 everyone for being here, I guess, after it's time for people to go home. So you must be really interested in audience. Okay, so, I guess what I will do is, as was in the discussion with Mr. Ahmed and so on, that I'm going to talk about two kinds of things. Basically, I'm first going to sort of define the term for the media with mining and social analytics. But also, I'm going to talk about the aspects of it from the research side as well as from the entrepreneurship side, right? So a little bit about my background, as it says over there, that I'm the research director for social computing at QCRI. And I've been here since about October of last year, so about six months. And in addition to that, well, before coming here, I was a professor of computer science at the University of Minnesota in the US. So I've had a long academic background over there for 25 years. But also, so therefore, I have been doing research in this area for quite some time, at least a decade, in the general area of web-binding and social analytics. But also, what I've done is that based on some of the research that was done, I, with a colleague of mine, started a company, which is based out of Los Angeles, okay? And so earlier, I was quite involved in it. But then when I took up the position here, I kind of stepped back. So I have more of an advisory for what I now as the chief scientific officer. So what I'm going to do is, I'm going to tell you about some general trends, in fact, so this is sort of the top outline, okay? But I'll give you a little bit about the background, okay? Also, a little bit about what social computing is at QCLI, what is it that you're doing, and so on. And then, I will talk about some systems that are there, some social systems. Like game systems and virtual worlds and social networks in general, that are actually becoming a great observatory for human behavior, okay? So you see all kinds of social and behavioral sciences. Whether it happens to be something like psychology, sociology, to even behavioral economics, decision making, all of these, and of course apply once, like marketing and so on. They all are about developing a better understanding of human nature and human behavior, okay? And what is happening is that with online systems, we are getting an opportunity to observe human behavior both at the individual level, at the pair-wise level, at the group, small group level, as well as at massive scales. All of these different kinds of behaviors have a very, very fine granularity. And of course, a huge amount of data is being generated. So because of the availability of this fine-grained, high-resolution and very comprehensive data sets, there's a real opportunity to gain a deeper understanding of human behavior, as I said, at the individual level, at the small group level, at the large group level, at the massive scale level. And from many different perspectives, that means how do people behave when they're given a stimulus of a certain kind or they're in a certain situation, so on. So the whole field of social sciences, social behavioral sciences is undergoing a massive revolution, okay? So that's about the first point that, and some of these kind of environments, all the social, online social and behavioral systems that are there, whether it happens to be Facebook or it happens to be Game System or whatever else it is, they are like a great observatory for human behavior, okay? And based on that, I will talk about two main topics. One is what is the kind of social and behavioral research we can do. And the research I've been doing is really part of a large team, and I will show you some examples of that. And in fact, when I was in Minnesota, I was part of a research program which spanned four different universities. There was a colleague who was an organization theorist at Kellogg School of Management at Northwestern University. Another one was a sociologist at Havana. A guy who's a media communications person at the University of Southern California and myself at Minnesota as a computer scientist. And together we worked, you know, as four different institutions, different disciplines across many, many, many people, right? So I'm going to pick a specific example from that, which my group was leading on something about qualitative study of online trust. And I will delve a little bit deeper into it to show you that how something as basic and fundamental as trust, right? Which is very important to all kinds of people. For example, you know, even business people talk about is, do people trust a brand? Do people trust an organization, right? And that's going to introduce things like financial organizations, you know, whether they have them to banks or insurance companies, then trust factor is very important. How does trust develop online? What are the fine nuances of trust? We will talk about things like how trust gets initiated, how it gets reciprocated, and how even revocation can happen. So this is what is often called the dynamics of trust. So we will go a little bit deeper into that. You can probably spend maybe about 20 minutes or so. Just to give you a flavor of how questions like trust or studies of things like trust, which classically have been done through surveys. You ask people, who do you trust? But now you can study it at a massive scale at a much finer than that. And then we will switch gears and move to the second part of the talk, which is looking at a concept. So now when you start, especially if you look at things like people trusting each other, so if you trust someone, then they have a certain influence on you, right? For instance, if they say something, they are more likely to listen to them because you trust them. So trust eventually leads to influence. And so we will look at the whole idea of influence. In fact, when we started studying influence, we realized that it has a great potential for monetization and referring to commercial sense. So what we did was we took some of the research that had been done before in this area and then transformed it into the concept of social influence and then we will sort of company around that. And that's what I will talk about. So here will be sort of more of basic research, very cross-disciplinary research from social science and computer science. And then here you will see how a startup was created out of it. What is it? Does it sell? What kind of funding is there? Is it on the standard or on the normal? And again, please, I'd like to keep this interactive. So ask me questions. Stop me if you have any clarifications or something. This is the social computing room. And you see this older picture. We don't have all people in here. In fact, Heather over there is one of the members of the team. We have roughly about 32 people, 13 full-time scientists, engineers, postdocs, we have three research associates. And actually, we also have 13 summer interns that are not counted. And in fact, well, I guess, I suppose I haven't advised this slide, Heather, since you don't even show up here. So I guess we have 14 nationalities. Heather is from Canada. So as you can see, it's a very, very eclectic group. And we really enjoy sort of, even we can think of it as a great cultural learning experience to have such a broad group of people. This is what it does. So actually, this is a busy slide, so I'll not go over all aspects of it, but let me just straight up tell you. So we study about sort of, we're studying sort of five different things. And in fact, these are the different topics. As you can see, there's a whole area called computational social science. And in fact, a lot of the talk later on, the research part of it is going to be about computational social science. How do you use computing as a new kind of tool to study social sciences? So we do a lot of that over here at QCRI. Then there's also a whole thing about social and behavioral aspects of health and wellness. In fact, one of the things that we have done is we have participated in a study of sort of health habits and healthy eating habits and so on, or eating habits in general in participation with Qatar University and Ahmad