- Share your favorite videos with friends
- Comment on videos and join the conversation
- Get personalized recommendations
- Enjoy exclusive offers
Purchased a FORA.tv video on another website? Login here with the temporary account credentials included in your receipt.
To download this program become a Front Row member. JOIN NOW >>
ZOOM IN: Learn more with related books and additional materials.
Sign up today to receive our weekly newsletter and special announcements.
Thanks to everybody for coming. It's a real pleasure to be able to introduce Scott Page tonight and I want to thank him in advance for making the trip here and it was actually quite a whirlwind trip because he is as you probably know, he has written not one but one and half books very recently. And he is on the this sort of frantic book tour with groupies and everything and it's all very exciting. So we are very excited that he was able to take the time out of that to come and and talk to us about this book and this work. Mr. Scott is going to talk about diversity and the importance and sort of the potential benefits that come from assembling diverse groups of people to tackle problems and to think about them, to find solutions to problems. The thing that I would like to say, and I think this is the thing that's really sort of exciting and interesting about the book, and I assume about the talk, is of course one of the problems that we have as a society right now when you discuss the topic of diversity is that its incredibly, emotionally and politically charged issue. And for most people it either you say the word and it either sort of invokes are moral imperative to create diverse groups or it creates this sort of specter of political correctness depending on which end of the political spectrum you happen to lie on. In any event both of those, you know that whole thing makes it very hard to have any sort of rational discussion about this. And one of the things that's really great about this book is Scott has really tried to sort of muster a whole bunch of information, arguments, models, anecdotes that are that are about sort of how do we really discuss this in a rational way where we say, "Well, what are the real the benefits?" And he also talks about the potential pitfalls and limitations of assembling diverse groups of people. And the other thing and I think and this I think will be very evident when he gets up here, is that it's a pretty funny book. And again this is a this tends to be a topic that has to become you know this, everyone gets very serious when they talk about it. And that again I think is makes it hard to have a sort of a really creative and interesting dialogue. And I think, for those two reasons, I think Scott and this book are going to be a tremendous benefit to sort of our cultural dialogue about diversity issues. The only other thing I want to say is just as you know, Scott has been associated with Santa Fe Institute for very long time. He and John Miller have been running Complex System Summer Schools for modeling in the basically competition modeling in Complex Systems in the Social Sciences for 11 years now. This summer they are going to do the 12th year. He was on the external faculty from 1999 to 2005. Our policy is after six years you have to go off for a year. And as soon as he was eligible to come back on he was brought back on, we couldn't wait to get him back. And and he resonates with Santa Fe Institute in three importance way I think. One that of course a lot of the work up there, whether it's in Physics or Biology or Social Sciences has to do with some sort of notion of emergence and I think this is something he is going to talk about a lot tonight of you know, when you get things together what is it that creates something that's greater or different than or in some cases may be less than the sum of the parts. Another that the type of diverse population that he is talking about is of course the model that the Santa Fe Institute has tried to build itself around, this notion that you bring together a small but diverse group of researchers with expertise and backgrounds in different areas. And that may this will lead to solutions to problems that are not very solvable by traditional means. But I think the most important thing and this is you know what, you know it really starts off right in the front of the book saying you know, that the take home message will be you know that we normally think that if you are going to solve a really hard problem you have to get someone really, really smart to solve it. And his message is, no actually a diverse group of relatively mediocre people very often will come up with a much better solution. And I think this is where he has really provided a benefit with Santa Fe Institute because he he provides an excellent post talk rationalization for most of our hiring decisions. And so with that I will turn it over to Scott. Thanks John. I want to start by saying that the bar for pretty funny economist is very low, getting started sort of lower some expectations. And also John Miller wanted me to say that I had actually only written 1.4 books and not 1.5. What I want to do is I want to start out with a few interesting facts. But before I present those I want to distinguish between scientific facts and social science facts. So scientific facts, like that a water molecule consists of two hydrogen atoms and one oxygen atom. That's true today, that will be do true tomorrow and that will be true next week. But social science facts, they change. So when one election Soccer Moms may vote Democratic, in one election Soccer Moms may vote Republican. If I would have given this talk just in the fall, like in October, it would have been the case that this side of the isle would have more people on it right, than this side. Right. So social scientific facts aren't really facts, but they are instead what we like to call empirical regularities, okay. But since that's too jargoning I am going to start out with I am going to start with some facts. So the first one and this is due to Jeffery West and also to John Quigley, when you double the size of a city, productivity per worker goes up about 13 percent. Okay. This is the style of fact that you know puzzled economist and a lot of people for decades. Second one, when you increase the number of contributors to an academic paper or to a Broadway Musical which is very similar to a lot of the academic paper and if you combine people who actually don't know each other very