The Challenges and Chances of AI in the Classroom

Pradnya: Think, debate, inspire, debate on pressing global challenges, a podcast of the Robert Bosch Academy. Hello and welcome to the latest episode of the podcast series, Think, Debate, Inspire of the Robert Bosch Academy. I'm Pradnya and I'm a senior project manager at the Robert Bosch Stiftung here. Our guest for today's episode is Matthew Rascoff, who is an education innovation leader and has worked across sectors to democratize access to knowledge and opportunity and since 2021, he's been associated with the Stanford University as a inaugural vice provost for digital education and as a lecturer in management at the Stanford Graduate School of Business.

Together with the Stanford Digital Education team, he has pioneered a model of mission-driven digital learning that accelerates social mobility for students who have been historically underserved by higher education. So in our conversation today, we are going to explore various facets around digital technologies, specifically AI and how they come into play when we talk about education. So, welcome to the podcast Matthew. Lovely to have you here.

Matthew Rascoff: Thank you so much to be here and I'm sorry I cannot be there in person that I'm so grateful for this exchange, which I see as a kind of continuation of all the wonderful conversations I had as a fellow at the Bosch Academy.

Pradnya: So to begin with, would you like to tell us a little bit about the project that brought you to Berlin and what you were working on during your stay?

Matthew Rascoff: I'll tell you a little bit about the project as it originated, but I think my time there also changed the project and it changed my thinking and that may be actually the more interesting conversation to have for what was the influence of being there on my ideas. I came in thinking about these questions of digital learning and AI and how it will affect the education system, how education systems will absorb, assimilate these ideas, make them productive for learners, make sure their benefits are shared fairly and equitably across the system, and the question of how that will play out on both sides of the Atlantic and different countries around the world is very salient to me.

Matthew Rascoff: These education systems were built by nation states. They're products of 19th century nationalism and yet the technologies that are developing today do not respect borders. Open AI does not really care what country you're in. They care if you're in a GDPR country or not, that's about it, and even then in many cases they just enforce GDPR regardless of where you are around the world because that's convenient for them. And that mismatch between the design of our education systems which was meant to support preparing a citizenry, which was meant to support preparing a military which was meant to support various goals, all of which applied to nation states, and now the world that we're living in, which does not seem to be purpose built for the needs of learners, where the technology really does not care.

Like that mismatch to me is it's both a source of friction because education leaders don't know what to do with it. They don't know how to think about these technologies, but it's also a source of immense opportunity because it means we have the chance to redesign these systems and to rethink how might they be more global, how might they be more European, how might they respect difference in a different way. And all of those to me are the possibilities that have been handed to our generation of education innovators to figure out.

And those were the questions that I brought with me to Germany. I would say in my time there, but the influence of my colleagues on my thinking, it was very much about the political aspects and the moral aspects. So I'm thinking of Natalia Gavrilyetsen and her experience of leading Moldova in a country under pressure from the East trying to figure out European citizenship. The questions, they're not theoretical for her. Like what it means to be part of Europe, what it means to prepare young people to be Europeans or not or something else to have pride in their country and in their culture, which is so obvious from any interaction you have with her, but also to be part of something bigger.

Is that something we should still aspire to? How will we prepare people to be part of it? That had not been on my radar, I would say, as an American showing up in Germany, but it very much is now. And I would say, you know, Barrett brought that from the Netherlands and the many lunches that we had. I think they widened my perspective certainly on the political dimensions of this. And education, you know, because it's a product of nation-states, it is also a political artifact. We don't usually think about it that way, we think of it as nonpartisan, but it was produced by a certain set of ideologies.

And those are also being questioned right now. And I think it's useful to frame these ideas in that larger context.

Pradnya: Yeah. So before we jump into the conversation on the interplay of AI and education, I would like to take you up on something that you just mentioned, which is education being a product that has been conceptualized by nation-states as a construct, and then you have the technology giants in a way, who have a very different reasoning and rationale for their business models and the way they operate. So do you see a space of constructive intervention between these two very solid conversations that are happening?

Pradnya: So, you know, like the purpose of education versus the purpose of technology in a way, because they are driven by two very different purposes for lack of education.

Matthew Rascoff: For lack of education. Yeah, I would say the current constructs for that interaction are one regulation. Like, we're going to keep this out of our schools, we're going to keep this out of our country, we're going to block it, we're going to limit it, we're going to come up with a custom version of it, and two, digital sovereignty, meaning we have to own this data.

