An absorbing conversation featuring Colin Camerer (CASBS fellow, 1997-98), among the world's most accomplished scholars in both behavioral economics and neuroeconomics, with economist Stephanie Wang (2024-25). Camerer discusses his groundbreaking work on the neuroeconomics of self-control and habit formation; offers insights on generating ideas for, building, then scaling behavioral models; and explains why neuroscience remains a wide-open field awaiting the contributions of so-far mostly reluctant economists and other social scientists.
An absorbing conversation featuring Colin Camerer (CASBS fellow, 1997-98), among the world's most accomplished scholars in both behavioral economics and neuroeconomics, with economist Stephanie Wang (2024-25). Camerer discusses his groundbreaking work on the neuroeconomics of self-control and habit formation; offers insights on generating ideas for, building, then scaling behavioral models; and explains why neuroscience remains a wide-open field awaiting the contributions of so-far mostly reluctant economists and other social scientists.
COLIN CAMERER: Caltech faculty page | Camerer research group | on Google Scholar | Wikipedia page | bio at the Decision Lab | bio at MacArthur Foundation |
STEPHANIE WANG: Pitt faculty page | Personal website | on Google Scholar | CASBS bio |
Works discussed or mentioned in this episode:
C. Camerer, Behavioral Game Theory: Experiments in Strategic Interaction. Princeton University Press, 2003.
C. Camerer, "Can Asset Markets Be Manipulated? A Field Experiment with Racetrack Betting," Journal of Political Economy, 1998.
C. Camerer, et al., "The Golden Age of Social Science," Proceedings of the National Academy of Sciences, 2021.
C. Camerer, et al., "A Neural Autopilot Theory of Habit: Evidence from Consumer Purchases and Social Media Use," Journal of the Experimental Analysis of Behavior, 2024.
S. Wang, C. Camerer, et al., "Looming Large or Seeming Small? Attitudes Toward Losses in a Representative Sample," Review of Economic Studies, 2025.
F. Ramsey, "Truth and Probability" (1926), published in F. Ramsey, The Foundations of Mathematics and Other Logical Essays (1931)
U. Malmendier, S. Nagel, "Depression Babies: Do Macroeconomic Experiences Affect Risk Taking?" Quarterly Journal of Economics, 2011.
M. Cobb, The Idea of the Brain: The Past and Future of Neuroscience, Basic Books, 2020.
M. Gaetani, "CASBS in the History of Behavioral Economics," CASBS website, 2018.
Also of interest:
S. Wang, et al., eds., "Mindful Economics: A Special Issue in Honor of Colin Camerer," Journal of Economic Behavior and Organization, forthcoming.
Narrator: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered.
Behavioral Economics, the study of how psychological and other social factors influence economic decisions, is now a well-established field, and CASBS followers know about the Center's role in its origin and development. In thinking about how or where to push the frontiers of behavioral economics, one promising avenue is in the direction of neuroeconomics, the cross-disciplinary and increasingly computational study of how economic behavior illuminates our understanding of the brain and how advances in neuroscience can help to get to the root of decision-making and inform the creation of new economic models. Among the few at the forefront of both fields is Colin Camerer, and today on Human-Centered, A Conversation with Camerer himself.
He's the Robert Kirby Professor of Behavioral Finance and Economics, and Director of the T&C Chen Center for Social and Decision Neuroscience at the California Institute of Technology. Among other things, he is a member of the American Academy of Arts and Sciences, a Fellow of the Econometric Society, as well as the Society for the Advancement of Economics, the recipient of a MacArthur Genius Fellowship, and has served as President of both the Society for Neuroeconomics and the Economic Science Association. In 1997-98, he was also a CASBS Fellow, and while here, wrote most of his landmark book, Behavioral Game Theory.
Notably, he was also the organizer of five luminaries who comprised that year's Behavioral Economics Working Group, assembled to take stock of the behavioral economics field and, hopefully, push it forward. It's likely you've heard of at least a couple of the group's other members, and will link to the CASBS article about them in this episode's notes. Joining Colin Camerer today is Stephanie Wang, a 2024-25 CASBS Fellow and Professor of Experimental and Behavioral Economics at the University of Pittsburgh.
Stephanie has enjoyed a few collaborations with Colin and in early 2025 she organized a fesh rift in his honor at Caltech. Stephanie's astute comments and questions throughout help elicit Colin's takeaways from his groundbreaking work on the neuroeconomics of self-control and habit formation, his insights on building behavioral models, the kind of data we're still missing for populating those models, what it may take to scale them up and exactly why many economists have been reluctant to take up neuroscience, making it a wide open field and one of the major themes of his upcoming book on neuroeconomics and social science. We'll also hear about a connection he made during his CASBS year that resulted in a productive reconnection 25 years later.
