Episode 298: Dan Meyer

Episode 298: Generative AI Hype and Classroom Challenges with Dan Meyer

[00:00:30] Fonz: Hello, everybody! Welcome to another great episode of My EdTech Life. Thank you so much for joining us on this wonderful day, and wherever it is that you're joining us from. Thank you, as always, for all of your support. We appreciate all the likes, shares, and follows. Thank you so much to all our new YouTube subscribers. We really appreciate everything that you guys are doing as far as engaging with our content—sharing it, listening, watching, all of those great things. As you know, we do what we do for you, so we can bring you some amazing conversations here in the education space, with the purpose of allowing us to listen to different perspectives, and helping us continue to grow. Education changes from day to day, but that's what we're here to do—bring you amazing content to help each and every one of us.

Today, I am really excited because I have an amazing guest—a wonderful guest that I have been following for a long while on LinkedIn and Twitter. He definitely has a great YouTube video out, and we'll talk a little bit about that as we get into AI in education, looking beyond the hype, and discussing what's really going on. We'll cover where AI may fall short, what might be promising, and more. But I would love to introduce my guest who's joining us all the way from California. Dan, thank you so much for joining us this evening. How are you doing?

[00:02:06] Dan: Hey, I'm doing great. I'm thrilled to be here. I think you always bring a really thoughtful lens to a lot of the tech hype, so I'm just pumped for the conversation. Thanks for having me!

[00:02:13] Fonz: Awesome! Well, thank you so much for joining, and I really appreciate you spending some time here with me today. I just want to understand the great work you're doing—you're putting out amazing content. I’ve always made it very clear on my podcast that I’m a cautious and careful advocate, trying to balance things out and bring both sides of any new EdTech to people. Especially now with generative AI in classrooms. So, Dan, for all our audience members joining us—whether watching on replay or checking out the podcast—can you give us a brief introduction about your context in the education space?

[00:02:56] Dan: Yeah, no idea if I'll be brief, but here we go. I taught high school mathematics to kids that did not like mathematics—freshmen on their second or third time through algebra—and that really shaped so much of my work, even today. I know the importance of engaging and motivating kids, what they're looking for in their education. So, I've been a part of the EdTech industry for the last 10 years, and I’ve seen how many of the solutions offered would work well for the top 5% most motivated students, but the rest—no. At Amplify, I helped build a core curriculum and digital lesson-building software that supports the 95% of students, zigging where a lot of EdTech has zagged. We're really excited about the results from students, teachers, and the market so far.

[00:03:49] Fonz: Excellent! Well, that's really exciting. I'm really happy to get to know more about that because, being a former math teacher myself, teaching high school math and even going down to elementary math, there are probably some points where we’d definitely find a lot of commonalities—especially now that we're talking about generative AI within the classrooms.

First of all, Dan, I would love to talk about your writing on Substack. You share a lot of great points and views both on LinkedIn and Twitter, which I’m a huge follower of, as are many of my friends. So really, it's an honor to have you here on the show. What I would love to do is get your perspective. Let's take a time machine back to November 2022. What was going through your mind as we noticed that things changed in the education space with the introduction of ChatGPT? Let me know your thought process and what you saw as the potential or the mishaps of this technology within education.

[00:05:16] Dan: Yeah, great question. I think I would dispute the idea that everything changed in the education space. I think you're absolutely right in the commercial tech market—wow, things changed instantly. Everyone’s pitch deck now includes some generative AI aspect. Everyone's creating new startups that are just wrappers around OpenAI API calls—that definitely happened.

My own reaction was, "This is neat." I'm not made of stone—I'm a nerd's nerd. Former math teacher here, I love technology, especially when it does new novel things that seem unexplainable. I just thought it was neat. I thought then, and think now, it's neat. When the words go on the screen, I'm like a seal clapping for more. I want more of it—it's just so interesting.

