David Desteno - The truth about trust - Talks at google

MING: Good afternoon, my friends. My mane is Ming, and I'm delighted today to introduce my friend the renowned social psychologist David DeSteno and the author of this book, "The Truth About Trust." I first got to know David-- thank you, Dana. I first got to know David through his highly innovative work in studying the science of compassion, which is a topic I'm very passionate about. David and his lab, they are renowned for devising very creative methods in studying how emotional states affect behavior. And they are also known for studying moral behavior in real time. Real time, not fake time like the other labs, real time. David is most interested to figure out how to foster prosocial behavior all around the world. And he told me that in summary his work can be summarised in three words, vice and virtue. Vice and virtue, there's one that I prefer over the other. And with that, my friends, let's please welcome my friend David DeSteno. DAVID DESTENO: And thank you for having me. It's a real honor to be with you here and share this work with you. And as you can probably guess by the title today, I'm going to talk about trust. I think it probably comes as no surprise to you that issues and dilemmas of trust pervade our lives. Trust determines who we want to work with, who we love and who would marry, who we trust to learn from, who we'll go to for support. Now, we all can remember the big stuff, the times trust really matters. Is a new business partner going to be trustworthy, or is he going to skim profits? Is a spouse being faithful or unfaithful? Is a child using drugs even when she swears, trust me. Trust me. Trust me. I'm not. But issues of trust aren't just about those potentially momentous situations. Issues of trust pervade our common daily life. Will my neighbor remember and I trust him to really feed my dog while I'm away, or am I going to come home and the dog's hungry? The mechanic, is he really being honest when he says my car need a new transmission? Was the salesperson when I bought this suit really honest when he told me it makes me look thin? You can tell me. I don't know. I won't comment on that. But whether it's big or whether it's small, what all of these issues have in common is a simple dynamic. They really depend on trust. And we know the more we trust individuals, we can gain a lot more by working together and cooperating. But in reality, as you probably can guess, trust is a double-edged sword. Yes, we can gain more by working together. That's why we have trust in the first place. But trusting somebody also makes us vulnerable to that person. It means that our outcomes are dependent on them being competent, on them having integrity, on them working with us. And so given that trust is so central to human life, you would hope, you would like to think that we really understand how it works, that we can make really good decisions about who we should trust or whether we're going to be trustworthy ourselves. But I'm here to tell you we're not really good about that. And until recently, the science underlying that hasn't been really good about it. And so in some ways, that's what led me to write this book. As a scientist, I really wanted to work on correcting a lot of the misconceptions that are out there to empower people to make better decisions but also so that we can work together to nudge us all to nudge society to become more trustworthy and cooperative overall. So to do that, you have to start like you would with anything else. You have to get rid of the misconceptions and figure out how trust really works. And so that's what I want to talk about today in general. In the book, I talk about lots of issues that I'm not going to talk about today. We talk about issues of how trust affects learning and academic success. One of the best predictors of a child's academic success isn't how much they like their teacher. It's how much they trust their teacher. And they trust that the teacher is competent in telling them and giving them information. We talk about how trust affects our relationships and especially romantic ones and how it can function to smooth out the bumps in those in ways that operate even below our conscious awareness to keep harmony with those we love. In the book, I talk about how trust is affected by power and money. There's great work out there showing that people's trustworthiness tracks socioeconomic status. This is work by my friend Paul Piff. He's a psychologist at Berkeley where he shows that higher SES correlates to increased untrustworthiness. But really it's not about being in the 1%. It's not a birthright of the 1% that makes you untrustworthy. It's simply about money and power relative to those around you. And so any of you, if we put you in a position even for 10 minutes where you feel elevated sense of power, it becomes a lot more difficult in some ways to actually be trustworthy. And also how and when can you trust yourself? It's already February. A lot of New Year's resolutions have gone by the wayside, so if it's a good thing to know. But today what I want to talk about is three broader themes. And the first is what does it mean to be trustworthy and how can we understand how trust operates within ourselves and our own trustworthiness? The second is can we actually detect whether somebody else is going to be trustworthy? In some ways, this has been the holy grail of governmental research and security research. And we've been pretty bad at it. But I have some new data I want to share with you that suggests we can do it. And then finally, the question that's probably closest to my heart in the work that I normally do, which is, how can we increase trustworthiness and thereby increase our own and each other's resilience around us? So let's start with the first question. Most of us, when we think about trust, we think about it as this stable trait. A person's trustworthy or they're not. But I want to convince you that that's probably not the best way to think about it. That's not how it really works. Growing up, we have this idea that it's a typical motif, right? You see it in cartoons all the time. There's an angel on one shoulder and a devil on the other, and they whisper into your ears. And if you grow up listening to the angel, well then you're going to be a good person. You're going to be trustworthy. Everybody is going to love you. Everything's going to be good. There's just one problem with that. And that is if you actually look at the scientific data, it doesn't really hold up. What we've learned over the past decade especially in psychological science is that people's moral behavior is a lot more variable than any of us would have expected. And it's a lot more influenced by the situation. And so if you want to control your own behavior and predict the behavior of those around you, you need to realize that it's not a stable trait. You need to understand how it's affected by the situation. And so my model for understanding trustworthiness, it's better to think of it as a scale, the old school type of type with the plates that go up and down