No summary for this episode yet.
You yourself are not going to sort of solve the obesity epidemic. You yourself are not going to sort of create world peace. You yourself are not going to sort of, you know, solve climate issues, right? Your brain just is going to be big enough. Collections of people, creating a larger ensemble model, actually have a hope of addressing these problems. Hello and welcome. I'm Shane Parrish, and this is another episode of The Knowledge Project, which is a podcast exploring the ideas, methods, and mental models that help you learn from the best of what other people have already figured out. You can learn more and stay up to date at fs.
blog.com. Before we get to today's guest, I get emails all the time from people saying, I never knew you had a newsletter. We do. It's called Brain Food, and it comes out every Sunday morning, usually 5.30 a.m. Eastern time. It's short, contains our recommendations for articles we found online, books, quotes, and more. It's become one of the most popular things we've ever done. There's hundreds of thousands of subscribers and it's free and you can learn more at fs.blog slash newsletter. That's fs.blog slash newsletter. Most of the guests on this podcast, The Knowledge Project, are subscribers to the weekly email, so make sure you check it out.
On today's show is Scott Page, professor of complex systems, political science, and economics at the University of Michigan. I reached out to Scott because over Christmas, I read a book that he wrote called The Model Thinker, which is all about how mental models can help you think better. And as you can imagine, this podcast is a deep dive into mental models, thinking tools, and developing your cognitive ability. It's time to listen and learn. You just wrote a book called The Model Thinker. And I want to explore that with you. What are mental models? So what a mental model is, is a framework that you use to make sense of the world.
So what the model thinker is, the book, it's a book that contains really three things. One is sort of a general philosophy of models. One is a collection of models that you can sort of play with and understand. And then a third thing is sort of this, some examples of how one in practice would apply a variety of models to a problem. So when I think about a mental model, as opposed to maybe a standard mathematical model, is in a mental model, what you have to do is you have to sort of map reality to the mathematics, right? So I may say, this would be one thing if someone would say, well, you should use a linear model here to decide who to hire, right?
Take your data and just put it on a linear model. well the thing is you have to decide what are the variables right so you know because a linear model contains things like maybe grade point average maybe work experience maybe personality tests so you have to think about um what are the things that the variables that i use to sort of attach reality into you know sort of connect reality to the sort of mathematical framework that exists out there so what i try and do in the book but also in my work is think about the mathematics is beautiful because it's logical it's right right but reality is kind of messy and confusing and complex and so what i see mental models in doing in some sense is mapping reality to the sort of clean logical structures of mathematics and we all have mental models whether we're conscious about it or not how did how did you land on this approach so what the The approach is this, is that when I was trained in school, I mean, even though it's starting in sixth, seventh, eighth grade, you'd learn a bunch of very simple models like force equals mass times acceleration or PV equals K in physics.
And in economics, you learn things like S equals D, supply equals demand. And these models are very simple. And sort of the whole idea was, I can explain patterns in the real world, or I can sort of make sense of the variation we see in the real world using a single simple equation. Then what happened is sometime in about 1990s, I went and visited, it was 1993, I went and visited the Santa Fe Institute, which is a think tank on complexity. And this is a place where they've been trying to encourage my advisors at the time who were very good game theorists. Roger Meyerson won the Nobel Prize in game theory, along with Leo Hurwitz, who's another of my advisors, and Stan Ryder was in that group as well.
And they were sort of these people who studied rational choice and how do people sort of optimize in social situations. And the Sanup Institute was all about the fact that the world was so complex, there was going to be hard to optimize. And so I wouldn't say that I had some sort of intellectual crisis. It was more the case that I found this intellectually fascinating. There was this disconnect. And the disconnect is that I'm trying to make sense of an extremely complex world using very simple models. And so what social science has done, I think, typically is sort of said, okay, the world's really complex.
Here's my model. And I can explain 30% of the variation. I can explain 10% of the variation. Or I can explain why these stocks went up in value or something. But that means you're missing the other 70% or 90%. And so what, not me alone, but a bunch of us have kind of happened on is this notion of collective intelligence, is the idea that one way you can make sense of complexity is by throwing ensembles of models at problems. So one of them may explain 20%, another 15%. And it's not that they add up to 100% that they're explaining everything. In fact, there's overlap. There's even sometimes contradictions in what they might explain and what they might predict.
But by looking at the world through an ensemble of logically coherent lenses, you can actually make sense of the complex world. And what's fascinating about this to me is there's a group of people who are some philosophers, some economists, some statisticians, some biologists. kind of playing in this space of collective intelligence, right? You might think biologists, what are biologists doing in this space? But if you think of ants, each individual ant has a mental model, so it's a map of the terrain of where the food sources are. And they can sort of aggregate that collectively within the nest.
And bees can do the same thing within the hive by doing these things called waggle dances, which sort of explain where the food is, right? So bee will come back and dance and say, look, I think there's food here. And another bee will come back and dance, think there's food here. And they can kind of aggregate their sort of crude maps of the world. At the same time that people were thinking about collective intelligence from a purely theoretical perspective, there was a set of people in computer science who were creating things like random forest algorithms and these giant sort of artificial intelligence algorithms that also were sort of constructing or creating collective intelligence by combining all sorts of very simple filters.
And so, and I think you've learned, I think there's sort of a growing consensus that our heads aren't big enough. No individual's head is big enough to sort of make sense of the complexity of the world. So you're going to have a set of models of how you think the world works. I'm going to have a set of models of how I think the world works. But collectively, right, in any one of us, it's just too small to make sense of the craziness, the complexity, the just sheer dimensionality of the world that sits in front of us. But collectively, we can kind of make sense of it. So let's take something outside of finance for a second.
Let's look at the obesity epidemic, right? Now you could blame that on infrastructure. You could blame it on food. You could blame it on the bacteria in our gut. You could blame it on, you know, changes in work-life balance, the lack of physical work, all sorts of things. And to understand any one of the dimensions that contributes to obesity, you'd probably need to have, you know, not necessarily graduate, you might take five, 10 years of study to just understand one piece of it. But if you tried to fix the obesity epidemic by just changing that one piece, by just sort of climbing that one little hill, you're not going to get very far because there's going to be probably systems level feedbacks.
And so there's going to be no silver bullet that's going to fix something like that. What you can do is by having a collection of people who each know kind of different parts and his knowledge overlaps and have different kind of models of how things work, you can get a much deeper understanding and you might be able to get just sort of, I think, a more holistic approach. And we can talk about this later. I think it leads to sort of a different way of thinking about policy when you think about going at these problems from a multiple model perspective. So to what extent is it fair to say that cognitive diversity is then a group of people who have different models in their head about how the world works?
It is. I think that, you know, this is where I occupy kind of this strange voice. The book I wrote before this was called The Diversity Bonus. and that book talks about the value of having diverse people in a room and the reason you want diverse people in the room is because different people bring different basic assumptions about how the world works they construct different basic um you know mental models of how the world works and they're going to see different parts of a problem so if you want to like just look at this you know if you look at sort of fluctuations in the stock market or if you look at the valuation of any particular company.
There's so many dimensions to a company like Amazon or Disney, right? That there's no way any one person can understand it. And so what you want is you want cognitive diversity. And what that cognitive diversity means is people who have different, you know, literally different sets of models or different information. And so one of the things that sort of leads off the book, and I use a lot when I teach this to undergraduates or general audiences, something called the wisdom hierarchy. And you want to think at the bottom out there's all this data, right? All this, you know, just whether you want to call it a fire hose of data or a hairball of data, you know, choose your favorite metaphor.
It's all just floating out there. On top of the data is information. What information is, is that this isn't some, how we structure the world. So you may say unemployment is up. What you're doing, you're shaking tons and tons of data about people having jobs and you're putting that into a single number. You're sort of categorizing both unemployment's up, inflation's up, and you're using those as your variables. Or someone else might have a very geographic view of things and say, well, Los Angeles is doing well, right? Texas is doing well or something, right? But there seems to be, you know, the Midwest, the economy is not doing as well.
Then what you do on top of information is knowledge. And what knowledge is, is understanding either correlative or causal relationships between those pieces of information, right? So if a piece of information is mass and a piece of information is acceleration, then knowledge is that force equals mass times acceleration, right? And if a piece of information is unemployment and a piece of information is inflation, then you might understand that by what unemployment is very, very low, you often get wage inflation. Again, that's a piece of knowledge. What wisdom is, is wisdom is understanding which knowledge is to bring to bear on a particular problem.