But when people say, "It's going to overtake everything," I'm like, "I don't know about that." It seems just neat, not something we should worry about. When people say, "This is going to transform and save everyone and everything," again, I'm like, "I don’t know about that either." I took a wait-and-see approach until the spring of 2023. Eventually, I came to the conclusion, which I’m still working on, that this doesn’t meet the core needs of teachers and students.

I made a prediction in June that we won’t see changes in teacher staffing, teacher attrition, or student learning that are discernible without a microscope. You won’t be able to look back in five years and say, "That was the moment generative AI came out, and teacher attrition went down, and student learning went up." In the same way you can with COVID—you could point to a moment when everything changed. But I don't think the same will be true for generative AI. I feel more confident in that prediction now than two years ago.

[00:07:35] Fonz: Excellent. It's a very similar approach to the way I felt. At the beginning, I was like, "Oh my gosh, this is really exciting—very novel!" I even interviewed ChatGPT on one of my podcast episodes, episode 164, when it first came out. I was cranking away, putting in questions, and getting answers. This was novel to me, and I love EdTech—this is the space I live in.

But in March 2023, I had a class for my doctoral program talking about AI, privacy, and security. That’s when I hit the brakes. We discussed the security aspect, data, and companies connecting to OpenAI’s API. It inundated the business space, and everyone wanted to get into education because there's a lot of money in education. There's a need for something that gives teachers time back and makes things easier—personalized learning, and so on.

But then I realized something wasn’t quite right. Many people say, "This is going to revolutionize education." So, let me ask you: How do you reconcile the possible upsides of generative AI and the potential downsides? Where do you stand?

[00:09:41] Dan: Yeah, look, no one would be more thrilled than me if generative AI provided major benefits to teachers and students. I have kids in schools, and I want their teachers to be the happiest possible. I want learning to be easier and more fun. So, I’m very interested in this space. But I note the difference between what teaching and learning needs and what generative AI offers.

A prominent tech investor categorized teaching as "free white-collar work." I thought, "I get where you and your cohort see this as transformative, but I see teaching not as white-collar work, but as social work and white-collar work." The social work takes precedence over the white-collar work. Motivating students, helping them see they belong in this discipline—that’s not what generative AI offers.

There are point solutions—teachers tell me that generative AI helps write parent emails faster, reclaiming a few more minutes of time every couple of weeks. That's great, but let’s not confuse that with the reasons teachers give for being unhappy in their work.

[00:11:39] Fonz: Yeah, there’s a lot to unpack there. I’ve had guests say, "AI will give teachers time back," because they know that’s a major struggle. But for me, a big weakness in many products is the lesson plan aspect. Companies say, "We’ve added standards, we’ve added this," but when you use it, the lesson plan falls short. They tell you, "Go create this, create that," and I wonder, "How is this saving me time?"

I remember watching your demo at the ASU GSV Air Show, and I thought, "Gen AI is really falling short here." It’s just giving you buzzwords, but not a whole lot of substance. What’s your take on EdTech companies and generative AI? Can they improve?

[00:13:57] Dan: Yeah, I’m trying to make sense of it. As learners, we’re curious—things don’t make sense, and we want to figure it out. What doesn’t make sense to me is the massive enthusiasm for generative AI on LinkedIn and among the EdTech crowd, and the empirical data from out there.

For example, RAND recently released data from its American math educator survey, where they asked teachers how often they used AI to support instruction. This was in spring 2024, well into the hype cycle, and 82% of teachers said "never." All my EdTech friends are excited, but there are these data points where it's like, "What is going on? 82%?"

The difference is a delivery problem. For example, Magic School has a lesson plan developer. You input standards, objectives, and context, press "Go," and it gives you a page of text describing a lesson plan. Investors might get pumped, but teachers know how far that is from an actual lesson plan.

Teachers think about context: what’s been done previously, curriculum, methods, vocabulary, even what day of the week it is. These tools frequently tell teachers, "Hey, we’ve done 80%, you do the remaining 20%." But it feels more like 20-80, or worse, because it’s so much work to finish what the tool started. It gives teachers more homework, and teachers don’t need more homework.