as opposed to a digital one. In any one moment, your mind, whether you know it or not, is weighing two types of cost. It's weighing costs and benefits in the short term versus costs and benefits in the long term. And those usually correlate with what's good for me in an expedient fashion right now versus what's good for me to do even if it costs me in the moment to built a reputation and to build social bonds in the long term. And depending upon the situation, which decision you choose can change from moment to moment. You can think about it. If my friend Ming loans me money, in the moment if I don't pay him back, well, I'm ahead. I've profited in the short term. But long term, it's probably a poor decision because he's not going to give me money again. I'm going to get a reputation as being a cheater. But if I can get away with it, my mind, unbeknownst to me and my own moral codes that I endorse, will try to push me to be a bit untrustworthy. And so I want to suggest to all of you who think this can't happen to me and that you are completely honest and trustworthy and wonderful, it can happen to any of us. And let me show you an example of how it happens and also why you probably don't think it's true of you even though it is. So the first issue is how do you study trustworthiness? I can't really walk around with a clipboard and say, Cindy, are you a trustworthy person? Because what people will probably say is they'll do one of two things. Either they know they're not and they'll, yes I am. Because who wants to say I'm not? But what happens more frequently is they think they are, and they predict they will be, but when push comes to shove, time and again our behavior isn't what we expect. And so the way that we have to study trustworthiness is not by asking people or looking at their past reputations but by staging events in real time as opposed to fake time where we can actually see when push comes to shove, what will people actually do when real rewards are on the line? So let me give you an example of how we do this. So we set up an experiment in our lab to look at this, and it's rather simple. We bring people into the lab. These are normal community members or even undergraduates from the Boston community all known to be trustworthy people. We bring them in and say, look, we've got two tasks that need to be done. One is really long and onerous and it's these terrible logic problems, and circling letters E, and random digit strings, and all the things that you would feel like is a big waste of your time. Or you can do a fun photo hunt on the computer. Here's a coin. I want you to flip the coin, and whatever one you get, it will determine whether you do the photo hunt or the logic problems. And whichever one you don't do, the person sitting in the next room is going to get. And we're going to trust you to do this the right way. Is that OK? They say sure. And then we let them go. What do you think happens? A lot of people just assign themselves to the good task. Any guesses for how many? 80%? Close, 90%. We've done this many, many times. So it's not a fluke finding. We've done it in our lab. Other people have copied the methodology. 90% of people-- well they do one other thing. Some of them don't flip the coin and just say, oh, I got the good task when they come out. Or some of them, because we have them on hidden video, flip the coin repeatedly until they get the answer they want, which is the same as not flipping at all. But they feel better about themselves. And so these are people who we asked them before, if you don't flip the coin, is that untrustworthy? They said, oh, it would be terribly untrustworthy. But they do it. And if you ask them when they come out, we have them rate on a computer how trustworthily they just acted. So here higher numbers mean higher trust on a one to seven scale. So when they're judging themselves doing this, they're above the midpoint. So they say, yeah, it was OK. I was trustworthy. If you take those same people and you now have them watch somebody else do this, they condemn the person for it. That person was not trustworthy. When I did it, it was OK. When that person did it, they're definitely not trustworthy. Now, the interesting thing about this is these were people who are normal people. And so when we see people like Lance Armstrong or Bernie Madoff, you think, oh, it's something wrong. Those people are morally corrupt and untrustworthy. No, well, yes what they did was untrustworthy. But the same process-- on a smaller scale, of course, and we can only study on a smaller scale in the lab-- happens with us. It happens with any of you. Now, the question is, well, why don't we realize this? Why don't we learn to stop trusting ourselves? Well, the reason why is our mind whitewashes our own behavior. So if you ask these subjects, why did you not flip the coin? They'll say things. They'll create stores like, well, yeah, I should have, but today I was late for an appointment. And if I'm not there, somebody depending on me. And so it was OK. So they'll create all kinds of justifications for why it was OK for them in the situation and how it doesn't reflect on the fact that they're an untrustworthy person but they can be untrustworthy. Now, in some way that's a good thing. It has to be adaptive, because if any of us felt like we couldn't trust ourselves, that alternative is much worse. Because it means we're not going to save money for the future because we know future us is going to go blow it a casino. We're not going to diet and take care of our health because we assume three days from now I'm going to gorge on ice cream or chocolate cake. We're stuck with ourselves. If somebody else is untrustworthy, we can stop interacting with them. We can't stop interacting with ourself, and so we need to trust ourselves even when we make mistakes. So that's OK, but what I'm here to do is to help you try and learn that so that you can decrease the probability that you're actually going to make those mistakes. But what I haven't told you yet is that there's any evidence that people actually recognize what they did was wrong. So let me give you an example. So in psychology we had this method which is called cognitive load. And it's a way to kind of tie up people's mental resources so they can engage in rationalization. And the way it works is we give them random digit strings of numbers, say like seven digits. And you have to remember these digits. So what we're doing is you'll get a string of numbers, and you'll have to say 7-6-5-4-1-0, 7-6-5-4-1-0, and then you'll have to answer a question, how trustworthily did you just act? And you have to remember these numbers because I'm going to have you enter them in a minute, and you've got to get them right. And so what this does is it ties up your mind. It prevents your mind from engaging in rationalization. So when we did this experiment again, and we have 90% of people who did cheat even though they said they wouldn't, what you find is on the white bars on the bottom, those who were under cognitive load, there's no difference in how you judge yourself or how you judge others. And