And sometimes that can be just selecting among the knowledge. Other times that can be a case where what you're doing is sort of combining and coalescing the knowledge. So let me give two examples from finance here. One of them involves my college roommate, Eric Ball, who has treasured Oracle. And this is one of my favorite stories in the book where he was, someone comes into his office and says, Iceland just collapsed. And two models sort of came in and said, one is you can think of the international financial system as a network of, you know, loans and deposits across banks and across countries.
Another model is just a simple supply and demand model. And so Munger has this wonderful quote about you want to sort of array your experiences on a latticework of models. And what Eric does in this situation, those are his two models, complicated network of loans and promises to pay and simple supply and demand. And he looked at the person who walked in his office and Iceland is smaller than Fresno. Go back to work. So that's his experience. It's a tidy country. It's not going to matter. bridge. Whereas if the person had walked in his office and had BlackRock just failed, he would have said, oh my goodness, I'm not going to use this supply and demand model.
I'm going to use this networks of contracts and promises to pay model. And so what you want to think about is you as an individual, and one of the fabulous things about your site, Barnum Street, is that it's all about all these metal models, all these ways that people have sort of making sense of the world. And one of the reasons people go to your site, one of the reasons people read business books, one of the reasons read, you know, gather knowledge is to in some sense accumulate knowledge in the form of, you know, ways of taking information, understanding relationships to it. And what we hope to gain is wisdom by having more knowledge to draw from.
But the point of the sort of core philosophy of the model thinker is even if you do the best you can, even if you're a lifelong learner, even if you're constantly amassing models, you're still not going to be up to the task of solving anyone, you know, you're not going to, you yourself are not going to sort of solve the obesity epidemic. You yourself are not going to sort of, you know, make, you know, create world peace. You yourself are not going to sort of, you know, solve climate issues, right? But what it's going to take, because you're just not going to, your brain just is going to be big enough.
But collections of people, right, by having different ensembles, almost can be creating a larger ensemble model. I actually have a hope of addressing these problems. Okay. There's so much I want to dive into there. Let's start with, in the hierarchy from data to information to knowledge to wisdom, it sounds like we're applying sort of mental models at the knowledge stage. And then wisdom is discerning which models are more relevant than not. Is that an accurate view of that? And if not, correct me. You know, I puzzle over this a lot. Every time I think I have an accurate view of it, I then reframe how I think about it.
So I was giving a talk the other day and someone said, I think the real space for mental models come in is in this move that's very subtle between data and information, which is true, right? Because when you think about, you know, how I might think about, like if I visit a city for the first time and somebody says, you know, tell me about Stockholm, you know, I immediately start putting it in categories. They might say, well, you know, it's a lot like London or something, right? Or you might, you know, you might sort of say, well, the people are friendly but reserved or something. So, again, you're taking all these sort of experiences and putting them in boxes.
So there is a sense in which just the act of going from your raw experiences into the information does, it's almost leaning on the models you already know. And so this is the thing that I've been puzzling over the last few weeks, which has been fun to think about is that if I have a set of like models in my head that we think of at that knowledge space, does that then bias how I filter the data into information? Probably does. Of course, because I mean, the models are helping you pick out which variables are, which variables you think will be more relevant and how those variables will interact with one another.
Right. So here's a, here's a really great example of sort of, of that. So there's phenomenon called the wisdom of crowds, right? The Sirwicky book, where you can have groups of people make accurate predictions. Well, the reality is that sometimes groups will be successful and sometimes they won't. And one of the reasons we write down models is to figure out what types of diversity will be useful and what types won't be. But there's been work by a number of people, Kayat Chen at HP Labs, Michael Sanchez-Burks here at the University of Michigan, where they sort of compared how to, you know, sort of suppose I have lots of actual data out there and I run a linear regression to try and predict something, and I have that compete against people, what you find is the regression does a lot better than any one person because of the fact that the linear regression just can include a lot more data.
It doesn't suffer from biases, all sorts of stuff. But oftentimes when you have groups of people compete against the linear models, the groups of people can beat the linear models. And when they do, what you find out where they beat them is when the linear model, the person constructing the linear model, because the linear model is actually constructed by some sort of person typically, doesn't have a way of including something in the model. So one example involved a consumer product was a printer. And the linear model said, this printer is going to sell, let's say 400,000 units. And when they used a crowd, the crowd was like, no, it's going to sell like 200,000 units.
That was a huge difference. And so they went back and they interrogated people in the crowd saying, why do you think this printer is not going to sell? It's handles the spinning pieces of paper, print quality is this good, the toner cartridge is easy to change, you know, all the sort of attributes that were in the linear line. And the first word out of the person's mouth was, but ugly. That is a but ugly printer. Well, there was no but ugly variable in your regression, because that's sort of like, you know, a design feature. And it wasn't a very attractive printer. But the difficulty with data in those situations and the form of the linear model is that it only looks backwards, right?
It can only look at sort of what's happened in the past, whereas people, when we're constructing models, often are kind of forward-looking. How are people going to respond to this new design protocol? Now, what's going to work best, ironically, in all these situations is a combination of the linear model and the model with people. And this gets to this sort of step from knowledge to wisdom that I find really fun. You could say, oh, so what you should do is you should average the linear model and the people. That seems to not be true. What you should do instead is if the linear model the people are close you know the predictions you probably should go to the linear model because it's really well calibrated right it's probably going to you know be better but if they're far apart if the linear model and the humans are giving very different predictions then you want to go talk to them i mean talk to the people and talk to the linear model now you could say how do you talk to the linear model well you look and you say what variables are in there what variables are the people using that the linear model is not what are the coefficients look like has the environment changed yeah is it right no that's the key thing the state the linear model is assuming stationarity and what's fun and this is like mit just started this this new school right this uh this new sort of data science school right that they're they get their first um i think it's the first one they're starting like 30 40 years they had like raised a billion dollars for this and one of the thing is they want people who are bilingual who can communicate between these sort of really sophisticated artificial intelligence models and the real world because the thing is people are afraid of sort of just throwing all this information into this giant AI model like it spits something out but if you're using relatively simple models you actually can it is easy to be bilingual it's easy to sort of look deeply at this you know whatever model you've got and say why is the model saying this?
I like that a lot. I want to sort of explore something with you, which is when we're talking about models and how we apply them, whether we're applying them at the information sort of like filtering state or data and information stage or the knowledge stage or the knowledge to wisdom stage, it seems to me like we can probably agree that the more models you have is a good thing in general, but only to the point where they're relevant for the specific problem that you're facing. Having extra models, if they're not useful, is not good. But the more tools you have in your toolbox, the more likely you can accommodate a wide variety of jobs.
I think that's right. But I also think there becomes this interesting challenge when you think about building teams or also building your own career. So in your interview with Atul Gawande, he made this fabulous point about his method of making a contribution to the world was sort of being able to communicate across different types of people in different areas, right? So he brought sort of a, you know, he'd been trained by doctors. So he had, you know, his parents were doctors. And so he, he'd sort of absorbed what the medical profession was all about. But at the same time, he had these, this deep interest in science and this deep interest in sort of political philosophy and literature and sort of public policy.
And that enabled him to fill what Ron Burt calls a structural hole, right? In terms of like, there's a network of people studying medicine. There's a network of people studying sort of politics and public policy. And he can sort of stand between those two things and make sense of them. Right. So one of the things I talk about in both the diversity bonus and also in the model thinkers, you could think of yourself as like this toolbox. And you've got some capacity to accumulate tools, mental models, ways of thinking. And what you could decide to do is you could decide to go really deep. You could be the world's expert on, or one of the world experts on random forest models or the Op-and-Off functions.
Or you could be, you know, one of the world's leading practitioners of sort of signaling models and economics. Alternatively, what you could do is you could go deep on a handful of models, right? There could be three or four things you're pretty good at. Or you could sort of be someone who, I think, in the financial space, I think a lot of people are really successful. like, you know, Bill Miller, a friend of mine, is by having just an awareness of a whole bunch of models, right? So having 20 models that you have at your disposal that you can think about. And then when you think, when you realize like, you know, this one may be important, then you dig a little bit deeper.
