Campus Spotlight: Hari Sundaram

>> Hi. Today we're here at the Siebel Center for Computer Science here at the University of Illinois. >> The Siebel Center was built in 2004. Thomas Siebel donated $32 million to the University of Illinois for this building. Currently, the building houses the University of Illinois computer science department which is one of the top five computer science departments in the nation. >> Great. Today we're here to speak with Professor Hari Sundaram. Hari's new to the University of Illinois. He holds joint departments, both in the department of computer science and also the department of advertising. Hari does cutting edge research on how digital tools such as smart phones can be used to change and influence human behavior. You ready Jolie? >> Yeah, let's go. >> Let's go. 

We're here with Hari Sundaram. Hari, thank you for the opportunity to visit with you today. >> My pleasure. >> Can you tell us a little bit about yourself and the work that you do? >> Sure. I'm an associate professor in computer science and I have a joint appointment with computer science and advertising and-- but my primary appointment is in computer science, so much of my work deals with analyzing collective behavior in online networks. So, an example of this would be how do you discover group formation. What's technically in computer sciences we call community discovery. How do you discover groups that form online? And some of the other problems that we have looked at is that how do you discover the rise of coordinated behavior. So, I don't know if you'd noticed that sometimes when on Facebook if there's a like on a particular post, the next time you log in there are literally hundreds of likes. It's almost as though all of this is happening simultaneously. So, how do you discover the rise of coordinated behavior? It'd be a building of mobile network laboratory. That is, a laboratory that runs on people's smart phones. So, the idea is that if you want to understand how people make decisions in the presence of social signals, that is, knowing what your friends are doing at this particular time, if you want to understand how people react to messaging in situ, like if you go to Walmart and you see a particular message, how will people make a decision then. You need to have infrastructure that allows you to measure reactions to different circumstances or contexts. So, that's the kind-- so I'm interested in networks in general, outside of Facebook and Twitter. So, for example, there's a lot of research in behavioral psychology that suggests that when individuals make decisions we are influenced by decisions of our friends as well as people with whom we can relate to. So, for example, as a trivial example, if I am walking in downtown Chicago and I walk past an Indian restaurant and I see the restaurant with a lot of other people from India eating there, then it's a cue that perhaps these people find value in this restaurant and I should join them, even though I don't know them. So, I'm trying to make use of this-- of these principles from behavioral psychology as part of my research in how people take decisions in social networks. >> OK. >> I was just going to ask if this primarily focuses on social media then because you had mentioned Facebook and Twitter and all these different-- >> Yeah, it does. So, it began-- a lot of my work begin with deep interest in social media. I was interested in understanding how people cooperate online. How do-- you know, how does meaning form through consensus online. So, that's how it began, but then I began to realize that networks play a much-- a really important part in our lives. If you begin to think of networks, not just Facebook or Twitter-- >> OK. >> But in terms of friendships, right? All of us have had friendships for millennia. I mean, human beings have-- the bases of human society are these strong relationships that exist between the individuals. If you start thinking of networks as preexisting outside of social media, then you begin to ask yourself how is it that individual decision-making is affected by these relationships. >> And you mentioned that these relationships may be formed around a common affinity, so could that be a brand, a product? >> Could be. It could be, for example, all those people who like the New York Times, for example. So, it could be that many of the people who read the New York Times have many other attributes, perhaps all of them are liberal, for example, or some of them have higher incomes that allow them to have affinity with each other. >> OK. How about things like automobiles, so things like Jeep is known for having a strong brand community, for example. >> Sure. Absolutely. So, yes. You can definitely form interest groups where you can join groups whereby, you know, you profess an interest for that brand and all the different dimensions that the brand entails. >> OK. So, I know you're doing this cutting edge research, do we see any firms that are utilizing these techniques to help market their products? >> So, many of the-- so all this stuff that I'm doing is still in its infancy, even outside. So, people are very interested in using social media for marketing purposes, for example. It's been widely-- for example, it's been widely used in terms of product placements on blogs, for example. It's been used in wire marketing whereby, you know, individuals go to bars or hotels or other places and they profess to be using a particular product. The big question that I'm trying to address is the question of how do we persuade millions of individuals to adopt behaviors that are beneficial to them, right? An example would be can one reduce their energy consumption. 