those are significantly less. You see yourself as less trustworthy than when you have the time to rationalize. So the second, the moment that you're committing the transgression, your mind knows it. You feel in your gut. You feel that pang of guilt. But what happens is you don't want to think of yourself as untrustworthy. And so your mind engages in this rationalization. The good you tamps away, tamps down the guilt so that it can create a view of you. Well, I had a reason, and it's OK. And I am trustworthy. So the point is to remember that all of us, even if we think of ourselves as trustworthy, I'm sure most of you in general are trustworthy, but your mind is making these calculations. Here when we gave them anonymity-- or at least they thought they were anonymous. They didn't know we have them on hidden video-- their mind's impulses for short-term gain created a story. It pushed them to say, well, I can get away with it now. Even not consciously, it just pushes them to make this decision as an impulsive way. And then they justify it because the long-term consequences they believe are not there, because they believe they're anonymous. Let's turn to the second question. The second question is, can I trust you? How do you figure, how do you determine that question about somebody? Now, as we all know, human society flourishes when we cooperate with each other and when we trust each other. The problem is if one person doesn't uphold his or her end of the bargain, that person can gain at the other's expense. And so what you have is a very dynamic yet delicate balance that we every day have to navigate through and optimize our outcomes. If we make the wrong decision over and over again, we're going to have a problem. So here what we try to do is we try to use people's reputation. Now, as I just told you, reputation isn't a great predictor, and so often we're wrong. But the problem that confronts us other times is sometimes we have to decide if we're going to trust somebody new who we don't know anything about. And we don't know their reputation, yet we're negotiating with them. What do you do there? You have the opportunity for establishing a long-term relationship or you have the opportunity for being screwed over in a way that you couldn't predict. And if you're wrong, well, time and time again that's going to cause you a lot of problems. It's a very non-optimal outcome to be wrong. So given all that, it would be nice if we could actually detect if somebody else was going to be trustworthy. Now, as I said at the beginning of this talk, people have been looking for the Holy Grail of what signifies deception or untrustworthiness for a long time. Is it a true smile? Does that mean I can trust you? Is it shifty eyes? Does that mean I can't trust you? And the TSA spent $40 million on this program to look for these single microexpressions that in GAO testimony before Congress has been shown to be utterly useless. And the problem is I think the reason why we haven't found how we can detect trustworthiness is we've been going about it in really the wrong way. There is not going to be one marker. There is not going to be one golden cue. Cues to trustworthiness are going to be subtle and dynamic. Why is that the case? Well, it's very adaptive if I'm standing here and you're looking and me and all of a sudden I see a major threat behind you to show fear. Because that lets you know even without turning around very quickly, there's something dangerous there. But trust isn't something that you want to communicate very easily or untrustworthiness. Why? I mean, imagine if you're trustworthy person and you had a clear tell. It's like walking around with a big T on your forehead that says, I'm trustworthy. What would happen? Everybody would want to cooperate with you, more of them so they could probably take advantage of you because they know they could. Or if you were untrustworthy and you walked around with a big U on your forehead, well, everybody would ignore you. And nobody would cooperate with you. And your outcomes would be poor. And so trust signals have to be played close to the vest. We have to interact with each other. I can get a feeling for you. You can get a feeling for me. And then we can decide and reveal our cards very slowly. So they're going to be subtle and dynamic. They're also going to be the context dependent. What signals trust in any one specific culture may vary. What signals trust in any one situation may vary. Think about it. There's different kinds of trust. There's integrity. So can I trust that you're going to do the best job you can to help me? Are you meaning well toward me? That's different than trusting your competence. If you don't have the competence to help me, all the intention in the world is going to be useless. And so the cues I look for for competence versus integrity may be very different, and we have to think about that. But the main reason why I think we haven't found the cues to trust is that they're going to occur in sets. I mean, think about it, right? If touching my face means I'm going to be untrustworthy, if I do this, am I doing that because I've got an itch or because I'm going to cheat you? Don't know by one thing. You can't tell. The only way you can begin to read cues to trustworthiness is to look for them occurring in sets so you can disambiguate the meaning of any single one. And that's what the field typically doesn't do. And so I'm going to quickly tell you about two experiments that we did to show how trust can be read. The first one is kind of exploratory. We threw out everything that we had known before, and we simply started to try and identify what cues actually predict real monetary trustworthy behavior and to demonstrate that they do this in an accurate way. And the second part was designed to actually confirm in a very tightly controlled, highly precise way that these are the cues that matter. And I'll show you what I mean by that in a second. We have an exploratory phase and a confirmatory phase. So how did we do this I will start with the exploratory phase. What are candidates for signals related to trust? Well, we brought 86 people into the lab and we put them into dyads, which are groups of two. The only requirement is you couldn't know the person with whom you were now going to interact. We gave them five minutes to have a get-to-know-you conversation. You could talk about anything you want. We gave them a list of topics, but they could talk about anything they wanted. And you're going to play a game for real money, a game that pits self-interest versus being trustworthy, communal interest. And I'll show you how the game works in a second. And then we gave them topics to start, but they could talk about anything that they wanted. So we brought them in. They simply sat across from each other at a table, half the subjects. And we had three cameras on them that were time locked so we could record every single gesture, every single cue they made. Now, we also had another group of subjects who conversed in their get to know you in separate rooms over Google Chat or Gchat-- any type of internet chat. And