But also those, that variety of models gives you, I think, two things. One is it gives you sort of a robustness that you think of in sort of a portfolio sense, that you're not going to make a mistake. But it also can give you this sort of incredible bonus in the sense that two models, rather than giving you the average of the performance of the two, often give you much, much better than the average, right? You get this sort of bonus from thinking about a variety of models. And so the book, what was super fun about the book and what's been really rewarding about it is what I do is I lay out this philosophy of like, okay, this is why to confront the complexity of the modern world, you need a variety of models, right?
Sort of using this data information, knowledge wisdom did. Then what I do is I take what I think are the 30 most important models that you might know, Markov models, linear models, Kernel-Blotto models, Lyapunov functions, systems dynamic models, simple signaling models from Gaither, just a whole variety of things. And it was just a great exercise was how do you write each one of these in seven to 12 pages in a way that everybody can understand them and then use them? And that's our real challenge. And the people at BASIC, Melissa Verna, who's the editor of the book, but she was the person who sort of wordsmithed this with me.
That was a real challenge because there were times when she would just say, no one is going to understand this. And that was a fun thing to do. I could pick up my book and pick up Markov models, for example, right? Markov models are these models where there's states of the world and then transitions between the states. And you read the book, it's like, you know, about 10, 12 pages. And I think most people can understand it. and all the math is in a box. If you go to the Wikipedia page and you type in Barker Models, you'll just go, wow, I should get a PhD in statistics. That's your only hope of understanding it.
And so I think that what I'm trying to do in the book, in some sense, is the same thing that Atul's time to do in his work life, is to say, here's a way to get kind of knee-deep in these things, to understand where they work. and no one is going to master all 30 models in this book because you could write a whole PhD on each one of them, you write 20 PhDs on each one. But the thing is, the awareness is really useful because you might say, you know, this Colonel Blotto game or these Markov models or these signaling models or these power law distributions, this is really interesting to me. And then you can go a little bit deeper.
And so I think that it's really meant to be, you know, just kind of a reference in a way, but also just, you know, sort of an awareness document where you can just sort of say, Hey, wow, I was just looking and here's a, here's a super interesting model I'd never even heard of. And it's really fun to think about. So I want to dive into some of those in a little bit, but before we get there, I just want to, I want to talk a little bit more about acquiring mental models. Like how do you pick which models to learn? If you're, you're working in an organization or you're a student, or how, how would you go about having a conversation with somebody about which mental models to prioritize and why?
Yeah. So I think one of the first things you want to ask is who are the relevant actors, right? So is this a single actor who's sort of just making a decision or is this a strategic situation where someone is taking an action and they've really got to take into account what the action is of someone else, what strategy, what choice somebody else is going to take, right? So if you're thinking of like, what action I'm going to take in a soccer game, I'm going to take in investing. I've got to think a lot about what other people are doing. Or you might want to ask, is it the case that I'm taking some sort of action and I'm embedded in a much larger structure where things are sort of moving and I'm taking cues from that larger social system?
So suppose I'm thinking about like, which book do I buy? You can think of that as like, or which album do I download? You can think of that as like, oh, it's just me making a decision. But it's not because you're really making that within a much larger sort of social, cultural, economic milieu where you're not even maybe aware of the fact that you're drawing signals about what other people are doing. So you want to think about what am I modeling? I'm modeling a person. I'm making an individual sort of isolated choice by modeling a strategic situation or modeling something that's much more sort of social ecological like that.
Second thing you want to ask is how rational is the person making the decision with the alternative not being irrational, but the alternative being sort of rule of thumb based. So there's a guy, Gerd Gigerenzer, who's a German social scientist, and Peter Todd have this book give this work on sort of adaptive toolboxes you know just this idea like i've just got a set of sort of mental models or tools and i apply those kind of like oh this is problem a i apply tool a this is problem b i brought tool b so there's a lot of stuff when i think about like where i'm going to go do my laundry or what coffee shop i'm going to go to but i probably don't sit around and rationally think about i just kind of like just follow some sort of routine and maybe i adapt that routine slowly maybe i learn a little bit but for the most part i might just follow rules.
So you want to ask, how are people having that? Then you can ask yourself, is my logic correct? So Colin Kammer, Richard Thaler, people studying behavioral economics would say, if it's repeated a lot, that should move you a little bit more towards the rational behavior because people should learn. And if the stakes are huge, that should move you towards rational behavior. If it's being made by an organization versus an individual, that could move you in Either way, if it's a big decision by an organization, you could imagine like, okay, this is going to be done rationally, right? So you have a committee of people think about it in a very careful way.
But if it's some sort of standard operating procedure within a large organization, then it could be way over the other extreme. This is just how we do it. This is how we've always done it in vain. This is how we're always going to do it in vain. So now you've got this idea of, okay, how do I, you have to, now you're thinking about, okay, how do I model the person? following a rule or optimizing, or maybe suffering from some sort of, you know, human bias, if it's a human, and then is it just a decision? Is it a game or is it some sort of social process? So those are sort of in some sense, the two main questions, you know, what context is the action taking place?
Who's making it? And then I think the real interesting challenge is if it's not a decision, if it's a game, or if it's some sort of social process, is making yourself really aware of the challenges of aggregation, you know, in the sense that like oftentimes things don't add up the way they're supposed to, right? Or there can be just a fundamental paradox in the assumptions you're making. So for example, Barabasi has this fabulous new book out called The Formula. And in this book, The Formula, he talks about how these are the lessons for success by looking at tons and tons of data. And in some of it is about some of the things we were talking about before, with Gawande is you want to like make sure you use your network really well.
You want to seize opportunities, those sorts of things. But there's a, there's a circular reasoning in there that in the sense that if everybody followed that formula, it's not clear that everybody would be successful. Right. And so the thing is oftentimes these systems can contain feedbacks within them, right. That make them logically inconsistent at the level of the whole. And in fact, his book, which is, again, it's a fabulous book. I'm not done reading it because I think he's right. And I think if you read his book, you'll be able to be more successful. But if everybody read his book, which he would like, then people wouldn't make sense.
So when you think about constructing a model, again, it depends on the context. You want to think through, is the whole thing sort of coherent? One of the things I love about, when you asked me the first question, what do I think of as a mental model? I think of it as, again, and I'm a real outlier here, I think of it as a way of saying, I'm going to use this mathematical model to make sense of reality. So one of the great examples of this thing is, I think the value of the formal mathematics is Markov model. So Markov model is, there's a set of states, like you could be happy, bored, sad, whatever, and then there's transitions between those states, right?
Or a market could be volatile, not volatile, whatever. And then if those transition probabilities are fixed, and if it's possible to get from any state to any other state, then that system goes to an equilibrium. And it's a unique equilibrium. So what that means is history doesn't matter, right? One-time interventions don't matter. There's just this vortex drawing it to one thing. What that model forces you to do then is if you want to argue the world is complex, if you want to argue for path dependence, if you want to argue that a policy intervention is going to make a difference in some way, you then have to either be saying, I'm creating a new state that didn't exist before, or I'm fundamentally changing these transition probabilities.
And so you get this idea that, so when somebody's constructing a model, oftentimes they'll say, well, I'm a systems thinker. And then if you have them write down their model, you'll say, wait a minute, that's a Markov model. That is the unique equilibrium. Do you think your system has the unique equilibrium? And they're like, no, no, it's very contingent, path dependent. And then you'll say, okay, well, then your model's got to be wrong, right? I mean, you're missing something. There's got to be some way of changing these transition problems. So I think that I view it as a very deliberative process within yourself of constructing a model.
First, you kind of ask, what is the general class of models? Is it a system? Is it a decision? Is it a game? And then once you write it down, then you can kind of go to the mathematics, and the mathematics will often tell you, given your assumptions, what must be true about the world. And then if it's not, you then have to kind of go back and say, well, let me rethink my model. Do models become a way of surfacing assumptions? Oh, absolutely. No, I think models force you to get the logic right. They force you to sort of say, what really matters here in terms of driving people's behaviors or firms' behaviors, how do those behaviors interact, right, in terms of, you know, how do they aggregate?