The way forward is embedding AI into platforms at the point of use, where it already has the context. I don’t see how text-input, text-output tools will be useful to teachers in the long run.

[00:17:42] Fonz: Yeah, I agree with you. A lot of what you’re saying resonates with what I’ve seen. I follow Dr. Emily Bender, and one of my favorite things she always says is that large language models are synthetic text-extruding machines, making paper mache outputs. They’re just predicting the next word based on the input. They’re not digging deep into understanding.

Often, when using these platforms, it seems like they fall short, even if they claim to have added standards. That’s one of my biggest questions: “Do you have the standards?” In Texas, we updated our standards last year, but the knowledge cutoff dates worry me. If a teacher goes in, thinking, “Let me pop in a standard,” and they're just giving it that context, are they really getting the true output they need?

As a teacher, I’d do my due diligence to make sure it fits, but it still feels like I’m redoing and reworking everything to ensure it aligns. Sometimes, I feel like the curriculum within our district does a better job than what I’m getting here. That’s one of the concerns I have.

One thing I wanted to ask you is about AI tutors. I know you wrote about it, and I’d love to get your thoughts on what Sal Khan wrote about reflecting on his vision of EdTech. What do you think is the most important thing we need to do to make the technology work better, not just for teachers, but also for students, beyond a chatbot?

[00:20:16] Dan: Yeah, great question. If I can dip back into the history of EdTech, circa 2010, I came of age alongside startups that believed computers were good for information distribution—giving you a video or text, and evaluating whether you understood it by asking multiple-choice questions. In doing so, it would separate the class into 30 little student-computer dyads.

But it turns out, that model—validated by RAND Research with the Gates Foundation—leads to students feeling isolated. They experience a lower sense of belonging, safety, and community, and feel like there isn’t an adult who knows them. It’s socially atomizing, all for minimal gains.

My cohort, on the other hand, envisioned computers differently. We saw how people used devices outside of class to connect and create things meaningful to them—photos and videos that go into their social networks, where they learn from feedback. We created a math platform and curriculum where teachers and students are together as a class, using devices to facilitate interactions. The computer's job is to awaken the thoughts students already have, give them game-like environments, and share their thoughts with the teacher. The teacher then highlights a few thoughts, and the class grows socially and intellectually.

When I see chatbots, I see a replication of the old, ineffective model, except now it's with AI instead of videos. It’s still one-to-one, the student and the chatbot, and students don’t enjoy it. K-12 kids don’t want to watch 20-minute educational videos, and they don’t want to interact with chatbots. They want to learn from people they care about, with people they care about.

That’s my main concern with the model—it’s not going to work at scale because kids don’t enjoy that interaction with computers in class.

[00:23:32] Fonz: Exactly. That’s what I see, too. Being in this role for the last seven years and visiting campuses, I see the same thing. I often tell teachers, "Maybe don't assign that 20-minute video—use bite-sized videos." I also tell teachers, "Hey, you know that content you recorded during lockdown? Keep it evergreen."

Students like seeing their own teacher explaining something in a short, bite-sized way, whether through Google Classroom or another platform. It helps maintain that connection. That’s one of the things I’ve noticed. AI claims to personalize learning, but in doing so, we’re siloing students away from communication and collaboration. It feels like a contradiction. We're saying, "Students need communication and collaboration skills," but we also want them to work with a chatbot instead of peers. It feels counterintuitive. That’s how I’m feeling right now.

[00:25:46] Dan: Yeah, I agree. Post-COVID, there’s been so much interest in tutoring models to help students make up for lost learning, and a lot of money has been spent on various models. Only some have been effective, and the ones that work usually involve consistent humans—tutors who are there multiple times a week. These tutors also have significant training. That’s what’s effective.

The AI chatbot model seems the least effective of all because it lacks those human elements. It’s not a consistent presence, and it doesn’t know the student personally. You can’t fit your whole personality into a context window, so it feels like starting fresh every time. It’s not trained on the individual’s context.

