Where AI Helps — Administration to Instruction

In this episode of Class Disrupted, Diane and I stepped back from our interviews to have a one-on-one conversation to reflect at the midpoint of our season on AI in education. I left the recording thinking “wow” after this one, because it felt like we didn’t hold back and explored a lot of different questions on this episode in a way most “education reformers” aren’t addressing at the moment.

We dove into the evolving role of AI in education and questioned whether AI is truly transforming the system or simply being layered onto outdated structures. We explored a framework of three school models that Diane posed (and I’m still noodling over!)—and discussed the challenges of meaningful innovation amidst existing accountability systems and educational policies. From these models, we then analyzed how one might expect transformational change to occur in K–12 schooling—through industrial schools incrementally changing and evolving over time or, as we both argued, through fundamental migration away from the existing system. I thought this was among the best episodes we’ve recorded, so I’ll be very curious your thoughts.

Diane Tavenner

Hey, Michael.

Michael Horn

Hey, Diane. It’s good that you came to Boston and in the freezing cold weather, no less, to hang out a little bit with me here and have a conversation.

Diane Tavenner

It’s really fun to be in person. We haven’t done this for a long time and the timing worked out perfectly because we are in the midst of this super interesting season where we’re exploring AI and education. And we’ve had several touch points where I’m like, oh, my gosh, there’s so many things that are coming up for me that I want to talk with you about. And so we get to have a conversation, the two of us, this morning.

Michael Horn

I am looking forward to it. And I’m sure you’re going to say things. I’m going to say, wait a minute, I think I know what you mean, but double click on that. Tell us more. And so I’m excited to go deep on wherever you want to go because the conversations, they’ve both been illuminating, but they brought up more questions for me, as seems to be constantly the case with this topic.

AI Disrupting Education Processes

Diane Tavenner

Indeed. Indeed. Okay, well, let’s dive in. And I had the great pleasure of spending time with you in your class yesterday. Thank you again, so much fun. And one of the topics that came up was this idea of. I think it turned out to be more provocative than I anticipated it to be. But this idea that I started said, you know, one of the things, a phrase I read almost constantly right now and hear everywhere is AI is changing education.

And I don’t believe that that phrase is true or accurate. And in fact, I believe AI is not changing education. And, and so I want to dig into that idea a little bit. You know, I would argue that it’s creating a lot of problems for folks in education who are sort of in the traditional model of schools. But I don’t think it’s changing education yet. And what do you think about that?

Michael Horn

I largely agree. So I’ve been thinking about this, but a different wavelength because I’ve been seeing over X and the various pundits. There’s a lot of conversation right now of banning cell phones in schools, as you know, and there’s a lot of conversation of not just cell phones, but screens, period, you know, Google Classroom, all the rest, because it creates access to all these other things, ban it all sort of things. And then you see the occasional commentators saying, does anyone ever believe otherwise at this point?

Diane Tavenner

Right.

Michael Horn

And I had this moment because I think I’m seen often as the tech guy in education. But if you read Disrupting Class, what we actually say is that just layering tech over the existing system is not going to do anything.

Diane Tavenner

Right. I think we’re going to get to that idea in a moment.

Michael Horn

So I think so I guess my instinct is, I agree with you. Like I think we’re layering a lot of AI over existing processes. It’s breaking, frankly, a lot of education. So the one push I might have on you is it may be creating the impetus to ask some bigger questions. And, and I’m not just saying I’m not going down the road of just because the world is AI, therefore this should be AI but like legitimately, you know, we have current assignments where you can now hack them through AI. That’s called cheating. And all of a sudden everyone goes in a tailspin.

Well, let’s ask some questions about the assignments and the work itself is sort of my take from that. So I think it might be an interesting push. But I agree most of what AI is doing right now is layering over existing processes. Some of them, I suspect it’s making more efficient. Great. Maybe some of them I think—it’s exacerbating problems that already existed. Is that what you have in mind or…

The Future of Education is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Diane Tavenner

That is what I have in mind. And you brought up, you know, the, one of the biggest conversations is about cheating. Right now we’re seeing all these distortions and strange behaviors and blue books returning. And I’m sure the company that makes those is happy about that. But you know, they might be, they’re…

Michael Horn

Still around or they have to resuscitate to really. Did we should look that up.

Diane Tavenner

Yeah, I think when I think about it, what’s happening with this idea is that everyone knows that they’re supposed to have an AI policy and strategy now, but most people don’t. And so this is confusing. And a lot of people, I think AI in education right now is very kind of one offy. Like people, individual people pulling it in and people, you know, and so it’s not coherent, it’s not a strategy. We see it in sort of, you know, lesson planning and assignment making, which is related to, you know, why are we even teaching what we’re teaching to your point? And if you can cheat on it, then what are we trying to do? And then it goes down the line to a lot of fear that I think it’s injecting everything from these very high profile cases we’re seeing of suicide that, you know, is potentially induced by the AI to, to big widespread data privacy. So all of that to say, I’m hopeful. I believe the technology itself, if deployed, can actually change education. But I think humans are going to have to do that redesign and that deployment in a really strategic, thoughtful way for it to change.

Otherwise, I just think it’s plaguing us with problems.

Michael Horn

Yeah, I think that’s right. And systems structures, models, matter and processes and, you know, they’re sort of automating or, you know, playing off the existing ones. We may have a small disagreement on one thing. I’m curious about this. So, like, we don’t have many disagreements, so I’m gonna lean in if we do. I do think, so the Blue book comment aside, I can imagine that there are things we want to do in the classroom that have no AI at all involved with them, because some foundational knowledge or skill that a student can hack using AI out of the classroom is something that they actually should still work on in an analog way to create automaticity on that.