All of us are interested in leading a more healthy lifestyle, right? So, if we can persuade people to adopt more healthy behaviors, it would have a huge impact on society, right? Not only would people lead healthier lives, they would lead more productive lives. As a direct consequence, insurance rates will fall, right? If you start thinking about reducing our energy footprint, you know, it will have a huge impact, cascading impact on sustainability. Right? So, that is the framing from which I come from. So, when you think about authenticity in that context, I'm interested in having people sign up for networks for behaviors that they want to adopt, but somehow can't seem to manage. I want to eat healthy, but for whatever reason I happen to be eating more steaks every week than I should, right? How do I persuade myself to be more healthy? So, what my goal is to embed you in a network of friends and people that you can relate to and you will start now in my network you'll start seeing messages. Promotional messages perhaps from businesses, but also messages from your friends saying that-- an example of this message would be that Hari, it's 11 a.m. today. I know you're about to go eat lunch, but 7 out of 10 of your friends have just gone to eat a salad. >> So, you'd get this message on your phone-- >> On your phone, yes. >> It would pop up on the screen. >> It would pop up on the screen and so these are the kinds of nudges that slowly help you attain the behaviors that you want. >> So, how does this work? How does this message on my screen nudge my behavior to eat healthy versus have-- >> So, the idea is that you're connected, so you're going to be influenced by the people that you see. So, if it is your friends who are adopting the behavior, then it's more likely that you will adopt this behavior. Or if it's people that you admire or have an affinity with, you're much more likely to adopt the behavior. So, there's a lot of research that talks about, you know, if there-- based on the social norms, knowing the social norms makes a big difference. >> So, the social norms, perhaps peer pressure? >> Peer pressure, but social norms make a big difference. So, this familiar towel study-- I don't know if you're familiar with this. In this towel study, what they did is that they put these messages inside hotel rooms that said that-- and these statistics were true, they said 75% of the people who have stayed in this room have reused the towel. >> I've seen these signs. >> That's right. This is partly due to the work of Robert Cialdini, who's worked extensively in persuasion. 

This is-- the adoption statistics roles. So, that's the idea. So, if you begin to understand what the social norms are, what is it that your friends are doing, what is it that people like you are doing, might you be more inclined to adopt? >> As you mentioned, this desire for collective action and belonging is, you know, is a deep human trait and that these behaviors and these social norms and peer pressure also, you know, quite old. So, what is new here in terms of the digital side? So, how are these digital tools like the smart phone changing these very traditional, you know, human traits and motives? >> I think the biggest difference is scale. So, so far, if you think about the role of computing in many of these complex human questions-- these are not computer science questions, these are human questions that intersect, you know, psychology, economics, and now technology. So, what has particularly changed in the last 10 or 15 years-- 2 things have happened. One is that the cost of communication has dramatically reduced. People have access to devices that are essentially full-fledged computers, a.k.a. smart phones, right? And I think the other thing that has happened is that technologies such as Flickr and Facebook, Flickr and Twitter, what they've shown us is the extraordinary desire for humans to be with each other, right? Be online. You know, this desire to connect with other people. We've always known this, but Facebook has shown to the extent at which we seek this. Right? So, I think the key to technology is I think going to play an increasing role as a glue in solving some of these big complex questions that humanity has found itself confronting. >> So, as you know, this course is called marketing in a digital world, we talked about the digital aspect just now, how about a bit about the marketing side? So, this as a tool for social good, like sustainability and healthy eating and energy usage. How do you envision firms using this or consumers using these tools to change marketplace behavior? >> So, excellent question. So, for example, consider this idea that, you know, I've joined the social networking app to promote-- so that I become more healthy. Right? So, I have a vested interest in becoming more healthy, right? And I'm-- now, let's say the marketplace knows this in terms of businesses know that, you know what, Hari has a tendency to eat healthy. He walks at Siebel Center. Can I incentivize him to come to my restaurants, make him aware of the kind of products that are available, for example, dishes that are available in my restaurant that he can partake that is healthy? So, first is promotional messages that can come to me when I'm nearby. Second thing is the cascading effect of being healthy, right? So, let's say I adopt these healthy behaviors and I've become more healthy. 