And that's just scratching the surface of this fascinating conversation. So let's dive in.
Stephanie Wong: Colin, thanks so much for joining us today. So you worked on your field-defining book, Behavioral Game Theory during your CASBS year. How did your time here change the book that you planned to write?
Colin Camerer: Not too much, actually. I mean, obviously, the main thing which everyone talks about and should, is having a really uninterrupted, very little constrained time to just think and write. And for a book like that, I knew a lot about the types of experiments in game theory I wanted to cover.
But it was useful to have time to go down a rabbit hole. And if I wanted to spend three days in a row reading about biology and game theory, I could do it.
Stephanie Wong: All right. So you want to talk about some of the people that you met during that year that I think led to work sort of later on that came from some of your discussions.
Colin Camerer: Yeah. So one person who was here was Mark Bouton, B-O-U-T-O-N. And he was from the University of Vermont. And he worked on reinforcement learning. There was another woman, Susan Mineka, who was, they were in a kind of an animal learning, animal behavior group. And Sue Mineka had done some really beautiful work on snake fear preparedness.
Turns out that if you have monkeys born and raised in a lab who've never seen a snake, when they first see a snake, they don't, it doesn't bother them at all. They just, it isn't built in evolutionarily. But if they see another monkey scared of a snake, then they learn to be scared of a snake too.
So it kind of shows you the subtlety of innateness. Anyway, there was things like that that were just fun to know about. Not that I was going to study snake preparedness and fear.
But Mark was different. He was very interested in reinforcement learning in animals, which is an old paradigm going back to Thorndike, later Tolman and others in the early 1920s, 30s, 40s in psychology. And at the time, I thought it was just corny and too simple.
It didn't seem to be, there wasn't a clear theory. The experiments had these very subtle differences in reward schedules, whether you give the monkey a pellet every minute or every five taps on a lever. It just looked somehow shallow or far removed from the things we were thinking about in behavioral economics.
But later, about three years ago, I became interested in modeling habit using ideas from neuroscience and from reinforcement learning. I realized, I should talk to Mark Bowden again, essentially 25 years later. So we had a bunch of conversations.
We had a little meeting at Caltech where Mark came and presented what he was working on. These experiments with rodents are really delicate. You know, you never quite know what the rodent, you can't instruct them, get them to read the instructions and take a comprehension check, right? Beep, beep, beep, beep, beep, you know? They're not very good at that. But so you have to just have other ways of trying to check that they comprehend what's happening. And also usually you're rewarding them with food or drink so they can get satiated. So there's a lot of limits on what you can do with them.
But his, and his research turned out to fit very nicely with what we've been thinking about kind of independently, which is what we call the Neural Autopilot Theory of Habit. And the idea is that what a habit is for, the function it serves, is to kind of offload behavior to things that can be done automatically and have been proved to be reliably rewarding, just like a neural autopilot in a cockpit of an airplane.
And the big controversy in this field is, are animals more strongly reinforced by things that happen regularly, like they get the same reward every minute, or by slot machines, where they have a chance of getting rewards, but they're never quite sure, and they're kind of pleasantly surprised. And what Mark calls the Vermont view, which is the same as our view about neural autopilot, is that it's the reliability of the rewards that's really crucial. Because that's kind of why you can afford to indulge in a habit.
Like if you're going to the same restaurant every Friday night with your kids, and the food is the same quality every time, then that licenses you to sort of build a habit, and you're still going to get reliable rewards.
Stephanie Wong: Where does sort of the slot machine type of evidence come in, and how do you reconcile that? There's been so many models of habit formation, right? What's really different about this one?
Colin Camerer: Yeah, so I think the slot machines activate, they do generate vigor and motivation, but I don't think they generate habit. Right?
So for example, there's a famous clip on the internet, you can see if people, a bunch of older people sitting at slot machines pulling the slots constantly. But they're looking at the screen, right? And it may be there's aspects of the motor activity which is habitual, right? But typically, like when you're driving a car and you're listening to music or someone's talking to you, you're not looking over, you know, to see, you know, to see their face as you would in a typical conversation.
So I think the slot machines are producing motivation but not habit. But again, this is quite controversial and a lot of people are disagreeing with us.
Stephanie Wong: So can you talk about some of the new data and evidence that you've brought to bear on this particular?
Colin Camerer: Yeah, so we have one paper on Weibo social media posting. That one is a little tricky because in order to take these models from animal learning and human learning as well and then look at field data, you have to have a concept of what is rewarding? What's the behavior that they're performing and is it rewarding?