Diane Tavenner

Okay.

Michael Horn

I don’t know if that’s Blue Books or what form factor. I’ll take the point there, but I guess that’s. I suspect if we break things down, there are still some foundational things we would want students to have to wrestle with that might not involve AI and be offline, if that makes sense. And then my take would be, okay, but don’t stop there. Now what are we going to use AI to create as opposed to consume with AI?

Diane Tavenner

I think that’s right. I really loved the conversation we just had with Laurence where he brought up some really interesting examples, to your point, of, you know, young people literally working together and in dialogue and, and then he talked about how AI could be supportive and enhance that. But to your point, the actual skill of having that conversation with another human and what you’re talking about is not about AI, so completely agree with that. My concern is when people are taking, you know, very old assignments and.

Michael Horn

And just dusting them off without any thought. Yeah. And I think I also think this gets the older you go, as in, I could be wrong about this. And this is, I’m sure, overly simplistic, but I think for a younger student, and, you know, I’ve got kiddos still in elementary school, so I’m still thinking a lot about that. I do think, like, that part of the landscape looks different from the older student in high school and college that, you know, it’s more problematic when you’re just dusting off that assignment, perhaps for that student.

Diane Tavenner

Right.

Michael Horn

But I do think, you know, developing number sense and automaticity with those things offline before you introduce the calculator and AI and so forth. That makes a heck of a lot of sense for a younger student. And so it’s as always with these conversations in education, I think we sort of make a statement and think it applies everywhere and there is nuance there.

Clarifying AI’s Role in Education

Diane Tavenner

That’s exactly where I’d like to go next because, so I think the dialogue around AI and education is complicated right now. And I hear a lot of people talking past each other and over each other because I think we’re using these very broad, sweeping general terms. So, for example, AI and education, and I was with a really great group of people a couple weeks ago and fortunately some really, you know, smart people noticed this talking past and talking over and called it out. And literally we went around this room and we were like, what do you mean by AI in education? And just within seconds we surfaced. Oh, well, you know, using LLMs like GPT and Claude and Gemini for instructional or operational support, using AI powered education apps, Khanmigo, Class Mojo, Magic School, AI policy development, you know, AI literacy lessons for students. And, people are literally using the phrase AI strategy, AI and education AI to mean all those things and more. And, and I’m finding that it’s very complicated to try to have meaningful dialogue when there isn’t a definition right now or people aren’t. We don’t have specificity yet.

I mean, I think some people don’t even know what AI is.

Michael Horn

Yeah, you’re probably right.

Diane Tavenner

Yeah, yeah.

Michael Horn

And it’s probably extremely fearful in those quarters. And the social media analogy is rampant right now as a result, probably because we’re not defining or breaking down. I mean, do you really not want AI to help an administrator better communicate or schedule or like really, that seems crazy, for example, on that end of it.

Diane Tavenner

And my sense is that what jumps to most people’s mind when they think about AI in education, we’ve sort of railed against this from the beginning, is literally how a student is engaging with it either in the classroom or at home. And most people have in their mind some version of some chatbot, generally speaking, which is incredibly narrow and limited, I think. And you just gave a good example of like, we could literally never bring it directly into the classroom with students. And there’s a million different uses for it in just running something as complicated as a school and a school system. And so, yeah, I guess this is just my plea for us collectively to start developing a more specific vocabulary, more intentionality. About what we mean. Let’s stop saying we’re doing AI.

Oh my gosh, everyone’s doing AI. What does that mean? And being really specific about it. And I think for me, I just want to flag as we go through the rest of this season because we’re going to have some really interesting conversations next. I’m going to push me and us to be really specific about what are people literally doing with AI. What does that mean?

Michael Horn

Yeah, and the conversation with Laurence, I think opened us up to that because it started to talk about very specific use cases. It does occur to me this problem has always existed in education since I’ve been in the field. Right. That we talk past each other or I remember, you know, there’s project based learning adherence to like an extreme degree. And they’ll say everything ought to be learned through projects. And then you say, well, okay, the kid learning to read though, in first grade, they’re like, oh no, no, no, that kid should get phonics and direct instruction and blah, blah, blah. And you’re like, okay, so there’s nuance, but we have to break apart, novice versus expert.

What’s the topic? What’s the goal? Right. Like, and so skill versus knowledge, as you know, that gets conflated, conflated all the time. And we don’t have precision. And so I think it’s a good plea you’re making, which is just like, let’s be more specific. What’s the objective? What’s the learner coming in with if that’s the level at which we’re talking?

Diane Tavenner

Okay, all right.

Michael Horn

Where are we going next?

Diane Tavenner

To one of my favorite topics, which is school models.

Michael Horn

Okay. Yep.

Diane Tavenner

So I’ve been reflecting on a number of conversations. I’ve been having a bunch of stuff. I’ve been reading dialogue that I know that’s happening. There’s a variety of people trying to think about the future and what it looks like with AI. And there’s. I think none of these are set yet. They’re all kind of rough, but they’re starting to fall into this pattern of where people are talking about three different models, if you will, of schools. And I want to come back to what is a model in a moment.

But, but this idea that there’s. I’m going to call it, I think generally people agree that we have an industrial model school at this point. And we have had for quite a long time. We’ve talked about this ad nauseam and that. So let’s call model one sort of current industrial model. And when with the emergence of AI, model one sort of stays industrial model, but you know, AI gets used in some of the ways we just talked about. You know, like there’s, you keep all your existing structures of grade levels and schedules and teaching roles, but you have an AI enabled tools where you’re helping it to grade student work or you’re using it to lesson plan and you know, instructionally plan. You’re, you’re doing some adaptive practice and feedback.