well, which is also interesting, you end up increasing the quality of the outcome. What does that mean? For Broadway show they need ticket sales, for an academic paper that means citations, right, not raises, but that's hard to get. The next one is the first 10 Nobel prizes in Chemistry went to ten people. It went to people like Madam Curie, you know famous scientist. But the last 10 have gone to 27 different people and 15 other people have sued claiming they should have won the Nobel Prize. Right. So the fact is great they used to be they used to think of speak of great scientist and now these great scientists have been replaced by teams of you know, sort of above mediocre scientists, okay. The next fact is that companies like Hewlett-Packard, Intel and Google and Ford have turned to internal prediction markets which are proving just as accurate and far less costly than having exerts. So Hewlett-Packard pays economists lot of money, takes several thousand hours from the economist to figure out how many printers are we going to sell of this type. They found that an alternative mechanism they could use they could just sort pool the people who work for Hewlett-Packard and they could say, "How many you think it's going to sell?" They could average those predictions and those predictions are just as good as the economist, they are cheaper and they are a lot better looking, in almost every case. Okay. There is also some work by Alan Blinder, who used to run our money supply but he doesn't; he is back at Princeton and somebody who is on our school board in Princeton is now running the money supply, raised an experiment having teams of people manage money supplies in artificial economies versus individuals. And what he shows is that these teams systematically outperform individuals at managing these money supplies, okay. And the last from the institute of Rosabeth Moss Kanter is that corporations with identity diverse boards tend to outperform corporations that have more homogenous boards, okay. So the central lesson from all these things is that diversity is really beneficial. So one of the things I joke about in the book is that all of us have some one in our family who can double the network and find someone who has a bumper sticker on their car, it says, "My child is an honor student at, blank", right. We don't have bumper stickers on our car although we could that say, "My child is different", right. And we probably should have those because that's every bit is important that your child is different as it is that your child in some sense, is some over achiever. Now the idea that diversity is beneficial isn't surprising to biologists, right. This is one of the central tenets of biology. But in biology we tend to think, how did we get how did we get here? How do we sort of rise from the primordial ooze into interesting people who have iPods and you know, how did BeyoncÃƒÆ’Ã†â€™Ãƒâ€šÃ‚Â© coming from these slugs, right. And what biologists will tell you is look its diversity it's this mutation and crossover and sexual reproduction combined with selection operators, right. But at the core of this lies a lot of differentiation and some ability to separate out what's good and what's not good. So what I am going to do tonight is I want to describe what I call logic of diversity and the title of this book was originally called The Logic of Diversity, but then it turns out when I signed the contract, Princeton had naming rights. So they renamed the book "The Difference" which has been a problem for me because a band, when I was an undergrad in Harvard was called The Difference. So I always think of this band as opposed to my book. They also renamed one of my children, because they thought it would market better. So the question you might ask is, "Okay, why?" right, there is lots of things I could do. I can be country music singer, I can clog. There is all sorts of things that I could do with my life, right. Why would I decide to try and work on the logic of diversity? And the reason why is this. One of the things we do with Santa Fe is we try to bring math to metaphor, right. So we have metaphors like, two heads are better than one which makes us sound like diversity is good and then we have get the other metaphors to say too many cooks spoil the broth, which sound like diversity is bad. If we don't have a science what we have got is sort of competing metaphors with people shouting at one another and we can't separate things out. Almost all knowledge is conditional, if these is true then that is true. And the only way we can unpack that conditionality is in some sense by "doing the science", "doing the math". One of my advisors use to say, it takes a hard chair and a bright light figure this stuff out. So this logic is going to let us figure out the conditions under which the diversity is beneficial. It's not going to be the case that diversity is always beneficial. Think of diversity like friction, friction is lousy if you want to get a good gas mileage; friction is great if you want to stop the car, right. So what we want to think about diversity is just something that exist, I mean we want to understand when is it good, when is it not good. Okay, so how does these logic work? The book fundamentally has two steps, right. The first step is to basically say, how is it that we differ? When most people talk about diversity, when the talk about differences, they mean what's sociologist call or psychologist call identity differences, differences in how we look to one another. So I mean bald white middle aged male or something like that, right. What I am talking about is really to think of it's inside the pumpkin differences, right. So when you open up our heads, how is it that we think differently, so what's different inside our heads as opposed to what's different outside our bodies. And I would think about four quick things. One, it's going to be perspectives, which is sort of how you see things, another is going be heuristics which are the tricks you use to try and solve problems. A third will be interpretations which are categories we use, ways we have of sort of characterizing and parsing the world. And the fourth would be predictive models, which would be things we use to sort of trying to understand what we think will happen next. Once I got this idea about how people differ, then I am going to look at two particular tasks, because we could sit around and say, oh we differ, isn't that wonderful that we differ and we can sort of tolerate and