Speaker 1:First of all, well I think we're going to have a question and answering period here at the end. I do lots of different things, so you can ask me lots of different questions. I'm associated with several universities, I'm on the Sustainable Development Global Partnership Board to help do the sustainable development goals, advise the American Bar Association, that should make you scared, things like that, and I have a couple books that are I think relevant here. One is honest signals, which is about individual expression, and particularly dyadic interaction, one-on-one interaction.
Then social physics is the larger version of that, which is essentially how culture develops, and how you can understand the social fabric. The phrase social physics people go, "What's that?" As I was just explaining, the phrase is actually about two centuries old. During the French Revolution, at the time when Natural Science was becoming Physics, and Alchemy was becoming Chemistry, there was a dream of using statistics to better understand the evolution of society. You have to remember that the French were going from a monarchy, to something that looked democratic, to the terror, to Napoleon.
So evolution of society was a very salient thing to be interested in, and that spread to England. Actually, the work in that is where the modern census comes from, and why we have Census around the world, but it had a problem. The problem was that it's very expensive to collect this data, and if you know about the US Census, you know that it's this huge effort, very expensive. When you get limited data it's also data that's biased in various ways, even though they try very hard.
Until just recently, they didn't have a lot of sophisticated statistics to analyze it. It was pretty basic stuff in many ways, but in the last decade, we've seen an explosion of digital data. Everybody has cell phones, that means somebody has data about where you are every minute of the day, who you call, who you call frequently, what you buy through credit cards, all the transactional data from your employment, and other places. The data is there, hopefully it's managed correctly, and not like some of the famous examples in the news right now.
Then there's been this revolution in statistics known as Machine Learning, or AI, which are really just fancy statistical methods, and methods of optimization. If you put those two together, you can really begin doing things that are very surprising. As someone who is trained at one point as a psychologist, I like to point out that most of the things that we know about humans, are based on very little data. The joke is, they're based on freshmen surveys in American universities primarily, but it's rare to have datasets that are very large and longitudinal.
Today, you can get things that are not only large and longitudinal, I'll show you several country scale things, but they're also very fine-grained. It's minute by minute, second by second, for millions of people for years, it's incredible. If you analyze that right, which requires combining the theories of psychology, together with the techniques of data analysis, you can get really interesting insights. I put together a little high-level Cruz over that, of the things that I've done, and I tried to pick things that will be surprising, and perhaps controversial, so that you will go home and think about them, and prove me wrong, or whatever.
So for me, lets see if we can get this to work, yes. I'm interested in how technology interacts with humans, and a lot of the terms of wearable computing, things like Google Glass came out of my laboratory. In fact the term Augmented Reality, came out of my lab apparently. What that meant, I was very interested and be quite familiar with what happens when you put sensors, and computation on the human body. This is an example of some people in my lab, they're wearing a little device that listens the tone of voice of people, who you're facing, where you are, who else is around, some body language, and records that with a timestamp, but it does that 32 times a second.
You have very fine-grained data about people, really for the first time, and we've done a whole variety of things like that and this is early on. Today what you see, is you see little bits of software that get in telephones, and you can do much the same thing. You're all carrying around a computer that has something like 16 sensors for position, for noise, for all sorts of things in it, and you can sample those things with people's permission, to be able to get a picture of essentially arbitrarily large numbers of groups of people. So we began doing this in the late 1990s, before there were cell phones really, and we began to get really interesting data out of that.
We wrote a paper about Computational Social Science, that was something that we named and since then, many of your universities, or many universities have set up Computational Social Science departments. I like this picture, this was from nature. One of the preeminent journals in the world, because the thing we found is, when you measured these things, you got things that were not dominated by cognitive characteristics. It looked a lot more like Jane Goodall, looking at apes, and I like that, tough. Okay.
That's where we come from, and what it does is, it gives a depth to the cognitive interpretations that is often missing. So the nature article for instance, at raising social science up to a new level, which is something I'd like to think that we're doing, and as I said in our science paper in 2010, we made the point that a field is emerging that leverages this capacity to collect, and analyze data at scale. It doesn't mean it replaces traditional social science, but it provides an orthogonal, and much larger in some ways, window on the things that we've been studying for a century, or more. I'll show you some of those things.