Medical Corporation. And basically, a bunch of children, they wanted to study some exercise and wellness camp, sports camp, and then they wanted to study, that is, the doctors around the university, they wanted to study as to what kind of impact it has on the... So putting the little kids, if you're not sure, maybe the middle school level, putting them into a sports and fitness program, what impact does it have later on on their lifestyle? So what we did was we went ahead and built some kind of an app which will sort of measure what is it that their daily lifestyle is. And in fact, we went ahead and built it and there's a data collection system behind it. Of course, we were going to all those IRB approvals and whatnot. And in fact, right now, something like 40 children are using... So we even gave them cell phones. And loaded with this application and they're using it and they're collecting lots of data. So there is the part about, let us say, the standard taking the blood pressure and the blood group and all those things for testing for or whatever, body mass index and all that. The biometrics part that hospitals do. But in addition to that, there's the whole behavioral aspect. Did it really change someone's lifestyle? And that is where things like social media and these applications come in. In fact, we are now... We have some of those pebble watches and things like that. In fact, you're waiting for buying Apple watches to start sort of instrumenting people and seeing what their behaviors are. There's a whole program behind social media and user analytics. In fact, of course, Al Jazeera being here, they're one of the partners. We do a lot of work in that part area. And sort of work with things like sort of analyzing the behavior of people on Al Jazeera's website. They're for building tools for their editors, for the newsletter of people, for helping them create new content and all kinds of things. Social behavioral aspects of urban mobility and sort of livability, if you will. So for example, here we're working a lot with organizations like CUMIC. We work with Mesharev and all these organizations. Here's a very simple example. Think about this way. For the longest time, I think for at least almost 100 years. So traffic science or the systematic study of traffic has been around for about 100 years. Since then, people have been testing things like study congestion. So what has been the standard approach for traffic in 18 years? Well, they will have sensors of all kinds and they will look at the data. And from that figure out, well, you know, if the amount of traffic is going up but the throughput starts dropping, that means there's some kind of congestion, right? The speed has dropped, there's congestion. Well, but now supposing condition happens, what they can't figure out is why it has happened. For example, let us say there's a condition in a certain area. They can figure that out because they can get the sensor data, traffic, speed and all that stuff. And analyze this, okay, the traffic is not moving, that means there's congestion. But is it a congestion because a vehicle broke down? Is it a congestion because there's a traffic signals are not working? Or is it a congestion because there's an accident, right? They all might have the same footprint from a traffic perspective, but the causes are very different and they will, for each of them, you need to have a different response. For instance, if traffic signal is not working, that means you need to send some repair crew. If there's an accident and you've someone not heard, you need to send some tow vehicles but also you need to send an ambulance, right? But if it's just a temporary congestion, you probably just need to do nothing. You wait for 15 minutes and you clean by yourself, right? Okay, usually they are not known so far what to do. But now, these days we have Twitter and Instagram, right? So, someone tweets saying, oh, there's an accident. Well, how do you bring that into this whole place? So, because now there are humans around who are doing the sensing for you, right? So, standard traffic sensors are measuring the numeric quantities but humans are cognitive sensors. We look at something that we tweet about and what we tweet is an interpretation, right? So, in fact, what we do over here is working with QMIC to be able to combine social makeup and transportation sensors to better understand things like radiation, incident response, industrial engagement. So, it's an entirely different way of thinking about traffic management. And finally, we have a huge program in front of our flagship program, this whole social computing room, which is social computing for crisis response. In fact, just today's, I think this is the for Kether Foundation, right? Yeah, so today if you look at Kether Foundation website, the front page is an article about the efforts that QCRI, especially his political group, has done in helping in the disaster in Nepal. So, there's a whole aspect. In fact, I always call it this way, right? So, the setting of the problem is the following. Let us say, I'm sure all of us saw on television all the devastation that happened in Nepal, right? For most people, it is, oh, this is too bad. Maybe if you have some friends or family, you call them and say, how are you doing, right? If you think a little bit more, you might say, where can I write some check? Where can I send some money? Maybe some people might even want to go back home. In fact, we had one guy. He was actually a coffee guy. He used to serve us coffee and cookies and so on in office. He was from Nepal and his family was, so he was so affected that basically he had to head back, right? But the point is that what our group has done is identified an opportunity as to if you feel bad, how you can spend anything from 15 minutes to 2 hours or maybe 3, 4 hours over a week. And what we have done is built a platform where the inherent volunteerism of individuals can be aggregated together and focused to solve some very hard to solve computational challenges but very easy for humans to do. So think of crowdsourcing now. And then provide the information from that to organizations like you and Ocha, the Office for the Coordination of Humanitarian Assistance or Red Cross or, in fact, for the Red Crescent because they were also very active in Nepal. And the idea is that doing all of that within something like an hour or 2 hours of a disaster happening. So that now organizations that are on the ground, they can actually do their job with much faster cycle time in a much more effective manner. And in fact, in doing so, something like 3,000 volunteers from around the world, almost what? Over 50 countries, I think? Over 55 countries. Over 55 countries, something like 3,000 volunteers helped and through this platform that they did. Yes? Yeah, this is awesome, actually. At the university, there's like a bunch of students doing the same startups. Like now they're trying to build this platform on the ground, like the crowd crisis thing. You know what I mean? So I think they will have a great opportunity to come to you guys and see exactly what they're doing because they are doing the same thing. Like they're building a platform to gather some volunteers for building anything in the world. You know, like help each other to... Well, you know, you should... I guess maybe you should talk to Heather. Yeah. She heads our program management aspect of this whole program. And this is, you know, we welcome students and so on. I think this is... It's an amazing initiative. Actually, one more thing. I don't know what I'm going to say now. It's included in one of these lists here. But I think there is one point that says now the real life is, I mean, like social media and virtual, you know, website and virtual life, now it's being like a black hole, right? So now it's taking everything from us that the real life work. You know what I mean? So how can we find a way to control the virtual life? It's really... It's like, really, it's been like a black hole right now. Yeah. So you raise a very interesting and important question. I'm kind of