Again, the center of our theory and also the slot machine analogies is how regular is the reward and is it the same amount of reward every time? Most of the time in the field data, we don't really get a good cut at that. With social media data which several people have studied, and you might think something like likes is a measure of reward, but who knows?
It's probably highly nonlinear. Who knows if it's likes or reposts or what? The better study we have is about tuna purchase.
The cool thing about that is at a certain period of time, the three of the big tuna makers, StarKist, Bumblebee, I forget the third one, all changed their prices and their can size within about a three month period. So what that meant was that people would go to the store and they literally see a six ounce can at one price and a five ounce can at the other price. And so you could see if they were kind of in the habit of buying the old big can, how much did that persist when this new can was introduced?
So that's a little bit better of a good stress test of whether they're forming habits or not. And it looks like in our data, we can classify people as being in habitual mode, maybe I think 20% of the time, which sounds about right. I don't think people develop tuna habits the way you might develop a habit for binging Netflix or what my mom, my Irish mom used to call Fish Witch Friday during Lent.
You know, it's just not habitized in the same highly automatic way as like driving your car or your morning routine when you get up and make coffee.
Stephanie Wong: So obviously, often people probably ask you when you're doing this sort of research how do you break bad habits and how do you form habits that you wish to form? What's your takeaway from your model and data so far?
Colin Camerer: I think for forming habits, the key would be, so there's one thing that's not explicitly in our model, but it's in our minds and it's in our writings, which is a lot of habit we think is initiated by a cue or some kind of environmental situational condition. So, if you're trying to build up a gym habit, it's probably good if you do that on as regular a basis as you can, like certain days of the week, certain times of the day. You might have a gym buddy who comes with you, who can each help each other build up self-control, and there's somebody to scold you if you're like, I'm too tired to go to the gym.
So, that's one thing that seems to be important. The second is reliability of reward. So, the way I would think about it is, if you're trying to build a habit for something like say gym attendance, you're going to have better days and worse days, in terms of lifting equipment or running, and so knows it will be raining or not raining or cold, and from the point of view of the mechanics of our theory, what you would want to do would be to either top up rewards exorbitantly in some way, if you had a bad run, and if you had a great run, you don't really actually want to reduce reward, right?
But you want to have as much stability and reward as you could possibly get. Breaking bad habits I think is a little harder because a lot of them involve self-control, where you're getting an immediate benefit, and the future cost is delayed, and the two things are really not in a common neural currency. You know, one is like the smell of Auntie Anne's pretzels, and the other is a figment of your imagination.
You know, it doesn't, it actually doesn't exist. So you have to work in the imaginability part. And I think there's some therapies and some progress that has been made on that, called episodic future thinking. You just try to really train yourself to think about the future in a way that kind of pulls it into the present and makes them comparable.
Stephanie Wong: So you mentioned some of your great pioneering work on the neuroeconomics of self-control, and you've just been finishing up a book on neuroeconomics. Tell us more about what you hope to, the message that you hope to convey with this book.
Colin Camerer: The working title is Neuroeconomics and Invitations for Social Scientists. And so ever since we started percolating on this, actually in the first conference that George Lowenstein and I organized with John Cohen, who's now at Princeton, was in Pittsburgh. And so it was around the time that we were just arriving here a few months before.
And we have not successfully gotten many economists to be interested in the details of how the brain works as a foundation for economic modeling. So this is kind of my last attempt to write in a book form. Here's a bunch of interesting things about the brain.
I have a whole chapter on theoretical economics that uses brain evidence in some way. There's only six or seven papers that I think are clear enough to be able to explain and write about, but the papers exist that do this. And what I mostly want to get across is two big themes.
One is there's a vision psychologist called David Maher in the 80s, and he was interested in vision, and he pioneered a kind of hierarchical idea, which was, you don't really understand a system, and you can't make it better or fix it unless you understand the function of the system. What is it for? What problem is it trying to solve?
The algorithmic specification, like what are the equations? Or if you're going to build an airplane, what's the blueprints look like? And then the mechanism, like actually how does it work in the brain or in flesh and blood?
And so habit is actually not a bad example because in our approach, the function of habits is to save time when repeating an action is likely to be rewarding. But in addition, you need to have a warning system, like in the airplane cockpit, so that if you've been going to the same restaurant week after week, and suddenly it's terrible one night because there's a new chef or something like that, then you have to have a system that says we should quit that habit and go looking at some other restaurants. And so that's part of the theory as well.
