Alison Gopnik is both a psychologist and philosopher at Berkeley, studying how children construct theories of the world from limited data. Her central insight is that babies learn like scientists, running experiments and updating beliefs based on evidence. But Tyler wonders: are scientists actually good learners? It’s a question that leads them into a wide-ranging conversation about what we’ve been systematically underestimating in young minds, what’s wrong with simple nature-versus-nurture frameworks, and whether AI represents genuine intelligence or just a very sophisticated library.
Tyler and Alison cover how children systematically experiment on the world and what study she’d run with $100 million, why babies are more conscious than adults and what consciousness even means, episodic memory and aphantasia, whether Freud got anything right about childhood and what’s held up best from Piaget, how we should teach young children versus school-age kids, how AI should change K-12 education and Gopnik’s case that it’s a cultural technology rather than intelligence, whether the enterprise of twin studies makes sense and why she sees nature versus nurture as the wrong framework entirely, autism and ADHD as diagnostic categories, whether the success of her siblings belies her skepticism about genetic inheritance, her new project on the economics and philosophy of caregiving, and more.
Watch the full conversation
Recorded October 30th, 2025.
Help us keep the conversations going in 2026. Donate to Conversations with Tyler today.
TYLER COWEN: Hello, everyone, and welcome back to Conversations with Tyler. Today, I am talking with Alison Gopnik. She is a professor of psychology and also philosophy at University of California at Berkeley. She is a very well-known expert in fields of human learning and also child developmental psychology, among other things. She has written extensively for the New York Times, the Wall Street Journal, where she was a columnist for 10 years, Atlantic magazine, and dozens of other places, all of which you will have heard of. Alison, welcome.
ALISON GOPNIK: Thank you.
COWEN: Now, one hypothesis you’re known for is this idea that the way children learn has a lot in common with the way that human scientists learn. What’s your basic model of how human scientists learn?
GOPNIK: This is interesting. When we started this out, one of the big puzzles that we had was, it’s fine to say that children are learning like scientists, but then the question became, how are scientists learning? When I started this project, a lot of the philosophers of science said, “Kuhn showed that there was nothing systematic you could say about that. It’s just sociology.” But interestingly, during the time that I’ve been working, there’s been this real change in the way that people think about philosophy of science.
We have some good computational models of how scientific theory change works. It turns out that those apply to children as well. The specific thing that I’ve looked at is, what is it that scientists do? Here’s this big, hard problem. All we seem to get from the world are a bunch of photons at the back of our retina and disturbances of air in our ears, and yet, children know about people and things, and scientists know about quarks and quantum phenomena. How do we ever get from the data to the theory?
One subcategory of that is, how do we ever get causal structure which is so important in science? How do we ever figure out what causes what just from a bunch of data that we have?
What’s happened is that philosophers of science and computer scientists have found some systematic ways that you could talk about that. Scientists — I think, mostly, not necessarily consciously, but just as part of what they do — and little kids are looking at data and systematically figuring out what kind of structure out there in the world could have caused this pattern of data. That’s not the only thing, of course, that’s going on in science. There’re lots of other things, too, but it’s at least one central thing going on in science that we’ve started to really understand.
Of course, it makes sense that scientists have the same brains that we had in the Pleistocene. Something in those brains must be enabling them to do what they do, in addition to all sorts of other institutional and social things. It’s something about this deep capacity to figure out the structure of the world — the world model, as the AI people say — from data.
COWEN: Are human scientists ever Bayesian? I think of them as not very Bayesian at all, that they’re mostly pretty stubborn.
GOPNIK: It’s interesting. One of the things that we’ve looked at is whether little kids are Bayesian, and you might be even more surprised to find that little kids are actually being pretty rational Bayesian, but a lot of it depends on how you ask.
If you asked a three-year-old, “Do you think that this pattern of conditional dependencies is giving you a confounding causal structure?” They would probably not give you a very sensible answer. Even when you ask scientists that, they don’t give you a very sensible answer. But when you look at their actual practice, what you see is that, in fact, kids, for example, are Bayesian, and so are scientists.
Now, the thing is that, in fact, in many respects, kids are better Bayesians than scientists, but a lot of it depends on your prior. If you have, as they say, a very peaked prior, you have a lot of experience, you have a lot of reason to believe that this prior assumption is right, then it’s rational not to change it when you just have a little bit of evidence. You should require a lot of evidence to overturn something that you have a lot of confirmation for.
It’s interesting that the kids, actually, are better at solving problems that involve unusual outcomes than the scientists are. I think what happens in science — we’ve just been doing some work about this — is that there’s also a social factor, where having a big distribution of people who are more likely to go with the prior versus people who are more likely to go with the evidence, which seems to be true in science, that collectively can get you to the right answer. There’s no arbitrary principle you can have about when should you abandon the theory and when should you hold onto it.
COWEN: I think of the kids as much more Bayesian than the scientists, and here’s what worries me about the scientists. When they revise their views, the direction in which they move is almost always predictable. If they’re, say, thinking over decades, how much does the money supply matter? They’ll move a bit in the direction of thinking it matters. Then they’ll move a bit more. Then finally, they might decide, well, it matters quite a bit. It shouldn’t be predictable if they’re Bayesian. It should be more like a random walk. Basically, they’re stubborn.
Then what I observe, if they’re proven to be wrong, they don’t usually admit it, whereas a child might. They just start working on something else. That, too, is a funny way of dealing with being wrong, right?
GOPNIK: There’s a beautiful idea that I’ve been using thinking about the kids that is in computer science. It actually comes from physics called simulated annealing. The idea behind simulated annealing is that you have some problem to solve. You have some space of solutions that you’re trying to get to.