And then how should people respond to that? There's a famous quote by Murray Gelman where he said, imagine how hard physics would be if electrons could think, right? and I'd written that in a paper and I um it was attributed to Murray and I and someone said I don't think Murray's ever said that and so I went to Murray and I said Murray have you ever said this and he read it and so then he just said imagine how difficult physics would be if electrons could think right and so he goes there I've said it you're safe um but one of the things about modeling and especially like I think in the space a lot of your listeners are in because they're people who who in some sense define the world is imagine how difficult physics would be if electrons could think and if electrons could decide on their own laws of physics.
So if you're running a large organization, or if you're a secretary of the treasury, or if you're, you know, if you're in any sort of policy position, you get to decide the laws of physics. You get to decide what's legal, what's not legal, what the strategy space is. So when you think about, when you're saying when someone comes to construct a model, what do they decide their assumptions on? One of the things you have to keep in mind is one reason people construct models is to build things, right to build buildings to build policies to build strategies when you do that you're defining in some sense this you know the state's base you're defining reality so if you tell your traders we aren't looking at these ratios you're defining the game for them right and so i think that it's i think the design aspect of models is often under looked overlooked right underappreciated so within the field i mean i'm in sort of an economics political science business whatever there's been, because there's so much data, there's this huge shift toward empirical research.
If you count the number of papers in the leading journals that are empirical versus theoretical, there's just been this massive shift towards empirical research, which in some sense is, you know, I applaud it. The work is much better, right? There's much more data. We can get a causality, huge fan of it. But I think there's a cost to that because what that, a lot of that empirical work is doing is really nailing down exactly what's the size of this effect? What's but slow right of that line. What's the size of the coefficient? How significant is it, right? And so we can suss out whether improving teacher quality matters more than reducing class size and by exactly how much, right?
So that's great. I'm all for it. However, that's taking the world as it is. And one of the really cool things about models I was trained by these people who did mechanism design is thinking about, can we, based on our understanding of how people act, redefine the world, construct mechanisms, institutions that work better? So if you look at the American government at the moment, it's kind of a mess. Everything from like sort of gerrymandering to the fact that, you know, we have this electoral college that made a lot of sense when states are all equal sized or roughly equal sized. You know, now some states that are tiny and still have the same number of senators as states that have 50 times as many people.
But even how we vote on things, what is under the purview of Congress? Why do we have a separate, in some sense, like a financial system? You think of the Federal Open Market Committee and the Federal Reserve System. That's quasi-governmental. The FDIC is quasi-governmental. But NASA and the NIH are not quite as quasi-governmental. You know, you'd like, you'd like, there's a deep question about what institutions we use where that is underappreciated. So Javis and Jenna Bednar and JJ Prescott are running a thing in February at Michigan called Markets, Hierarchies, Democracies, Algorithms, and Ecologies.
We're just sort of saying, look at all this stuff we have to do. We used to sort of think, okay, look, we can use markets, hierarchies, democracies, or just kind of let her go, right? You can see what happens. Like with the roads, we just kind of let it go. You decide to go somewhere, I decide to go somewhere, and then it's a total mess for the most part, right? But when we made these decisions about where we have markets, hierarchies, and democracies, that was made in a world where there was no data, no information technology, where we were exchanging beads as opposed to sending bits through the mail.
But now there's this fifth thing, right? There's these algorithms. And a lot of stuff, a lot of things can be done by algorithms as opposed to markets, hierarchies, and democracies. And there's a question, because the cost of change for all these institutions, should we be reallocating problems across these different institutional forms? That's a question you can't touch, really, by running regressions necessarily, other than to identify the places where it's not working. But you can use models to help you think through, what if we made this a market? What if we made this a democracy? What if we handed this out to an algorithm?
Yeah, it sounds like we're using multiple models to sort of construct a more accurate view of reality. We might never ever be able to understand reality completely, but the better we understand it, the better we sort of know what to do. And yet it strikes me as odd that we're often, one of the ways that we learn to apply models unconsciously is through school. And it's usually like a single model, right? Like you're reading a chapter in your grade 10 physics book on gravity, and then you get gravity problems. And then you know that I will apply this equation to this problem. It's almost an algorithm, right?
I know what the variables are, the school is going to give me the variables, and I'm just going to apply this. And we're taught with this sort of like one model view of the world. Why are we taught that way? And why is that wrong? I think it was right when we had a much simpler world, right? I think it was right when we thought, let's get taken in the context of a business decision. Like you might think, okay, here's how you make it a business decision. You figure out what the cost is going to be. And then you think about the net inflow of profits, right? And you think, okay, do the profits outweigh the cost, right?
Do the revenues outweigh the cost, right? So is it going to be positive cash flow? then now when you make a business decision, there's a recognition that there's environmental impact. There's an understanding that it's going to affect your ability to attract talent, right? Because it's going to be an interesting problem. There's a question of how does that position you strategically for the long run? There's a question of what it does for your capacity. There's a question of what it does for your brand. And so these decisions are just so much more complex than they were. There's so much just an increased awareness of the complexity of all these decisions, that there's no single model that's going to work.
So when you're in seventh grade, we're teaching you very simple things. And we're trying to teach you that there's some structure to the world. And so we want to say, look, here's the power of these physics models. They explain, not only they explain things that you see every day, like why objects fall to the floor, they also sort of explain things that you wouldn't have predicted beforehand, like the two things of different weight fall to the exact same time. Or they can predict things like the existence of the planet Uranus, which they didn't know was out there. So I think the simple models we teach people, because we thought, like Plato's famous quote about carving nature at its joints.
I think there was a belief that we could carve nature at its joints. And then for each one of those little pieces, you can sort of apply this model. And so people will sometimes say, oh, the many model thinker, it's like the parts of the elephant. And I'm like, no, no, it's almost exactly wrong in the sense that you want each model, you know, there is a sense in which, yeah, different models look at different parts, but you need that overlap, right? Because you can't carve nature at its joints. That's what we've learned over the last 50, 100 years, right, is that it's complex. The world is a complex place.
And so I think that the challenges to become a more nimble thinker is to be able to sort of move across these models. But at the same time, if you can't, like if that's just not your style, that doesn't mean there's no place for you in the modern economy. To the contrary, it means that maybe you should be one of those people who goes deep, right? Specialized. Yeah. So you need this weird balance of specialists, super generalist, quasi-specialist, generalist. I mean, there's even people who I've heard describe themselves as having that their human capital is in the shape of a T, right? In the sense that like there's a lot of, there's a whole bunch of things they know a decent amount about.
And then one thing they know deep. Or other people describe themselves as like the symbol for pi, right? Where there's two things they know pretty deep, not as deep as the T person. And then a range of things that sort of connect those two areas of knowledge and then a little bit out to each side. and I think that it's worth having a discussion with yourself I mean not you but your listeners is to think okay what are my capacities am I someone who's able to learn things really really deeply am I able to learn a lot of stuff and then think about a strategy for you know what sort of human capital you develop right because I think you can't make a difference in the world you can't go out there and do good you can't take this knowledge and this wisdom and make the world a better place unless you've sort of acquired a set of useful tools, not only individually, but also sort of they've got to be collectively useful.
Because you could learn 15 different models that are disconnected, that applied in different cases, and never sort of have any sort of gestalt, any sort of hole. And that might make it hard for you to sort of make a contribution. Or you could say, I'm going to be someone who learns 30 different models. But if you're not someone who's nimble and able to move across them, that may be more frustrating for you. I think as we're talking, one of the things that strikes me is if you're going to prioritize which models to learn, obviously the ones in your domain or discipline, the common ones are good to have an understanding of.
And then these general knowledge sort of models that apply across disciplines because those are less likely to be, other people are less likely to bring those to the table. So you can become your own sort of, in a way, and not to the extent that other people would, but your own cognitive diversity machine, almost, if you will. How do you go about iterating these models once you have them? Like, how do you put them into use? are you recommending a checklist sort of approach? How do you mentally store them, walk through them, pick out which ones are relevant and not? So this gets back to, you asked a really prescient question earlier, which was, how do you know how to model something and how do you think about what assumptions to make?
And I think what you do when you think about which models to use and how do you play them off one another, you want to again ask, What is the nature of the thing I'm looking at? And then not so much, you know, sort of a checklist, but you can sort of like page through the book or page through your collection of models and think which ones here might be relevant. So let me give an example that I find my students love to sit around and play with, which is there's two models that have to do with sort of competition between two high dimensional things. So one of them is a spatial model. And in the spatial model, there's an ideal point.