Just think about what a successful tutoring interaction looks like for you—whether it's with a tutor or someone else. Ask yourself, "Could a chatbot provide the same value?"

[00:27:10] Fonz: Absolutely. So, now that we’ve covered that, I’d like to ask you about your experience in EdTech, especially now with generative AI. What can EdTech companies do to better support teachers without overloading them with tech responsibilities?

[00:27:42] Dan: A couple of things come to mind immediately. First, make sure you have a lot of teaching experience on staff, ideally current teaching experience. When I first started at Desmos, we hired people straight from the classroom, and it was a great check on the magical thinking in the EdTech industry.

So many EdTech founders come from elite educational backgrounds, where they were autodidactic and found learning easy. They didn’t need the social context of learning. That experience isn’t representative of most students. Hiring current or recent teachers helps avoid those biases.

Once you have teachers on staff, make sure you’re in classrooms regularly. There’s nothing like watching your best ideas in action. It makes the rough edges apparent, so you can go back to the drawing board and refine things.

Lastly, realize that the classroom is a social space, far more than most people think. The reason students are there is for the social relationships they enjoy. Your solution needs to work with that, not against it.

[00:29:41] Fonz: I love that. Oftentimes, it feels like we’re customer support for these tools, and you’ve pointed out how some companies overlook the realities of the classroom. At a district level, decisions are often made from the top down, without considering barriers to learning—whether it’s the technology, the teaching, or something else.

EdTech companies often think their tools will solve everything, but who was the tool designed for, and what research was used? With generative AI, many companies still lack meaningful insight into the classroom realities. How can we slow things down, refine them, and ensure they really work for teachers? What do you think?

[00:32:23] Dan: That’s a great question. One risk is creating dozens of point solutions—50 different AI tools, each with its own quirks, that teachers must juggle. Whatever gains those tools offer might be offset by the challenge of managing them all. That’s one reason we see lower usage than the hype would have us believe.

Many in EdTech see teaching as a software-shaped problem, but teaching is a people-shaped problem. Teachers are vocal about why they’re unhappy—administrative burdens and paperwork are top complaints. If you’re in EdTech, it’s easy to think, "We have a tool that can do paperwork!" But the reality is that paperwork comes in 50 different forms—student locker combinations, field trip forms, and more. When you realize that, you see it’s not easily solved by software.

Some problems need to be solved outside of software. Schools need better planning around the teacher experience, which has nothing to do with software. That’s often lost on people who are used to solving everything with tech.

[00:34:52] Fonz: I love that. You’re hitting the nail on the head. Oftentimes, we do get these platforms that look great with the user experience (UX) and user interface (UI), but they can still be overwhelming. For example, one platform we use updated their features, and suddenly, the languages section changed. People didn’t know how to delete a language because there was no icon—it was hidden, and you had to hover over it.

It’s poor UX like that, but they did fix it after feedback. Still, it can be overwhelming when one platform has 75 tools, and the main menu looks like a pile of fire ants. That’s why it’s important to focus on people at the center, not just technology.

EdTech companies often move too fast, riding the hype wave, but all it takes is one of their plugged-in APIs to go under, and suddenly, everything falls apart. We’ll see where that goes. Looking ahead, Dan, what do you see as the biggest opportunity for EdTech to truly transform education, and what role should we, as educators, play in that?

[00:36:39] Dan: That’s a great question. I think the biggest untapped value right now is not technological but social. The opportunity lies in realizing the capacity and skills of the teachers at a given site and helping them network together to multiply their effect. Freeing them up from many burdens allows them to do what they do best.

Especially in math classes, the biggest opportunity is to realize what students bring to class every day. Students are pattern-seeking machines—they bring informal ways of thinking about math topics before you even start teaching them formally.

The classroom of the future is a social space where we’re finding ways to awaken what students already know and bring it to the surface, allowing the teacher to make more of it. That’s the future.