You know, I think the stuff that people probably are more familiar with because they see it. So, that’s kind of Model 1 still in the industrial world. I’m going to jump to model three before I talk about two, because two confuses me a little bit. So model three, let’s call that native AI education. I think most people I know would argue that this has not been invented yet. It doesn’t exist yet as a model.

Michael Horn

Do we know what it means?

Diane Tavenner

I think that the way people have started to describe it I’m not sure that I agree with. And so here’s where I am on this one, which is I don’t think we know what it looks like yet. I think we’re failing in our imagination right now of what’s possible. I think it’s a moment to go into the proverbial garage and do some real designing. Yeah, but let’s call that the post industrial model. I don’t like to call it the AI model because of the definitional problems we just said, but let’s just call it whatever the next school model, like the full model would be.

Michael Horn

Okay.

Diane Tavenner

So then there’s two, model two and this one gets kind of squeezed in the middle. I think some people are calling it AI integrated education. Okay. And basically the, the emerging definition I’ve heard is that it’s where you sort of modify selected structures where the sort of benefits justify the disruption. So for example, you know, you have much more interdisciplinary curriculum. You have competency based progression in certain places, you have flexibility in existing schedules in blocks or things like that. You might start seeing some of the time out of the building or, but you’re still sort of, I would argue, existing in the industrial model kind of box, if you will. Okay, but you’re, but you’re using an integrated AI approach to kind of hack some of those things.

Okay, yeah, so let me pause there before I start asking my question. See if like those resonate if you’ve heard about them, you know.

Michael Horn

Yeah, no, I haven’t thought about it this way. So I’m noodling as you’re saying it, this is real time. I guess I’m curious. Like models, like a Montessori, like a classical education or the new versions of classical education we’re seeing in microschools or you know, I don’t think Waldorf fits into your typology but like where would you slot. Like those are models too.

Diane Tavenner

They are.

Michael Horn

How do they slot into the schematic?

Diane Tavenner

Yeah. Well, let’s just take Montessori as an example. Right. So in some ways it’s still industrial. Most Montessori schools still exist Monday through Friday, kind of between 8 to 3 ish. They still have a teacher, you know, one to kind-of-many class. There’s, you know, they’ve sort of released or relaxed age grade bands, although I think society kind of imposes them on them. So you, you know there’s some sort of gravitational.

Michael Horn

I mean, you know my frustrations.

Diane Tavenner

I do know your frustrations. So I still think Montessori, maybe Montessori would be kind of a two.

Competency-Based Learning Discussion

Michael Horn

That’s what I was wondering is trying, it’s like it’s not AI, not AI enabled, but it uses the technology of the 1910s or whatever it was to have broken out of these certain structures. And so it’s a very competency based math sequence. Very competency based on the learning to read part of it and probably less so on everything else is your point. And there’s still some sort of, you were born in the year of the Scorpion and whatever it is, and therefore you’re going to learn this on this date with everyone else sort of element to it, I think is what you’re saying.

Diane Tavenner

I think that’s right. And, and one of the reasons I wanted to talk to you about this kind of framing is I’ve been trying to think about what sits in the model two category. Okay. I mean it feels very easy for me to identify, you know, almost every school as a Model 1 and many of them are starting to bring in these like AI tools if you will.

Michael Horn

Yeah.

Diane Tavenner

But they’re still clearly industrial models. It’s pretty easy for me to say I don’t think we’ve seen a model 3 yet with the infusion of AI. And then I think about like for example, what we did at Summit and Summit learning.

Michael Horn

Yeah.

Diane Tavenner

I think at the high school level that might be a model 2 without AI yet.

Michael Horn

Right.

Diane Tavenner

Where again we were sort of pushing the boundaries of that industrial framework of a model to try to, you know, reimagine or re-engineer portions or parts of what was happening with expeditions, for example, what kind of breaks the traditional five period, six period day, but all but doesn’t really break the calendar, if you will, or the, you know, eight to three kind of situation. So what do you think about that?

Michael Horn

That’s interesting. So I know we could probably geek out all day and create a taxonomy. So I won’t do that to our listeners, but I am thinking like you’ve seen almost different shots of goal, like. So I think of Florida Virtual School as an example. And I’m reading Julie Young’s draft. I’m not sure I’m supposed to say this, but draft memoir right now. And it breaks certain elements of that, but it’s still course based.

Diane Tavenner

Right, right. There you go.

Michael Horn

So the two things are interesting. And then I start to wonder. Everyone’s talking about Alpha Schools. We’re gonna have an episode on it, so stay tuned. Maybe we don’t get into it here, but, but things like that, where does that slot into your framework? Or I think about Acton Academy, probably falls into two is my guess. And so this is, I guess, what I’m trying to start to sort through as you, as you frame this.

Diane Tavenner

It’s why I wanted to bring it up today because we are about to shift to start talking with people who are either trying to redesign whole models or portions of it. And I think it will be helpful for us, for me for sure, to have this kind of framing in my mind.

Michael Horn

So you can say, pull it back. So we’re talking with an entrepreneur. Okay. You’re working in number one context. You’re working in two, three, maybe the frontier there.

Diane Tavenner

Exactly.

Michael Horn

Okay.