For instance when we put little badges on people, some of the properties that are very easy to measure from a signal processing point of view, are things that are closely related to brain properties. I used to do neurophysiology at one point in my career, I've done lots of things, so this is very natural for me. The oldest part of your nervous system is your autonomic system, your fight of flight system, and one thing you see is when you're alerted to something, or afraid of something, your behavior changes. You become quicker to move, your actions are more jerky, you do more actions, if you're talking you talk more frequently, and faster.
It's one of these things that's somewhat controllable, but largely an unconscious response. I can tell you interesting things about it, like for instance, we do this without even thinking, we judge people's autonomic nervous system without thinking about it. So for instance, if you think about a little dog, what happens when a dog gets excited? Well it barks, and the tail goes, right? That's how primitive this signal is, is that the activity level, the arousal of a dog is something that is instantly an effort at least perceivable by every human, and that's true for most animals.
Now as you get older, you can begin to suppress it and control it. For instance, we looked at a bunch of people who were semi-pro poker players, so in poker they have this thing called Tells, which is ways that people reveal their hand. One of the big ones is activity level, people with a good hand get very excited, and they talk more and stuff. A pro will shut themselves down, when they get a good hand, but it turns out people are not very good at estimating their arousal level, and they overdo it. When we did an experiment with semi-pro poker players, we could tell when they had a good hand, because they over suppressed their activity level, relative to normal people. Just to give you a sense.
The thing I have labeled here is influence, comes from your attentional system. One of the things that's characteristic of conversation is, it's almost like a dance between two people, if you have a two-person conversation. I stop speaking and you immediately jump in, and if you think about that for a second, you just notice there's something really weird. You've all seen experiments, or heard about experiments where the red light goes on, and you push the button. It typically takes 2 to 400 milliseconds for that light to hit your eyeball, go through whatever and turn into a push.
We regularly and effortlessly exchange who's holding the floor with a millisecond, or two delay. We do this all the time, and that's impossible except if you are in some way, modeling in your head what the other person is doing, and anticipating when they're going to stop speaking, and we're really good at it. What you can do is, you can measure small little delays and variations in that, to be able to tell who's most interested in this, who's pushing the conversation. The big version of it is you've all been on a call where somebody says, "What do you think John?" John doesn't say anything for like a half a second and you go, "[inaudible 00:11:37] John's reading his email again." Right? It only takes like a half a second, half that, half a second seems like an eternity to us, because we are so attuned to that.
Mimicry is another one, so you have these mirror neuron complexes in the human brain. It's supposed to be only humans, and maybe apes that have these sorts of things, but they're extremely effective. So this is the thing when you're within a fairly short distance from each other, and it only works actually over short distances, so you have to be able to almost touch the other person. If I nod my head, there'll be a high probability that you'll nod your head, if we're actually connecting. Doesn't mean you agree, just that we're connecting.
A friend of mine did an experiment, where he had a little computer graphic that was trying to sell you something. Tried to get you to behave in a particular way, and in one condition, he had a camera that watched the people, and if they nodded their head then the computer graphic character would nod its head four seconds later. People didn't know that that was happening, they didn't notice that they were being mimicked, but it was 30% more effective at selling, it was rated higher on trust, rated higher on empathy. This is massive effect, from this thing which people were unconscious about.
Consistency has to do with smoothness of performance, complexity of performance. Again something you can measure without any of these, are all without reference to content. All of these are the same things you can see in animals, and in fact we generally can read them effortlessly in animals, and use these same effects, they're very old. So a hypothesis that I investigated was, do we have a channel of communication, that's parallel to the linguistic structure that we have. When people talk about prosody, people talk about little bits of speech like, mm-hmm, that sort of thing, they often talk about those as being related to the linguistic structure.