One thing you can do, which is like what you’re describing about the money supply, is just make little changes to what you already know. That’s what you mean about moving in the predictable direction. You’re just changing things a little bit. Then seeing, “Okay, if I change it a little bit, is it doing a better job of accounting for the data?” That’s what people think of as a low-temperature search. The other kind of search you can do, the high-temperature search, is just bounce around the space. Try wild, crazy things. Exactly as you were saying, have just a more random walk.
The strategy that you see in computer science, this annealing, is start out with this wild, crazy, out-of-the-box, high-temperature search through the space, and then cool off and just fill in the details. If you think about your four-year-old, who do they sound like? Do they sound like the creature that’s just moving a little bit, or do they sound like they’re noisy and bouncy and random and doing all sorts of weird things? The four-year-old seemed to be a really good idea of this kind of random search.
But if you’re a scientist, of course, you have to balance those things. You can’t just think of crazy new ideas. You have to figure out how you’re going to test them. You have to get grant proposals. In science, you’re always going back and forth between, “Do I do the high-temperature, wild, crazy, out-of-the-box search? Do I think of ideas that don’t look like they would be very likely to begin with? Or do I fill in the details?”
I think you see both things happening. When you get big paradigm shifts, as Kuhn said, when you get big changes in science, a lot of times it’s because someone found an idea that looked like it was improbable. The nice thing about kids is, because they don’t have to worry about grant proposals, they can be off in the wild space all the time.
COWEN: Here’s a few things a four-year-old might do. I’m going to ask you which one is the most Bayesian. A four-year-old pulls his sister’s hair, a four-year-old tries to figure out how to use a fork properly, and a four-year-old tries to put together the pieces of a puzzle. Or pick your own nomination. In which do we see the child being the most Bayesian and then the least Bayesian?
GOPNIK: That’s an interesting case. It’s funny, when you said tries to use the fork properly, that’s a matter of conforming to a norm, but if you just —
COWEN: To get food into his mouth, right? At least to do that. I don’t mean proper manners in the Ann Landers sense.
GOPNIK: Yes. I don’t think any of those are really an example of the kind of exploration that I think is trying to get new evidence to change your theory. I think there are lots of things that you’ll see the kids doing. For instance, the fork is a good example. I have a lovely video of my son that I use to do teaching. He has a spoon, and he’s trying to eat an avocado with a spoon. Of course, eating an avocado with a spoon when you’re two years old is extremely challenging. It’s a full avocado.
Instead of actually trying to get it in his mouth, what he does is try all these different things that you could do with a spoon and an avocado. He bangs it on the side. He picks it up and turns it over. A lot of them don’t have the output of actually getting any of the avocado at all. That’s something that’s completely characteristic of what you see with little kids. You see them doing these kinds of experiments all the time.
With scientists, we underestimate how much that — we sometimes dismissively call it a fishing expedition — how much that very general experimentation is playing a role in scientific progress. In the grant, you’re supposed to say, “Here’s my three hypotheses, and here are the four experiments I’m going to do to test them.” But I think in practice, a lot of times, scientists are being like the little boy with the avocado and the spoon. They’re saying, “I don’t know, what will happen if I try this? What will happen if I try that?” Then they write the grant to get money to do the things that they’ve already done by doing all these experiments.
COWEN: There’s an alternative vision of how people learn, if learning’s even the right word, and it comes from Karl Friston. I think of it as a minimized surprise kind of theory, that you’re confronted with data from the world, and then you interpret the data and coordinate your actions in such a way to minimize surprise. That will imply a lot of cognitive biases, but that’s his basic model of humans. What do you think of that theory?
GOPNIK: I think it’s very underspecified. There’s this category of theories in science of things that have a good combination of intuitively being in the right direction and then have a lot of math, but it’s actually very hard to connect them to experiments and to an empirical research program. I think there’s something right about the fact that children and people in general, when they’re learning, are looking for violations of their predictions and something like surprise. But I’m a little skeptical about whether that formalism isn’t going to end up accounting for everything. I think it’s one of those kinds of theories.
It’s certainly very closely related to ideas that come out of the theory theory framework, which include things like, what you should be doing is making predictions, and you should be exploring things that are surprising and don’t fit the predictions that you already have. It’s interesting because everyone in science knows that experimentation is really crucial for science, and yet, we don’t have very good accounts of how to do experiments. Why and when would you see something surprising and decide that you should follow it up, or you should go out into the world and . . . ?
This is something that I think is a bit different from what Friston says and what I would say. I think a lot of the time, and I think this increasingly — it’s not just that you’re seeing something surprising and then changing your view. It’s that you’re seeing something surprising. You’re doing an intervention, something in the world, to try to explain or reproduce or find out the parameters of that surprising thing. Then that’s really the thing that’s driving your theory change.
COWEN: Let’s say you have $100 million unconstrained to run experiments on young humans and how they learn. Now, there’s an ethics constraint. You’re not going to do anything you don’t think is right, but what would you actually study with those resources?
GOPNIK: One of the good things about being a developmental psychologist is that it’s really cheap, like little chairs and a little table. The thing we use for causal inference is about a $20 toy. There are some really interesting attempts now to use big data with kids, and I think that might be something that you could use a lot of money for.
Someone like Mike Frank at Stanford is putting GoPros on babies, little GoPros that the babies don’t mind. Then you end up with this giant amount of data about what is it that the babies are actually experiencing. Of course, you’d want to do that for a very large group of babies because they’re all different. Then it would take some real computational power to analyze all of that data. That might be a direction that you could go in that you could use the extra money for.
COWEN: What do you want to test? Let’s say you have that set up. You hire the assistants. You have the quant people. What hypothesis do you want to look at?
GOPNIK: Again, what I’d really like to know is, how is it that the kids experiment? If you just took that GoPro data and you looked at what did the child do? What happened just before the child did that? What was the consequence of what the child did in terms of their data, their experience? I think that would be really interesting to analyze.