The algorithmic part of habit is a few equations that basically say when you're in habit, when you're in goal-directed mode, you just pick what's best for you. You look at all the things in the store, you look at the entire menu, and that's essentially the default, what we call typical rational choice economic thinking. But when you're in habit mode, you basically think what did I get last time when I was in this situation?
What's the reward reliability? If it's really good, it's above some threshold, and the threshold index is how habitual I am, then just do it again. That's it.
So if you were to build a kind of machine to do this for you, when it's in habit mode, it's saving a lot of time, so it can be daydreaming about other stuff and dealing with other issues. So that's the algorithmic level, some equations. The mechanistic level is you can do studies like my colleague, John O'Doherty, at Caltech in which you scan people's brains when they're making choices in which we think from the behavior itself, we can say if they're in habit mode or not.
We can see if they go faster, do they look at more things on the screen, so we can see the automaticity of habit. Then we can see if there's something in the brain that's deciding how to switch between habit mode and goal-directed mode. That's what John O'Doherty calls the arbiter.
Our model doesn't have an explicit arbiter, it has a threshold. So when you cross the threshold, you go from one to the other. As you form a habit, it builds up, the reliability will fall below the threshold, and so you're now in habit mode.
You can see people going in and out of habit. By the way, the standard view of habit in marketing and economics, and a lot of applied fields, is that habits are marked by what's called adjacent complementarity. Goods are called complements if having more of one makes you want to have more of the other, like hot dogs and mustard, hardware and software.
Gary Becker and others had this idea of adjacent complementarity that what a habit means is that I've been doing something a lot, and because I've been doing it before, I really want to do it even more. The utility goes up. We don't think that's really it for various reasons.
It just has a bunch of predictions, and I think it doesn't get all the psychology in. There's no concept of automaticity, of being able to multitask when you're in habit mode. None of those things are in there because there's not really a full neuroeconomic model.
I should add a fun fact, which is the philosopher Frank Ramsey, also a University of Chicago person, decades before Gary Becker and Kevin Murphy, had a very different view. And so Ramsey has a beautiful short book called On Truth, which was published posthumously. He was not a modest man apparently.
And in one of the sections, he actually talks about being in habit mode. He doesn't use the modern terms we do or the animal learning terms, but he talks about going to a particular bookstore in Cambridge, and he sort of automatically finds himself going there without consciously deliberating and thinking about what he's doing. And so if you put, so at least Ramsey is on our side.
Stephanie Wong: So talk about how you, your usual process of building a behavioral models. I think this habit one is a good example. If you were talking to students giving advice, how do you start and what's the process from which you build a model? What should be the starting point?
Colin Camerer: Yeah, that's a good question. So for me, first of all, I'm not sure I'd advise most students to try to do what I've done. You know, I've kind of ignored every, not ignored, I haven't ignored every norm, but most students in economics departments, still to this day, if they were doing the stuff I was doing when I was first learning and writing papers, their advisor would scold them that you're not doing it right.
Because I don't really care about being an economist or whatever. I'm a behavioral scientist or a behavioral economist. And I've happened to work at places like the University of Pennsylvania and Chicago for a while, and then Caltech, which was perfectly happy to have someone like me who's interested in combining ideas.
I'm sort of in the import-export business, like find something outside of economics that's pretty interesting that could be useful. Maybe it's hindsight bias, or maybe it's using eye tracking to see where people are looking in a game payoff matrix, and then sort of bring it in, demonstrate that it's useful, and the same with fMRI brain imaging. So where do ideas come from?
I have two kind of mantras which are somewhat conflicting, but Eugene Fama, who was a teacher of mine at the University of Chicago, a famous finance professor and Nobel laureate, used to say, do the literature view last? What he meant was, if you're thinking about habit, read a few papers, you can read popular books. I read fiction all the time and social media.
And just think about it, think about what you think habit might be. And then once you have a sort of framework, then you can go and look in the literature and say, does this fit into my framework? The downside of this is you might find out that you've reinvented the wheel.
So it worked for Fama because what he became famous for was arguing that markets are quite informationally efficient at a time in the 1960s when hardly anyone thought that was true. And he was right, mostly. Leaving some room for behavioral finance from Richard Thaler and others.
But I think their idea is right. A lot of graduate students, the mistake they make is to say, gosh, I want to think about habit and economics. I'm going to read tons of papers until I find my thesis topic.
The whole point about the thesis topic is it won't be in those papers. It has to be something somewhat new. Or it might be in Frank Ramsey's obscure philosophy book, which I just found out about a few months ago.