So let's suppose that you have like your ideal burrito, which that it's like weighs about a pound and a half. It's got this proportion of sort of meat and rice and it's hot, but not so hot that you've got to like, you know, you have a giant cup of water right next to you. So you can think of that as a point, like sort of four dimensional space, right? There's a size, there's a heat, there's an amount of, you know, beef and amount of rice or something. That's your perfect burrito. Well, then you can imagine all the burritos that are for sale in, you know, Toronto or Ann Arbor or New York. You could put each of those in that same space.
And then you're going to sort of choose the burrito that's closest to you. So then you're going to say, oh, this is the best burrito in Chicago. Well, if my ideal point is different than your ideal point, then I may think a different thing is the best burrito. Well, that same model you can use for, it's actually the workhorse model in political science for thinking about which candidate do I vote for? And if we aggregate them, like nobody's happy. Yeah. So nobody might be happy. But then there's another model that models the same thing called the Colonel Blotto game. And in the Colonel Blotto game, there's a whole set of fronts.
You can think of those as dimensions. But instead of it being a spatial characteristic, it's hedonic in the sense more is better. So if I think about buying a car, it could be like more miles per gallon is better. More legroom is better. Higher crash test scores are better. Less environmental damage is better. And so now when I think about comparing two things, I could just sort of say, it's not which one is closer to me. Like this burrito is better because it's near my ideal point. I can sort of go across all these different dimensions and say wins. But what's cool about both those models is that if there's a big set of people deciding in the first one, there's a whole bunch of people who have different ideal points and they're trying to decide.
There's generally no winner. So there's those sort of best things. So you think about, OK, I'm going to go to University of Michigan. I'm going to go to Northwestern. I'm going to go to Western Ontario and get a degree. And I'm going to apply for a job and I'm competing against seven other people. or maybe I'm, you know, up for a road trip, but I think, how did I not win this? I'm so great. The thing is, it may be that that's, you could think that's a spatial model and I just wasn't what people liked. Or you could think that's a hedonic model, a Colonel Bono model, like somebody just happened to beat me on some collection of fronts.
But one of the nice things that both those models sort of tell us is that there's kind of no best answer because like you're going to win relative to how someone else is. So it's, it's a strategy, it's more like a game, it's strategic. and there's no best thing you can do unless you happen to know where the other person was. And so the nice thing that comes out of that, there's sort of this calming sensation. My undergrads always feel like if you don't get a job, if you don't get a scholarship, you don't get into grad school, it's not because somebody was better than you. No, it just happens to be that they were positioned better than you.
And that's fine, right? It's just going to happen. But typically, when you think about maximizing your chance of getting one of those things, you need to think about, is this a spatial thing where I want to sort of make sure I've got the characteristics that are near what they're looking for? Or is it hedonic where I want to beat my competitors on as many things as possible, right? So I would have the most undergraduate research done. I would have the strongest letters. And a lot of things are sort of a combination of the two. But what's really useful is having both those models in your head, because what you're thinking about, and in the same story, if I'm an advertising firm and I'm pitching an advertising player, if I'm trying to be a supplier to a large auto company, it's multidimensional competition.
And so what you'd like to do is have both these models in your heads and say, let's think about this as a spatial model. Let's think about this as a purely hedonic competitive model and think about how would we position ourselves? Where are our competitors? And it gives you, I think, I think it's calming in a way, right? Because it gives you a way to structure your thinking. And it also lets you know that if you lose, it's not because you're necessarily worse. And if you win, it's not necessarily because you're better. So it's also, it's calming. It's also humbling, right? Because it's easy to think.
One of the things I deal with a lot in trying to present diversity, the value of diversity, is people who are successful, by definition, have always won. And they're in power. And they think, I'm good because, I'm here because I'm good. And they typically are. and they tend to think they're there because they've had a lot of ability. They have a lot of ability, which means that they've got flexibility in terms of what tools they've acquired. But the point is getting them then recognized is for the group to be better. You want people with other tools. So it's tricky because these people think that people in success will think that because they've won, because they're good, when in fact maybe they've won because they just happen to have the right combinations of talents at the right time.
I kind of think of that in an evolutionary sense, right? Where we have considered gene mutation today that might be selected as valuable, but a million years ago, the same gene mutation might have been negatively selected or filtered, if you will, because the environment has changed, the situation has changed. And we apply stories to these sort of random gyrations. And that's not to say that success is completely random, but there's an element of luck to everything. but how that's weighted varies depending on the circumstances right so you get into this really complicated view of the world and I find that really interesting when we're thinking about how to learn models and how to make better decisions and how to how to how do you teach your kids about complexity like how do you teach your kids uh not not necessarily university students but them as well but like how do you teach them hey the world isn't really this simple place and And here are some general thinking concepts that you need to learn about.
And how do you go about instilling that in children? It's such a fascinating question. I think that especially the article in the New York Times last week about how the upper quintile is spending so much more money on their children than those below with the idea of them being economically successful. So let's go back to a question you asked earlier about in school you learn what equals mass times acceleration. In the economy of 100 years ago, it almost depended, it depended a lot more on sort of you yourself being really, really good. Like you're a really good lawyer. You're a really good furniture maker.
Go back 200 years ago. Like you were successful if you ran your farm well. Right. So it's all about your individual ability and hard work. Right. So I was reading this fabulous book called The Rise of the Meritocracy, which is an old book like 50 years ago. So it talks about sort of like, you know, success is intelligence plus effort. right and it's it's actually where the word meritocracy came from if you imagine the world as a collection of individual silos and the you know sort of this the amount of grain in your silo depends on sort of how intelligent you are and how hard you work then it is all kind of about like your ability work hard get a's right in class develop these skills and that's a very instrumental view of the world but in a complex world your ability to contribute and again i'm to go back to your amazing interview with Atul Gawande.
In a complex world, your ability to succeed is going to depend on you sort of filling a niche that's valuable, right? Which, you know, as in Barabasi's book, it could be connecting things. It could be pulling resources and ideas from different places, but it's going to be filling a niche. And that niche could take all sorts of different forms. And so I think when I talk to undergraduates about this, when I talk to my two sons about this, what you want to think about is finding something that combines three things. You have to really love it. You've got to, you know, it's not to be your passion, but you've got to love the practice of it.
So if you're, you know, a great basketball player isn't someone of great ability, it's someone who loves practicing basketball. A great musician is someone who's got some ability there, but they love practicing music. So you've got to, you've really got to enjoy the practice of the thing you're doing. The second thing is you've got to have some innate ability, right? So my younger son's actually a reasonably good dancer when he was a younger kid. And there's not many adult male dancers. And the guy who runs the dance studio is that after I dropped him off one day, chased me down and said, is that your son?
And I said, yeah. And he goes, well, we need adult male dancers. And I said, yeah, that comes from the other side of the family. He says, no, no, it can't be coming. And he like watched me dance for like 30 seconds. And he's like, you're right. It comes from the other side of the family. and you know even if i love dancing my upper bar dancing is going to be pretty low right so you've got to have some ability there and then the third thing is you have to be able in some sense connect those things to something useful meaningful right some way you know of you think that it's going to make the world a better place right so the question in and of itself or the the thing you're going after has to have some meaning or purpose or value you've got to be that convince yourself of that and convince others of that.
Because otherwise, one of the things I find fascinating about the academy is that where people will be in small departments and they'll study something and it gets really interesting to them and they're the world's expert in that. And that's great because we're advancing knowledge. But outside of their small circle, no one may find that interesting. And I think that it's incumbent upon them to sort of think about, you know, are they using their talents in a way to, I think, are they making that interesting to other people or at least intriguing to other people? because I don't think you're adding that much value if only 30 people lead your work.
That's a great conversation to have with maybe like a 14 to 20-year-old. What if we go younger, right? Like how do we teach an eight-year-old about not only compounding, but power law distributions? And like how do we, we might not use those names and we might not use the mathematics behind them, but how do we start instilling models that are, like the way that I think about this is if the world is changing. There's a core set of models that are probably unchanging, mathematical ones that either cross human history and biology and perhaps all existence. Reciprocation is a great example of one.