My hypothesis is that you’d find that that was actually much more systematic than it looks on the surface. On the surface, it just looks like baby, kid crawling around doing a bunch of things, and a bunch of things are happening, but I think you’d find that there was much more systematic relationships between what happens, what the baby does next, what the outcome is, than you might think on the surface. A lot of those are going to be in the service of how could I figure out what’s going on in the world?
COWEN: What does it mean to say babies are more conscious than we are?
GOPNIK: I should say, first of all, my view about consciousness is that it’s very unlikely to be a single thing. There’s a famous analogy that I think is right about consciousness and life. Life is an example of something that we thought, back in the 19th century, was going to turn out to be a simple thing, and there were whole research programs about what is it that makes a living thing living? It turned out that was just the wrong question, that actually there’re many, many different kinds of processes that are all involved in different things that we think of as life. I think that’s likely to be what will happen with consciousness.
But it is interesting — and maybe not surprising — that the thing that professors first thought of as consciousness is that experience you have when you’re sitting at your desk, and you’re introspecting, and you’re trying to solve a problem, and so on and so forth. But I think that’s really different from awareness, from sentience, from your experience of the world around you. In fact, in some ways — I think professors all recognize this — it’s intention with those things. While you’re sitting at your desk trying to solve this problem, you are notoriously not paying attention to all the other things going on around you.
I think if you look at the science, what you see with babies is that they are conscious of all the things that are going on around them. That the way that even their brains work is to take in lots and lots of information. Their brains are very much what neuroscientists would call plastic. They’re surrounded by novelty. They’re not just focusing on one thing at a time, the way we do when we’re grown-ups.
I think if you think about the context as grown-ups that we are in that state, like if we travel to a new place or we’re trying to do something new . . . It’s not like we’re unconscious when we go to Paris for the first time. On the contrary, it feels like we’re full of experience. We’re vividly experiencing the world around us. I think that’s what it’s like with babies.
COWEN: People who have very weak episodic memories — as you know, that’s a thing — are those people less conscious?
GOPNIK: No. I think arguably, if you think about the children, for example — they don’t have as much episodic memory, especially the little ones, like the two-, three-year-olds. I think that may actually be making you more conscious in the sense that, again, this adult tendency is to try to . . . You could think of it as trying to reduce, to compress the information around you into a particular narrative, like the kind of narrative you have in an episodic memory. A lot of times, as adults, what we’re trying to do is take all this information, all this data, and just compress it into a single narrative, which is a very useful thing to do.
I think babies, before they develop a lot of episodic . . . They have some episodic memory even from very early on, but it’s around three or four that they develop the kind of autobiographical memory that grown-ups have. I think that actually makes them more conscious in the sense that they’re more focused on the present, and they’re experiencing the present more.
COWEN: If someone has aphantasia, they don’t retain images very well, are they also more conscious for that reason?
GOPNIK: Aphantasia is a whole complicated story. Aphantasia is a puzzle about why it is that we have the kind of imagery that we have. It’s not that people who have aphantasia aren’t vividly experiencing a visual world; it’s that they don’t use that experience when they’re trying to, say, imagine something or when they’re trying to generate an image. I think that what the aphantasia work shows you is that business of generating an image when you’re trying to solve a problem is epiphenomenal, that it’s not that you’re solving the problem by looking in your head and seeing the picture.
People with aphantasia, like my friend Ed Catmull, who’s the co-founder of Pixar with my husband, has aphantasia. He’s an animator who doesn’t see pictures inside of his head. I think it just shows that there’s a big gap between what we think of when we think about, say, an animator having a picture inside of his head and what’s actually going on cognitively. I don’t think it would be relevant to the question about consciousness as we usually experience it.
COWEN: If someone has aphantasia, is it somehow that the top-down processing is broken, and thus the image is not retained because some part of the digesting of the image doesn’t happen?
GOPNIK: No, I don’t think that’s what’s going on. I think what’s going on is that all of us, in our ordinary experience, when we’re doing something like making an image, are doing this cognitive work of generating a visual representation. The interesting, and I think contingent thing, is that some of us also have this feedback to our actual visual system, to our actual visual cortex. I think there’s some neuroscience evidence that supports that.
What happens is, almost as a side effect, when we’re generating these ideas or generating these pictures, we actually activate the parts of our brain that are usually activated when we’re really, genuinely seeing something. It’s interesting that it doesn’t seem to have consequences for all these other abilities, like being able to generate a picture or being able to do a lot of cognitive capacities, like being an animator.
COWEN: Aphantasia — it is pretty strongly correlated with autism, right? So, it ought to have some micro foundations in common with autism.
GOPNIK: I don’t think so. I don’t think that’s right. It may be that people with autism . . . No, I don’t think that’s right.
COWEN: You don’t think it’s correlated?
GOPNIK: I don’t think it’s correlated.
COWEN: You mentioned Pixar. From graphics and animation, what is it we learn about how babies process the world? Babies love animation; young kids love animation. What do we learn?
GOPNIK: It’s interesting. I think we all thought, to begin with, that when we were talking to, say, the animators at Pixar, there’d be great insights into things like facial expressions. One thing we do know with babies is that from the time babies are born, they’re incredibly tuned into facial expressions and the way that facial expressions indicate emotions, for instance.
Of course, any animator — that’s their stock in trade, being able to take a bunch of polygons, have them have a face, and have it express emotion. An example I like is in Ratatouille. There’s this wonderful moment where little Remy, the rat, looks very embarrassed and proud at the same time. You can just tell that from his facial expression.