We've been thinking about habit for years. Because his discussion of habit, even though he's revered in economics for Ramsey pricing and for all kinds of other things, and he died young at age 28, that view of habit didn't get into the economics canon. I'm trying to shove it back out there so at least some people will know that Ramsey, this is not a crazy idea we had.
Ramsey had the same idea, really, that the psychologist and neuroscientist later came to think it was important.
Stephanie Wong: You said there were two conflicting views on it.
Colin Camerer: Oh yes. Oh, I think the other view is just read and listen and think about a lot of stuff. And a lot of the cool things we've done have involved either…
Well, so I'll tell you one very fun example. So I used to go and bet on horse racing a lot. I don't go anymore because the horse racing is tragically kind of a dying areas. And I didn't do make money. It's very, very hard to make money betting against the other bettors basically in horse racing.
So one day I went to Santanita, which is a beautiful, fantastic race track, really, really wonderful. And they had introduced this machine where you could put money in and get a voucher. And then you put the voucher in and you make bets on horses. So you have a voucher and then a bunch of paper slips. And at one point I put in a bet that I'd made previously. I put the wrong piece of paper in. And so I put in a bet, it was the 7th race. I put a bet that I had made for race 9. And it said, oh, this race hasn't been held yet. Do you want to cancel your bet? And I thought, oh yeah, okay. Oh yeah, no, no, I don't want to cancel.
Canceling bets. What if somebody went and bet $1,000 on a horse 20 minutes before the race? You can bet up to half an hour before the race, actually more now. You can bet on the Kentucky Derby for next year. And then I canceled the bet. And so what the crowd would see was the odds going way down. And then when I canceled the bet later, a lot more money has come in, so the odds wouldn't move very much. Right?
And there's a very interesting, essentially the essence of market efficiency is, if you see somebody moving the market, do you think that they have information and you bet the way they do? Or do you think that they're just idiots and they've kind of ruined the odds on your horse and you ignore them? And so I just went and did this.
And so anyone who hadn't gone to the race track and hadn't learned that you can cancel bets would never have written this paper. And thank you, Lars Peter Hansen, Nobel Prize winner from Chicago, editor of the Journal of Economic Economy. He did not make me write down a fancy-pets theory of exactly how this worked because I don't know how to do that. And he thought this is just a really interesting experiment.
And what we showed also was that the crowd mostly ignored these bets. It's as if people knew a big bet that early, way before the race, so everyone would see it was probably just some person fooling around, which it was.
Stephanie Wong: You know, I always teach that paper in my behavioral economics class, and someone always asks, like, how did he come up with that idea? And I'm glad that you finally…
Colin Camerer: That's the answer.
Stephanie Wong: Now I have the answer. Now I can tell them how...
Colin Camerer: I can also tell you that if you try to cancel... If there's a long line at the machine, and you try to go to a live human teller and cancel, they're not happy. They really… They use every muscle in the face to make you feel ashamed and guilty.
Stephanie Wong: So this is a great example of what you're renowned for, which is finding sort of new data, embracing new methods for gathering data. So what kind of data do you think we're still missing in the...
Colin Camerer: Well, so a couple of years ago, we wrote up a paper with two graduate students called the Golden Age of Social Science. And the idea was we have more data than ever before, and not just censors that we have that, and not just social media that we have that. But a lot of governments, like New York was the first one that I know of, LA, a lot of them have data that's easily seen.
Like if you want to find out subway crime in New York, you just look on their website. It gives you monthly counts. So a number of governments have, I mean, in North Korea is different, right?
So a number of governments have really put data out there because they want people to look at it and help them, and they want to have this transparency. So there's a lot more data of all types. And in big data, they often talk about three Vs.
The velocity, amount of new stuff, the volume, huge, and variety. Right? It's a lot of different types of data.
The second thing about Golden Age is the problems we deal with are really multi-scientific. So tragically, there's no better example than pandemic. So you might think a pandemic like COVID or avian flu, which we're facing now, is purely a biological medical problem.
But we learned from the pandemic vaccine, the vaccine was developed in rapid time, thanks partly to Michael Cramer, who had this idea of advanced contracting, essentially ensuring the company that if they dealt with the vaccine quickly, there would be a market.
But then you had the vaccine and a lot of people didn't want to take it. So they did not want to take it, similar with climate change and with obesity and social media, is clearly a social science problem. It's a combination of political science and just conspiracy theory and all kinds of things.
So clearly, after the pandemic, vaccine has developed, the next question would be could people take it? And that's the social science challenge. All right.
So that's part of this Golden Age of Social Science. I think the thing I would love to see more of from a somewhat narrow point of view, like for the study of habit, would be sensorized data. So we have a project with Aura, which makes a ring that's kind of like a Fitbit, but it's a ring instead of a Fitbit, which can measure accelerometer.