I call my insurance company and say look, I've really adopted a healthy behavior, healthy lifestyle. Cut my insurance. So, naturally the insurance company's going to come around and say where's the proof? It's like, I'm not going to believe you. And a very important aspect of my work is behavior verification. So, I claim that I've done something and can I prove to another person that what I claim is accurate. Now, if the insurance companies know that Hari is very healthy, leads a very healthy lifestyle, goes to bed early, sleeps adequate number of hours and so on, they might cut my insurance rate. So, you-- I think you'll start to see-- provided individuals are willing to expose their behavioral history in return for some sort of a, you know, cut in prices, you'll see a very interesting dynamic taking place between businesses and individuals now. >> So, going back to what you were saying before, I think privacy might be one of the issues with social media and digital marketing and all of that, but you said people would be migrating more to perhaps show, you know, companies their behavior. Why do you think they're going to make that change? >> Well, only in response for some sort of an advantage, right? So, this-- we're already seeing this. For example with Progressive, an insurance company, car insurance company, now puts in small monitors in individuals' cars, right? Now, in response for good driving behavior, Progressive says that it'll give you a cut in your insurance rate. So, only if I believe that the company that wants my behavioral profile will give me some sort of a price cut, but I totally agree with you. In return, it also has to say that it will honor my privacy. It will not resell my data to somebody else without my permission. >> Can you give us a demonstration of how this works? >> I'll be glad to. So, what this dashboard allows us to do is that it allows us to configure and generate apps that have a very specific focus. So, this infrastructure allows us to create social networking apps that are across platform, across Android, iPhone, and Windows Mobile, and what it allows-- this particular dashboard shows the experimental configuration. So, you know, experimenter could create a database, could select classes to be created, could select things like what kind of network topology they want and, you know, and add any special cases as they need. For example, in this case a friend. They could also configure things like what to show for any user. So, you know, you could have thresholds. What kind of thresholds do you want to use to trigger positive or negative messages, number of messages per day, any particular time messages ought to be displayed, and things like if you're trying to understand the relationship between messages and influence, so as I discussed earlier, we need to start thinking in terms of positive and negative framing of information. So, begin figure out which users are part of the experimental group who get positive messages, who get negative messages, and the control group that gets neutral messages. And there are some other features here that allow an experimenter to think about message display based on granularity including, you know, time, including space, as well as network. So, the first app that we've created with that infrastructure is an app that focuses on healthy living. So, the idea is that you want to persuade people to have healthy eating and exercising behaviors and we want to run a controlled experiment that begins to understand the effects of different kinds of information framing, the different kinds of network structures that persuade people to eat healthy and exercise, right? So, we created this app. So, what you do is that you login with Facebook and it's a pretty standard procedure. You'd find this on many different apps. So, this is an example of Sanjana who is enjoying a apple Danish. So, if you go down, this is my breakfast for today. I'm also eating the same apple Danish. You can like if you want, you can enter a comment, and, you know, and the friends in my network have entered these comments. So, if you go down, you'll find that you'll see activities of people in your network who have posted things about where they've gone and what they have-- photographs of what they've eaten and so what we also have are these informational messages that attempt to persuade you to subtly change your behavior. So, for example, this one says if you're a serious grad student, this would be of most relevance to you. It says four out of seven serious grad students have eaten healthy over the past week. You get some other statistics like 63 out of 100 of the people in the network have eaten in North Urbana, so this is an example of a special aggregate. So, all these messages that you're right now are positive healthy messages and that's because I'm part of the control group that gets these positive messages, right? 