We thought that if we talked to those animators, we could get an idea of how is it that babies and children and grown-ups are so good at doing that. We’re incredibly good at inferring emotion from just tiny little indicators in someone’s face, and we know that babies, in the first few months of life, can do that.
It turns out that the way the animators do it is completely intuitive. They have no explicit idea of what it is they’re doing. They’re like actors. In fact, they have a lot of the same characteristics as actors. An actor can do the same thing. They can give you a piece of information about the way that the world works. They can make their faces have a particular emotion without really understanding it. So, that didn’t seem to help very much.
On the other hand, an idea that a lot of people have had now and are formalizing is the idea of the visual system, vision as what people sometimes call inverse graphics. The idea is that the same way that in computer vision, you can take a bunch of data and get a picture of the world, that maybe something like that is happening for humans. That what we have — some people sometimes describe it as being a game engine in the head. The same thing that lets you generate pictures in a video game is the thing that’s actually letting you figure out how the visual world works for your visual system. That has actually been really productive.
COWEN: Is there anything in Freud’s understanding of childhood that’s really held up?
GOPNIK: That’s a good question and a complicated question. When I first started doing the work that we do now, Freud was — and Freud still is — I think rather surprisingly to the intellectual world in general — it just doesn’t show up in modern psychology. Even having someone teach a class about Freud would be unusual. I don’t think anyone does in Berkeley’s department, which is the best department in the country.
On the other hand, some of the ideas . . . some of the Freudians have been very enthusiastic about the work that people like me and my colleagues have done because, I think, the intuition that there was more going on in even little infants, that even small babies and young children could make inferences about the social world around them or the psychological world around them, and that was influencing how they grew up — I think that idea has turned out to be right, and it wasn’t obvious that that was going to turn out to be right.
COWEN: What in Piaget has held up the best?
GOPNIK: Piaget, on the other hand, still is, I think, the big theoretical foundation of what everyone has done since in cognitive development, and all of us — our attitude about Freud is, “Yes, well, of course, here’s this little bit of something that turns out to actually be right.” Whereas with Piaget, we’re all trying to claim his legacy, that what we’re doing is what Piaget was trying to do.
Here’s what I think has held up. Well, two things that have held up. One thing is, I started out talking about this problem of how is it that we could ever know as much as we do about the world around us, given that we have such a small amount of data.
Well, going back to Plato and Aristotle, Plato and Aristotle talk about that problem, and the two solutions that they come to are . . . one is, okay, that structure couldn’t have been learned from the data. It must have just been there innately in a past world for Plato, through evolution for someone like Chomsky or Steve Pinker or some people like that.
One solution has been, okay, it’s all just innate. The other solution has been, it looks as if there’s all this abstract structure, but really, it’s just statistical combinations of the data. If you think about the current AI deep learning approach, that’s like that. It looks as if there’s all this knowledge and intelligence. Really, if you just put enough data together, you’ll get the same results.
What Piaget thought, and I think what most developmentalists thought was, neither of those is a good account of what’s happening with kids. If we look at kids from the time they’re very, very little, literally from the time they’re babies, from the time they’re born, we see a lot of abstract structure. We see a lot of coherence, a lot of inference, a lot of generalizations. We also see that change as a result of the kid’s experience. Neither of those options, either the nativist or empiricist one, is a good account of what’s happening with kids.
Piaget just had this word constructivism about what was going on. I think what a lot of us since then — like the Bayesian approach. The Bayesian approach quite explicitly people have called rational constructivism. We want to try and take that idea that you’re really building a world model from data in a rational way that Piaget had, and give it some modern juice.
The other thing that’s interesting about Piaget — and there’s a bit of a narrative about this — is his observations. It turns out they probably weren’t his observations, mostly. If you look at his books, the observations and the theory have quite a different tone. His wife, Valentine, actually did a lot of the actual observations of the babies.
Those observations have held up remarkably well. You can take your average nine-month-old and do one of Piaget’s experiments with them, and you’ll get the result that Piaget had, even though his interpretation of that has changed a lot over time. He didn’t think that, and I think the way that change has gone is, there’s much more going on in the baby’s minds. It’s much more abstract. There’s much more of this representation than Piaget thought. It’s much more theory-like.
COWEN: Some of my colleagues — they’re big fans of using twin studies to try to separate out nature from nurture. Do you find that enterprise convincing?
GOPNIK: Well, any science that people do is interesting. I think if you look at someone like Eric Turkheimer’s work, one of the things that comes out of that is that your intuition to begin with, “Okay, well, we’re going to look at the twins, and if there’s a high correlation between what one twin does and the other twin, it’s going to be nature. If there isn’t, it’s going to be nurture.” I think increasingly, that kind of model has turned out to just be wrong. That’s the wrong model. It’s just too simple.
COWEN: What’s wrong with it?
GOPNIK: Let me give you the example that I give when I’m doing my developmental psychology class. We know that there’s a particular kind of disorder in a particular gene that means that you can’t metabolize a particular substance in your food, and the result is that you have mental retardation and a lot of difficulties. We actually know pretty much what the gene is. We test babies. They prick their heel when they’re born to see if they have this disorder. If they do, then we make sure that they don’t eat anything that has that enzyme or that piece in it, and then they’re fine.
The question is, does this come from nature or does it come from nurture? How much is it nature? How much is it nurture? In one sense, it’s 100 percent nature. It depends on this particular gene. In another sense, it’s 100 percent nurture. If you get rid of the material in the environment, then you won’t get the syndrome.
That’s a very specific case, but the same thing’s true in Eric’s work with something like SES [socioeconomic status]. What Eric found was that you had much more convergence in twin studies in upper-class contexts than in poorer families. The explanation is that if you’re in a poor family, small differences in your environment can make a big difference in how the rest of your life goes, and that’s less likely to be true in a richer family.