appreciate, even celebrate our differences. But what I am going to be do is actually give a very pragmatic version of diversity. I am going to talk about, okay, so we have got these differences, how do they function, how do they work? Can they actually enable us to do things better? And I will talk about two very specific tasks. One is going to be solving problems, like coming up with renewable sort of energy or coming up with a good kind of ice cream, lets say and third will be second will be prediction, right. So it could be who is going to win the next election? It could be who is going to win the NCAA tournament; it could be just about anything, okay. So let me talk to the how does the people different? So what is the perspective? Now one thing that's really fun about doing this is, almost every corporation of America has a brochure that says, "We seek out people with diverse perspectives", right. And universities say this as well, "We have students to bring a variety of perspectives to our campus". But then if you ask them what that means the will it's circular, they say well, perspectives are perspectives. And they often say it in quotes, they will say perspectives, you know what I mean. But you don't know what they mean? Well, if you go on to a Computer Science Department, they will tell you what a perspective is is it's representation of the set of possible solutions. So what does that mean? This is going to be the terrible part; we are going to do some math. So in like eighth or ninth grade we would learn, most of us live in what we think it was a Cartesian world, right. SO if we have some points sitting there, we can think of the there is an X coordinate and a Y coordinate. So is the horizontal coordinate and a polar coordinate. I would walk over there I will walk over I am going to go off screen, so people watching this on tape, I am not disappearing. There is an X coordinate and a Y coordinate and that's the point. But alternatively someone could use polar coordinates right, when I was a math major, we used to say polar coordinates are cooler, right, than Cartesian coordinates. But at the same point you are going to give there being an angle, theta and a radius R. Now these are two ways to identify the exact same point right. So it's two different languages to uniquely identify this point. Steve Martin actually said this better. He said those French; they have a different word for everything, right. And don't worry about writing that down; it's actually in the book, okay. So what is the value of diverse perspectives? The value of diverse perspectives is they actually simplify problems, right. So if you read history of science what you will find is most great breakthroughs in science come from people with really bad hair and if you have been to the institute you will notice that there is hope, right. And if you look at Newton's theory of Planetary Motion and Mandalay's periodic table, what you will find is here is people who sort of rethought as well. Mandalay is the better example here. Mandalay have tried to think about that he knew there must be some structure to the elements, right. So he tried all sort of ways arranging the elements in order to find structure. Eventually he organized them by atomic weights, right. By just how much they weighed, right. Well, weight is sort of a silly way to organize the elements. It would be like organizing the animal kingdom by weights. So it would be like saying those alligators weigh almost as much as horses. So let's put them next to each other, right. Now, it makes sense because now you know about electrons and protons and neutrons. So atomic weight makes sense, but at the time it didn't, right. But once he did it, suddenly there was the all the structure that became evident. So great breakthroughs actually come not so much from having this huge engine you turn incredibly fast, even though some do, more often it's the case that someone sees the world in a way so that it suddenly becomes simple. But I am going to do something that's more down the earth. This is a game called Sum to 15 and it's a game developed by Herbert Simon who won a Nobel Prize in economics and he was famous for sort of talking about how people weren't as smart as economist said they were, which is pretty smart. So the set up of this game works as follows. And after this talk is over, if you want to go make money with this at home, feel free. So these cards numbered one through nine face up on a table. And players take turns alternating cards. And the object is just to hold exactly three cards that add up to 15, okay. One of the things I am worse at in the world is false modesty and so I am just going to flat out I say this, as a child I was the number one rank player in the world at Sum to 15, okay. And here is a match I played against Scott (Demarkey) in 1998, on St Patrick's Day. So here is the cards one through nine, and I went first and picked four. Now remember you want to get exactly three cards for Sum to 15. So Demarkey picks five which is an obvious choice in this game, right because that's right in the middle. I now pick six, now six is an odd choice because four plus five plus six adds up to 15. I seem to have "stacked" the deck against myself. But what I am trying to do is confuse Demarkey so he doesn't know what I am doing, right. Demarkey then picks eight which is a fatal error and here is why, because I can then pick two. Now what does this do? Four plus two plus nine is fifteen, so if I pick the nine so if the next period, if I pick the nine I am going to win. But six plus two plus seven is fifteen, so I could pick the seven next period and win. Demarkey has five plus eight which adds up to thirteen, but the only card he could pick to let him win would be the two. So I have got it, right, I win Demarkey concedes, right, and I retain my championship, okay. Let me remind you of something from probably eighth grade math. There is something called the magic square. And in a magic square you put the cards, you put numbers one through nine on a grid like this and in every row and every column and every diagonals sums up to 15. So eight plus three plus four sums up to 15, four plus nine plus two sums up to 15 and if I do the diagonals six plus five plus four and eight plus five plus two also adds up to 15, okay. Let me show you my match with Demarkey again, okay. Remember I picked the four, he picked the five, I picked the six. So I take the four, Demarkey takes the five, I take the six, Demarkey takes