If you're using the Aura app, we know where you're walking around. It measures heart rate, sleep. And so you can answer a lot of questions like, if you could define statistically what it looks like behaviorally for people to be in a habit mode or not in a habit mode.
You could have this time series of habit, not habit, habit, not habit. Sleep is a very interesting thing. It's not a Bayer Economics question alone.
We've read a lot about this, and I think amazingly, I don't think the question of whether exercise generates better sleep or better sleep generates more exercise is actually understood. And one reason is it's hard to do a natural type of experiment. You might be able to do it with non-human primates or something
So I think measuring a lot of things about individuals going about their daily activity will be extremely interesting.
Stephanie Wong: Well, how do you think about the concern that will people behave differently or think differently about their behavior once they know they've got that aura ring on or there's some sort of surveillance or this is being recorded in some way, that people's response to that?
Colin Camerer: Yeah, no, that's a legitimate concern. You can imagine some situation in which, let's say you put a bunch of cameras in a classroom and you want to be able to see where their kids are looking and maybe they have eye trackers on so you can see if they're paying attention to the teacher. Maybe for the first day or week that they're very self-conscious and they behave in a highly artificial way.
But I think for most of these things, it would wear off very quickly. And roughly speaking, people don't seem to be that hung up on loss of privacy in kind of everyday scenarios. Like I don't think kids in school would mind, particularly if you have to also make of course a use case that by censorizing in this way, we're going to figure out how to run the classroom better.
It's going to help some children somewhere, maybe your own kids, if you're getting like parental permission.
Stephanie Wong: One other thing I'm interested in is your thoughts in scaling some of the ways you've been gathering data. You mentioned seeing where people are looking, gathering all these ways of understanding visual salience. What are ways in which we can sort of scale that up and really understand at a massive scale what's going on?
Colin Camerer: I'm glad you brought up salience. It's something we've been keenly interested in. Again, I think it's a poster child for where you get ideas from all across natural and social science.
Salience is a word that's used a lot recently in economics. It's used to refer to salience of race in intergroup conflict, salience of high and low prices in consumer organization, salience of like a rare event being overweighted, like winning the lottery or a catastrophic failure of some kind. But salience also has a very precise meaning in neuropsychology.
So the visual sense we studied, we used an algorithm developed in Italy called SAM, Salience Attentive Model. Basically, you feed in any picture, color, black and white, any 2D picture, and there's similar algorithms for video, but through any 2D picture, and it will produce a predicted heat map, which means there's colors that show how likely it is that people will look in certain spots of the picture. And it's extremely accurate.
It's extremely accurate. So you might be able to use stuff like that to look at, say, videos from a presidential debate or something like that. Like we do know that perceptual contrast is a driver of salience.
You know, if there's a– if there's– Mike Pence's hair is white and if there's a black fly on his hair, first a fly is very unexpected. That's what's called Bayesian surprising. And the fact that if it was a white fly and a white hair, you wouldn't notice it as readily, right?
So I think you could do a certain amount of visual salience analysis about stuff like that. And it may be very interesting to look at movies, advertisements, for example, and things like that. We're thinking about one project recently, which is there are databases that show you credit card mailers.
So, you know, even I get these, I don't know what triggers the mailing, but credit card companies mail out a lot of credit card offers, a lot. And so there's data that show you several thousand of these mailers. And if you put them into this saliency algorithm, it tells you what order people look at different things on the mailer, what order and for how often.
Now, again, this isn't people using eye tracking actually looking at those mailers, although we probably will do some of that to see what happens. And, you know, if you want the answer to a question, what is the right font you have to put your annual percentage rate in if you want people to notice it? That's a question we can scientifically answer.
The battleground, of course, has to do with politics and economics and Washington and banking, where, you know, I mean, you can do the math.
Stephanie Wong: So coming back to, you know, the topic of your two books, one on behavioral game theory and this one on neuroeconomics, how do you see the future of the intersection of these two fields and understanding sort of strategic interactions, social interactions with a particular cognitive perspective?
Colin Camerer: There's been maybe 20 to 50 really good neuroscience papers on strategic thinking. It's a little harder to do because typically you're going to have, say, one person being scanned and they're playing another person who's not being scanned. And this self-consciousness, as you mentioned, comes up.
Maybe the scanned person is doing something different, so they're really not matched ideally. Actually, Reid Montague and me and others a few years ago did use something called hyperscanning, where we actually have two people being scanned at the same time and they're doing things with each other. And you can study something called neural synchrony, you know, are their brains, like if we're having a conversation, the listening part of your brain and the speaking part of my brain, if they're not co-active, then that means people aren't having a two-way conversation.