One that I particularly have thought about, and I think is really interesting, is thinking about the effect of nurture on development, so the effect of having caregivers. As you probably know, there’s a sort of people who’ve said, “No, we are overestimating how much what parents do influences kids’ development.”
But the argument that I and others have made is that the effect — and we’re doing this more and more in the context of this big caregiving program — is that the effect of having a protective caregiver is that it allows more variability. When you don’t have to worry at a particular moment about what your particular situation is, you can do many more different things, and you can have more variability. That’s true in biology and ecology, for example.
The example I give in my book, The Gardener and the Carpenter, is if you have a garden that is protected and has a lot of possibilities, you’re going to get a much wider variety of plants than if you have a much more restricted circumstance in which only a few plants will thrive. Okay, what does this have to do with kids?
Well, when people say that there isn’t an effect of the family environment on development, what they mean is that they don’t see high correlations between kids in the same family and some measure of what an adult is like. You get siblings — sometimes it’s called the non-shared environment. The things that don’t seem to be the same things are playing a larger role than the shared environment. So, what you might think is, okay, if you have two kids and they have the same parents, those kids are going to be more similar to one another.
But if this picture about what caregiving does is allow variability is true, then you might expect the opposite. You might think if you have a really caring family, what that means is that the siblings are going to have more opportunity to develop in really different ways, so you won’t see a correlation. Again, part of the problem is, if the effect of nurture is not on the mean but on the variation, on the standard deviation, you’re not going to see it in any straightforward way in a twin study.
COWEN: If the Turkheimer hypothesis is correct and societies are growing wealthier and wealthier, over time, virtually everything will be genetically determined rather than environmentally because we’ll all be higher SES, we’ll all have better environments, less likely to be crippled by polio in our youth or whatever, and it will just be about genes.
GOPNIK: Let me give you another example, Tyler, of the Turkheimer phenomenon that I think is interesting. If you look at smoking, for example, it turns out that when smoking was very available, when you did a twin study, there was a relatively small genetic contribution to whether you smoked or not. As smoking became less and less likely, and as there were more and more restrictions to keep you from smoking, then you start seeing what looks like more of a genetic effect, and it’s because the only people who are smoking are people who have a very strong tendency to smoke in the first place.
Exactly what happens, how the caregiving, how the nurture is going to affect and interact with the genetics is going to be really complicated and unpredictable. So, you could argue just the opposite. You could say, as we develop a more and more safe environment, that actually what’s going to happen is that the potential for variability is going to get to be greater and greater. The possibility of doing something new, doing something different from what you did before — those are the kinds of things that are going to become more important.
I think the nature-nurture is one of those examples of something that intuitively seems like a good way of characterizing things, but is, again — like the example of life — just isn’t the right framework, that what we have to do is say, “Okay, here’s a particular trait, here’s a particular capacity. Let’s track all these complicated ways that the environment and genetics are going to interact to bring a particular kind of outcome.”
COWEN: If it’s something like height, where there is clearly an environmental component, especially if the child is not well-fed, but it seems perfectly fine to say above a certain dietary level, it’s mostly genetic, right? No one says that’s ambiguous, and more and more traits will become like that.
GOPNIK: Well, first of all, I’m not sure that’s true. To a striking degree, the traits that people have looked at, like educational attainment, for example — we haven’t found consistent relationships to genetics. I think the reason for that is exactly because there’s this very complicated developmental process that goes from the genetics to the outcome.
Even if you think about fruit flies, for example. I have some geneticist colleagues who work on this — fruit fly sex determination. You’d think, “Well, that has to be just the result of genes.” It turns out that there’s this long developmental — long by fruit fly standards — developmental process that goes from the genetics to the proteins to the morphology, and there’s lots of possibility of variation throughout that. I think that hasn’t turned out to be a scientifically helpful way of understanding what’s going on in development.
The other thing, of course, is, from my perspective, the common features of, say, what kids are doing are much more interesting than the variations. What I really want to know is how is it that anyone could have a brain that enables them to accomplish these amazing capacities? Thinking about, is this child smarter than the other one, given how unbelievably smart all of them are to begin with, I just think it’s not an interesting question.
COWEN: But say, what you would call the lay belief that smarter parents give birth to smarter children, at least above subsistence — surely you would accept that, right?
GOPNIK: Again, what does smarter mean?
COWEN: How you would do on an IQ test.
GOPNIK: What does genetics mean? It’s interesting, Tyler, that IQ tests, for example — they have their own scholarly and scientific universe, but they’re not something that we would teach about or think about in a developmental psychology class, and there’s a good principled reason for that. The good principled reason — this has come up a lot in AI recently. There’s this idea in AI of artificial general intelligence, and that is assuming that there’s something called general intelligence.
Again, I think, a lot like consciousness or life, it’s one of these lay ideas about how people work. When you actually look at children, for example, what you see is not just that there isn’t a single thing that’s general intelligence. You actually see different cognitive capacities that are in tension with one another. You mentioned one about the scientist who’s trying to think of some new idea versus the scientist who’s looking at a more specific idea, right? A classic example of this tension that I’ve talked about and studied is in computer sciences: exploration versus exploitation.
What do you count as IQ? In fact, most of what IQ is about is how well do you do in school? How well do you do on school tests? That’s actually, in many respects, in tension with how good are you at exploring the world around you? The kinds of things that you need to do to have particular goals, to accomplish them, the kinds of things that we emphasize a lot, say, in a school context, are actually in tension. This gets back to the point about babies being more conscious than we are — are actually in tension with the kinds of things that will let you explore.
Think about the Bayesian example. If you have a flatter prior, and you pay more attention to evidence, you are probably not going to do as well on an IQ test.
COWEN: Say, in five-factor theory, openness is positively correlated with IQ, right?