the eight, I take the two. Now have I got him either way. It turns out Sum to 15 is just tic-tac-toe, that's why I was the best in the world at a very young age, right. But the thing about Sum to 15 is it's actually a difficult game and if you play Sum the 15 like of a a number of people play this in front of crowds like this and you tell one person it's tic-tac-toe and you give them the magic square, the other person is doing his incredibly complicated calculations in their head and losing for about eight or nine times until you eventually figure out this very sophisticated set of rules that turns out to be things like, take the middle, right. Try and get two things in the corner. So the point I want to draw from this is that there is something that you can call the Savant Existence theorem, which for any problem there is going to be some perspective to makes it easy. So any problem that sits out there, there is going to be some way to transform what's a Sum to 15 problem that seems fairly complicated into tic-tac-toe which is very easy. And so what we do in science a lot of times that is trying to figure out what way can I look at this? How do I conceive this problem? How do I think of it? So that I can turn this problem into some thing simple and this is why diverse perspectives as such a useful tool, okay. The next thing I want to talk about now is heuristics. So heuristics are tricks we use, little tools we use to try and solve problems. One of the most famous heuristics that was described in the book "Why Not" by Nalebuff and Ayres is called "do the opposite". So everybody has been doing one thing try do the opposite and see if that work. This was immortalized in a Seinfeld episode where George realizes, every decision he has made in his life has been the exact wrong decision. So what ever his brain tells him to do he just does the opposite. And by the end of the show he is the Assistant General Manager of the New York Yankees, right, which is great. Now one of the things I am pushing here one and I will talk about this much of the end of the talk is that we tend to think of people is having an intelligence level, so like an IQ, somewhere between you know 80 and 280 or some thing like that, right. What I am pushing is not to think of people as having some little you know, ability level that can be measured on a single scale. But in stead to think of people as being tool boxes of skills, to think of peoples mind as being filled with ways of representing problems, perspectives, heuristics, tools they have from solving them as well as these interpretations and these predictctive models. But let me show you how heuristics work. So here is some IQ test questions, so in doing research for this book I took a bunch of IQ tests to sort of see how they worked. And these are some of my favorite questions from the test and I felt that my IQ is in this these tests are really impressive. So I took probably 20 tests, and my IQ lies in this thin range between 87 and 246, okay. It depends on what I had for breakfast I think, okay. So here is one thing they will ask on this test. They will say fill in the blank, one, two, three, five, blank 13, right. And this is the famous series called the Fibonacci series and the answer there is eight and the trick is it's differences, thirteen minus eight is five, eight minus five is three, five minus three is two and three minus two is one, okay. All these exams they will often ask the following question just that you see how the questions work, so it's one, four, blank, 16, 25, 36 and the answer here is that its squares, right. So this is a very easy one, each one gets one squared, two squared, three squared, four squared, five squared, six squared. Here is the hardest one, right. So this is what when I was at Caltech we would say this separates the Caltech students from the MIT students. One of my friends who went to Harvard said, oh this is easy; these are the years that Boston won the playoffs. Now if you look at this, this is flat out hard, right. This is a flat out a hard question and we can all laugh and think, I think I have got it, I think it's 14 or something right, but the answer is it's 42. This is also sort of a culturally biased question because its Douglas Adams Hitchhiker's Guide to Galaxy has a universal ultimate question and the answer of that is 42. It also happens to be answer of this one. But where does the 42 comes from? It turns out, it's differences of squares. So six minus two is two squared, two minus one is one squared, 42 minus six is six squared, and 1806 minus 42 is 42 squared. Why do I bring this up, because differences of squares is just a combination of the first two heuristics. The first one which we could all do was subtract. The second one like we could all do is square. The third one just subtract and square, okay. One of the main points of the book, in fact may be the main point of the first part of the book, is the fact that when people talk about diversity they tend to use an insurance model, a portfolio model, they say, you want diverse people for the same reason you want diverse stocks, because you want people that pay off in different states of the world. It's as though there is a problem where we suddenly need the bald white guy to solve it, so we trot him out, right. I think the way to think about it instead is to think about it in terms of people having there being super additive benefits to skills, right. So we saw if we have one heuristic which is differences and another heuristic which is squared we get for free a third heuristic which is differences of the squares. So you literally get when it comes down to these heuristics we have in our heads, the super additivity. You get that one plus one literally does equal three and that three is the actual benefit from diversity. And this is why; again we want to write down sort of a logic, okay. So again you get the super additivity, the benefits of diverse heuristics shouldn't be confused with these portfolio insurance effects, its actual super additivity, okay. Now here is sort of the most fun part, these are interpretations and interpretations is a bit of a misnomer, psychologists tend to call these categorizations. But these are meaningful structured categories we use to make sense of the world. Because one psychologist, a friend of mine told me, we lump to live. So what we do is experience and we have things we see, we put them in categories. We will say that is a truck or this is a tall person or something like