AI Tools: Improving vs Innovating

Diane Tavenner

And I think there’s a couple of reasons why this is important. The first is back to that, talking past and over each other. One of the things I noticed is there are a lot of people who are gravitating to sort of the AI, you know, enabled tools that will definitely improve, you know, Model one industrial model, if you will. And they’re very passionate about that. They have really strong arguments about, like, there’s kids in schools today who need things to be better. And so we should be, you know, deploying these tools as best we can to do that. Then there’s a whole other group of people, smaller, who are like obsessing about designing Model three, a post industrial model. I don’t think anyone who’s been listening will be confused about where my kind of passions and interests lie.

So I’m definitely, you know, my attention goes to this question, and this, my energy is in that direction. And I really caught myself because I can be dismissive of that first group. And I think that is really problematic for me to do that because I. There. Well, here’s my question.

Michael Horn

Yeah.

Diane Tavenner

Do you think if those models are true in the way we’ve sort of laid them out, is the theory of action or change that you progress from 1 to 2 to 3? Because some people believe that.

Michael Horn

I strongly don’t think so.

Diane Tavenner

I don’t either. Okay, good. Say more because you’re the expert.

Michael Horn

Yeah, no, well, so. So my energy is also in three, as you know. And no one listening will be confused about that. But I think it is prudent from a systems perspective, like thinking about the country, that 80% of the dollars in energy are going into the number one. I think that is from a like sound strategy perspective. Makes a ton of sense. Right. It’s where most of the students are.

It’s like classic sustaining innovation. If I’m running a company and I see the new thing coming that I think is going to upset the apple cart, I don’t push stop on what we’re doing today.

Diane Tavenner

Right.

Michael Horn

I start to test and learn what we talked about on the fringes. And then like, I start to move things out there. Okay. So that’s where I go to the statement that I don’t see any cases where number one morphs into number three or we learn stuff from number three. And I had a guest in the class say, how do we pull it back into number one? I’ve never seen that work. You’ve never seen that number three replaces number one

Diane Tavenner

So then it has to be effectively designed from scratch, grown from scratch. It’s not, you know, evolving. No. Okay. Well, some people think it’s gonna.

Michael Horn

No, I know. And I just, I. And I think it’s totally rational to be putting bets and have a portfolio strategy that are in all three buckets. And I think you can learn lessons between them. Absolutely right. I mean, we know a lot about cognitive science from number one. We also don’t know a lot, I think, because. Take growth mindset, for example.

Right. My read of the literature is incredibly powerful. And if anything in the environment undermines the message of growth mindset, it pulls the kid back into the fixed mindset view and undermines all of that intervention. And basically every structure in number one does that.

Diane Tavenner

Right.

Michael Horn

So we can have our lesson on growth mindset. I don’t think that’s the best way to do it. But like we can have our lesson on growth mindset. We might see a temporary bump on some sort of assessment and then like immediately you get the C grade in the class and you’ve been labeled because you can’t take the feedback and do anything with it. You’re not even reading the feedback and you no longer think that.

Diane Tavenner

Yeah, well, and this is the point of growth mindset not being permanent. It’s not. You don’t either have one or you don’t.

Michael Horn

Right.

Diane Tavenner

It’s a continuous state that you’re in and you can fluctuate from in and out of that state regularly. Okay, so. Well, that’s an interesting conversation to have with folks who believe that the theory of change is that progression versus what we just.

Michael Horn

And I guess stay with it one more second because I remember when we came out with Disrupting Class, a lot of people would push us and say, well, we’re talking about systems change. What are you talking about? And I think we were talking about systems change too. But my theory of system change is system replacement.

Diane Tavenner

Well, there you go.

Michael Horn

And I think it’s really hard in the US for all the reasons we know. And one of the reasons I’m in some ways more optimistic than I have been is I actually see a path for that change, that replace or disruption of systems that I haven’t seen because.

Diane Tavenner

The technology is so.

Michael Horn

Well, and the ESA policies.

Diane Tavenner

Oh, and ESAs.

Customized Education Choices Rising

Michael Horn

Right. And so we see a level of entrepreneurship, a choice and I would argue now a family increasingly, if you’re in Arizona, Florida, Arkansas, wherever. It’s not just like the free public school or I pay money, it’s like, oh, if I just default to the free public school, I’m actually foregoing 8 to 12, $13,000 that I could be spending on my kids education in the way that’s customized for what they need and what they have shown interest in, et cetera, et cetera. That’s like a very different decision set now where all of a sudden it’s actually expensive to default to the free.

Diane Tavenner:

Well, and to your point, it might take a little bit of time, but it really changes people’s, you know, mindsets around everything.

Michael Horn

And I was shocked. I. I have to look deeper into this. But Ron Mattis at Step up for Students in Florida sent me this report they did. He said the number of learners in Florida who are now doing a la carte learning. So not they don’t have a primary school five days a week. It’s a billion dollar market is going through that and I was like, I have to like sit with that.

Right. Still. And I haven’t fully digested it because that’s, that seems like a lot. But he, but it basically, if that’s true, over the course of a decade or so, whatever the choice landscape in Florida has been, people went from, okay, I have education, savings accounts, I choose a school.

Diane Tavenner

Right.

Michael Horn

To your point, with technology and a lot of entrepreneurship and a change in the landscape, to all of a sudden saying I can unbundle and do a whole set of things with this, that’s a, that’s faster than I would have expected.

Diane Tavenner

That is faster. Oh, I’d be so curious.

Michael Horn

I want to dig in all sorts of things now.

Diane Tavenner

Let’s do that at some point. Well, and what it suggests is that individual families are essentially crafting their own personal model. Now is it AI native?