going to talk a little bit on that. But I think that's a maybe a later on coffee discussion. Yeah. Which is the following, right? Each and every time through technological evolution and evolution or whatever, right? Each time you have introduced something new, okay? It has always been... The one thing we have not been able to invent is more than 24 hours in a day, right? So all of us, you know, since the beginning of time I've had 24 hours a day and that's not changing. But each time we are creating new things to engage us and therefore take away the time. There was a time... So where have you been getting that time? That is by technological advancements that have caused the essential time to decrease, right? So for example, if you can grow... If the number of calories you need to eat if you can grow that with a lesser amount of time that means you have more discretionary time, right? So it's always been a question of of mandatory time versus discretionary time. So a lot of advancements of all kinds have had... been focused towards reducing our mandatory time, okay? And then each time you get some more discretionary time but then we go ahead, we have free time so we go ahead and create new kind of toys to engage us, right? And then those toys start eating into our discretionary time. Till now we have reached a stage... There is no free time, right? Well, so now we have sufficient... More time, I mean there's more toys than we have time to pay attention to. In fact, I'll tell you, when I became first aware of that when I was a child, probably about 11 years old is when television was introduced in my hometown there was exactly one channel. Then a few years down the line they introduced three channels. Suddenly I realized that because each channel goes for 24 hours there's no way I can watch all the TV that's there, right? So because this whole thing, in fact economists have nowadays started talking about what is known as the attention economy. So the real finite resource is attention that we have, the time that we have. And a lot of activities whether that happen to be physical activities like going and playing soccer, they happen to be adding activities like going to a restaurant and eating or they happen to be playing video games or whatever else it is or watching a movie that is information, they all are vying for our attention. And that's the real resource that is out there. So if you think about it today from a certain if you take the attention economists perspective what all these free sites like whether it is Google, whether it is Facebook or any of these where you spend time what they're after is your attention. How much of attention can you have from how many people? That's it. Now yes, the thing is that overall a finite resource I mean there are about what? Seven and a half billion people in the world times 24 hours that is the total attention that exists. And that's not growing in a very fast way, right? And if you know it, it's not about the fact that while you know people need to sleep and also not everyone has access to computers not everyone has access to smartphones all those kind of things. And then actually if you start doing that you come to the conclusion that some Facebook already has something like 10% of everyone's possible no. I think it's roughly about 3.5%. Yeah, Facebook is at about 3.5% of all the possible attention that is out there. 3.5%. Which is a huge amount. Of course. Google has also closed about 2.5%. So these are the huge players that are already learning a few percentage points of all the attention that exists out there. In fact if you take again that perspective why did you know that when Facebook bought WhatsApp they paid about 20 billion dollars. And people are saying you know WhatsApp is a trivial application. You can put 2 engineers around like 2 or 3 engineers or maybe 5 engineers for 6 months you can build WhatsApp. But the real reason was that what Facebook realized and well first of all realized where people have called it WhatsApp is that Facebook attractive though it is there are 2 conditions in which it does not work very well. One is that if the population does not have smart phones. So the population's economic level is such that they can only have those basic feature phones. And secondly the bandwidth the connectivity is either too costly or too unreliable. So what you needed was a stripped down version of a social network. And that's why in places like India or Brazil and so on many of these places such as large populations but do not have the economic where without everyone to have a smart phone like an iPhone or Samsung or any of those and the bandwidth is not so reliable. Suddenly WhatsApp became very popular so what they paid for was not the software but they paid for the fact that half a billion people are already signed up for it. It is their attention and people who are using WhatsApp they have half a billion people spending an hour of WhatsApp every day. So that's half a billion hours a day over a 30, 40 year period. That is the real value of WhatsApp. It was not the software the software you could probably build for a million hours but they paid 19. something million dollars. So that's a whole entirely different so this whole attention business is the real, I mean that's the it's an important consideration and there are many ways to look at it from you know there's a sociological perspective that what is happening to society what is happening to people there's an economic perspective on it. In fact it can best be summarized by the following saying that you see one of the you know it feels like physics and all that theorems feel like sociology doesn't have theorems so they have sayings. So one of the closest or the most accurate sayings in sociology has been that geographical proximity drives socialization. So in history it is always in the case that you made friends with people who were next to you. In fact it is also the famous saying that if you cannot be with the one you love you better love the one you obey. If you cannot be with the one you love you better love the one you obey. Which you can translate that to friendship that if you cannot be with your friends who are far away you better develop friendships locally and that has been the driving factor of sociology of socialization in general. But now what has digital medium done? It says that now no matter how geographically far apart you can choose to socialize with people who are not even next to you. Sure people do that. So that means when people are physically moving around they are still maintaining their original social groupings. So one of the things is that moving around has become easier because now you don't pay the penalty of not interacting with your friends from the past. But now what is happening is you move into a new environment but your old friends are still around you interacting with them because the technology allows you to do so. But that is eating up in your time to form new friendships. Because that background comes in, right? So there are all kinds of interesting questions that people are asking that how society will be changed. And that can be entirely different. That's a very, very interesting question that you have to think about. So I'm going to talk a little bit about how you know we talked a lot about how the environment is changing. But there is actually a little bit of a historical aspect to it. Which is that one of the trends which has again happened throughout history is that whenever a new kind of instrumentation is introduced it fundamentally changes the way the science in that discipline is done. So for example, you see in 1950s the electron microscope was introduced. Electron microscope was the first time people could actually see what atoms look like. Atoms and molecules. Before that, they could never actually see it. And that actually led to a fundamental change in the way we did chemistry. Because now, it feels like structural chemistry started. Just to give you a simple example, electron microscope was invented in 1950s. In 1943, Watson and Crick they got the Nobel Prize for