GOPNIK: Again, even if you’re thinking about what’s happening in the personality literature, that’s another example of reifying things that are going on. Again, I think an advantage of doing stuff with babies and young children is that we can actually look at what are the cognitive capacities specifically? How are kids learning? How are their computations affecting what they do? That’s just very orthogonal from the whole discussion about IQ.
COWEN: Let’s say you’re called into a typical American K–12 school. Not a top one, but say 70th, 75th percentile, a pretty good school. They just say to you — maybe this has happened — “Well, you’ve studied all this about children. How can we improve how we teach our kids?” What is it that you tell them?
GOPNIK: I’ve written about this in my book. I think there’re two different things to say. If you’re thinking about young kids in particular, like early childhood, before 7, say, I think we have a very good model of inquiry-based, often play-based education, where you have a warm caregiver, and you get lots of opportunities to play and explore. I think we have pretty good reason to think that’s the right thing for these very young children.
But for what we think of as the school-age children, it’s a really different kind of model. It’s a different kind of thing they’re trying to do. If you think about development — as I’ve argued — as being this shift from mostly about exploration to skills that you’re going to use for exploitation, how do you actually develop those skills?
I think we have good reason to believe that some kind of intuitive apprenticeship model is how you develop those skills. You do something that you think is going to be important. You have a teacher who gives you feedback. The teacher also shows you examples of the skill. That kind of interaction. Sometimes, the teacher could be quite mean [laughs] about telling you when you’ve done something wrong. I think that’s a really good way for school-age children to learn.
I think we have good reason to believe that some kind of intuitive apprenticeship model is how you develop those skills. You do something that you think is going to be important. You have a teacher who gives you feedback. The teacher also shows you examples of the skill. Sometimes, the teacher could be quite mean about telling you when you’ve done something wrong. I think that’s a really good way for school-age children to learn.
I think it’s not a coincidence, for instance, that so many kids really want to do music and sports, even though we all say, “No, learn how to code. That’s the thing that will actually be helpful to you.” Because music and sports are among the few examples where we actually do this kind of apprenticeship. You do the thing, you get feedback, you try and do the thing again.
One of the things I say is, imagine if we tried to teach baseball the way that we teach science. How do we teach science? What we would do is, we would tell everybody about great baseball games when they were little. Maybe when they were in high school, they could throw the ball a lot to second base. When they were in college, they could reproduce great baseball plays, but they wouldn’t actually get to play the game until they were in graduate school. If you taught baseball that way, you wouldn’t think that people would be as good at baseball.
It’s funny because, of course, what kids end up — I don’t know if you’ve had this experience, but I think a lot of faculty have had this experience — because kids are so naturally tuned to this model, what do they get to be good at at school? They get to be incredibly good at going to school. They get to be really good at taking tests. The ones that we get in our first-year college classes are just these masters of all the things you need to do for school.
Then we faculty are worried because we say to them, “Think up a new experiment.” They say, “Oh, I don’t know how to do that. Nobody ever taught me to do something on my own or do something creative.” There’s a wonderful idea. Do you know about Goodhart’s law?
COWEN: Sure. We use it in economics. Yes.
GOPNIK: Yes, that’s right. It’s funny because I learned about it from my son-in-law, who’s actually the chief quant for the Seattle Mariners. He was telling me about his difficulties in that job.
Of course, Goodhart’s law — what happens in school, in our current school system, is a really good illustration of Goodhart’s law. Goodhart’s law is the idea that when you try and optimize something that you think is a signal for something else, very often, what happens is that the people that you’re trying to choose just try and maximize the first signal. It ceases to be correlated with the thing that you’re really trying to measure.
I think the current way that we do schooling is a good example of Goodhart’s law. We teach kids — because kids are so good at wanting to be skilled — we teach them how to be good at school, which we think is going to be correlated with the ability to do a wide range of things as an adult. Then it ends up being a separate kind of skill. Sorry, that’s a long answer.
COWEN: Now that we have generative AI, how should that change what the K–12 schools do in terms of which qualities or features they want to bring out in the kids?
GOPNIK: My view about generative AI — and I’ve actually written about this in a paper in Science with Henry Farrell, who I think you know is —
COWEN: Yes, I know Henry.
GOPNIK: — a political scientist, and James Evans, who’s a sociologist. Again, our intuitive lay conception of how AI works is really misguided. We very much have this golem view about here’s this non-living thing that we’ve given a mind to, and that always works out badly. It’s going to either be for good or for ill. It will be superintelligence. That’s the narrative. We think the right narrative is to think of it as what I’ve called a cultural technology. It’s a way of getting information from other people.
The way that generative AI works is that it’s trained on all the stuff that very intelligent humans have done. It’s not surprising that a lot of times it will simulate what intelligent humans would do. I think it’s analogous to things like print or writing, or internet search itself, libraries, where one of the things that is characteristic of humans and has always been — and as some people have argued, I think rightly, is our superpower — is that we can get information from other people, and we can use that information to make progress ourselves.
Generative AI is the latest technique for doing that. What generative AI tells you is, here’s a summary of what all the people on the net have said in this context, and learning how to use those cultural technologies. Again, if you want a new, genuinely intelligent agent in the world — if you have a kitten, that will be a genuinely intelligent agent — probably won’t change the world too much. You change a cultural technology, you introduce print — that really does change the world in radical ways for good or for ill.
COWEN: That seems wrong to me. In your piece with Henry, you don’t consider reasoning models. Reasoning models, in some way, “think.” They can now prove some mathematical theorems. Almost every day, there’s some new, albeit often minor, scientific discovery that comes from AIs that was not previously on the internet. Isn’t the actual model of AI now — 2025 — quite different?
