: How do you wrap around that with students, who, like all of us, are going to try to essentially avoid doing effortful thinking for sustained periods? Over and over again, technologists just assume away that problem.

In the case of something like large language models, or LLMs, how do they approach this problem of effortful thinking? Do they just ignore it altogether?

It’s an interesting question. I’m almost not sure how to answer it, because there is no thinking happening on the part of an LLM. A large language model takes the prompts and the text that you give it and tries to come up with something that is responsive and useful in relation to that text. And what’s interesting is that certain people — I’m thinking of most prominently — have talked about how amazing this is conceptually from an education perspective, because with LLMs you will have this infinitely patient teacher. But that’s actually not what you want from a teacher. You want, in some sense, an impatient teacher who’s going to push your thinking, who’s going to try to understand what you’re bringing to any task or educational experience, lift up the strengths that you have, and then work on building your knowledge in areas where you don’t yet have it. I don’t think LLMs are capable of doing any of that.

As you say, there’s no real thinking going on. It’s just a prediction machine. There’s an interaction, I guess, but it’s an illusion. Is that the word you would use?

Yes. It’s the illusion of a conversation. 

In your Education Next essay, you quote the cognitive scientist , who says LLMs are “frequently wrong, but never in doubt.” It feels to me like that is extremely dangerous in something young people interact with.

Yes! Absolutely. This is where it’s really important to distinguish between the now and the real and the present versus the hypothetical imagined future. There’s just no question that right now, this “hallucination problem” is endemic. And because LLMs are not thinking, they generate text that is factually inaccurate all the time. Even some of the people who are trying to push it out into the world acknowledge this, but then they’ll just put this little asterisk: “And that’s why an educator must always double-check.” Well, who has the time? I mean, what utility is this? And then people will say, “Well yes, but surely it’s going to get better in the future.” To which I say, Maybe, let’s wait and see. Maybe we should wait until we’ve arrived at that point before we push this out.

Do we know how often LLMs are making mistakes?

I can say just from my own personal usage of Khanmigo that it happens a lot, for reasons that are frankly predictable once you understand how the technology works. How often is it happening with seventh-grade students who are just learning this idea for the first time? We just don’t know. [In response to a query about errors, Khan Academy sent links to two on its site, noted that Khanmigo “occasionally makes mistakes, which we expected.” It also pointed, among other things, that Khanmigo now uses a calculator to solve numerical problems instead of using AI’s predictive capabilities.]

One of the things you say in the EdNext piece is that you just “sound like a Luddite” as opposed to actually being one. The Luddites saw the danger in automation and were trying to push against it. Is it the same, in a way, as what you’re doing? 

Thank you for asking that question because I feel my naturally contrarian ways risk painting me into a corner I’m really not in. Because in some sense, generative AI and large language models are incredible — they really are. It is a remarkable achievement that they are able to produce fluent and coherent narratives in response to just about any combination of words that you might choose to throw at them. So I am not a Luddite who thinks that we need to burn this all down.

There are methods and ways, both within education and in society more broadly, in which this tool could be incredibly useful for certain purposes. Already, it’s proving incredibly stimulating in thinking about and understanding how humans think and learn, and how that is similar and different from what they do. If we could just avoid the ridiculous overhype and magical thinking that seems to accompany the introduction of any new technology and calm down and investigate before pushing it out into our education institutions, then I think we’d be a lot better off. There really is a middle ground here. That’s where I’m trying to situate myself. 

Maybe this is a third rail that we shouldn’t be touching, but I was reading about Thomas Edison and his ideas on education. He had a great quote about movies, which he thought would revolutionize classrooms. He said, “The motion picture will endure as long as poor people exist.” It made me think: One of the underlying themes of ed tech is this idea of bringing technology to the people. Do you see a latent class divide here? Rich kids will get an actual personal tutor, but everybody else will get an LLM? 

My worry runs differently than that. Again, back to the Willingham quote: “Our minds are not built to think.” Here’s the harsh reality that could indeed be a third rail, but it needs to be acknowledged if we’re going to make meaningful progress: If we fail in building knowledge in our students, thinking gets harder and harder, which is why school gets harder and harder, and why over time you start to see students who find school really miserable. Some of them drop out. Some of them stop trying very hard. These folks — the data is overwhelming on this — typically end up having lives that are shorter, with less economic means, more dire health outcomes. All of this is both correlated and interrelated causation.

But here’s the thing: For those students in particular, a device that alleviates the cognitive burden of schooling will be appealing. I’m really worried that this now-widely available technology will be something they turn to, particularly around the incredibly cognitively challenging task of writing — and that they will continue to look to this as a way of automating their own cognition. No one really needs to worry about the children of privilege. They are the success stories academically and, quite frankly, many of them enjoy learning and thinking and will avoid wanting to use this as a way of outsourcing their own thinking. But it could just make the existing divide a lot wider than it is today — much wider.

How is education research responding to AI?

The real challenge is that the pace of technology, particularly the pace of technological developments in the generative AI world, is so fast that traditional research methods are not going to be able to keep up. It’s not that there won’t be studies — I’m sure there are already some underway, and there’s tiny, emerging studies that I have seen here and there. But we just don’t have the capabilities as a research enterprise to be doing things the traditional way. A really important question that needs to be grappled with, as a matter of policy, potentially as a matter of philanthropy and just as a matter of society, is: So, what then? Do we just do it and hope for the best? Because that may be what ends up happening.

As we’ve seen with and in schools, there can be real impacts that you don’t realize until five, 10 years down the road. Then you go back and say, “Well, I wish we’d been thinking about that in advance rather than just rolling the dice and seeing where it came up.” We don’t do that in other realms of life. We don’t let people just come up with medicines that they think will cure certain diseases and then just say, “Well, we’ll see. We’ll introduce it into broader society and let’s figure it out.” I’m not necessarily saying that we need the equivalent per se, but something that would give us better insight and real-time information to help us figure out the overall positives and not-so-positives seems to me a real challenge that is underappreciated at the moment.