figuring out the double helix structure of the DNA. They got the Nobel Prize in chemistry. Why? Because people used to use these ball and stick models to figure out the structure of chemical bonds and so on. And it was a pure heat and trial. It was amazing that they figured out through heat and trial about the double helix structure of the DNA. They got the Nobel Prize. Frankly speaking, if they had waited 10 years they could have seen an electron microscope and the instrument changed within a decade a Nobel Prize double discovery to a high school... Well, anyone can actually look through the electron microscope. So that's the power of new instrumentation. Gene sequencing, it has completely changed biology. Because now, things like targeted therapies and understanding bases for different kinds of diseases and so on is entirely things like when the Hubble Space Telescope completely changed the way we look at understanding galaxies and so on. So all these are entirely new areas. Now things like social networks are entirely radically new instrumentation for social science. That's another way to think about it. That before this the only instrument social scientists have had is either observational or inquiry. So for example, you can see this is done a lot by child psychologists. They will go to a little daycare center and see how kids are interacting with each other. Who's fighting with whom and who's making up and they are crying or whatever is laughing. Someone slaps someone and they actually sit down and kind of take notes on that. And people have studied social networks through that. Or they will go and come to a group of adults and ask questions about attitudes. But now of course we have new instruments. Now there is this whole one special kind of instrument that I want to talk about and these are called massively multiplayer games or like World of Warcraft or even immersive environments like Second Life. These are special kind of instruments for data collection. Because how many people here are familiar with any of these? Some people are. How many people are spending lots of time? No one raised their hands. They are all familiar. What are you doing? We used to do it. By the way I used to always get students like I would say that for something like data research assistant to come play video games you get a long line in front of them. They get to play video games and you pay us for doing so. Long line of people standing outside the door. The interesting thing is that these environments are familiar. There can be some realistic ones too. Especially if you look at those mafia wars. They don't have fanciful characters but they can be part of a gang war so there are all kinds of things like that. Point is they have three important elements. All of them are driven by one element in fact these three elements are what make them attractive and addictive. One is called achievement motivation. So you always will achieve something. You will have some kind of scoring mechanisms and when leader board you go to this level then you go to the next level. So there is always a sense of achievement. Then a second one is socialization. It's only achievement. How many people have spent hours playing Tetris? Selling somewhere playing Tetris. There is no socialization which is great from an achievement perspective. A Tetris, Cube Runner, all those little handheld device games. Mobile device games. But the second is socialization. For example you can form teams, you can form associations and go together and so on. And finally the third one is known as immersion. Especially if there is an unfolding story, a complex plot you can become one of the characters. You can pick an avatar. So usually it is achievement, socialization and immersion that drive sort of that is what makes it attractive or addictive. Here is a picture. Lots of people playing. I think this one maybe I don't know where it is from but lots of people playing these games. And of course you know every attraction has been captured. It's all been attracted. Many of them will be attracted to each other and of course for example one of the biggest ones is something like Xbox Live. Many millions of people you can join Xbox Live or going games. So you will be connected to anyone in the world and you can start playing the games. Now the key thing is that by studying these what can you find out about the real world. So let me give you a specific example from social commerce on the social networking side. So for instance here is Levi's the jeans company and what this says is connect Levi's with Facebook to a friend on the site and share on Facebook what they are saying is let us say you go to Levi's online and open an account. Now this is you have seen all those social sign in sign in with your Facebook account. How many people have signed up with their Facebook account? Well you know what as soon as you sign up with your Facebook account you know what you are doing you are actually giving them permission to collect your Facebook data. So now you are giving them permission. Now what happens? So for instance in a sense that say it was Levi's you are connected your friends and then first coming this way and whatever you are doing here is potentially getting published to a content board. So for instance let us say you have a picture and publish it on this thing saying hey I got this pair of jeans. Well now what will happen? Suddenly some of your friends say oh great I love those jeans some of us say I hate those jeans most of your other friends totally ignore it. So just by posting that as a stimulus and how your friends reacted to it Levi's now knows which of your friends even noticed it which of them liked it which of them disliked it and what they were doing versus there are some who look like half a paragraph thing about you know gushing about it right by analyzing all that they can figure out the strength of how much influence you have on them positive negative things like that so all this stuff is going on. So let me go and what I will try to do is probably I might not even go over all the slides and again if anyone is more interested I can provide the full set of slides for 7.2 we will read later on because we are going to look at a study of trust and just to give you a sense for how real serious science can be done using all these video games and things like that so what are some of the big picture questions about trust so for example how do we express trust so for example in different social context sometimes you have a cooperative setting that means in this game environment they are known as a whole bunch of players get together and they are fighting the environment so for example there is a big monster that you are fighting so that's a cooperative situation or it could be adversarial player versus player 2 players are fighting against each other I mean they don't necessarily want to fight each other chess is a player versus player so it's a competitive environment in different kinds of social networks so again most realistic setting will not just have a single kind of social network it will have multiple kinds of social network so for example it could be that if one person is trading with another person so that is one kind of social network or let us say there is a mentoring connection there is a mentor and there is a mentee that's also social network I will talk about a special group called housing then there is something called room also they can be chat so in fact in these environments there are multiple kinds of social networks which really mimics real life because in real life we interact with different kinds of different types of people in different modes and why it is important from a trust perspective is because there are certain interactions which does not require much trust but other interactions are due so for example to chat with someone you don't require to know much about them anyway you can just see someone say start chatting after a while you might realize that you don't really find much in common then you know but another extreme is that would you give someone let us say you are