Matthew Rascoff: You may be a global company, but you're going to put your servers inside our borders, and then we will control the data, and we will decide what gets built and how it gets built, and the limitations on the data. And, you know, both of those, I would say, are very much in the discourse in Berlin right now, and in Europe right now. And, you know, part of the reason that technology is behind in Germany, I think, is because of this ideology of digital sovereignty. Like, if the government can only procure technology from German vendors, then you don't get to use the cutting edge, and like that's going to limit the pace of innovation, not just in education, but in AI writ large.

And I think that, you know, it's not really a question for me to answer. It's a question for, you know, the citizens of Germany to decide, like, are they willing to make that trade off? Are they willing to live with one technological generation behind, so that they can have sovereignty over their data, and that their government procurements will only come from government- you know, approved vendors from, you know, German companies.

Pradnya: Which is the way of regulating it.

Matthew Rascoff: It is exactly, but both of which are trying to kind of reinstate national thinking and layer it on to this kind of global technology giants.

I don't think, as you can tell, I don't think either of those are going to be quite successful. And I also don't think, like, education needs that. I don't think it would benefit from that. Because, to me, like, the fundamental questions in education are actually universal. Like, brains are pretty much similar around the world. And at the root, at the kind of fundamental root, when you're building learning technologies, you're designing supports for learning, you're designing supports for brains for human beings who are fundamentally the same.

The education system attaches a lot of, you know, cultural and linguistic and political elements to those fundamental cognitive processes. And those are unavoidable. Those are part of the education system. But really, at the root, we should be able to build great learning experiences and send borders, and then maybe figure out how to adapt them for different languages for different cultural contexts. And, like, that to me is kind of a more hopeful take for how we could be, you know, more global in this space. And there aren't very many examples of that.

Even the global, you know, education companies, like, they don't tend to think along these lines. I don't think they're thinking, you know, about these kind of fundamental questions of neuroscience and neurodevelopment. But there are some examples, you know, of global education concepts that really play out. And then it's happened in the past also. One of my favorite education books is Biography of Maria Montessori. It's called The Child Is The Teacher. And it's about how her idea spread from the slums of Rome to all of Europe and then to the United States and her, you know, became ubiquitous around the world.

And she built kind of the first global education concept with a toy manufacturing operation as the commercial, you know, means behind it that allowed her to spread her ideas. Like, that happened in late 19th, early 20th century. And so there are models for that. And, you know, certainly early childhood education, Pestalozzi, that all originated in Germany and in Berlin. Like, those ideas spread around the world. So there have been movements like that. Some, you know, in Maria Montessori's case, it was enabled by manipulatives.

There was hardware technology. That's what she was spreading around the world. That was kind of the enablement. Now it's software. But there are definitely some examples of how this could work in a positive way. I would say open educational resources are probably the current example of how this could play out. And that movement is also kind of being rethought and reimagined in the context of AI. But I would say I am hopeful about the universalist potential of positive transnational education, you know, thinking.

Pradnya: So there is a tendency, especially here in Germany or in Western Europe, to keep technology specifically AI out of our classrooms, because of challenges around regulation and so on and so forth.

Pradnya: Because we also want to be controlling the kind of data that we are providing and putting out there. So in your work outside of this fellowship, you have also kind of looked at the positives that access to digital technologies and how that can help us bridge gaps in terms of access to education. Some of your work at Stanford has also focused on social mobility for students from underrepresented groups. So even if we are to be cautious about, you know, bringing in technology into the classrooms, there is a big chance that we are missing out on if we don't do it.

Matthew Rascoff: Absolutely. I will share one example from Stanford and maybe one from outside as well. The work that I have done at Stanford is about reimagining digital learning, repurposing it, to serve the goals of equity and access to education. What we have done is offered courses from Stanford for credit to low income high schools across the United States that are hybrid. So there is a digital component that is offered by, you know, the university with content, assessments, and live instruction, teaching fellows, offer discussion section every week, help with problems, office hours.

And then there is a local in the high school component with teachers who we train and support. And they can provide the accountability. They add a social element. They bring the students together. They ensure they are staying on task. And that hybrid model has allowed us to offer Stanford courses in this model. It is called dual enrollment. I don't think it exists in Germany where the students can earn both high school and university credit at the same time. That serves as a springboard for them, you know, psychologically and academically, in terms of social capital, to increase their chances, to increase, to set their sights higher.