that. If each event that we experienced we treat it as an individual idiosyncratic event, we would basically slobber, right. Because we couldn't in for any causality, we would have no idea what's going on? So just basic statements we make, like Joe Biden is a liberal, someone is a Soccer Mom, my brother one time called me and said, oh my gosh I am NASCAR dad and we talked about it, he is okay with it. People will look at stocks and they will say they will quantify them by price earnings ratio. Artists of modern arts, ska music, you might be walking down the street and see there is a band playing ska music and just go in because you know as all of us know, ska is good. So you go, right. And so what I want to do is I am going to talk about the following sort of thing that cultural anthropologists do. And this is called the pile sort and this is worked by William Batchelder who is at Irvine, they will people list of food items like this. Or things in the environment and they will ask them to put them into piles. And so these piles are the categories that we use to make sense of world. So here is a list of things you might find, like fresh salmon, camp de sole, spam, canned beans and so on, right. So the particular cultural group that I belong to, David Brooks referred to as Bobos which are sort of bourgeois or Bohemian something or so, right. So we would put these in three categories vegetables, fish and meat, and then canned stuff, right. Canned stuff that we would never eat, sort of thing, right. Couple of weeks ago when Vice President Dick Cheney went over to Pakistan, there was some problems with Air Force Two and so they send him in a cargo plane, right. And but they just couldn't put the Vice President in a cargo plane, so they loaded an air stream trailer inside the cargo plane and then strapped it down and Cheney to his credit, according to rumor any way insisted on wearing cowboy boots and drinking whisky if he was going to be in an air stream trailer, which I think you have to give him credit. So you can think of an air stream sort of these same things and they can say well, there is some vegetables there, there is some fish and meat, right. And notice they have fresh salmon and canned salmon in the same category here and then there is weird stuff, the people from Santa Fe eat, right. It goes like wherever they like, arugula, fennel ahi tuna, in fact for dinner, ironically, tonight I had Ahi Tuna and Arugula which would all go in the weird stuff category here, right. Now the point of these is that people categorize the world differently. And these different categorizations we use imply that when we make predictions about what's going to unfold, they were going to be predict things differently, okay. So these predictive models that people have make inferences from the categories that we create. So my older son when he was three said, red cars go fast, right. So he had actually categorized cars by color and he figured out that red ones go faster. I have been told by someone from the insurances associate system in America that this is in fact true. Right, that red cars do go fast but there is better categorizations than color, okay. Now the point of this is that diversity begets diversity in some sense. If we lump differently, if we put things in the different categories then we predict differently, right. So one person might say Amazon is a delivery company, so when Amazon had a public offering some people might have said, look Amazon is just UPS with a warehouse or something. Someone else might have said Amazon is a high tech firm. The first person is still working in Starbucks, right and second person has lots of money, right. How you categorize things has big implications for what you think its value is or what you think is likely to unfold, okay. So that's how we differ, we differ in how we represent the world, we differ in the heuristics we have, we differ in how we lump things, we differ in how we predict things. The next question is are these differences good, are these differences bad, are they indifferent, right? How does it play out? So what I want to do is I am going to talk about why is that the case that diversity actually improves stuff. So first when we talked about problem solving and this will actually could be fairly brief. So when I was working in Caltech, at one point I was constructing a model, I was trying to figure out what skills the people acquire in an economy. At that time I was what they called a mainstream economist, now I am considered a Santa Fe economist, which is not to be confused with a mainstream economist. And one of the things I did, is I created this little artificial economy on my computer and I created how much of problem solvers. So a bunch of people that the economy consisted of a bunch of problems and these people went out and tried to solve these problems and if they solved these problems, they got money, right, little computer money. What I decided to do is rank these agents by how smart they were, basically how well they were doing at solving these problems. Now all these agents had to be smart, right. So I wasn't filling in dumb agents. Every agent had some heuristics and some perspectives and stuff like that and I was just ranking them. And to test for sort of errors in the code I did the following experiment, I created group one which I call the Halberstam group which was the best and the brightest, right, so the 20 best agents in my little artificial economy. Then I created a second group that was 20 random agents. All these agents are smart but they are random. Then I had each group work collectively to try and solve these problems in the economy. So when one agent in a group would get stuck, another agent would come in and try and help them. Right, so these agents were little computer programs and one of them sort of using the using the Santa Fe metaphor got stuck on the peak of their little rugged landscape, another one would come in and find a better peak on the landscape. And when everyone was stuck on the same peak they would stop, okay. Well what happened in this experiment is if you look at this terms of IQ, like this first group that the group of best were all these big IQ people. The second group the diverse group had some big IQ people and they had some people like the size of the thing they represents or the pumpkin represents their score they had some people who were sort of pin headed people like this nine - right, who wasn't very good at solving the problems, okay. So