the logic for this is the same amount of information is being exchanged in the conversation, but in one condition you have access to the person's nonverbal cues. In the other you don't. And then we brought them into separate rooms if they weren't in separate rooms already. And we said, you're going to play this game. We gave each of them four tokens. And the tokens are worth $1 to each of them but $2 to their partner. And so this game is called the give some game. And it's a nice analog for self-interest versus communal interest. Because if you want to be selfish, you can try and get the other person to give you all of his and give nothing. And that means you'll have $12 and he'll have nothing. But the most trustworthy thing to do if you really implicitly trust each other and want to benefit each other is to exchange all you have at the same time, because then you all started with four and now you have eight. And so we had people making real decisions and we paid them accordingly. And we also had them tell us what they thought their partner was going to do. Now, the nice thing about it was whether or not you talked to your partner over an internet chat or face to face, the amount of trustworthy behavior didn't change, which is nice. I think it's because people are now becoming very used to communicating over internet mediated platforms. And so it's not like being face to face made people more trustworthy. There was people who were cheating and being cooperative at equal levels in both cases. But here the axis is error, the amount that you were off. And so lower bars mean accuracy in terms of absolute value. If you were in the presence of the other person, your guess for how much that person was going to be trustworthy or cheat you in absolute dollars was significantly greater. So what this tells us is that people are picking up on a cue. There is some information there that your mind is gleaning from body language, whether you know it or not. And so what we did next was we ran models of all these possible combinations of cues to see what would matter. And the model that predicted untrustworthiness the best consisted of four cues, touching your hands, touching your face, crossing your arms, and leaning away. If you think about it, what does this really mean? Well, we know from the nonverbal literature that fidgeting with your hands and touching your face repeatedly is usually a marker of anxiety and not feeling comfortable. Crossing your arms and leaning away is a marker of I don't want to affiliate with you. Put them together, what does it mean? It means, I don't really want to be with you. I don't like you. And I'm nervous because I'm going to screw you over in a minute. And so none of these cues predict it on their own, but together they did. So the more often you saw a partner show this set of cues, the smaller number of tokens you expected that person to share with you, which meant the more selfish you thought that person was going to be. And the more often you yourself or any subject emitted this four set of cues, the less trustworthy you actually were, the more tokens you kept and tried to get from the other person without sharing. And so in some sense, what we're showing is ground truth here. These cues are predicting actual financial cheating versus cooperative behavior. Now, the most interesting part about it was if you asked our subjects, so what did you use? They had no idea. Or they would suggest it was other cues that didn't predict anything. Yeah, I showed you they were more accurate when they saw the person. And they adjusted their numbers and guesses accordingly. So what this means is your mind is using these cues even though you're not aware of what they are. It's still building intuitions with them. But how do you know those are the right cues? People are doing lots of things. How do I know that when I'm crossing my arms, it's not that my left pupil is dilating and that's the magic cue. Well, being scientists, we needed to have precise control. No matter what actor I had, I couldn't get them to have exceedingly precise control of every expression they're emitting. So what do you do? You need a robot. So this robot, her name is Nexi. She was designed and created by my collaborator Cynthia Breazeal at MIT's Media Lab. And so we simply used Nexi. And I'll show you a video of it in a moment. But the experiment was simple. We repeated the same thing we did before, except we replaced one of the people with the robot. So now you're talking to this robot. And the robot will emit the cues and express the cues that we think signify untrustworthiness or not in a very, very precisely controllable way. The robot was controlled by two people, one who was the voice of the robot, the other who would control whether or not she made the cues. Because you don't want the same person doing it, because they might give cues in their vocal tone. And because I know a lot of you are engineers, I'll give you a quick idea of how this worked. The one person here you can see who is sitting in front of the computer, there's a webcam on her face. As she moves her head, it's gotten by the webcam. The robot's head moves in real time. She's wearing a mic here, so as she speaks, it picks up the phonemes and the robot's mouth moves in real time. The next person controls whether the robot gives these untrustworthy cues or other similar cues. And the third person is our robot mischief person who basically controls and monitors the system. Because every once in awhile, it would go haywire and the robot would like it's doing something crazy, like it's possessed or something. It would break down. But normally it worked wonderfully and just fine. And so we had participants, 65 participants from the community. We brought in 31 of them. We showed that we had seen the cues that Nexi meant untrustworthy. The others didn't. Here's a picture of it crossing its arms. Here's a picture of it touching its face. I'll show you a video in a second. So the first part is people have to get used to the fact that they're talking to a robot. So we had this kind of part where they just acclimated to it. And here's what that looks like. [VIDEO PLAYBACK] NEXI: So my name's Nexi. What's your name? KIM: My name's Kim. NEXI: Kim? It's very nice to meet you. KIM: You too. NEXI: To get started today, why don't I tell you a little bit about myself? KIM: OK. NEXI: I was born and built at the MIT Media Lab two years ago. So I guess in human years, I'm pretty young. But in robot years, that's more like being 20. KIM: (NERVOUS LAUGH). DAVID DESTENO: So you can see she's a little uncomfortable. In fact, we had to put that black barrier on the bottom because people were afraid it was going to go Terminator on them and kill them. So we needed that little barrier. But they quickly acclimated to this, as you'll see. [VIDEO PLAYBACK] KIM: That's basically all I do for fun though. I don't have a lot of time. NEXI: Did you grow up in Upstate New York? KIM: Yeah, I did until I was 18, when I moved out here. NEXI: It seems like that must have been a big transition for you. KIM: It was. It was a really big