You know, low income students we know from data in the US systematically do not apply to selective colleges. Even the ones who would get in, even those who are in the top 10 percentiles of grades and test scores. The majority of low income students in that category never apply to any selective institution in the US. So they don't show up at places like Stanford. That the term for that is under matching. Meaning wealthy students in the top 10 percentiles of grades and test scores generally apply, get in and attend and graduate from selective colleges and universities.

And that doesn't happen for low income students. So this intervention is really about building a different and better pathway for low income students from what you'd call high school gymnasium or in Germany students would be tracked. They might not even be in a gymnasium. They might be in a comprehensive school. But talented students are systematically left behind by our systems that are not designed really for identifying and cultivating their talent and supporting them and bringing them in to higher education. That's what our digital strategy at Stanford has been about for the past four years.

And I'm very proud of that model. It's not fundamentally about the technology. It's enabled by the technology. But it's technology and people working in a hybrid where we've seen much higher rates of student success, course completion, grade performance, then you would in a MOOC, in a fully self-paced, fully online course, without a community learning alongside you. And in those cases, there would not be academic credit provided because the rigor is not there. So what we've done is something that's rigorous that holds up to the standards of our Stanford faculty with the same learning outcomes and the same assessments that we would offer on campus.

But we've been able to provide this to thousands of high school students and low income communities across the United States.

Pradnya: So would you go to the extent of saying that it has democratized access to education in the higher institutions?

Matthew Rascoff: Absolutely. Absolutely. It has. And we follow these students and their trajectory, you know, what comes after. So 13 of the students that we supported in high school are now matriculating as undergraduates in Stanford. We've sent 35 of them to the University of California, 19 of them are in Ivy League institutions.

These are all students that we identified, supported, cultivated, and then set on this path of selective higher education. So we have a research study that's tracking their longitudinal performance. So we see their university performance has improved. Their attention is going up. Their likelihood of getting financial aid increases. So we've got a study in partnership with Johns Hopkins University that's tracking the longitudinal outcomes of this model. And we've got, you know, 15 other institutions across the United States have joined us in a consortium in providing courses on this model as well, including many of our great peer universities.

So yes, I do think it has democratized access, not enough. There's a lot more work to be done. You know, the problem is on the order of hundreds of thousands, we're shipping away at it. Across our consortium, we will serve 25,000 students in this past academic year. That's good, but we need to keep going. And my concern is that this needs to remain the focus. And, you know, the policy environment that we're operating in is putting a lot of financial pressure on institutions, including Stanford, and it's making it difficult to impossible to sustain efforts that are oriented towards outreach, that are oriented towards extending opportunity to those who've historically not had it.

There's a massive narrowing of mission and focus to just ourselves and our own interests. That's the response to a lot of the pressure right now, and that is extremely dangerous. I think when our public value is being questioned by society, we need to reinforce that. We need to be more aggressive and more open and more porous and invest more in public goods and the creation of opportunity for everybody across the country. And there cannot be a closure. There needs to be a reopening. And that to me is a really...

Pradnya: There's a very big contradiction at play here.

Pradnya: You are trying to democratize access to education for a larger group of people, and on the other hand, the policy is kind of making it very difficult to make it equitable and for people to have easy access.

Matthew Rascoff: Exactly. I mean, we have not come under direct attack ourselves from the right, but if our broader institutions are, and we have to live in the context of those institutions, it's going to be very hard to sustain this work across those places. And great universities do hold a place in society. They represent a set of ideals.

Matthew Rascoff: The reason students are willing to work so hard, the reason their completion rates are so high for our courses, is because they aspire to Stanford. They want to earn the credit from Stanford. They want to get an A in one of our courses and get a transcript from the university that says, Leland Stanford Junior University at the top, that will help them. And their goals, if they want to come here or to another institution, that will help them. So no other institutions play that role. Like a startup can't do that. It doesn't have that motivating power, but we do.

So we have a role in society that's really about aspirations. It's about motivating people to work harder to achieve, to challenge themselves, to be part of great learning communities. And if we don't do that, who will?

Pradnya: And the second example, which was outside of Stanford, that you brought up earlier?

Matthew Rascoff: Well, I wanted to share a little bit about this kind of model of AI that I see emerging. I shared a little bit with the foundation staff. And the framework that I think might be useful for your listeners is to think about three steps of the development of digital learning in AI.