here is what happened. The diverse group almost always outperforms the group of the best. So this diverse group is almost always out performing the alpha group. Now when I say almost always, right what happened here is when you get a result like this its so counter intuitive, you got to "do the math" and in this case the math got hard enough that I brought in my friend (Lu Hong) who helped me do the math. Right, so Lu and I basically proved a theorem like it's a set of conditions and I talk about these in the book, under which this will be always be true. And the conditions tend to be, the problem has to be hard, right. The people have to be diverse and the people have to be smart, right. And the groups you have to choose have to be reasonably large. So if you have groups of size one then obviously the best person does better than a random person. But once you get to groups of size 20 or 30 the result always holds. Well here is why that's true. If you think of the alpha Group, let's now sort of picking the people in terms of IQ's, let's pick them with having tool boxes. So the alpha group let's suppose that the best tools for this problem are tools that were sort of early in the alphabet, like 2A, 2B, 2C. So the alpha group is going to have all these tools, everybody is going to have 2A, 2B, 2C or may be some D's and E's and stuff like that. But there is tremendous overlap in their tools. This diverse group, yeah, there are some A's and B's and C's and D's in there but there is also, if you walk way over here, right, and this just gets some I met this wonderful woman from Illinois today and I felt terrible, right. This I am really sorry. Yeah, this Illinois person over here who, in principle, I mean who looks sort of pinheaded on this, right, actually he is bringing his I tool and his L tool that nobody in the alpha group has. So even now on his or her own that person isn't very useful, in the context of the group because they've got a different tool; they actually end up being beneficial. Okay. So let me turn to prediction and which is the second task and here you will have a little bit of fun. In Jim Surowiecki's book "The Wisdom of Crowds", he begins with this amazing anecdote. And he is it's the 1906 West of England Fat Stock and Poultry Exhibition, beautifully titled events. You know they haven't thought of Lollapalooza yet or something like that. So this is what they call it. People had to guess the weight of a steer and the average guess was within one pound of the weight of that steer. So that steer weighed 1198 pounds 1197 pounds, the average guess was 1198 pounds, okay. Surowiecki finds this amazing, all right. I used to own cattle and guessing the weight of that steer isn't that hard. If you live in New York and you like to hear what Surowiecki totally impressed. If you live in Iowa cows are just big people, sort of more interesting skin tones, right. So it's not that amazing that they were with in 50 pounds, 60 pounds, stuff like that. But it's incredible that they were within a pound. So what I am going to do is trying to understand why is it, that these groups of peoples could be so accurate in these things. And with that we want to try and understand that why is that the case companies like Google which they sort of if this punish way of referring to their group called profit, which is that they sort they sort of poll their employees, use their employees to predict what's going to happen. So would HP does, Intel does, and that sort of stuff, okay. So, let me do something incredibly important. The NFL draft, see how this works. So this is the 2005 NFL draft and here is the first six players picked in the draft and here is six experts or seven experts and I just coded them A, B, C, D, E, F, G. These are people from like ESPN, Sports Illustrated, places like that. And they only predictions about where these people would be drafted. So all everybody picked Alex Smith number one, Ronny Brown some people had him two, some people had him four and the crowd's guess so the average guess for Ronny Brown was 2.7, right. As we can see, each of these individuals guesses and then you can see the crowd's guess. Well we could ask, because we could say, okay, how do these "experts" do versus the crowd of experts, which does better? Well the way you measure how well someone is you sort of say, how far off they are from reality. But then you know I will talk about this in a second, we actually have to square those differences, because if I am positive one and the negative one, I don't want to have those things cancelled out. So we used this technique squared errors. The bigger the squared error the worse you were. So here is the squared errors for each of those people and here is the squared error of the crowd. Now, (Dan Katelan) who was an undergraduate in Commerce at college, spent a summer working for me where he did like 15 NFL drafts and NBA drafts. And this happens every time, right. I think at one instance the crowd took second. But every other time the crowd took first. And when the crowd takes second you don't know who's going to take first priority, right. You couldn't tell for sure, "Oh, I knew person C would be the best". So what's amazing here is the crowd is better than anybody in it, okay. Let me do a very specific example sort of slowly to explain how this works. So I had another undergraduate Michael Forest go in and then look at NOAA and then weather.com, and look at at their predictions of sort of the low temperature in different cities over for a particular day, right. So, NOAA had New York at 12, Chicago at 6, L.A. at 40, weather.com had 10, 14 and 46. And so these are predictions about what the temperature the low temperature would be. So we could then take the average of those. So we can think of this average as sort of being what the crowd is predicting, right. So the average of 12 and 10 is 11 and the average 6 and 14 is 10 and the average of 40 and 46 is 43. Now the actual temperatures are quite different. But I want to remind you this is weather prediction, it's not science, okay. So they are off by quite a bit, that's okay, we all make mistakes, move on. How do we find how well they do? So what we are going to do is we are going to take NOAA's prediction of 16, the actual temperature is 18, what we do is we take the sixteen minus the eighteen and square it, so it becomes instead of just two, a negative two, it becomes a positive four. When I used to teach statistics the way I would explains this is as