transition. But I kind of decided that it wasn't the life that I wanted. DAVID DESTENO: So they would self disclose. We heard about pets dying and all these things. One person kept asking the robot if it believed in God. That person was hard. But for the most part, people behave. And just so you can tell what it looks like face on, I'll just show you a 10-second clip. [VIDEO PLAYBACK] NEXI: We all share a big, open room. There are lots of cords and gadgets. So it's probably not like your house. But it's home for me. Why don't you tell me about where you're from? MAN: Well, I was born in Lawrence, Massachusetts, and I had a residency in Somerville right now. I've been doing residential the past four months. DAVID DESTENO: And so we then had them play this game with the robot. We told them, look, the robot's got an artificial intelligence algorithm that it's going to decide how much money it wants to give you and how much money it thinks you're going to give it based on how the interaction went. It didn't, but that's what we told them. And then we asked them questions about how much they trusted the robot, et cetera. So what happened? So to make a long story short, what happened is these are, for those of you who are mathematically inclined, these are standardized regression coefficients. When Nexi made the cues that signaled untrustworthiness in the human to human interactions, people reported trusting it less. Now, the important thing is they didn't report liking it less, because I was worried, oh, they just might think it's doing something weird. No, they liked it equally, but they trusted it less. Now, that's important, because to me that makes it real. Because we all have friends that we like who we wouldn't trust with our money. And so OK, and the less they trusted it, the fewer tokens they predicted Nexi would give them, basically meaning they thought Nexi was going to be selfish and cheat them, and the smaller amount of money in that game they actually gave it. And so what this tells us is that we know these are the cues because we manipulated them with exact precision here while nothing else was happening or things were happening that we could control. And so cues to trustworthiness can be imperfectly assessed, but better than chance. And so the TSA starts to need to look for cues in sets in a context-dependent way. But in some ways, the more interesting part of this is that what it suggests is that technology is now good enough that the mind will now use these cues to ascribe moral intent to robots, or to avatars, or to virtual agents. So while you may not get it from R2-D2, you will probably get it from Wall-E. See, I'll hear aww. Wall-E is not human in the least, but he has enough human characteristics in the eyes and in the hands that he can move them in a way that pings our mind's mental machinery to make us feel trust, or warmth, or compassion toward him. So what does this mean? It's a whole Pandora's box, because in some ways it's good. So for people like Cynthia who want to design these robots and she's working on the smaller ones so that they can actually accompany kids for medical treatment where parents can't go. Think radiation treatments for kids with cancer. They can go with the children. They'll seem more trustworthy, more comforting. But like any other science, it's not good or bad. It depends on the uses of the people who want them. We all know trust sells, so if I'm a marketer, what does this mean? It means that I have the perfect trustworthy or untrustworthy person that I can show you. Because in a human, stuff leaks. No matter how much we're going to try and control it, which is why we could pick up on untrustworthiness. There is no leaking here. We can control everything. And so as we're conversing more and more with automated agents and avatars, our trust is going to be manipulated in ways that we could never have known before or that our mind is not ready to defend against. OK, finally, the last part of the talk, how do we go about enhancing trustworthiness and enhancing the compassion and resilience of each other? To let you know just how powerful this can be and how quickly trust can change and compassion can change, let me give you of my favorite examples. Some of you may know this story. It's called the Christmas Eve truce of World War I. So it was 1914, and the British were fighting the Germans outside of Ypres, Belgium. And it had been a long and a bloody battle. And they were each in their trenches separated by the no-man's land in between. And on Christmas Eve, as the Brits looked across the no-man's land, they started to see lights appear. And then they started to hear songs. And at first, they didn't know what they were because they were in German, and they didn't speak German. But then they soon recognized the melodies. And what they were were Christmas carols. And what happened next was amazing. The men came out of their trenches, and they started celebrating together. They started exchanging trinkets. They started talking about their families, showing pictures, celebrating. Now, these were men who hours ago were trying to kill each other. And no one would have ever trusted if I walked out, was an open shot, I couldn't trust that you weren't going to shoot me. They always had shot each other before, but here they were celebrating with each other in a very communal way, by their own words, very amazing. Here we were laughing and chatting to men who only a few hours before we were trying to kill. Now, if that's not a big change in how trustworthy somebody can be, I don't know what is. So the question is, how do we display such trust and compassion in one moment and such cruelty the next? Because if we can understand that, then we can do something about it. But to answer that question, what we have to realize first is how do we address a different one? How do we identify who is worthy to help? The world is full of more people than we could possibly help. Not that we don't want to help them, but it could be overwhelming. And there's this phenomenon that we know of in psychology called compassion fatigue, which is when you're confronted with people over and over and over again who need help, you begin to dial it down. I have this experience that I'm not proud of when I go with my daughter to New York, and we're walking by, and there was a homeless person. She was like, daddy, help this person. And then I realized that in that moment, I'm completely ignoring this person, because it's a common thing that I face all the time. And if I stopped to try and help every person, it would be overwhelming. And so we have to understand how our mind goes about deciding whose pain is worthy to feel, who it's worth to help, and who it's worth be trustworthy toward. And once we understand that, then we can figure out, OK, how do we increase the number of people to whom we should feel that? Well, one way that I think our mind does it is it uses a simple metric. And that metric is similarity. So it comes back to this is Robert Trivers, who was the discoverer of reciprocal altruism, which is basically the idea, why do we help people in