And education. The step that we're in right now, I call it one-to-one, meaning we've conceived of these tools as kind of consumer tools for individuals. It's helping me with my project, with my research, with a question that I'm asking. And the heuristic is like, good tutor, who can ask you good questions, who can prompt you, who can challenge you. But it's really about you as an individual learner. Step two, we call many-to-one, meaning a tool is there to help a teacher understand what is going on in her classroom.

Can she aggregate information from many different source systems? Can she identify a student who's at risk? Can she assess students more effectively? Can she provide information and gather information back more productively? And the many are the students, and the one is the teacher, and the AI can facilitate that. But the highest tier of this technology and the goal that I think that builders should be focused on is what I call many-to-many. And many-to-many means the AI is there to facilitate human interaction. It's there to enable people to work more effectively together in small groups.

The role is more akin to like a teaching fellow, or we call it section leaders, a really great enabler of a conversation, of a partnership, a collaboration, a design exercise, whatever it is, that many-to-many model. That is the kind of emerging technology. That's the new possibility in this space that I'm most excited about. And there are some providers that are working on this, both in research labs, but also in commercial education technology. So one that I really like is called OCO Labs. And it's a K-12 education startup, and it basically uses this AI to support small group turn-taking.

And the AI is there to basically call on different people and say, okay, now it's your turn, give a little bit of feedback, semantically analyze the response, can determine the accuracy of a response. But it's basically there to manage this very fragile process of small groups of working together that's very hard for a teacher if they've got 25 or 30 students in the classroom. It's very hard for them to manage five groups of five. But the AI is there, not as a replacement for the teacher, but as an enablement for them to do more small group learning to allow these students to work better together.

And I think you're a question for me about this model. What is the role for the teacher? The role for the teacher is to manage these technologies. It's to ensure that they're working effectively, it's to kind of distribute them, and then gather information out of them, because they can produce relevant kind of assessment information about who's doing what, how much are students speaking, are they getting a fair shot, and what are their struggles, and what might be a better group to put them into next time. And that many to many concept, that is a very exciting direction for this technology.

And it's not what most people think about when they think about education, AI. They think, chat, chat is one to one. But to me, the real possibilities will emerge in many to many.

Pradnya: Yeah, I think it resonates with me on multiple levels, because on the one hand, I'm also thinking about the implications of using that in higher research, because when you have a broader platform where you can put up your findings and your experiments for other researchers to learn from your mistakes, I think there is an equalizing force at play, which kind of transgresses boundaries.

Pradnya: You know, there's global south, global north boundaries, access to resource boundaries. I think there is a huge potential there.

Matthew Rascoff: Absolutely, and there is a group of researchers that are working on that idea. There's a mathematician, Terence Tao, at UCLA, who is working on AI technologies that will facilitate research collaborations in mathematics that allow people to work together in labs around the world, to do kind of big science-style research in mathematics, which has really struggled with co-authorship and doing the big collaborations that you've seen in physics and biology.

Matthew Rascoff: Typical physics paper or biology paper might have dozens of authors on it. Math papers typically don't, because it's very hard to work together. And what he sees the AI is the kind of the enablement for a large research collaboration by allowing people to check each other's work, by supporting kind of this turn-taking model on the research side rather than the teaching and learning side. So I completely agree with you, and there is actually a lot of progress that's being made. If you look up, maybe we can put in the show notes what Terence Tao is doing.

He's given a few talks on this topic, and I think it's very inspiring.

Pradnya: But having said that, the question that arises out of all of this conversation, I came across a course that Wharton has designed together with OpenAI, where they wanted to roll out AI training course for teachers, how to use generative AI in classrooms. And just the fact that OpenAI is on board with the designing of that particular program would not particularly create a sense of great confidence amongst a typical Western European mindset because then they would have issues in believing the motives of OpenAI behind designing this course.

Pradnya: They probably are not being upfront about the slippery slopes that go along with it. So is that a space that you also come across during your work, and how have you navigated that so far?

Matthew Rascoff: It is a question that comes up all the time, and I think it's a legitimate one to ask, like who made this, and can I trust it? That's a good information literacy question to ask. It gives me some confidence to know that Wharton is involved, also that it was not entirely the product of OpenAI. There is some peer review, I don't know what the course in particular, I don't know what the process was.

Matthew Rascoff: We at Stanford, my team actually built something similar in partnership with Google, and we took some content from Google, but we also built some of our own in the context of high schools, critically, and the idea is not to just use the technology naively, unthinkingly, and unquestioningly, but to challenge it and to think about, how is this produced, by whom, with what bias, and what are the limitations of it? That to me, real AI literacy is not just about how to use these tools or why in what context, but it also allows you to ask those more fundamental questions about its politics, its production, its funding, and who is behind it.