follows. If you are doing archery and one arrow goes low and the other arrow goes high you cannot then therefore say, bulls eye, right. What you have to do is you go to say, that was off by this much and it was off by this much, again. Again a way to turn everything into positive is just to square them. So NOAA's squared error, it's a 105, if we take web.com squared error it's a 117, if you take the average it's 111, if you take the crowd it's 77, okay. So again the crowd is better than anybody in it. This wont always be true but it's true in these two examples and two is data, right, for here is what's totally fascinating. If I look at how different they are from each other, so I think how far is NOAA's predictions from the crowd's difference predictions. So if we go back to the predictions, right, NOAA's prediction is sixteen and the crowd's is thirteen, so they differ by three, right. So sixteen minus thirteen is three and if I square that I get nine. So if I just take their squared difference, from the crowd's prediction both of them differ from the crowd by 34, okay. Why 34, it's not like I am a big Walter Payton fan or something, the reason this is interesting is that 77 happens to equal 111 minus 34. So the crowd's error equals their average minus their diversity in some sense, their diversity of predictions. Well that's pretty funky that that works here right, that's pretty cool. But again I set this up, right, so shouldn't be surprising Let's check the other example. In this example the average error of the people on the the NFL predictors is a 137.3, their diversity is a 102.9 and if I subtract those I get 34.4, okay. So again I get the average minus the diversity equals the crowd error. This is always true. It's so true that three different people proved it; a statistician proved that, computer scientist proved it and an economist proved it, all using different notations, right. So none of them knew the other proved it. All three got 10 year typically, okay. I want to pause on this slide because this slide is incredibly interesting, in American society especially but I think in some sense to globally, we really prize ability, right so I started out talking about these sort of bumper stickers that say, my child is an honor student. The proxy for ability in this equation is average error, that's how smart the people are. The proxy for sort of thinking different in diversity which we tend to think of it on political terms is that diversity term. But if I look at this equation I want to say how well the society do it, figuring things out, how well the markets figure out prices, how all the democracies figure who they are going to vote for. That's the crowd error, right. So in this weird sense our collective ability is equal parts, individual ability and collective difference, right. It's not that it's 5000 times ability, right, plus a tiny, tiny, tiny little bit diversity. It's not like ability brings the cake and diversity sort of multi colored sprinkles on the top that makes it look sort of cute, right. They are equal partners in this thing. And again this isn't a set up, this is math okay. So remember where does this comes from? Where does this diversity come from? The diversity comes from the fact that the people have different interpretations, these different pile sorts, if the bow-bow and the air stream person see the world differently, they make different predictions and if they make different predictions that's a good thing, okay. Let's go back just to put a bow on this thing, in Galton's steer example the average squared error was 2956, the diversity was 2955.4, the crowd error was 0.6, right. So they were smart but they were diverse, okay. So these theorems and so here is the funny thing I am a mathematical economist by training, right, I am not a Civil Rights activist or something, right, although I am quickly becoming one. What's interesting about this is that these theorems, that when solving problems there exist a set of conditions under which diversity can trump ability and that when making predictions diversity matters just as much as ability are not political statements, right. There is no intentions for these things to be political statements, what they are is just mathematical truth, they are just math, like the Pythagorean Theorem, right. But the power of math is that it helps us by sort of formally defining the terms, right, let's just understand how these processes work and how things aggregate and what we see here is that diversity is a very powerful thing. Okay, so I am going to finish with six points. And these six points I am going to go as follows. First one is the pumpkin has limited space. When you hang out at a place like Santa Fe and especially and if you ever get chance to talk to (Murray) you realize the pumpkin fear the pumpkin, right. The pumpkin is incredibly powerful; it's amazing what the pumpkin can do. But it's still the case that we can only fit so much in there, right. And every time I listen to the hip hop I am like the normalization group theory is getting crowded out, right, just a little bit. The second point is I think it's very, very important that we think in terms of tool boxes not IQs and SAT scores. What's very intriguing is when you think about letting students into colleges, we think about grade point average and SAT score, that that's so important. We have never looked that a faculty hired, any place I have been in, at Caltech, at Michigan, at Iowa, we never looked at anyone's IQ or SAT score when making a faculty hiring position. At that point all we are looking is what tools do they have, right. So we are really looking at the tool box and not at sort of, sometimes IQ. IQ may be a crude measure for how many tools we can put in, right. But what we really care about is what tools people the selected and are those tools going to be useful for the rest of this? Which leads at the third point is even though the pumpkin is constrained, collective diversity is nearly unlimited, right. There is a lot of us and we are all different and that can be good, right. And what we need to do is sort of recognize that we can leverage our differences to be more productive. The fourth point is this that, it really is I think diversity that drives innovation, right. So this is notice the cow motif here. So I remember learning I think as an undergrad that the word vaccine comes from the French or Spanish of "vaca", right which means cow and the reason why is when the smallpox was sort of