I mean, there's a scientific definition. And that's been used increasingly to study some cool things. But I don't think we've learned a lot from the FMRI studies about strategic thinking.
Mostly they're supportive of work that me and Vince Crawford and Rosemary Nagel and others have worked on called level K thinking, which is the idea that a level 0 type does something kind of automatic or salient. And then a level 1 type playing a game thinks they're playing level 0s. And level 2s thinks they're playing level 1s and 0s, etc.
And that hierarchical structure seems to work very nicely. And you see markers of that in the brain. Like there's a brain circuit involved in what's called theory of mind or mentalizing, which is one human thinking about another purposive organism, like another person.
Like, do they like me, what are they going to do, are they a friend or foe? And you see markers in dorsal medial prefrontal cortex and temporal parod junction when people are doing higher order strategic thinking. But it'd be great to see a lot more studies on that.
There's a lot more interesting stuff to do. One thing, obviously, I think about a lot is, what are the exciting new things going on in behavioral economics? Is there stuff that's not done?
I would love to see a lot more in economics where the product is just not selling in the sense that graduate students in economics departments who are trained in theory often express some mild curiosity about neuroscience, but they're not really professionally committed to plunge into that area. And it is hard. I mean, I think if you want to win a Nobel Prize 30 years from now or 40, this is the thing to work on.
Because you could also work on AI in economics, or you could work on, you know, diff in diff. I mean, there's a lot, but there's a lot of competition, right? If you're working in neuroeconomics with someone like me, or with the handful of people who are really excited about this, mostly in neuroscience groups, there's not much competition and there's just so much interesting stuff to study.
But I want to mention one other thing. Another area that's not quite as kind of radical as neuroeconomics is to use more biological data and more evidence from other fields. I think there's a couple of reasons why economics, the young tigers in economics coming out of graduate school and getting tenure and publishing in competitive journals are not very interested in neuroeconomics.
And one is there's a lot of risk aversion, at least many people think so. Others have said they think the same way about the topics people choose. And if you were at many of these types of departments, even except for maybe Zurich, which has a deep interest in neuroeconomics, kind of Caltech style, and said I want to do some neuroscience and apply to economics, I think you would really be discouraged by your advisers on the grounds that it may not get you an economics job, which is true.
But if someone in an economics department wanted to hire a neuroeconomist, you would be the entire job market, right? So it's essentially the risk of matching, that is, do I graduate in a year when someone thinks, hey, this is a good time to give it a try versus congestion. So if you're studying the flavor du jour, in whatever people, economists are obsessed about recently, maybe it's complexity and experimental economics, which is like, you, there's going to be 22 people graduating studying complexity and experimental economics.
And a lot of them are going to be from a higher ranking school from you. And there's a certain amount of elitism and rankingness, which I think is, in fact, economics profession. Sometimes it's just screening for talent in a healthy way.
But a lot, I think some of it's really a kind of sociological, like social reproduction of the elite, essentially, which is a phrase I never understood until I became an economist. So I think people are nervous, number one. And I think number two, there are simpler techniques like eye tracking.
You know, you can get portable eye trackers even, or a $5,000 eye tracker, which is pretty good. You know, we have these very high-end ones. You can get eye trackers that are not that expensive.
I mean, there was a student from Penn State whose name I can't remember, who bought one himself. So he was on a Zoom interview at Caltech, and I said, where is the eye tracker from? He said, well, I just bought this myself, and I was like clapping.
He was blushing. So if you're ambitious enough, you can do it, right? Even if you're a relatively poor student.
So that hasn't taken off either, even though you can think of so many questions in economic theory in which knowing what people are looking at is useful data. Like are they eliminating your dominant strategies? Or are they looking more at the probabilities or dollar amounts to understand preference reversal?
There's a long list of questions. Once you open your eyes to what can we study better with eye tracker? And not only that, a lot of the papers have been published pretty prominently by Ariel Rubinstein, by our group, by Vince Crawford.
So it's not that no one succeeded in publishing in these procedure journals. It's out there. I think economists are allergic to running their own machines.
They think it's too hard. They're really naive about cost and benefits. Let's take eye tracking.
Almost every psychology department in the US probably has some sophomores or juniors who are volunteering in a psychology lab and learn how to do eye tracking. So if you can pass the prelims at Princeton or Stanford or Caltech, you can learn how to use an eye trigger just like those 21-year-olds, right? And my shero in this is Ulrika Malmendier.