To me, it preserves a role for independent education institutions that is quite helpful. It means we still have something to offer this conversation because we are outside of that system. We can use it, we can understand it. Obviously, I don't think it serves students well to ignore it completely, but let's use it with the critical lens. When I was at Duke University previously, I taught a course that was about examining education technology as a text. Let's really analyze the terms of service and the privacy policy, and understand what is it asking us for before we dive into this unthinkingly and with so much enthusiasm, let's really understand what are we doing here?

It was a close reading of education technology from a critical perspective. That to me is absolutely necessary, and it doesn't happen very often. Usually, the model is, I will adopt it. Adopt it. It's my child. Now I owe it forever, and I am responsible for taking care of it. That's not the right way to think about technology. We don't adopt technologies. Technologies are there to serve human needs, and they're built by humans, and they have constraints and limitations. That term to me, the term user also leads me quite cold.

I don't think that's the right model for education. Students are not users. They should be interrogating it. They should be...

Pradnya: Co-creating it.

Matthew Rascoff: Co-creating it. Exactly. This should be a creative process. The user sounds so passive. It sounds so limiting. It sounds so commercial. It's very top-down approach. All that, yes, I agree with the limitations. To me, that suggests two things. It suggests that there's a role for independent education institutions to build models of AI literacy. AI fluency is what Anthropic calls it.

I like that model. They can preserve some autonomy in that process. And second, I think it suggests that there will be an ongoing role for education-specific AI's that are sensitive to the differences that respect the idea that a student is not a user and neither is a teacher. And that good educational AI will come out of something that feels like a creative process, a generative process, a classroom process. So there are companies like PlayLab AI that are trying to build kind of an education-first model that I'm very hopeful about.

And it's different than what you might get from the big frontier labs. And I would encourage educators who are listening to this to check out PlayLab, for example.

Pradnya: Okay. Again, building up on what you just said. Two questions come to mind. On the one hand, there is this genie that you've let out of the bottle, which is AI and technology in a broader sense in classrooms. And a recent study that came out of MIT kind of tried to draw attention to the fact that young adults who are increasingly using chat GPT or generative AI in a broader sense in their classrooms are more prone to erosion of critical thinking skills because they have stopped asking the critical question.

Pradnya: So as an educator, it's a very, very thin line that you have to walk on. So on the one hand, you want to bring in technology into your class. But at the same time, you also have to put in the stop gaps. Into the picture. And I think right now, a lot of conversations, at least that I follow, which is not from the perspective of an expert educator or anything, just as someone who's interested in the topic, is that it's a very binary conversation. Either you have technology in your class and then you just let it take its course or you don't introduce it at all.

And I think there is a nuance that is missing in the public discourse. And where do you think that is coming from?

Matthew Rascoff: I agree. I think the nuance is missing in the public conversation. It's also missing in the education policies. Like so schools are saying we've blocked it or we're using it. But use it to what end and what context and what classroom, what students and what courses, and that's actually where the interesting questions come in. That's why educators need to get involved because those are far too nuanced, as you said, for a district-wide, lend-wide, nation-wide policy to handle.

Matthew Rascoff: So to me, I think there's three different layers of nuance. One of them, I would say, is about mastery. So when you have mastered a certain skill and you've practiced it, you've developed it, you've perfected it, it's okay to offload that. You don't need to keep doing arithmetic in your head. Once you've figured out arithmetic, once you've learned the multiplication tables, it's okay to offload that to a device that can do that work for you. And the spreadsheet, you know, that you've got in front of you to do your accounting, it's not harming your mastery of mathematics.

It's just helping you do your job more effectively. So mastery is one. Two is developmental, developmental meaning, like I would not introduce certain technologies to certain children at certain ages. We need to have a theory of when it is appropriate for them to use different tools. Regardless of their mastery, you know, like it's not appropriate to introduce screens to the very youngest children. We know that. And so, you know, I don't think we yet have a developmental framework for AI. But we need that. And I think that should be a research focus.

And then the third is disciplinary. And I've seen this in my own teaching. So, you know, we have a policy at the Business School at Stanford that says we can't stop the students from using AI. You know, we have to allow them to basically bring it in. They just have to cite it as a source. And to me, the question is, you know, what are you trying to teach them? And if they're using it, you know, in, you know, a marketing course to do all the marketing thinking for them, then they've offloaded the creative work. Like that to me is very dangerous.