Michael Horn

Probably not.

Diane Tavenner

Probably not yet. But I bet they’re starting to use some of, you know, the AI enabled tools as part of that. Yeah.

Michael Horn

And they’re probably making also some of these trade offs in terms of like when is it analog because they control the home environment. When is AI a tool to create something? They’re probably making a bunch of these nuanced choices on the ground that like you couldn’t dictate from a central planning curriculum standards perspective.

Diane Tavenner

Right. Although that might be a feature of whatever the new Model 3 is. I mean, my hope is that it is that it is personalized to that degree within the context.

Michael Horn

Yeah, great point.

Diane Tavenner

Yeah.

Michael Horn

And so now we’ve just blown both of our minds.

Diane Tavenner

I want to go back to Model 2 for a minute because I had this really fascinating conversation with your, you know, former colleague and collaborator Julia Freeland Fisher. And she said, huh, I wonder if this model two is akin to what happened when the steam powered ship was sort of invented and there was this period of time where the new steam powered ships had to be outfitted with sails because the new technology was so unreliable. And she suggested that maybe model two was that. And what the interesting point she made is she said those were the most expensive models because you had to have both technologies on them. And this hybrid version is really expensive. So I, what do you think of that?

Michael Horn

100%. I agree. I, I hadn’t framed it immediately into that typology, but that’s almost every industry, when you see disruption, you see the old players take the new technology, right. Like there’s sort of a line, oh, they ignore the new technology. Not true. They layer it on the existing structure. Right. And the sailing ships are the perfect example.

I think the first sail ships to navigate the US was like 1819 or something like. Or 1803 and then 1819, the first transatlantic ship, the USS Savannah. And they had sales and they had steam bolted on. And I think only I’m going to get the numbers wrong but like 80 hours out of the 600 or whatever it took to cross were powered by steam. Basically every time that wind went the wrong way, they fired it up and kept going. Right. And so it’s a classic sustaining innovation on the old paradigm.

Diane Tavenner

Okay. But it’s still. Those models do not get us to model 3.

Michael Horn

They don’t. Yeah. It’s, you know, the story is that it was a 100 year disruption.

Diane Tavenner

Yeah.

Michael Horn

Where still ultimately the steamship native companies, shipbuilders ultimately upended the sail ship. And it was around 1900 I think.

Diane Tavenner

And it’s a different model ship.

Michael Horn

It’s a completely different model. Right. You don’t have the same components. You can do things differently in terms of construction because you’re not outfitting around an aerodynamic sail. Right. Like a totally different set of things you can do. So.

Diane Tavenner

Okay, I have a question. Now, you said you felt comfortable with the field sort of spending 80% of its resources on Model 1 improvements, leveraging AI. Is there a risk that we over invest in Model 1 and undermine the emergence of Model 3 because we kind of keep this old industrial model going, breathe new life into it and there isn’t a sense of urgency around model three creating three. Yeah.

Michael Horn

Two thoughts. Clay used to always say this. The best experts in a field, like you’re a very strange anomaly. The best, deepest experts in a field are almost always consumed with the toughest problems in, we’re going to call it Model 1 at the edge of the existing paradigm.

Diane Tavenner

Interesting.

Innovation Beyond Traditional Expertise

Michael Horn

And it’s these people who are almost less expert in some way or for some reason have taken their expertise and brought it out that invent the future. But like it’s very hard to persuade the people who are dealing with the hardest, most intractable problems in the first paradigm to be persuaded to design out there. It’s why I think like, you know, when you and I met for the first time and you actually liked Disrupting Class, that was like a bit of a revelation because like we couldn’t get all these people to sort of like actually engage with it. Right. And so. Or, or they thought they were engaging with it but missing the point. Right. And so I don’t know where that goes.

Except, like, in some ways, I’m not surprised that that’s the current moment we’re in. I think the danger is if those individuals then block off our avenues to pursue three, I’m okay with them being consumed with one. I think it’s great. There are a lot of underserved kids there that need better education. And I think if they use that as a justification to block off three, through policy change, through blocking entrepreneurship, through blocking families making these choices, that would be deeply concerning.

Diane Tavenner

So glad we’re having this conversation. There’s two places where I have fear about that and.

Michael Horn

Well, you’ve lived it.

Diane Tavenner

I did, yes. Continue to, it’s my life. And there’s two places that I just want to raise here. And at the risk of how, you know, these are sort of controversial and they’re very nuanced. I often am misunderstood, so I don’t talk about them out loud very often.

Michael Horn

But thanks for doing it here.

Diane Tavenner

Here we go. So the first is the big assessment and accountability system. And you know that my belief is that that structure, which is well intended and people are deeply passionate and invested in making sure that we have real data and know what’s going on. I just spent time with a parent advocate who’s like, those tests are the only receipts we have of what’s happening with our kids. Right.

Michael Horn

There’s a great article recently around how people are just shocked because the tests have gone away and they’ve been relying on grades, which are even more worthless measures. Yeah.

Diane Tavenner

Right. And so there’s a lot of energy going to. How do we bring those back? How do we reestablish them? And, and my belief is, and my lived experience is, and most people don’t like hearing this, who believe in them, is that the existence of that accountability structure, I truly believed deeply dampened innovation and the move towards now would be model three. And I’m super disinterested in hearing about waivers and all these things. And. No, it really has an impact.

Michael Horn

Let’s get into how, because I’ve moved toward you a lot on this one. But in one standpoint, it’s like, well, it’s just focused on outcomes, frees up the inputs. You get there however you want. Like, how does it actually restrict the innovation? And is that a. And why is that a bad thing?