leaving town and you want your plants watered are you going to give them the keys to your house that requires a very heavy trust so you can see that something like chatting with someone trading with someone loaning someone money letting your children go to someone's house to play or giving the keys to your house where they are gone they all have different levels of trust so trust is something that is at the fundamental basis that is why a study of trust is very important ok that there are some roles for instance what role can features drive from trust type of playing prediction tasks for instance lane prediction formation, leakage chain, trust propensity since it's prediction and so on there are three specific things that we are going to study one is trust initiation so what role does social interaction play in trust initiation and what role does trust play in socialization so in fact let us say a good marker of trust is that I give you the keys to my house if that happens now these people really trust each other and before that all kinds of interactions are socialization so one of the questions we are interested in is that how much of socialization is needed before a trust will develop for instance it will be the very rare person who will not know their neighbors and just randomly pick a neighbor and say ok here are the keys to my house that's not likely to happen then of course the reciprocation when is trust is ok for example the standard kind of things that let us say if I leave so supposing I know I two or three times I go on travel and I give you the keys to my house and then later on you leave in some sense there is an expectation that you will also trust me but supposing you don't and actually I get to know that you are giving the keys to someone else then I will feel a little bit of betrayal right so even though trust is a unidirectional relationship but there is an expectation that it will turn bi-directional now let us take something even an activity that requires less level of trust for instance you know someone moves into your neighborhood yes we are talking about all my chat giving away your critical information yes it requires trust I guess that's like a common practice exactly I'll give you the specific marker of trust when you say that but I'm doing sort of real life examples to set some context for instance let us say you invite someone to your house for dinner and then you to the thing once twice after that there is an expectation otherwise you know there is always the social protocols that are involved and all of them have to do with trust and also your vocation so in fact so here are some of the relationships in this environment so a specific environment we are using is called EverQuest2 there is something called chat so you can communicate in-game as the nutrition of the players or there is something called mentoring actual relationship called housing and this one what happens is that you can actually build a house by actually spending some real money you can buy some you spend real money and you buy you give it to Sony the creator of this game and you get some virtual currency we use it to build a house and then you put stuff in it and then you can give someone access to your house and when someone access to your house they have the permission or they have the ability to enter your house when you are not around and potentially use the things you have and sometimes even take them away so therefore when a person gives access to their house to someone else we say that trust has actually developed and it is a one way relationship that means I give you access to my house so you can access my house but that does not mean I can access your house because you give me the permission so it has that whole here is just to give you some and here is showing you the kind of data for instance actually the amount of data we were studying it was 675,000 players every single click over a 10 month period about 3.5 terabytes of data that is what makes it a big data problem in fact as you can see this is only a one month period out of the 675,000 players almost 350,000 of them are participating in the chat in that month 87 million chats trading 295,000 players participated in that or 28.6 million over a nine month period so much less frequent than this one this also mentoring only 86,000 over 9 months even less rare and finally over 9 month period only 64,000 nodes participated in this trust relationship and only 128,000 so one of the important thing this says is that you see this is the easiest activity to do this is a little bit more difficult this is a little bit more difficult in fact you can plot them along these axes here is what we will call low familiarity threshold and a very high familiarity threshold and this is instantaneous and this is long period in some sense instantaneous can be thought of as an activity while long period can be thought of as a relationship so chatting with someone is an activity but if you chat with them constantly then you can say that's a relationship this is a whole activity versus relationship and familiarity threshold versus low versus high here is another way to look at it that if you think from a familiarity threshold perspective over here you like you require a very so this is actually the graph density so if you draw the graph on the same graph if you look at the housing relationship it's very low density and the chat is more than that and the density and the familiarity threshold goes opposite directions for example you require low familiarity threshold on this side you require a low familiarity threshold here for chat to happen and you require very high familiarity for housing to happen and therefore density here is why density is low as you can see there are some of them and some of them are of 6 and 7, 5,000 nodes you can look at a big plot you can see this partnership network instant messaging trade e-mail chat and housing plus as you can see this is far far less dense compared to many of the other ones and chat of course is very very dense So, this actually is a little movie that we made, looked at the 9 month period in 2 minutes, how trust forms. In fact, there is some interesting thing you will see that trust leaks will form first and then they will connect. So, here is what happens right. So, for instance if I know this I trust this person, I trust this person right. Eventually these two people also trust each other. So, trust bees become triangles and then what happens is that they grow and then they eventually you will see that they connect. So, for example there was this group here, then some connections happen now this more stuff starts forming. You see that? This is how and this is leaks answer, this is exactly how it happens in real life. So, this actually is a good visual picture to show that the dynamics of this environment is not that different from what we expect. Therefore, conclusions are drawn from this environment can potentially apply in studying many of the real life situations. In fact, let me tell you some of the interesting things. So, this one right that supposing I have one friend here, I have one friend here right. So, I have this bee situation, bees convert to triangles or bees convert to deltas right. This is social science theory, it is called the theory of social balance. Why or what is the theory of social balance? The theory of social balance is that if I have two friends, they do not know each other, but I am friends with both of them and I like both of them. Then eventually they will also become friends. I will actively try to do that and there is a social psychologist or actually psychologist will say the following that look what happens is I like this persons company and I like this persons company. Right now the two of them do not know each other. Therefore, I have to spend separate amounts of time to get the happiness from these relationships. If they also become friends, then if I go to one lunch, I will get the benefit of both the friendships. So, I am really trying to maximize my return on investment right. That is what I am really cognitively, unknowingly