It works on human and social systems. It's also a physics concept. How do we teach our kids or should we teach our kids is maybe even a different question on this, but should those models be learned in school as models so that you start developing this lattice work or this mental sort of like book in your head where you're flipping through pages and going, oh, this model might apply here. Nope, it doesn't go to the next model. How do we instill that in our children, even if they don't understand the mathematics behind them, so that we start understanding the world is more complex than single models.
And part of your goal is to just what you said, right, to fill this niche. But one of the ways that you're going to fill that niche is your aggregation of these models and how you apply them is going to be more valuable or less valuable in a group setting in a particular company. And your understanding of how other people are applying models is also going to be a key element of strategy in the future, right? So if we can anticipate that our competitors are following models that they learned in business school, well, now we know how they're likely to respond to what we're doing and we're not likely to be surprised.
And we can use that information to make our business or our company more competitive. Yeah, it's a great question. I think two things come to mind. One is, I think we could do a little bit more of sort of meta teaching in the sense that one of the things that people really like about, I did an online course called Model Thinking, which is a MOOC. And one of the sort of tropes in there is that when you, there's something that's called, that I borrowed from my colleague, Mark Newman, when he talks about distributions, which is logic structure function. So if you see some sort of structure pattern out there in the world, there has to be some logic as to how that came to be.
And then you also want to ask yourself, is there some functionality to that structure? Does it matter? Right. So when you talked about like normal distributions versus power law distributions. So we'll teach kids the bell curve. What we won't do is sort of say, here's the bell curve, and this is a structure. Think about all the other structures you can draw. But now we want to ask, which structures do we actually see in nature? And we don't see that many. We see bell curves. We see sort of stretched out bell curves, which are log normal. We see power law things. But we rarely see things that have like five peaks to them.
So why is that? And so then we need a logic that explains the structures we see. So what logic underpins normal distributions, which are you adding things up? What logic underpins log normal distributions? You're multiplying things. And what logic gives you power laws? Well, in the book, I said there's a bunch of them, right? There's preferential attachment. There's self-organized criticality. But there's logics that will give us those powers. Then you want to ask, does it matter? Do we care? And then that's sort of an easy thing to convince kids of because you could say, well, if incomes like heights are distributed normally.
So that's nice and predictable and seems fair. But if heights were distributed by a power law, there'd be 10,000 people as tall as giraffes. There'd be someone as tall as the Burj Tower. And then be like 170 million people in the United States, seven inches tall. And they're like, whoa, that would be pretty bad. It would be really hard to design buildings as well, right? Because you'd need them for little tiny people in trips. And so I think this logic structure function thing is really important. I think the other thing that we need to do is give them experiences of using the same broad idea across a variety of disciplines.
So one of the things I did in class that I'm hoping to teach again, because the students just absolutely loved it, but it just didn't work out this year, called Collective Intelligence, where we just sort of did a whole bunch of different sort of in-class exercises to sort of explore where Collective Intelligence comes from. So here's one example that was just, I think, just most- Before we go on, what is collective intelligence? So collective intelligence is where sort of the whole is sort of smarter than any one individual unit. So you can think of that in a predictive context. This could be the wisdom of crowds sort of thing where people guessing the weight of a steer, the average, the crowd's guess is going to be better than the average guess of the person in it.
That's just a mathematical fact. But here what we're doing is we're looking at sort of collective intelligence in terms of solving a problem. So here's the setup. Really fun. I had a graduate student make up a bunch of problems defined over 100 by 100 grids. So imagine a checkerboard that's 100 by 100. And each one of those cells has a value. So one of the problems was really simple. It was like what we call a Mount Fuji problem. It was just one big peak. Not necessarily right in the center. It was just kind of in the upper right. It was a huge peak. And that had the highest line. Everything kind of fell off from that.
Another problem had like five or six little peaks over the landscape. but with one being higher. And then another one was really, really rugged. So he just created a bunch of problems and I didn't know what they were, right? That was part of the key. No one knew what the values were. So I created three teams. One team was the physicists. And what the physicists did is they got to sit around and first say, okay, which eight points do we check? And then they would get the values from those points. So it's kind of like that game Battleship where they would say like, you know, D7 and they'd come in and we'd say, this is the value from D7.
And then they'd get another. So they got like five rounds where they got to check. Here's 10 points. So five rounds where they get to check 10 points. And the goal was to find the highest point. Another group was the Hayekians, the decentralized market, where each person went to pick a point. And then they'd just come back and they would say, here's the point I picked and here's the value. But there's no coordination. So the idea was that you could see value by comparing those two. Because you could kind of see where other people picked and you might want to go near where they were. but you also wanted to build information for yourself and the group by trying other points right so there was all sorts of cooperation and competition going on that group the third group was the bees and so the bees would point at a square they couldn't give a number like they couldn't say you know a26 they just could point somewhere on this big square we would approximate we thought that was we would show value and they had to go back and waggle dance right now the thing is it turns out undergrads won't waggle they won't like waggle dance they're just too insecure so we had them just dance with their hand what they did do is they had to like sort of like point the direction it was in and then the longer they waggled that was kind of the better the value okay and then we compared the waggle dancing bees to the hackians to the physicist and on the two easy on a easy problem and on the problem with five peaks the bees did just physicists right infection on the five peaks they did ironically just a tiny bit better than the physicists And so we're talking about this afterwards.
And someone says, well, that's because the bees can take a derivative. And everybody's like, what? He goes, well, no. Like, to solve this, you just kind of got to take derivatives. So they can sort of find the highest, they can find the highest point. And then they can take derivatives because they can see who is wagging longer, right? And it was only on the really hard problem. Physicists did the best. And so what you learn from that is that bees, markets, and problem-solving teams are all deal with high dimensional problems with like you know things that they saw right if it's not super hard and finding p finding food isn't super hard then these are just as good as physicists because it's as if derivatives and markets are just as good as physicists but when it gets super hard the market's not going to work right because you need all sorts of coordination so that's going back then to this thing we talked about before about when to use a market when to use a democracy, when you use a hierarchy, when you just kind of let it rip, that probably depends on the difficulty of the problem.
But what's cool about that, and this is where you can do things with young kids as they see, well, here's this idea, collective intelligence that spans disciplines. If you want to teach, like, so in the same class, I'll just give one more example because it was just so fun. There's this amazing game called Rush Hour. I don't know if you've ever seen it. We have little cars and trucks and you slide them around and you've got to get this red car out. So, yeah. Yeah. So you get the kind of gives you a configuration. And these configurations are like easy, medium, hard and very hard or something.
And what happens is so here's what here's the experiment I do in class. And again, the numbers are too small to say this is any sort of scientific result. But it's it's always worked. It's always been really fun. is I have people play rush hour and they play like an easy, a medium, a hard, and we time how long it takes them to do each one on average, right? And the harder ones take a lot longer. Then I have them write down models for how to play rush hour. So one model might be solve it backwards, which is think about how that car is going to get out. Another model is like get the big trucks out of the way, right?
Another model is move forward, then move backwards. So move forward as far as you can, right? And then move it backwards. And then what I do is I have another set of people read the mental models from the first set of people and then play not the same games, but different games that have the same difficulty and compute how long it takes them. And what happens is they're just a lot better. And what you see is that this is something where it's not tacit knowledge. It's actually learnable knowledge playing rush out. What I've been struggling with, and you probably have a good idea, this is I'm trying to come up with a game where you can't, it's purely tacit.
Like I can't communicate. So my friend John Milleroy's jokes that like this weekend, he's going to read a couple of books on tennis and then go become a professional tennis player. You know, you can't. Yeah. So writing a bike or what? Yeah. Right, right, right. Yeah. So I'm trying to find like a really cool example to juxtapose with Rush Hour. Maybe one of your listeners will email it in. You could do it as a new game where people could learn it, but then there's nothing they can... It'd be nice if it didn't involve physical skills too, just mental skills. That's what makes it hot, right. So these are other things...