the biological sense? It's I scratch your back today. You'll scratch mine tomorrow. In some ways, that's what similarity is. When there's a lot of people who need my help, going back to that equation of short-term versus long-term gain, who should I help? Who is it worth it for me to help? What your mind does shaped by evolution is it decides the person who is more similar to me is the person that it's worth helping, because that's more likely the person who's going to pay me back and be around later. At least initially that's how it works. And so what we wanted to do was to see how deeply embedded this bias is. If I said to you, if an American soldier's on the battlefield and he comes across an American soldier and a member of the Taliban and both of them are suffering the same wounds, who is he or she going to feel more compassion for? And if I said the American soldier, you might not find that surprising. But what I want to argue is that it's not dependent on longstanding conflict. It's this unconscious computation that your mind makes. And so we tried to strip that down to as basic a level as we could. And we did that by using something called motor synchrony, which is people basically moving together in time. You see it in the military. You see it in conga lines. You see it lots of places. You see it in lots of rituals. And the idea is that if two people are moving together, that's a marker that for here and now, their outcomes, their purposes, their goals are joined. And so we wanted to see if we could actually show this effect at that level. So we brought people into a lab, and they thought it was a music perception study. So they sat across from each other. And there was sensors on the table, and they had earphones on. They didn't talk. All they had to do was as you hear the tones in your earphone, tap the sensor. And so it was constructed so that the two people would either be tapping in time or completely randomly and out of time. That was it. Then what happens is they see the person they were tapping with get cheated. This party is staged, but they don't know it. They believe it's real. They see this person get cheated in a way that makes that person have to do a lot of extra work that they shouldn't have had to do. And then what happens is we give them a chance to decide if they want to go and help that person and relieve that person's burden. And that's what we look at. So what happens? We asked the people, how similar were you to that person in the experiment? The simple act of tapping your hands-- they didn't talk. They didn't do anything-- made them feel that they were more similar to the other person. Now, if you ask them why, they'll create a story. They'll say, oh, I think we were in the same class. Or I think I've met this person somewhere or we share the same goals. They don't know. They never talked to this person before. None of that was true. But because they had this intuition that they felt more similar, they had to create a story for it. How much compassion did you feel for this person when they got cheated and got stuck doing this onerous work that they weren't supposed to do? Remember, in both cases, the amount of suffering is exactly the same. Yet they feel more compassion for this person if they were just tapping in time with them. How many wanted to go help this person? This I found truly amazing. 6 out of 34 people would say, oh, I'll go help that person who was harmed and cheated versus 17 of 35. We had a threefold difference. When you tapped your hands in time with this person, 50% of them said, I want to go help this person who was wronged. That's a huge effect if it's scalable. How much time did they spend helping? These are seconds. So if you tapped your hands with this person, you spent a lot more time knowing that everything you did would relieve that person's burden. And if you look at it-- again these are regression coefficients-- if you tapped your hand in time with this person, yes, you felt more similar to them. And yes you like them more. But what actually predicted the compassion you feel? Not how much you like them but how similar you felt to them If you tapped with them, you felt more similar. That predicted how much compassion you felt toward them even though the level of suffering was the same objectively. And the amount of compassion you felt for them directly predicted how much time, how much effort you put into relieving their pain. Now, what this suggests is that compassion and trustworthiness are flexible. Because if you're going to be trustworthy to me, that means you're going to sacrifice your own immediate outcomes to benefit me like these people did here. Can I trust you to help me? Can trust you not to shoot me back with the Brits and the Germans. Where I live, what this means is trying to solve some of the more contentious things we have in Boston, which is Yankees versus Red Sox. But what that means basically is not thinking about your new neighbor as the guy who hates the dreaded Yankees. Think about him as the guy who likes Starbucks as much as you do. If you can actually retrain your mind to find similarities that you have with people, it will increase your trustworthiness toward them and the compassion that you feel toward them. When you think about social media, there are tremendous ways to do this. We can use the computational power of social media in ways to connect people that have never been connected before. Think about things like profiles on Facebook or other things. We have vast knowledge of what people like and don't like, what they've done or haven't done. Perhaps what you can do is find what people in conflict have in common very rapidly in the background and surface that information to them. And if you do, then it should function in just the same way as tapping your hands. There's nothing magic about tapping your hands. We've done it with wearing the same wristband colors, et cetera. Anything that you can do to highlight similarity with someone will make your goals seems more joined, which will increase the compassion you feel to them if they're suffering, which will increase how trustworthy you are toward them even in ways that don't involve compassion. So at Google, I'm really interested and open to talking with any of you about if you have ideas about using the computational power that you all have to kind of nudge trustworthiness and compassion in the world. But that's all kind of a top-down way. This is a way that we have to remind ourselves, OK, think about this person as similar to me or not. It will be nice if we had a way that could make it automatic, a way that works from the bottom up so that we don't have to stop and remind ourselves. And one way that we can do this-- and I know it's an idea close to Ming's heart, and I'm really honored to be here to be able to talk with him about it-- is mindfulness. If you read the paper or know anything about mindfulness, what you'll know is that it is enjoying a renaissance. And we know it does all kinds of wonderful things. And probably many of you have more experience with it here thanks to Ming's course. It will increase