So, you can go down and you can see, you know, other messages. For example, this particular message says that if you have a very large network, if you and your network on average walk 62,000 steps you'd have collectively walked a full circle. Of course, here my network is extraordinarily large, so that is actually possible. So, you go down and you know you get other kinds of statistics as well. You also, for example, you also get tabular information. We selected tables in part because graphs are very hard to read on a smart phone. It's very easy to generate a graph, but it's very hard to read the statistics and, you know, we allow you to generate graphs here. So, what else can you do? You can obviously-- again, it's a social network, you can take photographs of food that you've eaten. So, in this particular application, the idea is that you decide-- so you decide to-- what you do is that you several things. In this particular application you do several things. You figure out how much money you want to spend. OK, so in this particular case, if I want to spend only $9, I would have this setting. And I have this setting that says vegetarian and so the minute setting is very interesting. For example, if I set it to 60 minutes, it means that the application should give me restaurants or suggest restaurants to me where I can find food for less than $9 and I can be back in my office within 60 minutes. Right? So, I save these preferences and then when I do a search, I can-- immediately I have restaurant suggestions that are based upon my preferences and as you can see many of these restaurants are close by. There, you know, for example, Don Juan Cocina is within five minutes and the average cost of something that I can eat there is around $6, right. Now, for example, then if choose Don Juan's Cocina Mexicana and Cantina, I get this nice map that I can just post and say that I'm going. OK? So, once I say I'm going, then my entire network gets to see this information. It says that Hari is going to Don Juan Cocina Mexicana and Cantina and so my network is informed, right? And then once I got to my network, I go there and in this application you're supposed to take a photograph of the food that you've eaten which you post, right? So, in our network, you get points only if you take photographs of what you've eaten. Now, of course it's very hard to take-- to understand from a photo if the photo is healthy or not, so at the back end we have another framework that farms out these photographs to other members of the network who don't really know that it is Hari who has posted, so they will anonymously rate this food, that actually gives me a healthy rating of whether I've eaten healthy or not. So, the key point about this infrastructure is that it allows us to rapidly prototype a wide variety of applications, not just for health, but we are also developing an application for public safety. You can also have an application for entertainment. The key idea is making apps that have a specific purpose that solves a problem that people actually have. 

Once you create such apps that are engaging, at the back end, the experimenter can run small experiments to understand the effects of different kinds of messaging, different kinds of topologies, different kinds of, you know, of settings such as, for example, what happens if at 11 a.m. you get a push notification on your smart phone that says 7 out of 10 of your friends are eating a salad today. So, it's very hard to evaluate these questions on existing social networks like Twitter or Facebook because we don't really have full control of what members of the network see and when they see it. 

>> It's 2014; where are we in 10 years down the road, 2024, 25? >> It's hard to predict the future, but I'm going to use a quote from one of my favorite authors, William Gibson, who said that the future has already happened. It's just unevenly distributed, right? So, I think the biggest thing that we will see is that we will see technology transform mundane things. We'll see transform things like adoption of large-scale behaviors like becoming more healthy. >> Looking at technology, here you have a demonstration of, you know, the smart phone. What other tools, digital tools, might play a role in this future that you've laid out? >> So, I think an important part of it is that when I say technology is infused everywhere, I mean that it could be in the very fabrics that we wear. So, I think wearables will become much more important. They may not become prominent necessarily as prominent as the, you know, forthcoming Apple watch. It could be, for example, some of the fibers woven into your shirt and that could be tracking things like heart rate, it could be tracking things like, you know, whether you're breathing properly. Right? Obviously, the World Wide Web will continue to play a very important role in terms of connecting people. 

>> Certainly. >> Very fascinating stuff. Where would someone go to find out more information about the work that you're doing? >> So, the best, best place would be my website. Would be sundaram.cs.illinois.edu or they should get in touch with me over email. >> Great. Thank you so much! We appreciate your time. >> Thank you. Thank you so much.