It was Adam Robinson actually told me that Rush Hour was one of the best games that he knew of to teach thinking skills to young children. Oh, really? Yeah, and we spent the summer playing that this year on vacation. and we would take it. It's a great game to take to like restaurants and stuff. And my kids were at the time, um, eight and nine. And, you know, we would sit there for, you know, they would sit through a whole two hours and just play this. Um, and it was a totally awesome game. It was fascinating. I mean, and as a parent, I just promised them 30 minutes of iPad. If they got through all 40 problems in like three weeks, they were like, Oh my God, this is amazing.
it's amazing how hard they'll work for that 30 minutes of us right get the incentives right no but it's it is funny though how I think it's because it's a physical game when I'm doing this in class I'll say you only have 10 minutes sometimes I'm just extracting the game from my students hands you know and then I'm like look you can take it home yeah back the next day because it's so much fun let's talk about a few of the models that you have in your book before we finish up here. I want to, can you actually, how about, I'm going to mention three of them and you walk me through sort of how you present them and how you use them.
Right. Let's start with power law distributions. Okay. So power law distributions are distributions that have, so let's start with what they're not. So a normal distribution is something like human height where the average person's 5'10", there's some people 5'8", there's some people 6'0", and it falls off really fast. A power law distribution, most events are very, very, very small. And there's occasional huge events. So if you look at earthquakes, there's thousands and thousands of tiny, tiny earthquakes. And there's an occasional huge earthquake. If you look at the distribution of city sizes in most countries, there's tons and tons of small towns.
There's an occasional New York, London, Tokyo. If you look at book sales, music sales, right? Most books sell 300 or 400 copies. There's occasional books that sell millions of copies. And there's a question of what causes power laws. So unlike normal distributions, which come from just kind of adding things up or averaging things, power loss have a bunch of causes. So what I do in the book is I go back, let's go back to this logic structure function. It's a structure we see a lot, right? This long tail distribution. The question is what causes it? And so I talk about three models in the book.
One is something called the preferential attachment model, where imagine things kind of like walk. You can imagine that there's a set of cities or there's a set of books, and the probability I move to a city or the probability I buy a book is proportional to the number of other people living in that city or buying that book. We can see right away there's positive feedback there. More people move to New York, right? More people move to New York. Or there's more people buy the tipping point. Ironically, more people buy the tipping point. So the tipping point sells a million copies. There's 10 million people in New York.
If nobody buys somebody's boring book, then nobody buys the boring book. But another way that power laws form is through random walks. So imagine that each firm, a firm starts by somebody, you know, joining their firms. They've got these startup firms. They're one person. Now, suppose that they're equally likely to sort of fail or hire a second person. And then suppose they're equally likely to go back down to one person or a third person. Well, the life of that firm, the firm's going to exist as long as there's a positive number of workers. So if it's a completely random walk, like a coin flip, you can imagine that most firms are going to die really quickly, right?
You add an employee yourself, then you fold. You had two employees, then you go down one, up one, down one, down one, you die off. So that would say that the lifespan of firms, some of them can be really short, but if you happen to get really big, you're going to last a long time. That should be a power law, and it is. It's also true that the lifespan of species, phylogenera, in ecology, you can think of that as perfectly random, also satisfies the power law. And then a third way to get these power laws is from something called self-organized criticality. So if I drop grains of sand over a desk, eventually I get a big sand pile.
And then if I look at how many grains of sand fall off the floor, most of the time when I drop a grain of sand, once the pile is formed, it'll be very few. But occasionally I'll get these giant avalanches. And so what's happening there is the system is sort of aggregating to this critical state. So think of like traffic in Los Angeles or traffic in Toronto or New York. What happens is it gets itself organized despite where cars are spaced pretty close. and all of a sudden there's one accident, boom, there's three hour delays. So most of the time things are kind of fine, but one accident can really delay.
So now what you've got is you've got, okay, now we have a logic that explains the structure. Why does it matter? Well, it clearly matters in the case of things, you know, for you think of things like, you know, book sales, music sales, those sorts of things. It means that there's going to be some people who are wildly successful and a whole bunch of people who are not that successful. And that may not be, we may decide that's not fair, right? We may decide that like, if I'm Malcolm Gladwell, I shouldn't necessarily think, wow, I'm amazing because I sold 4 million books. It's because no, you just happen to be the New Yorker books because you benefited from those positive feedback.
So it actually could change how we think about, you know, how we attack people. If you thought, no, this person sold these books because they're just so much better, right? That's a very different story than if you say, no, just the natural process of people buying books leads to big winners. Then you start realizing, you know, the big winners are as much luck as they are skill. that's really interesting let's go to the next model i want to talk about which is something that when i was reading it in your book i was like oh first year physics which was concave and convex yeah walk me through that one so what this is i got i got these wrong on my physics like the first assignment i got them mixed up and my like my it was hilarious it was yeah All these memories came back.
Yeah, no, this is a challenging thing because there are certain things you almost, you have to cover. Otherwise, you're sort of, it's a disservice, right? And so the basic idea of sort of linearity, right, is that something has the same slope always. So the next thing is worth just as much as the previous thing. And fundamental to so many models throughout the book is some assumptions of either concavity, which is sort of diminishing turns, or convexity, which is increasing returns. So we just talked about preferential attachment. That's a form of convexity. The odds that somebody buys your book increases more and more people buy your book.
So the odds that the first person buys the tipping point are low, but then the odds that the million and first buyers are much higher, right, because so many people have bought. So convexity just means that the odds of something happening or the payoff from something increase as more people do it. So many things in the world are the opposite. They're concave. So if you think about like, so concavity means that the added value of the next thing diminishes. So for example, like the next bite of chocolate cake, the next scoop of ice cream, right? There's just diminishing returns to that. So if you think about like adding workers to a firm, right, as you keep adding workers, like the value to those additional workers go down.
And that's true of people in teams as well, right? So when you think about, suppose I decide I've got an important decision to make, you know, the second person is going to add a lot to the first. The third person will add a lot to the second, right, and so on. But at some point, you're just not going to add much value. And so there tends to be in sort of team performance on a specific path, a certain level of concavity. these i think the challenge for me in writing that was like how do you make concavity and convexity even remotely exciting right and um because it's it's just kind of like mainstream math and the easiest way to teach it is almost in terms of derivative right so linear function has a constant derivative and a concave function there it falls off but so what you should you try and make the the case that these are in some sense fundamental to not recognizing in particular concavity can lead to really flawed assumptions.
In the 1970s, Japan had this really fast growth. And there are all these articles saying Japan's going to overtake the United States in eight years. But the thing is, if you construct the model, you realize that as you sort of industrialize, you're going to grow pretty fast, but there's going to be diminishing returns to that industrialization. The same is true of China, right? So if you do a linear projection of China, you know, five years ago, you'd have said, oh my gosh, you know, by 2040, China's economy is going to be just enormous. But the reality is growth is going to fall off Because what the model shows, in order to maintain anything even close to linear growth, you have to innovate like crazy.
There's massive levels of innovation. So I think that the idea behind the concavity and convexity chapter was to try and get people to recognize that there's just diminishing returns to so many things, right? That linear thinking can be dangerous. Linear projections can be really dangerous. and the last model i want to talk about i guess it's actually more than one model but local interaction models yeah so these are fun these are like super fun so local interaction models uh the there's some simple computer models not that convexity and cavity aren't fun but they're fun for a smaller set of people i think so these local interaction models are models where you think of like people first off like maybe on a checkerboard but eventually can put them on a network.
And what you imagine is, is the set of things, my behavior depends on the people around me. So one of the examples, a simple example I give often is sort of like, how do you greet people? Right? So do you shake hands? Do you bow? Do you fist bump? Right? It doesn't matter what you do, but you do the same thing that other people do. Right? So if you go to bow and I go to shake hands, I'm going to poke your eye out. So it's not important. It's not going to work. So what you want is, these are in some sense, what we call a pure coordination game. What I'm trying to do is I'm trying to just coordinate with the people I'm interacting with.
This happens on so many dimensions. So in an earlier book I wrote called The Difference, I talk about where you store your ketchup. Do you store your ketchup in the fridge or do you store your ketchup in the cupboard? Again, it doesn't matter what you do. No, it does matter. Just always the fridge. Yeah, and the cupboard people think the fridge people are crazy. And once a doctor said to me, Scott, I think you may think this is funny, but you have to store ketchup in the fridge because it has vinegar in it. And I said, where do you store your vinegar? He said, in the fridge. And like the whole room is like, what are you, a crazy person?