ravaging Europe, someone figured out that milkmaids weren't getting smallpox. Now this requires someone having milkmaids in their own category. As my yoga teacher used to say, label it a thought, put it away, okay, just let that go. Someone had milkmaids in their own category, realizes they are not getting smallpox, right. And then figures out, may be they have got some interims that they had the other disease called cowpox that naturally vaccinated them, right, from getting smallpox and this is where the origins of vaccine is. So somebody putting milkmaids in their own category in some sense leads us some great breakthrough. It's an oversimplification, but I am just trying to make the point. The fifth one is and this is may be the most important, is we have to learn to ride the bike. And what do I mean by this? In the book I talk about something like the what I call the parable of the bicycles. One of the places I work is called the Institute for Social Research and there the people actually sort of do empirical social science research. So one of the things that they might like to do is, they might like to figure out which is a better mode of transportation, running or riding a bike? So let's suppose I go to some town, I round up all the five and six year olds and I have them all run as far as they can in 10 seconds. Law of modern statistics is right, what I should get is that I should get a nice bell curve with a mean of about 50 meters. But that distribution should be pretty tight. Some kids may run 70 meters, some may run only 30 but it's going to be pretty tight around 50 meters. Alternatively, I can take those same kids and put them on bikes and ask how far can they go in ten seconds on a bike. Now some of the kids, I may walk off the frame again, who can ride bikes are probably going to do like a 150 meters. Others, right are going to be wipe out, right. This experiment hasn't been done because you couldn't get it past human subjects because of the fact that the kids would wipe out, right. But if you do these experiment, you think okay, if I just look at the means I would say well biking isn't necessarily better than running, right, they look like they are about the same. And if I thought variance was a bad thing I would say, well you know actually running may be better than biking because its we know what's going to happen. Why do I bring this up? If you look at the literature, there is a massive empirical literature on identity diverse groups and functional diverse groups, right, in terms of how they perform. If you look at if you put together sort of all the studies on identity diverse groups what you find is that this is what the homogenous groups look like and that's what the identity diverse groups look like. Identity diverse groups on average do about the same as homogenous groups. But the best groups and the worst groups are identity diverse groups. So that leads a lot of people to say look, there is no benefit from identity diversity, right; there is no benefit for bringing people together or from functional diversities as well. There is no benefit from doing it because the means are approximately the same, right. But what I want to say is you could look at this in another way, you could look at this and say, may be we should learn to ride the bike, right. So if we learn to ride the bike we can walk way down the tail, right. And that's where we want to be, as we want to be way down the right end of that tail, okay. The sixth point is this, where you keep your ketchup isn't essential okay. This isn't a philosophical Zen like moment. What do I mean here? Essentialism is the belief that somehow if we look different on the outside, if we have got sort of genetic differences that those translate in some essential way, to behavioral differences. So that when we look across cultures, we look across groups even within a society that those differences are some how essential. So one of the questions I ask people all the time is where do you keep your ketchups? So how many people keep their ketchup in the fridge, okay. How many people keep it in the cupboard, okay. So the people keep it in the cupboard, a lot of them are probably from England, right, or New Zealand, I see some people nodding their heads, a lot of African Americans especially from the south tend to keep their ketchup in the cupboard, right, it's a general rule. Essentialism would say this that we are going to find some little protein there, there is going to be a little ketchup in the cupboard gene, right. Now we laugh at this and we should laugh at this but I guess, when you meet someone who stores their ketchup in a different place you think they are freak and weird, right. I gave a talk when some guys said he got it up on the doctor and you can't keep ketchup in the cupboard because it has vinegar everybody is like nodding their heads go on. And I said, where do you keep your vinegar? Right. All right and the last point is that I was going to end sort of a big, happy SFI one of the things and I think John hinted at this one of things that we try to do to at SFI is we try and bring in people who sort of have the "big academic resumes" that have the big ability thing. But what we were also trying to do is bring in people who are different, right and try and create these reproductive synergies across disciplines. And one of the reasons why is when I talk about these formal ways of thinking, these perspectives, right, these ways of representing problems with these heuristics, like these tricks we have like, here is a way to approximate this function, right. Here is the little trick I used to make you know, random numbers on my computer, that's sort of stuff. It's these diversities in tool boxes that all the different scientists at SFI bring together that allows sort of collectively sort of hold our own against (Murray) right And that's why SFI is such an amazing place. If you look at sort I sort of in the book I sort of end on this, is that a lot of modern universities are sort of balkanized, it's like little silos and no one ever leaves those silos and there is places like the Santa Fe Institute, the Center for Advanced Studies at Stanford, they really try to bring in people right, from across disciplines and allow them to share their tools because if we can share those tools and if we can do this at society level as well then we will find is that we can not only leverage our ability but we can leverage our diversity to everyone's benefit. Thank you very, very much.