that is what I am trying to do. But there is a problem, supposing these two meet and supposing they do not like each other or let us say initially they like, but after while they do not like each other. Then you know what will happen? Both my relationships are trouble because each of them will say stop interacting with the other one. That is called the theory of balance. So, in any triangle, all three pluses is a stable state, two minuses and one plus is a stable state because these two like each other and both of them hate this one. That is perfectly fine, but two pluses and a minus is an unstable state. In fact, one of the best examples of this is the following. Let us say there are two couples and the ladies are friends and the guys are friends and of course you know each case one man and one woman is married to each other. Let us say one couple gets a divorce. The effect of that starts going into these things because like well it is your friend who is you know the wife is telling the husband that it is you know it is your friend who is the problem and the husband who is the wife is your friend who is the problem. So, in fact these are the kind of things that people call, I mean sociologists have studied this and they call them theory of balance. So, some of those things you see over here. Is this the real time move on? No, no, this is ten months behavior compress into a two minute move. Is it ten months behavior of real time people right? Yes, yes, yes. But I am seeing that one group is not given having one single connection with the other group. No, so over a period of time if you teach interactive. No, ok, there is nothing that happens. I stopped moving in between, but it is entirely possible that ok. So, here is another interesting thing that happens right. So, every group you will see like this one is a complete tree. Like this one is a complete tree right. Everyone has given access to everyone else to the house right. In this case they are not. Eventually it will become like that ok. But here is another very interesting thing. So, I talked about this whole concept that if this situation like this it will become a tree right. Eventually everyone knows everyone else. But here is another problem that eventually how large can a tree grow? There is a finite bond. Why? Because each person has only a certain amount of time to socialize and that is known as the social bandwidth. So, cliques grow. But beyond that what happens is that if a clique becomes too large then the strength of connectivity of each head becomes. So, people start feeling oh this group is not cohesive anymore. And then there is a natural tendency. Whenever any organization where there was a small group of people that interacted a lot with each other and they really enjoyed each other's company and then lot of other people start coming in. Sometimes the resentment say no we are not that close anymore because of all these extraneous factors. That is why very often when a clique gets to a certain size then adding every new person requires approval from everyone else. And therefore becomes harder and harder. And that is where we get the term cliqueish behavior from. Because when a clique has saturated the means that whole group has reached a they are all maxed out of the social bandwidth. And then they are done. Then they won't allow anyone else to come in. And that is the behavior that you see. It happens that you like. I mean in some cases it is a standard constraint. For instance if you have a group of people who go and take off together. It will typically be a force up. It is very hard for a fifth member to join a golf force up because the game of golf doesn't allow a fifth person to get in. So let's say you have a tennis partner. Either you play singles with one person or you have four doubles. So there is no place for three or five. So that is a natural clique. So there are some constraints like that but even in natural. So these are constraints imposed by the appropriate team or a basketball team or whatever. Sports have a very good example of that. But even in the general selling cliques will grow to a certain point. After that they will not connect for the simple reason that saturation has happened. Of course as I said this is not physics or math. This is general guideline rather than actual theorems. What do you see it? You see that. But one more question. What is the effect of the circumstance? You see the old previous slide. There are a lot of groups in the circumstance now. Why it is so like this? Okay. So first of all the geographical topological layout doesn't mean anything. Whether the circumstance is centered in a resistant layout. Okay. Let's go back to the previous slide. You are not from this part. What this says is that there was this large network of people who over a nine month period had positively traded a lot with each other. While there were these people who hardly ever traded or maybe just with a small amount of trade with each other. That's all. There is some sense. These people that you see are very infrequent traders. They are very infrequent traders. And they only trade a little bit with only a few people. While these people have the trade over all of them. So in that sense they are more social and these are less social. Here are some more interesting things. You can draw things like degree distributions. For instance this is law of frequency and law of degree. As you can see that there are very small there are lots of such people with very small degree and there are only a few people with very high degree. In fact this is a law of plot and in mathematics there is something called a power law distribution. If you plot this on a standard non-law of plot it will look like and these are again known as power law distributions or things like that. They just says that none of these is so uniform. There are some people who are very, very highly connected and lots and lots of people who have very low connectivity. That's just the way it is. The nature of behavior, nature of social connectivity, socialization is like that. There are some people who are social butterflies and there are very few of them. And they naturally are over here. And there are lots of people who are just quickly so this is about the picture of let us say there are two people A and B need to trust each other. Either A can initiate a trust or B can initiate a trust to get to this stage. The reciprocation can happen. The evocation can happen. Further evocation can have cascading effect. One interesting phenomenon is that if you revoke trust so one person gives, A gives B trust and let us say a week or two after that people get A trust. But if anyone of them takes away the trust we have found as you would expect that the other person will revoke right away because taking trust away from someone or taking the access to house away is offensive. So negative signals propagate faster in any social network. That's the way to think about it. So positive signals which means I do something positive that has a certain barrier to movement but negative signals they move fast in a social network in a temporal perspective. So we studied all these kind of things and in fact let me just say that how do we go all the signs but only give you some conclusions? For instance we started off with this hypothesis that obviously socializations you can observe and in general trust is not measurable. But what we said is well we will use housing to measure this right because it's a good marker of proxy for trust. And our question was to see does socialization drive trust or does trust drive socialization? That means if two people start socializing will they eventually trust each other or is it that once they trust each other they will socialize more? What do you think? What do you folks think? Socialization drives trust. But supposing people trust each other will they socialize more? Why? For example in real life there are certain friends that you can actually trust but are too worried to spend time with and then you have friends whom you actually trust you don't trust but you're really good