Like, you know, you don't store vinegar. The same is true as soy sauce. There's soy in the fridge, soy in the cupboard people. And it doesn't matter what you do, but whatever you do takes on a lot of importance. It defines who you are. So one of the fun things I do in class, I also talked about in the book, is you can imagine you're actually playing a whole series of local interaction models. And that collection of local interaction solutions you can think of as comprising a part of culture. So my wife, Jenna Bednar, who I mentioned before is a political scientist, she had some papers on this where you can think of cultures like a set of sort of coordinated behaviors across a variety of settings.
And so I'll do this in class. I'll say, OK, do people use their phones at the dinner table? Do people take their shoes off in your house? Is the TV on? Do you hug your family? Right. Just a whole set of things. And then I'll have people sort of I have the students vote, like, you know, using a Google form, like, you know, which ones they do. And we sort of have like, you know, here's the modal response across all these things. Here's the people who are correct, which are the people who do what I do. Right. And I'm like, you know, these are my people. You can move to the front and you all get A's.
and but the best part about teaching this is one time this is like 10 years ago this this kid comes up after class and he goes oh my god oh my god this explains it and i'm like explains what he's like my girlfriend's family and i was like what and he goes everything i do they do the opposite and what's weird about it what's great about these local interaction models is that prior to that he had thought intrinsically they were just weird people right right he just thought These are crazy people who have, he goes, they have their own napkins. Like they always have their own napkin with a napkin holder.
They take their shoes off. Like, you know, they always have the radio out of the house. Nobody's, it was just a whole set of things that they did. Like they hugged each other, right? And he's like, well, they're hugging me, right? All these things that they did, he thought that they were just like part of their genetic makeup, an essential part of their character. When in fact, it was just a series of coordination problems that their family had solved, right? The other example I have in this space that was great is somebody told me this story about how at New Year's Eve one year, she'd been married into this family for 20 years.
She said, you know, look, I love I love the family. They're great. But I hate the boiled cabbage and beet soup on New Year's Eve. You know, 20 years. And I think I can say that. Turned out everybody hated it. Nobody mentioned it. And it turns out, like, I guess somebody like dead for like 15 years, supposedly, like, did they think, right? Yeah. And then they decided, like, going forward that they would make like one ceremonial beat or something. But then they had to work. So I think that, like, you don't realize, though, is that you can, a lot of who we are and what we do comes to these local interaction models.
Now let's make this serious for a moment away from, like, the ketchup and the bowing. when I go work for a firm or if I'm working in an organization, you know, stock analyst, psychologist, whatever I'm doing, mental models that we use are like local interaction, right? I mean, like, it's like, oh, you're using that mental model. It's easier for me to use that mental model. And that then works against this diversity, right? So there's, it really becomes a super important thing that, and what's also very funny is that your mental model is better than mine, but it's still worth it for me to hang on to my mental model because it's giving that diversity right so collectively it's worthwhile but there's going to be so again back to the point you raised earlier about evolution and this is where the many model thinking become fun is you realize like so i've been i go work in some organization i'm working in some community of practice and i've got a collection of metal models i'm using it's it just becomes easy for me to start coordinating on other people's metal models right using other people's terminology because it's more it's it's more efficient and i know you know how to appeal to them how to persuade them how to interact with them, how they see the world, and then they're predictable.
This kind of goes back to, have you read Ender's Game? No. One of the key moments in Ender's Game is like Ender, who's this kid who ends up saving the world, totally fictional book by Orson Scott Card. We just read it with my kids. And one of the key moments is like, he's like, I can defeat my enemy, but only when I really understand them and how they think and how they view the world. And I always thought that that was really interesting, because I'm trying to teach my kids that one model that I want them to have, if you want to call it a meta model, is perspective taking. What does this problem look like through the lens of this person?
What does it look like through the lens of this person? And mentally walk around the table and then have a hierarchy too. What does it look like to shareholders? What does it look like to the government? What does it look like to all the people that interact with the system? And through that, you can get this more nuanced view of reality. And if you see the problem through everybody else's lens, you know how to talk to them in their sort of language or in a way that might be more able to appeal to them. Yeah, that's such a great point, because one of the things that I struggle with in this whole space, and I think it's a good place to struggle, is as you move from sort of very formal models, like I'm fitting some sort of hierarchical linear model versus some like abstract perspective taking versus sort of some notion of sort of like a disciplinary approach to a problem.
So let me give a very specific example that I find cool to think about, which is the drug approval process. So if you look at a company like Gilead, Genentech, right, somebody constructs a molecule, and then they've got to decide, okay, is this molecule something that we can use to improve people's health? One perspective to take on that is just purely a pharmacological perspective, right sort of body chemistry how does it work right just pure science but then there's also sort of a sociological perspective in terms of like you know will people you know how will people take this how will this get passed on will it get abused could it be abused what uses it there's also sort of a purely almost organizational science business school perspective in terms of how do we if it's complicated to explain how do we educate the doctors in terms of how to use this right um then there's also people who understand just the political process which is like you know what's the likelihood it'll get approved, even if it works in all these other dimensions?
Can we get this through the government approval process if it's somehow something different, given that they've got boxes that they use? So what you've got is you bring all these different disciplines to bear, and you've got to, just like you're saying in this book, if I'm the CEO of Yulia and I've got to make the call, do we take this drug to market? I actually have to hire people who can take all those different perspectives, right? Otherwise, you know, I probably won't be CEO for very long because I'm not going to do it. But then you realize, just let's make things just a tiny bit less abstract for a moment and think about traditional arguments for a liberal arts education, right?
The reason you want to read literature from a whole bunch of different vantage points, like, so the reason you don't want to just read sort of the great man view of, you know, Canadian history or US economic history or something like that is because there's all these other people who experienced that same thing and saw it from a very different perspective. But what's funny here is that I'm kind of making this point, when you think about many models, you could think like, I'm making this point like, oh my gosh, people should be spending more time learning technical stuff. And on one hand, that's kind of true, people should be learning technical stuff.
But the core argument I'm making is very similar to the argument that people at the other extreme are making in terms of like, the reason why liberal arts education is so important is the ability to do perspective taking, right? To sort of learn to see the world through different eyes. I think where the difference is, is that I'm a, you know, I think I'm a pragmatist in a way, right? I mean, I just see so many opportunities. And so I feel like I'm coming at from a much more sort of pragmatic perspective in terms of going out there and making a difference in the world, as opposed to just purely appreciating all these different ways of seeing things.
And the reason that distinction matters is if in literature, it could be that every perspective is worth considering, right? And engaging and thinking about because there's no, there's no end game, ironically, given the name of the story you mentioned. But the, but if I'm making an investment decision, if I'm worried about drug approval, if I'm trying to write a policy to, you know, reduce inequality, if I'm trying to think about how do we teach people, there is an end game, right? There are things we can measure. There is performance characteristics. And so it could very well be the case that you could say, I think we should think about it from this perspective, or I think we could use this model.
And then we can kind of beta test that perspective and that model and think, no, we shouldn't, right? So there's a difference, I think, in the approach I'm promoting in that, yeah, you throw out a whole bunch of models, but if the spaghetti doesn't stick to the fridge, the spaghetti doesn't stick to the fridge, and you let it go, right? It may be something you, it's probably something you keep around because there's going to be other cases where it does work. But the point is, there's going to be cases where it doesn't. And so. You don't want to force it. No, you don't want to. So there's the limits of inclusion, right?
In the sense that like you only want to be inclusive to things that are actually going to help you to do whatever it is you're trying to do. I think that's a great place to sort of end this conversation. I feel like we could go on for another few hours. But I want to thank you so much for your time, Scott. This has been fascinating. Thanks. This was, it's really fun to have these open-ended conversations and I really appreciate the format, you know, as opposed to simple, you know, answer, respond, but to give me time to elaborate on the book and what I've been thinking. Thank you. Awesome. We'll have to do part two at some point.
Thanks. Hey guys, this is Shane again. Just a few more things before we wrap up. You can find show notes at farnamstreetblog.com slash podcast. That's F-A-R-N-A-M-S-T-R-E-E-T-B-L-O-G dot com slash podcast. You can also find information there on how to get a transcript. And if you'd like to receive a weekly email from me filled with all sorts of brain food, go to farnamstreetblog.com slash newsletter. This is all the good stuff I've found on the web that week that I've read and shared with close friends, books I'm reading, and so much more. Thank you for listening. Thank you.