When I see these things and signals, what I'm really seeing is something that seems to be much older, fatally genetically, much older and evolution works by adding things to existing structures. It doesn't work by getting rid of the older structures, so what I seem to be seeing there is that there's this signaling behavior, the same signaling behavior you would see in other animals, and on top of that you have a linguistic structure. That might imply that language evolved as an elaboration of signaling, rather than the signaling being something that's adjunct to the language.
Notice that it's not emotion. I've made no reference to emotion, no arousal is often seen as being part of emotion, and that's fine, but things like mimicry are not necessarily emotional at all. Attention, is that, I don't know, you decide that's emotion, I don't really care. I don't have any dog in that fight but, it's this other thing which is, we've evolved along with all sorts of animals for communication. You can use this in various ways, so we've done some interesting demos, experiments of for instance people, mid-career people negotiating salaries in a mock exam, where their grade, and a course depended on how well they did on this negotiation. You can predict the outcome of the negotiation with great accuracy, by just looking at the signaling behavior at the beginning.
We went to a bar, where they were doing speed-dating, you can predict when they're going to exchange information based on the signaling behavior, independent of who the people are, or what they say to each other. Indeed, it's the woman signaling that matters a lot more than the man's signaling, or this is not the great picture, but we looked at mid-career people pitching real business plans. These are people in the business school, who are about to graduate, this is what they were going to do after they graduated. They were rated by the other people who are mid-career people, and we could do as well, at rating the business plan as these mid-career executives did, without listening to the words.
That's scary, and what we were doing basically, there was measuring this fluency, so how well were they producing their pitch when they were pitching their business plan, and how excited were they. What was their arousal level, and it's as if the people, the humans, were saying, "Well, I don't know much about it, but they seem excited, and they seem what they are talking about, know what they're talking about, so it must be okay." That simple rule, predicted the ratings of the business plans. Since we did this, it's turned into a commercial product, some of the students went off and commercialized it.
They do things in call centers, where they're trying to help people avoid getting into fights on the telephone, and that of course makes Custer's more satisfied, but the interesting part is, it also makes the call center, or employees a lot happier. Right, because it's really bad to get into fights with people time, and time, and time again. Pretty interesting. That view of things is where on that I go for the next two or three examples. It's a view that's not a cognitive science view, it's a view that we evolved as a social species, to be able to harvest information in our environment, and come to a consensus about what to do, and that's a major factor in our fitness, and why we're here today.
That's the underlying substrate, on which our society is built. That's not the only factor that matters, but that's the foundation, and that foundation would also apply further to ape troops. If you take that view on things, and you'd say well look if you're going to decide what to do, having a lot of different information is probably a good thing to do. So diversity of contributions, would almost certainly dominate group performance, and there are mathematical theorems and signal processing and things like that, that say this.
This is diversity of ideas incidentally, not diversity of features of the individuals. We had a paper in science that found that, you probably maybe you're familiar with it, Anita Woolley, myself Tom Malone, and we put badges on groups, and had them go through a whole series of tests, and develop something that was analogous to an IQ test, but for groups. A group IQ test, and we found that these objective features of the group, explained about 40% of the variance, about the same as IQ, about as strong as an effect across the different groups.
What was interesting to me, and part because they're my badges and so forth, was that these signaling things, and in particular they're just the equality of contribution to the group discussion, did everybody contribute to it, was the dominant factor. It was three and four times more powerful than things like personality, things like feelings of identity with a group, gender, on, and on, and on. The very simple thing that's consistent with our performance in a group, having its evolutionary roots in a pre linguistic period, where you want everybody to sit around and express interest in different options, and maybe have very limited linguistic competence.
We've done a lot more with this over the years. One of the interesting ones, I thought I'd drop on you here is, we recently did one with what I called Sloan Fellows, at MIT. These are mid-career people who come back for a one-year program, and they do all their work in groups like this, and so we badged up all of the groups. They do a lot to make sure that the groups are diverse and so forth, but one of the big problems they continually have, and many of you will relate to this, is the women continually complain that they're not let into the conversation, they don't get enough air time. How many people have had that experience? Okay, right.