friends you don't mind spending a lot of time with them but you will have a problem in case as they say right? Why girls in college or high school love to spend time with bad boys? Also funny because Friends who are fun to party with you won't trust them but you want to spend time with them Right? It's also funny because last week someone gave me a key to water plants so I was laughing when they left the country So which means when they left the country they think you're born here but they're not born here Okay You guys are absolutely right In fact here's the interesting picture So we plotted the degree of socialization over here and on this side we plotted time and this is when trust has actually been formed So what you find is there's some interaction happens it leads to a certain point and then trust formation happens actually the socialization highest when trust formation happens but after that actually drops and stays up to a certain level Okay So the best way to think about this is this It's like this is the courtship before marriage So you have to be very enthusiastic but after that you can say Now you can take that as an example Another example is think about like in a business setting that the sales person has to work much harder before they win a contract but once they've signed the contract you just need to have a certain level of relationship which is a sustainance now It should not drop to zero because it does like even your mother you have to call them at least once a year or mother's day So you need to have any relationship to survive it has to have a certain sustenance level but the sustenance level need not have the same level of order as the winning level and in fact this was observed in all kinds of things So you know these are found in the confusion so I should not just skip any of these things but here's an interesting thing that actually socialization does lead to trust but trust does not high level trust does not lead to increased socialization you can actually settle down and stay at the same level Now I'm going to skip a lot of these other social science things and let's go to the monetization part because that is the reason So I guess the main point about that was that there's a all these social networks virtual environments game and so on so a new kind of instrument to measure human behavior and human interaction in a very fine manner it actually is providing opportunity for us to study social science and models and predictive models in a very fine manner So let's look at the monetization aspect So now we talked about this whole business of trust Now the thing is that trust essentially leads to influence So what is the monetization of influence So what we said was and let me give you a specific example Now I was just assuming that everyone here is familiar with Amazon's type e-commerce So think of it this way I buy books and I buy movies and I have a friend she also buys books and movies from Amazon Now Amazon knows my book and movie buying behavior and knows her book and movie buying behavior as I said And they can build pretty good models of how much I like to spend on books and on movies and on music and how much she is going to But it also happens and it does happen in our family that in general she takes it is hard to admit that she takes my advice or anything but let us say in general I am one of the exporter of books and she is one of the exporter of the music Okay So if I will tell her that here is a book and she might consider buying it Okay So that means part of her buying of books is driven by my influence and similarly maybe the other way for music Now why is this important for Amazon to know because supposing one of us quits So if I quit then my amount of music I bought but also her buying of books my reduce Amazon by just knowing my interaction history with them or her interaction history with them cannot figure that out But if Amazon knew our social network and influence through that then they can figure that out So that was the whole idea that what we did was we said that the total customer value is what they spend because what their social capital is So for instance in this case my value when Amazon is judging my value they should judge it by how much I spend on books and movies but books and music but also how much I cause my wife to spend on books That is my real value because if I leave the system if I turn away that is the total system value loss that will happen to Amazon That is the core idea So that is my understanding So here you are sort of individual people but now if you understand the social connection between them you can understand what the real influence is And then what can we do from that we can figure out customer value, predicting customer loss the influence of social value So there are some people who drive on a lot of social value and those have been from social values and social commerce and things like that And for example you have just a picture So what is social value? Let us say someone has the usual life time value of $43 and but they have a social value of $53 So this person for example they only spend $43 and let's say it's over 1 quarter that's the amount we are saying in 1 quarter they will spend $43 but they can cause this user wife to spend $53 That means the actual value of this person is $96 that's the $43 So now the question is how do you figure out this whole business out how do you calculate this and that's you know sort of the you know this just shows a little dynamic version of it Then you start So the point is if you don't look at the connectivity you get a certain picture you can say okay life time value of this person 100 whatever right then you add you know social value then suddenly you realize that there's more going on and you can you know So it's a question of how do you actually do things and why do we care because for instance if this user turns then it's okay this right now when this person is around this much is being caused but if this person goes away in fact not just that we also have what are called contagion models of churn that means if this person quits if many of them are good friends they may also quit right So quitting which is also called churning has a contagion effect because usually someone quits but they're dissatisfied with something right and they will also their friends they interact with friends so that means they have a negative feeling about the service or game or whatever it is and they that negative feeling will spread so that's why understanding all this is very important it's not just spent but also churn because once churn happens then all the spend is completely gone this is the case of these people are still around but like in this case this person is still around but there will be a reduction in their spend in this case these people could actually go away and this is an actual data this is an environment the model gave about 97% accuracy for what was predicted what was actually observed and so as we calculated we first estimated that this person's social value is about $243.66 we turned out that this person actually churned and the actual drop in the spending was $2.36 you can see this number compared to this number is almost 97% accuracy which is a highly accurate model of what is going to happen let me skip some more of these so what the model of the story is that if you look at a community that means now that we have social networks and interaction behaviors and strength of ties and things like that you should not be thinking about your customer base as a set of people you should think of them as a network of people you saw some pictures of how those networks evolve how those networks can be analyzed and all those things right what this is saying is you can start developing metrics of how those networks drive the time spent the dollars spent the amount of items bought you see time is also important because especially if you think about a social media site then the time spent is directly proportional to the amount of added and regenerated right there is a direct connection between time spent and ad generation