Diane Tavenner

Yeah, I think that it’s. Well, let me share a quote that I hear very often.

Michael Horn

Okay.

Diane Tavenner

Which is, look, I’m not opposed to measuring different things but we don’t have those measurements yet. And so until we do, give me reading and math. And you know, I’m going to judge schools on reading and math, basically, which is effectively what we test in this country. And first of all, I think the problem is we actually do have those other assessments and they are crowded out. They aren’t accepted as, you know, mainstream, valid, reliable. No one is moving towards adopting them because it’s all about reading and math. And so I think it is really, you know, you measure what you value, you value what you measure. And there isn’t.

The system is not saying, no, completely unacceptable that we’re literally measuring our entire system on these two Important. Yeah, very important. Please do not misinterpret me. People always accuse you don’t want kids to read.

Michael Horn

Well, by the way. But I’m curious what you think of this. This is a classic case where I think defining the age span is important because I am strongly in favor of not losing the measures to families. Note how I said it, by the way, but measures to families on can your kid learn how to read, get those skills through, hopefully third grade. But you know, I’m. I’m actually willing to live with some variants in the age.

Michael Horn

All the reading tests after that are really knowledge tests.

Diane Tavenner

Correct.

Michael Horn

And so I would be much more comfortable, frankly, with every school picking like the. Or student, hey, you just did a deep dive on X. Go show your competency in X. I think that’d be a much more interesting. It’d be super jagged, students showing all sorts of deep dives on a variety of things and so forth. I think that’d be way more interesting. Math, I think, is a little different.

Diane Tavenner

Yes.

Michael Horn

And I don’t know where it stops. Probably around algebra, but. Yeah.

Diane Tavenner

Well, you just said a key point that really bothers me the most, which is the accountability and testing framework that we’ve had in this country is not about informing parents. And it’s not actionable data. It’s not timely data. It’s not what we would call that feedback, honest, actual timely data.

Michael Horn

No. And in fact, it’s negative reinforcement cycles.

Diane Tavenner

Exactly. And so let’s just take reading as an example. The oldest assessment technology is a reading record. I mean, schools could literally choose to assess every single kid that way and put resources towards that. It might not even be that many more minutes than they already spend on stage.

Michael Horn

By the way, AI can really do that now.

Diane Tavenner

Well, and I’m not even getting into…

Michael Horn

What technology can do.

Diane Tavenner

So why, why these old assessments. Right. And so anyway, I’m deeply concerned that there’s so much good intent there and so much potential.

Michael Horn

But you’re arguing that it’s crowding out a ton of these other measures that either are there or could be developed more robustly.

Diane Tavenner

Right. And in the same way that I can be sort of dismissive of efforts around Model one, I think a lot of folks focused on today and now in kids in school are very hand wavy and very dismissive of the impact this has on the potential for innovation. So I’m, you know,

Michael Horn

Super interesting. Yeah. Okay.

Diane Tavenner

The second one is

Michael Horn

You’re taking a breath, you’re giving me a look for those that can’t. We’re not a video this time.

Diane Tavenner

No, we’re not.

Michael Horn

Yeah, go ahead. Where are you going?

Diane Tavenner

Special education.

Michael Horn

Oh, okay.

Diane Tavenner

And I want to say up front, my belief is, are we, by the.

Michael Horn

Are we at the 50th anniversary of special ed at the IDA, the federal level?

Diane Tavenner

We might be.

Michael Horn

I think we are, yeah.

Reimagining Education for Every Child

Diane Tavenner

Okay. Yeah. The intention is right. So many amazing people working on behalf of kids here and most people who’ve spent so much time in schools like I have with families, you know, it’s a system that is about compliance more than it is about children, is. I don’t believe it gets young people what they need. And I think that has a really challenging impact on our ability to educate all of our children. And this is one of, in my view, one of the biggest promises of a post industrial model is that truly every child gets a personalized education.

Michael Horn

Because everyone’s now getting an ILP as a good. Exactly right.

Diane Tavenner

Exactly, exactly. And my worry is that in both the assessment case and special education, that new models, model threes, will be judged and held accountable to the current accountability systems and the law, which completely compromises their ability to design completely new and better approaches.

Michael Horn

Yeah. And my colleague, or I guess former colleague at the Christensen Institute, Tom Arnett, has written a lot about this one, about how when you apply the standards to the new system that were for the old, you hamstring and often stunt it completely. I think that’s very fair. My pushback historically has been. Yeah, but the existing system is all input driven and then it has outcomes layered over. If we strip out the inputs, which by the way, people are trying to put back on for the Attempts at Model 3 right now as well. Right. Like accreditation, really.

Michael Horn

I think you’re pointing out even though these output measures, I don’t even think they’re outcome measures, but output measures have been layered on, I do see where they could pull model three back in some unfortunate ways for design. And I think those are to me, that’s where the fears are really. It’s. It’s less the effort question in dollars and more the are we hamstringing it to actually just look like the existing thing we already have in slightly modified?

Diane Tavenner

Right. I’ve certainly learned from you the most, you know, how disruption happens is that people take it outside of the existing system. They have different expectations. You know, they look at it fundamentally differently. And so maybe this is the importance of ESAs. And I mean, as a person deeply invested in public schools in America, I would be very sad if we’re going to push all the innovation out into the private sector because we can’t welcome it into the public sector.

Michael Horn

Yeah.

Diane Tavenner

And maybe that’s what we’re gonna see.