GOPNIK: I don’t think so. If you look at the way that the reasoning models work, they work the same way that all the other models work, which is that they look at patterns of text on the web. One of the things that is a pattern that you have — again, this is positive — one of the things that you have are patterns of reasoning. What you have are patterns of, here’s someone who was trying to solve a math problem. Here’s the steps that they took to solve that math problem. Can I find a general statistical pattern in those steps and reproduce it in this other context?
COWEN: They do much more than that. It’s clear they look at data on the web, but the people I speak to who build them say it’s not transparent, even to them, exactly how the thing works, but as they apply more scaling, it gets better and better at its own reasoning. There’s 01, there’s 03, now there’s GPT-5. GPT-6 is on the way. The scaling just seems to give it more ability to do actual reasoning of a unique sort that’s not just copying the reasoning of some human.
GOPNIK: Well, that’s the question. I think the fact that they’re so good in this mysterious way at picking up patterns and reproducing patterns — that’s clearly a really important thing that these technologies can do, but that’s not what humans are doing, and it’s not even what animal agents are doing. So, I would be impressed if they were actually designing experiments that would tell you something about something new that was going on in the world that all the other people around them didn’t know before.
COWEN: They do that in biology already, right?
GOPNIK: The way they’re reasoning is like a kid in a multiplication class who learns, here’s what the formula is that you need to do for multiplication. It’s interesting that they’re not good at things like basic arithmetic often. You get this weird combination of it’s reproducing a reasoning process that you could imagine in math, but it’s not doing basic arithmetic. I think another place where they’re really falling down is the novel relationships that humans are very good at generating.
Those are the kinds of things where the generative AI is not working. In any case, in terms of the issue about K–12, every time we have a new way of accessing information, it’s really important to teach. That’s why we teach kids how to read. We teach kids how to calculate. We teach kids how to use the internet. A lot of that teaching is going to be, here’s how this cultural technology works positively to give you some information that’s true, and here’s how it works to give you information that’s completely false.
It’s funny that these reasoning systems are still hallucinating. I think the reason why they’re hallucinating is because their objective function isn’t about truth in the way that it is for even a little baby. Their function is to produce something that is going to be the sort of thing that a human will like with something like reinforcement learning from human feedback.
COWEN: Hallucination rates are plummeting. If I use GPT-5 Pro edition, it will outperform a very good human lawyer for a typical query. It will do better on a medical exam than, say, a prospective doctor would do. I’m not saying there’s zero hallucination, but they’re already ahead of humans. If I took an economics test, GPT-5 would beat me.
GOPNIK: Right. That, again, gets you back to this point about what’s going on in schooling. What we’re doing is we’re working out . . . Here’s an analogy. Who knows more, you or the UC Berkeley library? Think about that example of the law case. All that legal information is there in the legal code. It’s just that you don’t really think about the legal code. You can access it, and you could access it better. You could have more of that legal information than any individual lawyer is going to have. You don’t really think about that legal code as being a lawyer.
The legal code is a place where you can combine and generate lots and lots of information that’s come from all the lawyers who’ve done work before, and a great lawyer. Again, the fact that you have a legal code that’s written down makes it much easier to do the law. The fact that you can quickly access what previous lawyers have done makes it easier to do the law. The fact that you can answer the medical test makes it easier to do medicine.
The human capacities are the ones that go beyond just extracting the information from other people. Again, having ways of extracting that information more effectively is great. That’s really going to be transformative.
COWEN: If I write out a unique economics problem, it will beat most human economists in trying to solve the problem, a problem that no one’s ever seen before. I create it. I write it down. I give it to the beast. I give it to some humans. Mostly, it beats the humans.
GOPNIK: Yes. Is it going to actually get a novel insight about economics that isn’t there before, as opposed to it’s just using the kind of apparatus that you already have?
COWEN: The demand curve will still slope downwards, but it gets the answer, and maybe the humans don’t. There’s something unique about that.
GOPNIK: It depends on the humans. The other thing to say is, we don’t know. We’ll actually see what happens. We’ll see what the outcomes are going to be. I’m pretty skeptical just because we have been trying to figure out how two-year-old babies go out and solve the kinds of problems that they solve in the world, and they’re solving problems.
Here’s a really good example. If you go to robotics, for instance, there’s a giant gap between what the LLMs, the large models, are doing, and what robotics is doing. In robotics, which again is the thing that the two-year-olds are doing, is going out, experimenting with the real world, getting data from the real world, doing something new in the real world. That’s the thing that the LLMs are really not good at doing.
Some people have argued — I like this idea — that they’re like Derrida’s revenge, that the postmodernist idea that all you had to do is have text, and as long as you had more and more text, you didn’t have to worry about whether the text actually was making contact with some external reality. ChatGPT is like Derrida in practice. What it’s doing is putting together lots and lots of text and generating more text, without ever quite making that contact with an external reality.
COWEN: There’re some papers in experimental economics. They indicate that autistic people — they’re more likely to be Bayesian, and they’re more likely to do better at game theoretic exercises than non-autistic humans. Should we infer from that that autistics have theory of mind?
GOPNIK: Well, this is another example of something where there’s a simple, intuitive account. People have autism or people don’t have autism. Again, I think the reality beyond just thinking about the spectrum is that there’s a whole incredibly wide range of ways of interacting with the world, some of which you think of as being in the normal variation, some of which cause trouble for the people who have them, to the point of kids who can’t speak, never learn how to use language, need to be taken care of all the time.
It’s not at all clear that there’s any single underlying fact about how that variation works. There’s some evidence, even in people who wouldn’t be diagnosed with autism, that there’s this tension between being very socially involved and engaged and being good at this abstract pattern induction, but I think it’s just too simple.
Again, to use another analogy, it’s as if, in the 19th century, you said, “Do people with dropsy have some characteristic or not?” It turns out dropsy is not actually a thing. Dropsy is a symptom. I think the same thing is going to turn out to be true and has turned out to be true with autism.