But if it's a policy analysis course, let's say it's a data analysis course. And they've done all the data analysis themselves. And all they want help with is writing up the results. And they've done, you know, independent creative thinking. And they want help writing up the results. And maybe checking the grammar on that. That seems perfectly fine. And the disciplinary context of an analysis course, offload by all means, you know, the work that has not relevant.

Pradnya: You're bringing in a sense of accountability into the conversation.

And they're not stigmatizing it. Because then you stigmatize it. You say, oh, you know, like, oh, I use chat GPT, but I don't know how to tell my prof that. Because I judged.

Matthew Rascoff: Exactly. So I think the policy has to respect, like, what are the learning goals of this course? And if the learning goals of this course are about data, then it's okay to offload some of the non-data focus. And, you know, similarly, like, I was using AI to help with German translation when I was in Germany constantly. It may not have helped my linguistic development in Germany so much.

That was important. I was trying to improve my German. But in some context, it was, you know, my goal was different. It wasn't about perfecting my der, die, das. It was about, you know, trying to produce some paper, some idea gets some idea into the conversation. And it's like, you know, you sort of need to answer the question with respect to the learning goal at hand. So that's, I think, the framework for thinking about this. It's about mastery. It's about development. And it's about disciplines. Like, those are the three layers of nuance that I think we need to decide when it is appropriate to bring this technology in and on what terms.

And I don't think I haven't seen that yet emerge. But I think that's what teachers will need in order to make these finer grain distinctions.

Pradnya: Mm-hmm. Okay. So before we wrap up the conversation on this part of the podcast, two questions for you. The first question is, if you're aware of any particular projects, however small scale they are, which are working towards diversifying the LLMs that a lot of these technologies are based on because the construction of LLMs is also kind of driven by a particular power dynamics at play.

Pradnya: So that's the first part of the question. And the second part of the question is, if a policy maker who is cautiously interested in introducing technology in education policy, but doesn't know where to start or, yeah, is looking for inspiration. What would be the three things that you would tell him or her?

Matthew Rascoff: Looking for inspiration for building in this space or...

Pradnya: No, for introducing technology in education or in classrooms to be specific in a responsible way.

Matthew Rascoff: I've got some resources for that. I would say on the diversifying question, it is an important one to ask.

Matthew Rascoff: It's not just the models themselves, but it's also the training of the models, which has been done in many cases by humans from very particular backgrounds, cultural contexts. There's a lot of outsourcing in that to low-income countries. And one of the reasons why you sometimes see the linguistic quirks in the English is because of the places where the training has happened through reinforcement learning. So that's kind of an artifact, the peculiarity of the globalization that we were talking about before. There is a startup that I think is quite exciting in this space.

There's a research group at Oxford called Cosmos Institute, which is both a nonprofit organization, but they also have an investment arm. And they just funded a new group that's building custom AIs for every individual. And the idea is rather than you putting your information into one of these big platforms and then letting its memory improve and understand you better and better over time and then kind of build a model of you. So you're hooked on it forever. Once it knows enough about you, it becomes more and more sticky, more and more kind of addictive.

Their model instead is that everybody should have their own. And you'll train your own and you'll own it. And it'll be part of your toolset. And it's a different perspective. Forgetting the name of the startup right now, I think it's...

Pradnya: We'll find it and put it in the show notes.

Matthew Rascoff: The Cosmos Institute is a very interesting organization at Oxford and they call themselves philosopher builders, meaning you can't go into this space without having some fundamental ontological and epistemological views. And you really get down to those very quickly, actually.

And that's part of what's so fun about this area. I studied philosophy myself as an undergraduate and I feel like I'm now putting it to work 20 years later in unanticipated ways because you really do need to have a theory of the world. When you're designing in this complementary way, you need to have a theory of what will the humans retain and what will they not? And those really basic questions are very exciting and they have not typically been asked, certainly not in Silicon Valley, where it's often just about what's the next feature?

How will we get people to stick with us for longer? How will we get them to come back?

Pradnya: That was the next water looking.

Matthew Rascoff: Exactly, exactly. And I like this group at Oxford because they are definitely embracing the challenge of thinking bigger about these technologies. I would say among the big tech companies, Anthropic is the one also that is visionary, it's bigger thinking, and they have grand ideas for what these technologies can do in the world that go well beyond financial returns. And that's exciting to see as well.