Michael Horn

Yeah. I’ve always felt like the public officials ought to be responsible not for the institutions, but for the constituents. Right. And so the models may change. And by the way, look, in Florida, you have districts now launching their own microschools and creating certain services a la carte. And like, like they’re spinning off autonomously. Let’s see where it goes.

Michael Horn

Right. I mean, I don’t think we know the final thing yet. And the conversation I was having with one of my students yesterday as well was, you know, no one’s cracked yet, I think, in these. So they’re not really model three attempts because they’re not AI native. But let’s just call like this sort of emerging ecosystem. We haven’t seen a lot of high school models.

Diane Tavenner

Nope.

Michael Horn

And I think part of it is because disruption starts as primitive, able to solve simple problems, not the most complex. Identity formation becomes much more important in high school. Right. And all these rituals that we may roll our eyes at around Friday Night Lights or prom or whatever else, they’re part of this identity formation and asking who am I in relation to others? And these small, you know, I think, you know, Tyler Thigpen, Forest School, Acton Academy, he’s done a good job of creating rituals, but most high school attempts have not yet built that. And so I kind of wonder, is the upmarket, if you will, solving for all of those things with very different traditions that don’t look like Friday Night Lights, but are actually more meaningful for the current time around identity formation?

Diane Tavenner

Totally. Well, and now you’re getting at the heart of what I’m trying to contribute to with Futre, which is how do we support some of that positive identity formation and search for who I am and the life I want to lead, both in the digital world and then connect that to real world experience.

Michael Horn

Well, I think it’s interesting though, that your market is the traditional industrial Model one, largely. And so I’m, I mean, I’m curious how you think about that.

Diane Tavenner

I’m living in a bipolar world. Yeah,

Michael Horn

yeah, yeah. Okay, okay, okay, okay. Well, I. You’ve built it with a modular interface, as I understand. Right. So it can exist in both, I think is part of your answer. And I, I imagine you’d say a native model 3 would actually answer a lot of the future questions as part of the design of the model itself.

Building Toward Model 3 Framework

Diane Tavenner

I think so. And I do think, you know, yes, I hope that what we’re building can live in both worlds and is one of, you know, the early ideas or components of what a Model 3 will look like. And I certainly will be engaging with folks on pushing that area, so hopefully we’ll talk more about that. I think where this is all leading for me is the next part of our season. So we’re gonna talk to a bunch of different people and I’m gonna be really. I’m gonna be in the back of my mind thinking, all right, well, where do you sit in this imperfect framework, this developing frame? But, but sort of, where is your effort sitting in that? Are you literally a whole school model? Are you an element to a model? Are you, you know, an AI enabled tool? Are you really trying to push the boundaries of designing for Model 3? Are you an interesting model two? And what do those look like? So.

Michael Horn

Yeah, well, and that’ll be interesting because I think as I look at the guests ahead, we have a lot of folks in Model 1 who are working with that system. And I’ve been wondering, given the hypothesis that we have fleshed out over the last couple of seasons of AI, like how that fits with the things that we’re interested in. And this is good. I think we’ve given a good framework on the importance, frankly, of all three of those elements and the work that they need to be doing and the dangers of crossing over perhaps, assumptions from the worlds across the different models.

Diane Tavenner

Perhaps. Awesome.

Michael Horn

This got interesting. A little spicy.

Diane Tavenner

A little bit spicy. Well, super useful for me and helpful for me to think about things. Any last things on your mind?

Michael Horn

I have one last thing. Hopefully we won’t get cut out of the studio, which is, I thought a lot about what is the world into which people are going and how does that map back to what is still core and what is not core and so forth. And I just want to float an idea by you and have you attack it.

Diane Tavenner

Great.

Michael Horn

The reflection I’ve had is we know there’s a considerable amount of cognitive science that suggests we learn best through story stories, narrative arc, and we don’t actually deliver most learning or offer learning opportunities in that. And so I guess I’ve been wondering as we think through, you know, we had the back and forth of do they need to memorize state capitals? And we both said, probably not. But I do think they should know that there’s a thing as a state capital. And so my thought about it is almost like Montessori has the I’m gonna mess it up, the great lessons or something like that. Right. And it’s a narrative arc. But I almost can imagine narrative interactive arcs where you’re like sort of, okay, how did the country’s governance evolve over time? And these thin layers that would build a lot of common reservoir of knowledge. And I think I’m largely talking K5, maybe K8, that that could be a big part.

And like in, in the various disciplines, if you will. Right. Civics, a variety of deep dives in history, et cetera, et cetera, science. I think it should be active. I think it should be multimodal. It’s not clear to me. It’s the teacher delivering the story.

Diane Tavenner

Say what you mean by multimodal, because a lot of people are using that term and I don’t think many people know what it means.

Michael Horn

Yeah, yeah. So I guess I see it as being like, you can imagine it being some of these lessons being video based through an AI. You can imagine an auditory sound. Right. You can imagine interactive where you’re actually answering questions both verbally and written as you’re working through something, you can imagine, like the state capitol one. So you have a lesson around how did state capitals evolve in state government?

Diane Tavenner

I mean, it could be VR, like literally immersive.

Michael Horn

Right, Exactly. And then you could almost imagine then like you pop out and like, my kids still draw maps. I actually think that’s really valuable. But I don’t think that they then have to drill memorizing every feature, but they don’t know what question to ask Gemini or ChatGPT without like sort of that thin knowledge base. Right. And that’s sort of where I’m wondering if you’re. We evolved to something like that that recognizes the importance of some knowledge.