your creativity. It will increase your productivity. It's good for your health. It'll lower your blood pressure. It'll even increase your scores on standardized tests. These are all good things. But if you think about it, it's not what it was originally designed for. If you look at what Buddha said or many of the other ancient meditation teachers-- well this is a quote by Buddha. I teach one thing and one only. That is suffering and the end of suffering. There weren't LSATs and GMATs back then. So all these other things that meditation does are wonderful, and they're great. But one of the main purposes was to foster compassion and end suffering and to increase our being good to each other and being trustworthy to each other. And so what we decided to do was actually put that idea to a test. And so we brought people into the lab. These were people who had never meditated before. They were members of the Boston community. And they were all equally interested in meditation, doing a class for eight weeks. We assigned half of them to actually take a mindfulness class led by a Buddhist lama. And they also would go home during the week with MP3s created by the lama that they would practice. The other half were put on a wait list. So this way we had groups that were equally interested in medication, because you might imagine that if we just recruited people who wanted to meditate, they might have been different types of people in the first place. So both groups were equally interested, but only half of them actually got the course. The other half got it after we did the measure. After eight weeks, we brought them back to the lab. Now, they thought they were coming to have their memory, and their executive control, and all these cognitive measures tested, which we did. But before we did those, what we really were interested in is what was going to happen in the waiting room. And so in our waiting room, we had three chairs. Two were filled by actors and the third was for the subject. And so when the subject arrived, what did the subject do? Well, all them except one sat down. We couldn't get that other guy to sit down no matter what. But most of them sat down. And then a third actor would enter the room. This person was on crutches, had one of those foot boots you wear when your ankle is broken. And as she'd walk down the hall entering the room, she would kind of wince in pain and looked noticeably uncomfortable. And she'd enter the room, and there weren't any chairs. And so she'd lean against the wall. And the question was, what would the person do? The actors were told to busy themselves in their iPhone and to not pay attention. Now, in psychology we call this a bystander effect where this really limits helping. If you're in a situation where you see somebody in pain and other people aren't helping, that tends to decrease anybody's odds of helping, because you say, oh, it's not a big deal or maybe I shouldn't help. And so this situation is one that makes being counted on that you're trustworthy to come and help the lowest it possibly can. So what happens? So people who were in the control group, only very small percentage of them helped, like 16%. Among those who meditated, 50% of them helped. That's a threefold increase. And that's a threefold increase in the situation that is designed to actually work the most against your willing to help. Now, if that can happen after eight weeks and if that is scalable, that is a huge, huge effect that you can count on other people. You can trust them that they're going to help you. Now, why does it work that way? It works that way because one part of mindfulness is this idea of equanimity. And that means realizing that I am similar to you, and you are similar to me. Friends are enemies, are enemies can become friends. And what that does is it trains the mind to see us all as valuable and interlinked. And it breaks down the categories that we put on each other of we're different in religion. We're different in sports teams you like, et cetera. And I think that's why it works. And then it becomes automatic. It does the same thing that my little tapping example was doing. And so when it comes down to it, really what I want to say is that in the end, it's trust or it's dust. And what I mean by that is without trust, our ability to be resilient as a society is exceedingly low. And so how can we build it up? Anything we can do to nudge it up is important to being resilient. In the fall of 2012, I don't know how many of you remember out here, but on the East Coast, super storm Sandy hit New York. And it was a devastating storm. And there are neighborhoods that still aren't recovered. But the AP did a great study. Controlling for the amount of damage that occurred, they looked at what was the single most important predictor of a neighborhood's resilience. The single most important predictor was how much neighbors trusted each other. How much they know that the other person they could count on them, that that person was going to have compassion for them, that they were going to work together. The neighborhoods that were higher in trust were the neighborhoods that got up and running in terms of commerce, and support, and social services faster than anything. And that's why I say in the end, it really is trust or dust. If we don't trust, we're harming everybody. But of course there were people in neighborhoods who price gouged and who did things they shouldn't. And so really my message in the book is, yes, trusting is good. We should all trust. But trusting wisely is better. And so it's my hope that any of you who read this or come into contact with this work, it will empower you to think about the way trust actually works and the forces that impinge on it to make better decisions about who you can trust but also how to foster trustworthiness in yourself. And I thank you so much for listening to me. MING: Thank you, my friend. We have time for questions. Anybody have any questions? AUDIENCE: Two fairly related questions, in the earlier test about looking for cues, so I'm just wondering how you came to the domain of different things you were looking for that could conceivably be a cue in your analysis. Because you could say whether the pinkie is touching the hand and the hand's touching is or not. And then along with that, with whether you were going into it starting off focused on physical cues or if you were also considering the types of issues that were brought up in conversation, which could then play into the talking with a robot or talking over the internet. DAVID DESTENO: So let me do the second one first. We were primarily interested in physical cues. There's lots of work out there as well on linguistics and the type of phrasing that people use as well as vocal tone. We weren't looking for those. It doesn't mean that those don't matter. And so I'm not saying these are the only cues that matter, but these are sufficient to predict. The more that we know about, our accuracy will go. But we were interested in the actual physical, biological motion cues. How did we get them? We simply started with the brute force method, was