[00:38:03] Fonz: I really love that. Excellent, Dan. As we get ready to wrap up, here are the last three questions I always ask my guests. First, if you could leave my listeners with one piece of advice or insight on how to navigate this world of AI and EdTech, what would it be?

[00:38:26] Dan: I’d say whatever your role—whether you’re a teacher, technologist, EdTech funder, or something else—generative AI has thrown the gauntlet down and offered us a challenge. It makes us ask, “What is the work of teaching?” Many people think, “Teachers will do the motivational stuff, and AI will do the cognitive stuff.” But the work of teaching intertwines cognitive and social aspects in ways that can’t be separated. The challenge is to help teachers do both simultaneously. Make sure you know what teaching is really about.

[00:39:10] Fonz: That’s great advice. Thank you so much, Dan. I appreciate you taking the time to be here today, sharing your views, and amplifying your voice and your work. For our audience members, how can they connect with you?

[00:39:31] Dan: Yeah, please hit me up at danmeyer.substack.com, where I post weekly on teaching, technology, and math learning. You can also find me on Twitter—well, X—at @ddmeyer, and on LinkedIn if you can get past all the hype. I’d love to connect.

[00:39:56] Fonz: Excellent! I’ll make sure to put all that information in the show notes. Now, let’s move on to the last three questions I always ask my guests. Question one: We know every superhero has a weakness or a pain point. For Superman, it’s kryptonite. In the current state of education, what would you say is your current “edu kryptonite”?

[00:40:35] Dan: That’s a fun one! I’d say I’m overly focused on the product over the process. I love that moment when you see something amazing happen in the classroom—the student and teacher both say, “Whoa, I’m awesome!” I love those transcendent moments. But the work for me—and I’m grateful to my partners at Desmos and Amplify for helping me—is figuring out how to systematize and package those moments, making them something districts can procure through purchase orders and processes.

I get that the best people in EdTech love the full stack, from instruction to interacting with superintendents of curriculum. I just get really enamored with one part of that—the moment of instruction—and I need to keep growing in my appreciation for the rest.

[00:41:34] Fonz: Great answer! Thank you for sharing that. Now, question two: If you could trade places with anyone in the world for a day, who would it be and why?

[00:41:45] Dan: I think I’d love to trade places with one of my kids, just to see what they’re doing at school all day. Maybe it would be too terrifying, but I’d love to know more about what schooling looks like from a kid’s perspective. Even better if I could access their inner monologue!

[00:42:04] Fonz: I love that answer! It’s a very common one when I ask it, and it’s a great perspective. Finally, question three: If you could have a billboard with anything on it, what would it be and why?

[00:42:19] Dan: I’d probably put one on Highway 101 here in California, right outside Silicon Valley, for EdTech developers. It would say, “Have you used your stuff?” Just a reminder to everyone building for teachers—use it yourself, try it out, see if it works.

[00:42:42] Fonz: Love it! I am all for that sign for sure, Dan. Well, thank you so much again for being here and sharing your time with us today. I’m really excited to release this episode, and I’m a big fan of your work. Thank you for all the thoughtful writing you do, especially in a space full of hype beasts. Your content is seen, and I always look forward to your Substack posts on Wednesdays.

For our audience members, thank you so much for joining us today. Please make sure to visit our website at myedtech.life, where you can check out this amazing episode and all the other 296 episodes. Actually, after this one, it’ll be 297 episodes. You’ll find creators, educators, entrepreneurs, edupreneurs, and more. There’s something for everyone, and I promise you can take some knowledge nuggets and sprinkle them into what you’re already doing.

If you’re not following us on socials, please do! We’re on all platforms as @myedtechlife. And don’t forget to jump over to our YouTube channel, give us a thumbs up, and subscribe so you can continue to get amazing interviews and content. As you know, we do our best to bring wonderful conversations to help us all grow in the education space.

As always, my friends, until next time, don’t forget—Stay Techie!