Diane Tavenner

Yes.

Michael Horn

We could have mastery assessments where we thought it was really important.

Diane Tavenner

Yes.

Michael Horn

We don’t have to have it for everything, frankly, it’s just exposure is probably good enough, especially if it’s interactive. I don’t know. What do you think of that idea? What are the flaws? And sorry. And then creating the space then for like, hey, you’re interested in this? Okay, here’s your project. Go deep, right? Like, and that’s where the deep explorations of learning how to learn and developing the skills would really be.

Diane Tavenner

This feels very fun to me to think about this. And these are the types of thoughts I’m constantly playing with and that I think should influence the design of Model 3. I love that you brought up this idea of memorizing the 50 state capitals because I think maybe we are misunderstood when we both say we. We don’t necessarily think kids should memorize the 50 capitals. That’s not because we don’t love America, believe in America, think that they shouldn’t. I think what we’re both more interested in is literally having them have like a deep story about each of the capitals and really internalizing. I mean, I will tell you, we get to travel a lot. Do you, do you like how I frame that? We get to travel a lot.

And when I travel, I love this country so much. It’s so fascinating. There’s so much.

Michael Horn

It’s so much fun to dive in, right? And take the, like you’re in, you’re in wherever and you go to the Alamo or whatever it is. And like, it’s so much fun.

Deep Learning Over Memorization

Diane Tavenner

It’s so curiosity driven. And so what if young kids didn’t memorize 50 capitals? But what if they went deep on a couple of them, like in a story based way, in an immersive way, and they got the idea of state capitals and what they mean and the importance. They got very cool stories about, you know, a few of them at that age. And then they got a lifetime of like, oh, I could, there’s so many more I can learn. And there’s so many interesting stories about them. And they’re not just a name on the page and, you know, on a flat map, but they’re real places that have real significance and they’re different from each other and because they have such access to knowledge now, if they really need to go look it up, they can go look it up..

Michael Horn

They can do the deep dive. Right? And I think the knowledge conversation, I’m a big believer in the importance of a fundamental knowledge base and the depth at which those occur. I think we don’t have a nuanced conversation around.

Diane Tavenner

Right. And I also am okay with it, I’m gonna call it the Swiss cheese of knowledge.

Michael Horn

Yeah, so am I.

Diane Tavenner

That you don’t have. Every fourth grader in America does not need to know the same facts.

Michael Horn

Yeah.

Diane Tavenner

It’s okay if we learn them at different points and different times and that there’s, you know, sort of regional differences around that. I’m much more committed to everyone having a common set of really important skills, at least at a baseline level. And then ideally spike lots of people spiking in the different skills in different places because we need all those.

Michael Horn

But when you say the skills, you’re thinking that it’s been developed through them working in different domains and areas repeatedly in deep dives. Right. And so

Diane Tavenner

Because you need content to practice skills.

Michael Horn

Exactly right. And you create that integration. I think a lot of times in school it goes the other way where like, oh, we learn how to think critically about what.

Diane Tavenner

Exactly.

Michael Horn

And so again, these crosswalks extremes, I think are right. Yeah. Anyway,

Diane Tavenner

Yeah. And so, you know, and this is why we both like a project based environment because it’s the integration of the two and there’s such power in what AI can do now where you can really do personalized learning on, in the content to bring to those, you know, engaging, collaborative, communal type, project based experiences. So I mean, I love what you’re saying in the direction you’re going. It’s very nuanced as you know, it’s.

Michael Horn

We should have some more fun later on and. But I just wanted to float the general idea because I had this moment in our conversation with Alex where I was like, at what level are we thinking about difference and what does stay the same? And I think part of my reflection has been there’s actually a fair amount that stays the same, but how we’ve done it probably changes pretty radically.

Diane Tavenner

Indeed.

We’ve been recording pretty frequently and I know we’re both feeling a little stretched on thinking about new books and things we’re reading. We’ve maybe exhausted our list so I thought maybe we’d take a break from that list only today. Thank you. And replace it with this will make this episode a little less evergreen. But for those who are listening, we’re actually recording this right before the week of Thanksgiving, and I thought I would end with some gratitude.

Michael Horn

Oh, I like it.

Diane Tavenner

So one of the fun moments of yesterday’s engagement with your class and then the office hours afterwards was there for so many young, amazing people who so many of their questions were very personal yesterday about, you know, how to be a mom and lead and how mentorship and all of. And, you know, my relationship with my husband over the years. And I’m so. I’m appreciative that they were thinking about that. And one of the things that came up was just our friendship. And I think you know this. But I am so grateful for our friendship, and it is truly one of, for me, the big, you know, if there are any highlights coming out of COVID the fact that we decided to do this, it gives us time together. It’s just so much fun, and I’m so grateful.

Michael Horn

You know, I’m a crier, so I’m trying not to right now. Thank you. I feel the same way. And it’s one of those things where I feel like, how lucky am I that we get to have this conversation? Even though I moved away from the Bay Area over a decade ago, which is wild, 12 years, but. Yeah. And I think it’s. So when this comes out, it’ll be after the new year, I think, and so forth. But I always tell my students, because, as you saw, like, 55 or so percent are not from the U.S.

I say take the time because how cool is it to have a day when you get to say thanks? So thank you as well. Yeah. And thank you all for joining us through the sentimental moment. But also on Class Disrupted. And just keep your questions and curiosity coming. We suspect there’ll be things you disagree with that we said here, and we can’t wait to learn from you. So thank you, as always, and we’ll see you next time on Class Disrupted.

The Future of Education is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.