COWEN: Diagnosis of autism doesn’t pick up or reflect anything?
GOPNIK: It’s like dropsy. Diagnosis of dropsy in the 19th century was picking up something, which is that the person had swelling in their legs. I guess that’s what dropsy was about, but it didn’t actually track something that was going on in the external world about what was going on inside of that person. I think that’s the situation with autism as well.
COWEN: How do you understand ADHD?
GOPNIK: I think that’s another example where you have a lot of variation, and a lot of the variation is just variation in individuals. Some of it is variation in what it is that the culture thinks is important. Sometimes that variation is just variation. Sometimes that variation gets to be dysfunctional or cause difficulties.
I think one thing that’s interesting is, again, to go back to the consciousness point, we know and we just take for granted that little kids like two-year-olds — we say that they don’t pay attention, but what we really mean is that they don’t not pay attention. They’re paying attention to everything at once. That’s why two-year-olds are really distractible.
Then, as we get older, we get this more and more focused kind of attention. People vary in how much they end up within that state of focused attention. I think there’re lots of reasons to believe that an industrial schooled society really pushes people in the direction of having very focused attention. We really want people to have very focused attention. What that means is that the kind of variation in what people are like becomes dysfunctional in a way that it might not have been in the past.
I think that’s true in general, that we have to think about ways that there’re these interactions between the environment that you find yourself in and something like whatever the combination of genetics and development is that lets you have a particular phenotype, and you’re trying to figure out how those two things interact.
COWEN: Andy Warhol — overrated or underrated?
GOPNIK: As you probably know, Tyler, I have a brother who wrote the definitive biography of Andy Warhol. I don’t know if you did know.
COWEN: Of course, that’s why I ask.
GOPNIK: Exactly. I wouldn’t say that he’s underrated, but he’s definitely not overrated. He’s someone who made a much wider, broader difference to the way that we think about artistic practice than other artists have. He kept doing interesting things. He kept doing different things. He kept doing things that ended up having consequences for what was going on with other artists.
He was someone who was really shifting the way that people thought about art. I’ve talked enough with Blake and also with various members of my family who are interested in art to think that that’s really part of the thing that you want an artist to do.
COWEN: The movie Tár — overrated or underrated?
GOPNIK: You know, I haven’t seen —
COWEN: Your brother’s in the movie.
GOPNIK: I know my brother’s in the movie.
COWEN: He’s great in it. He’s tremendous in it.
GOPNIK: That’s what everybody says. Interesting question about why I didn’t actually feel moved to see it, but I haven’t, so I don’t have a good opinion about it.
COWEN: There’s you — you’re tenured at Berkeley, you’re famous. There’s Blake, The Definitive Warhol Biography, and Adam, who’s amazing, writes for the New Yorker, and you don’t believe in heritability and IQ being very concrete things? I just don’t get it. I think you’re in denial.
GOPNIK: Actually, I think that example is maybe partly why I don’t believe in that. In fact, what I do believe is that the effect of caregiving is to increase variability, is to increase variation. Our family, our care — there were six of us in 11 years. My parents were graduate students, and even before they were graduate students, they were that great generation of immigrant kids.
We had this combination of a great deal of warmth, a great deal of love, an enormous amount of stuff that was around us — books and ideas. We got taken to the Guggenheim, when Adam was three and I was four, for the opening of the Guggenheim. We both remember this vividly. But we were also completely free. We were just in regular public schools. As was true in those days, in general, we came home after school, and we basically did whatever it was that we wanted. I was involved. The kids were taking care of each other a lot of the time.
The result is that you get a lot of variation. It’s an interesting example in our family where we have six kids who presumably all have somewhat similar genetics, all in that 11 years grow up in the same context, and they come out completely differently. They come out with really different strengths, really different weaknesses, things that they’re good at, things that they’re not good at. Even if you think about what Blake and Adam and I are like as thinkers, we’re all foxes instead of hedgehogs. We’re all people who have done lots of different things and thought about lots of different things.
So, my view is that what nurture will do is let you have variability. That’s the thing that, in a sense, is heritable. That’s contradictory, the idea that what’s heritable is the standard deviation instead of the mean, but that’s my view about that. I think my childhood did have the effect of making me suspicious of those simple nature-nurture oppositions.
COWEN: Very last question: what will you do next?
GOPNIK: The big thing that we’ve been working on now, as I mentioned a couple of times, is this project about caregiving. This speaks to the issues that we’ve been talking about with nurture and nature and what the effect of caregiving is. Caregiving’s fascinating because if you ask people, “What’s most important in your life? What are your greatest moral issues? What’s most meaningful to you?” They’ll say something about “Taking care of my kids, taking care of my elderly father, taking care of my spouse.”
Yet, economics is a really good example where that doesn’t even show up in the GDP. All that work that we do, taking care of each other — it’s just completely invisible from an economic perspective. Even the fundamental structure of caregiving, which is that you’re giving resources to someone else to accomplish their goals exactly because they don’t have resources that are very different from the usual social contract power kinds of relations. Yet it’s been very, very understudied compared to all the other kinds of social relationships that we’ve studied.
Obviously, since I’m someone who thinks about children and about taking care of children, that’s true, but I think it’s also true about taking care of elders. One of the things in the new book that I’m writing is going to be about thinking about elders and the role of — perhaps a bit autobiographical — but as we have more and more elders around too, what’s the role of elders? How are they working? How do we take care of them? How do they help take care of the rest of us?
COWEN: Alison Gopnik, thank you very much.
GOPNIK: Thank you, Tyler.
Photo Credit: Rod Searcey
This episode was made possible through the support of the John Templeton Foundation.