In terms of inspiration, one resource that I would think about is a teacher who's been a fellow on my team, his name is Mike Taubman. And he recently started his own newsletter, a sub-stack newsletter, but he's been posting almost daily about different exercises that he's built for his own high school classroom in New Jersey. For how he is teaching with AI, he helped us create the AI curriculum that I mentioned before that uses some of the Google resources but examines it critically. He helped us do that. He is one of the best thinkers in this area.

So rather than trying myself to provide this inspiration, he's so creative in this area and he is himself a teacher in a high school who's working through these ideas and trying to understand how to support students with the technology and how to overcome some of the constraints. And he's doing it around like an individual exercise. He's got a lesson plan for one day, and he publishes the lesson plan. He says, you know, here's how it worked with my students. Now you try it. He's kind of building a community from the ground up that's really organized around a day of learning or an hour even of learning and how an assessment question played out.

How a challenge from the news came into the classroom and how he presented it to the students. Mike Taubman is the teacher and let's put it in the show notes. So some of our audience can listen in on this and really get inspiration.

Pradnya: Cool. Thank you so much. But my question still remains partly unanswered. Three things that you should, that a policy maker who was not very familiar with the space of technology but considers himself or herself to be in a position to make policies on education and technology. That they should be factoring in and then that would help them be a little more open-minded towards introducing technology in classrooms.

Matthew Rascoff: That's a scary prospect. A policy maker is trying to work in this space where they don't have experience.

Pradnya: Unfortunately, that is the reality.

Matthew Rascoff: It is the reality. And, you know, I think one kind of fatuous response that I often hear is like, well, go familiarize yourself with this technology. Like, that's not actually the right answer in this context. Because I think the most exciting in eyes are really the ones that are particular, not the ones that are general. They're the ones that are really for education. And I think they have like a teacher-centric model, a student-centric model that's very hard as a non-teacher non-student to understand and to get the value.

So I really do think we've over-indexed on kind of taking these consumer technologies and assuming that that's how it's what's going to play out in the classroom. And then we end up reinforcing really bad pedagogy. And, you know, there's an AI company that I don't want to name. That's the fastest growing ed tech company in history. It's the first one. They went from zero to a million users in like a space of a few months. But what it actually produces is worksheets. It produces basically the same kinds of materials, the same exercises, the same on we in education that you could have had with the previous generation of technology.

But it just does it at light speed. You know, like the more efficient production of bad teaching and learning is a very scary prospect. And I don't, we don't need to scale what we did in the past. Like, let's figure out what we need to design for in the future. So, to me, like, I think that, you know, the challenge you just presented, it's really not a technological challenge first and foremost. Like, what do we...

Pradnya: It's a mindset challenge.

Matthew Rascoff: What do we want to equip our students with? And what remains for humans? What are our goals for them?

How are we going to support those goals? And the technology will play some role in that. But it's really there in the service of that larger educational mission. You know, here's a challenge for you. Like, these AI's can pass all the tests. You know, the Abitur can be aced by an AI at this point. So what is the point of the test? If the AI can do it better than a human already can. Like, that is a much...

Pradnya: I see the fallacy of my question. So maybe that deserves a own podcast episode of it.

Matthew Rascoff: Well, I just mean like, like, it's...

Those are the questions that I think we should be asking. Not... The purpose of education and why... And the purpose is evolving right now, as the scope for human beings is changing in response to this technology, it is going to have to level up in a way. And the things that we educated young people for before are no longer sufficient for the challenges that they are going to face in their lives. And that is a very scary prospect, but that is actually what we need to confront.

Pradnya: True. So before finding answers to the big questions we need to find answers for the fundamental questions.

Pradnya: So...

Matthew Rascoff: Exactly.

Pradnya: On that very highly philosophical note, coming back to the lighter part of the conversation, we always ask our fellows to at the end of the conversation if there is a particular highlight, low-light moment of surprise from your stay in Berlin that comes to your mind. When you look back at the time you spent in Berlin.

Matthew Rascoff: Okay. So, you know, I've been to Germany many times. I've been very fortunate to come at different points in my life, at different points in my career. I was there in my 20s and my 30s, now I'm in my 40s, and I got to experience the city in the light of a family, because I brought my two children with me and I put them in the JFK school in Zehlendorf.

Matthew Rascoff: And, you know, I thought I understood something about the education system, and then I put my kids in German-American school, this bicultural school, and it opened my eyes in an entirely different way. And so, I would say-

Pradnya: That was the moment of surprise.