looking for individual cues that had some predictive ability at all. Because even if they're not predictive greatly on their own, they have to have some predictive power on their own. And then we would begin to do is to assemble different subsets just trying to maximize the amount of accuracy that we can predict. So it was a very bottom-up approach. And then once we have those four sets, those predicted the greatest amount of variance in people's selfish monetary behavior. Which is then why, again, it was really important to use the robot. Because you're right. This was a correlational method. Who knows what we could be picking up. Maybe on every time I cross my arm, it was my pinkie. And so we could actually manipulate it with precision with the robot to validate it. MING: If you could put a pinkie down here, it's not trustworthy. That's what I learned from a movie. AUDIENCE: Let's see if I can word this the right way. It seems like the general conclusion from the research or your conclusion is you're saying we should be more trusting of others, like sort of the hope for the better world. And the question is, is there also then some drive for people themselves to be trustworthy? Like in the example you gave in the beginning about people in a marriage, one cheated on the other, is it to say, put that aside. Trust that person. Or is there some other conclusion in that sense? DAVID DESTENO: It's a good question. What we know from all the-- so people like Martin Nowak at Harvard is a straight mathematician, evolutionary biologist. And so they run these fantastic models. And what we know is that if you are untrustworthy in the short run, you will profit immensely. But over time, that profit then starts to go down. And so in the long run, people who are trustworthy profit the most in terms of everything and even as a group. And so we know that's the better outcome. But if you can be untrustworthy and not get caught, you're going to profit. So how do we try to balance those? And so what we're trying to do is to make everybody want to be more trustworthy but at the same time also make better decisions. It will be impossible to have a world where everybody is trustworthy. Because if everybody's trustworthy, you stop even looking and caring. You just automatically say, yes, I'll trust you. And then if there's a mutation or whatever that causes people to be more untrustworthy, they're going to profit like crazy. Until then everybody starts caring, and so it's always going to be in an equilibrium. The question is, can we increase the set point for trustworthiness to a higher level? And so it's about increasing your own trustworthiness but about deciding if you can trust somebody else wisely. So yes, if you know absolutely nothing, it's better to trust than not trust in the long run in terms of quantifying the benefits that can happen. But it's certainly not as good as making an informed correct decision. And so my hope is to try and open people's eyes to how trust really works so that you can make better decisions. AUDIENCE: So similar to that question, trust over time, have you done any research into how analysis of trust has to change, how much it needs to be dependent on data changing? Like the first study was on the initial get to know you, how much I trust you as a person. And then the question is later on, something happens. How much should future events be added into that evaluation? DAVID DESTENO: You mean at what point will I change my judgment of whether I can trust you? It's dependent on a lot of things. It's often dependent on the magnitude of what you have held up your end for what you haven't. But I guess my argument is that you need to look at each situation if it's important and new. Because even somebody who has been always trustworthy, if the costs and benefits change-- the reason they're always trustworthy is the cost and benefits are rather stable. Take that person and change the cost and benefits either by dangling a reward that is immense in front of them or giving them anonymity so they won't get caught, like our people, and they'll change. So I think there's not a clear time frame. I think we all adjust at different rates depending upon the magnitude. But my message is no matter what you think, consider the situation. If it's somebody you always trusted, consider has their power changed? Has anything else changed? Because they may not want to be untrustworthy just like our subjects did, but they will be and construct a story for why. AUDIENCE: Hi and thank you for coming. I have a question actually kind of related to that. So have you done any study in how people's relationship long term potentially changes how sensitive or not sensitive they are to social cues? I can imagine someone who perhaps does all of the social cues that you mentioned in terms of lying, but perhaps like a brother and sister, for example. They've gotten numb to it over time, and maybe they can't pick up on it anymore. Or do you have a sort of sense of how long-term relationships can change how people pick up on that? DAVID DESTENO: Two things on that, we haven't done that work, but what we know from the nonverbal literature in general is that people have a what are often termed accents. So there's a kind of panhuman way of doing it. But then different cultures or even different families or individuals will have modifications of that. And so the longer you are with someone, you can learn that for this person, this is that person's tell in some ways. And it will be some combination of these. But other things added to that will increase your accuracy for that person. But another thing in terms of long term, what does trust do, it's beneficial. So there's great work done by Sandra Murray. She's psychologist who studies romantic relationships. And one thing that the trust does in our relationships is it smooths out the bumps, as I said. So we've probably all had times when our significant other does something and it makes us go hmm, whether it's you think the person's flirting with someone or they're working late or what is it. Well, if you inherently trust this person, at a very nonconscious level, that trust erases that hmm. It just gives you an intuition that everything's fine. And if you trust that intuition, that's good. Because lots of times we'll do something that we're not trying to be untrustworthy. It's just an inadvertent thing. But if a person interprets that as, oh, you're untrustworthy, it can start you into kind of a death spiral. And so the good thing about trusting someone over the long time in a relationship is that it helps smooth out those bumps so there aren't mistakes made. So that one person doesn't interpret the other person as flirting or doing something with somebody else that they shouldn't have. Now, if they keep doing it, well, then you're going to know it's real. But that's a benefit of trust long term. MING: Thank you, my friend. So the book is "The Truth About Trust" available where books are sold, also available at the back of this room. And David will be around to sign books. And my friends, David DeSteno. DAVID DESTENO: Thank you all.