so you can think in terms of the community structure and the health of a community as a profit center so the point is that so far we have had CRM managers customer relationship managers right whose goal are CRM programs whose goal is to think about here is a valuable customer how can I keep him or her around how can I make sure that they don't leave how can I make sure that they spend money whatever it is right what is being said here is that people are not individuals they are part of communities and if you can understand the community structure and the strengths of the communities you can start thinking of CRM as more of community relationship management so you say well here is a manager who is not responsible for these 50 people but responsible for these 3 communities that are up to 50 and they have to understand the community dynamics and target target their efforts towards communities so here is a specific example for instance if I think of an individual by themselves my retention strategy might be to give them a discount of some kind but if I understand that they are part of a community I may say look if you bring 2 of your friends we will give everyone a discount so now what you are doing is that you are not only incentivizing the individual you are actually incentivizing group activity and we know that socialization leads to trust so going shopping with your best buddy I think those with double X to start that better than those with X and Y leads to for the socialization so that is the whole idea ok well I think you have already seen the slide here is a picture of we have actually done a lot of as I said there is a whole academic side to this project it is called the virtual world's observatory we have a better URL but that is the one we have 4 people this guy is a non-contractor he is at Northwestern University out in Chicago expert in networks and organization theories this person is called Paul at Illinois that is me when I had hair and this guy is the future Williams he is a social psychologist and you have had interactions with lots of people from sociologists at Indiana at Venice Lady she used to be at Michigan now she is one of the people who have some of the groups at Facebook she has moved to Facebook since then so we have a tie with most of these people this guy is pretty well known Ron Bird is a sociologist at the business school at the University of Chicago he came up with a well known theory of structural holds which is a game for so quite a few interesting people and of course you know in the labs at Bungie and all kinds of data providers anyway so this is the academic side and then this is the commercial side so two of us we decided we will do something commercial with it which is the second part that we saw and he is he is the CEO of the company I used to be the CTO then later on I kind of stepped back and so I am the chief scientific professor in fact, in the previous life on the previous sabbatical I used to be the data mining architect for Amazon in Seattle so that is what I got many of these experiences and the company right now has for 15 full time employees this is some of the funding history of it initially we got about a million and a half from various US defense agencies and NSF and so on through which the basic scientific research was done and some of the core algorithms are also created and then good thing about this money is that they don't ever ask for any equity so from an entrepreneurship perspective you can get free money take it okay the only people who give free money are from your parents so maybe that is a string attached then after that we raised a million dollars in series A okay and then we kind of did a series B but there was more as you can see it is a much smaller series B usually very often it was a proper series B it would be like almost three times as much but we didn't have to raise quite as much in technology terms we actually filed three patterns one pattern is already been granted in the core area of the social value influence one is on social contagion models put on and things like that right now the revenue status is that the company is spending about almost 200,000 dollars a year it's bringing in about 135,000 and 40,000 so it's not revenue shall yet it does have runway till the end of this year so either they have to reach the revenue neutral stage or they have to go and raise more funds or get acquired so I think those are the potential paths there are conversations with various people including Facebook because this would be a perfect technology for Facebook news for example but there are a bunch of conversations going on and in terms of markets gaming is obviously a natural one then also entertainment and e-commerce it really isn't this order it's significant because gaming and entertainment are quite close to each other especially these guys my friend he's a professor at USC University of Southern California and the company is based on LA they have a lot of connections with Hollywood and all those guys and Disney for example so they're doing quite a bit of work with those guys gaming and entertainment is the main one and e-commerce is a much bigger space but the challenge is also that e-commerce also has many big well funded players so only if it succeeds in these two then eventually it will go to e-commerce and maybe it will just get bought thank you folks well I'm around for a while if you're willing to talk are there any questions? are you on a slide where social networks are changing the way human interactions and behaviors are measured and you also showed a slide where the trust is formed what was that slide based on on this particular game that trust factor and then you also said a real-life situation were in sales which were like up to a certain level you perform well and once the trust is there the interactions that one was also in the same game environment so those interactions were like trade and chat and all those activities these are all in-game activities and then after that the the question was that at that point that trust part was when someone gave access to their house again it is all in-game and then after that did they continue to interact with each other like trading and so on it dropped a little bit I just want to know based on this like based on this matrix are like the real-life human interactions and behaviors are studied and since you talked about social psychology and I don't know if you're tracking human behaviors so in case you are measuring real-life human behaviors based on online behaviors I'm sure that there is a difference and what is that difference is it like small difference like 10% or 20% because who we are online is not mostly who we are offline so I don't know if that comes under part of your work I'm sure like QCRI also does some humanitarian things okay so I would say that the the whole question of how much of our online behavior is reflective of our offline behavior right of course I mean potentially do different kinds of things different kinds of activities sometimes we are doing the same thing which is talking to people like we are talking to face to face but we are talking so obviously the kind of activities that we are doing like let us say sitting together having a behavior there's a different kind of socialization different context than let us say talking to the phone about a homework or something so obviously the activities are very different that we do online and offline obviously the behaviors are very different but the most of it has been studied that online offline comparison between the two if in the context of similar activities and of course the one that people have studied the most is communication so there's a whole concept called presence if you talk to any of the people who design even from user interfaces and so on they could talk about the world concept of presence and the communication people call it telepresence the idea is so in real life we have a certain presence you can see how I look, how I hold my hands and how I speak or just articulate and all that business so I have a set not only am I speaking