Matthew Rascoff: There are many moments of surprise. And most of them, pleasant surprise. You know, one of them very simply was that, you know, young kids in Germany, including, you know, sometimes first-second graders, get themselves to school in the public transportation system.

And my son, who's a third-grader, built up to that, and by the end of our time there, you know, the end of the school year, he was navigating two buses and a transfer beginning from our house in Dahlem to his school in Zehlendorf in 25 minutes away with no hesitation. And, like, that-

Pradnya: That's a level of independence and autonomy.

Matthew Rascoff: That's not part of the formal education system, but it's very much the informal education system. Like, that's what students are actually learning. And that is very exciting. The German curriculum requires swimming in the third grade.

So, you know, he already knew how to swim, but every single student, it's life skills, but it's also- it's embracing a social role for the school that is very rare in the United States. It's basically saying the school's responsible, not just for your academic well-being, but in this more holistic way for your entire being. And, you know, every year, the few dozen students in the US drown because they did not learn to swim. They thought they could get into the ocean, but they couldn't. And if they had a curriculum that required them to learn how to swim, and to become Sea Horses, it would be different.

Pradnya: It would be different.

Matthew Rascoff: And, like, that aspect of, you know, the role of a school in the life of a child, in a family, in society, that was quite different, and very interesting to see. And it was part of my own broader perspective on Germany, but I could not have achieved just from, you know, coming to lunches and reading policy, but talking to smart people. No. In Germany, don't even know that this is something that's special about their system. They don't even know to talk about it as something that's unique. You really have to come as an outsider to experience this aspect of the schools and to understand its value.

Pradnya: And we always close by asking our guests what inspires them, apart from the work and the passion for your, for the topics that you are very passionate about. There are some pieces of inspiration all around us. So sometimes it can be a book, it can be a podcast, sometimes it can be a TV show that makes you think that, okay, it's maybe worth fighting for, so anything that comes to your mind.

Matthew Rascoff: I mean, my experience there was so positive, it was so welcoming. I think it was really the conversations that I had. It wasn't one particular, you know, thing that I read, but just the spirit of the academy, of the foundation, of the Richard von Weizsäcker program, it was about learning, it was about exchange, it was about openness, like the world feels like it's getting narrower and smaller and more small-minded and this group stands out for representing liberal values, for representing universalism, for representing the possibilities of exchange and the values of openness that are under so much pressure around the world and like that just that in itself is a statement.

Pradnya: It was a breather enough for you.

Matthew Rascoff: It was, and I was so grateful to participate in it. It was a daily source of inspiration for me, the staff there, the other fellows there, and the values of the foundation, which I think are so important for me, they are truly the educational values, the idea that we can learn from one another, they were not fundamentally in competition with one another, you know, that there's-

Pradnya: That was the show notes as well. The values that the foundation stands for.

Matthew Rascoff: And I embrace them, and I hope others will join in that too.

Matthew Rascoff: I'm so grateful for the opportunity to have done this fellowship for the leadership, for the staff, for you, and your colleagues who are just so welcoming and so warm. Berlin to me, it's represented many different things and different historical periods. But today is Berlin. What I experienced of it represents openness, represents cosmopolitanism. It represents people coming from all over the world. And that is an amazing turnaround, you know, given the history of the city, given the many different layers, you know, and anguish.

Pradnya: And maybe also it stood out to you from the context that you were coming from the US. I think it was a good breathing space for you to see that.

Matthew Rascoff: Absolutely, but I think the same is true. Well, here from my colleagues, the other fellows, I think the same is true for those who are coming from other parts of the world. And to me, I wrote an essay a few years ago about the spirit of freedom in Berlin and what it represents. And that to me is alive and well in the Robert Bosch Academy and in the fellowship program there.

Pradnya: Well, I couldn't have asked for a better end for this conversation, Matthew. So thank you for joining us today during your thoughts on a vast range of topics. And thank you for the listeners for tuning in. As always, you'll find the details about the references that came up during the conversation in the show notes. And yeah, we hope to see you next month. Subscribe to our channel on all platforms and tell your friends about us. And if you have any comments, questions, please feel free to drop us a line at contact at Robert Bosch Academy.de.

Matthew Rascoff: Thank you so much. Thank you. Think, debate, inspire. A podcast of the Robert Bosch Academy presenting inspiring ideas to address major challenges of our time. Subscribe to our podcast on all platforms.