So Ulrika has worked in behavioral finance and she has a very famous paper on depression babies, which includes my parents. Well, my parents are not in her paper, but they were depression babies. And the idea is that if you have an early, and maybe not even early, but if you have a life experience which traumatizes you, or really changes your outlook, maybe your brain activity as well, how long does that last?
And if it's PTSD, how do you get rid of it? And things like that. And so the Depression Babies paper just shows that people who are young during the Depression are much more cautious about investing in stocks, and they take in less debt, you know, as if the Depression left this long-lasting stamp in their brains, that this could happen again, so I have to be very cautious and save a lot and things like that.
And that's exactly why dad and mom are like... So anyway, in a couple of months, Ulrika manages a brand new center called the O'Donnell Center for Behavioral Economics, and she and I are going to put together a one-day list of essentially who in the world knows about childhood scarring and trauma and how long the effects are and are there therapeutics and ways to kind of deal with that. And some of that involves psychiatry.
Some of it is people studying childhood development who have to be interested in resilience and trauma. I think that's going to be extremely interesting. But I think the big theme that Ulrika and I and many people agree on is that Behavioral Economics 1.0 was kind of fueled by cognitive psych and Tversky and Kahneman to a large extent.
Not only that, because like George Ainsley was a psychiatrist who was very interested in self-control and he influenced David Labes' thinking and George Lowenstein's thinking a lot. So, you know, there were multiple tracks. But it's certainly true that there's a lot of stuff in social science that could be useful in integrating into economics models that's not just framing, right?
And so, childhood experience is one, what happens to the brain is one. There's been a lot of interest in norms and things that are really traditionally sociological as being, we know they're very important in economic settings in lots and lots of ways like workplace norms and also revolutions and things like that. So I hope one of the waves of economics will be that memory, attention, the brain, children, and the entire life cycle, children, adolescents, middle age, aging, that whole ontogeny path is extremely interesting.
Even, let's take something like studying gender, right? Besides the non-binary-ness, puberty and menopause are tremendously important things that happen to people. So let's take a woman, because both happen.
So a young girl, prepubescent is one thing. In puberty, she can be quite different, right? I mean, literally biologically and also in attitude and hating her parents and whatever else. Poor teenage girls. I only have a son, so I don't know firsthand. And then young adulthood is a different phase. After age 20 or 25 or so, you know, sensory systems start to decay a little bit. Aging is a different thing. You know, so lots and lots of economics papers will just include age. Like age or log of age, right? But if you're, if you're somebody who studies aging and studies human ontogeny, it's kind of ridiculous, right? Why don't you put in pre-pubescent, post-pubescent, post-menopausal?
Sometimes you can measure those. You know, it's not as easy as just measuring age from some database. But again, as we get better and better data, all these bio variables will start to be things we can think about.
Stephanie Wong: Okay, so almost every time I have the pleasure of talking to you, you're talking about some new book or article that you've recently read that really interested you. Can you share with us one of these that you've read recently?
Colin Camerer: Well, Matthew Cobb has a book called The Idea of the Brain, which is about the history of neuroscience and which is really magnificent. And man, a lot of scientists made a lot of dumb mistakes for hundreds of years. And it might not be over.
Right? So I can't remember the details. I'm not good at history and my memory is decaying.
But I think Aristotle or Plato thought that the heart was really the crucial thing. The brain was just this spongy gunk. I mean, and then there was a lot of Ramon I Cajal was the person who introduced what's called the neuron doctrine, that the neurons was the building block. And that was controversial. You know, there was so many false starts, even though you could. And you could cut people open, and you could do experiments with cats and electricity and things like that for years before humans.
Anyway, it's really, really good. I also think that it used to be in economics departments, there was always a course on the history of economic thought. And now they mostly dwindled, partly because there's a lot of other important things you'd need to learn.
But if you don't know the history of your science, you're missing out on something because you get a lot of perspective about how humble we should be now. And often, like with Frank Ramsey, it turns out you read something somebody wrote 100 years ago, and it was exactly what you've been thinking about. So there's vestiges of really important ideas in the past, as well as some false starts that are instructive about how people can go off the rails for a long time.
Stephanie Wong: That's a great note to end on. Thanks so much for taking the time.
Colin Camerer: Thank you so much, Stephanie and CASBS. It was a fantastic experience, intellectually and very fun when I was here in 1997. It's really a terrific place.
Narrator: That was Colin Camerer in conversation with Stephanie Wang. As always, you can follow us online or in your podcast app of choice. And if you're interested in learning more about the Center's people, projects and rich history, you can visit our website at casbs.stanford.edu.
Until next time, from everyone at CASBS and the Human-Centered Team, thanks for listening.