Sinan Aral: While there has been misinformation and disinformation throughout human history, we’ve never had a technology that has essentially rewired the central nervous system of humanity within one decade, that accelerates the spread of information, as much as social media does, in an algorithmically controlled fashion. 

Noshir Contractor: Welcome to this episode of Untangling The Web, a podcast of the Web Science Trust. I am Noshir Contractor and I will be your host today. On this podcast we bring thought leaders to explore how the web is shaping society and how society in turn is shaping the web.

My guest today is Sinan Aral. You just heard him talking about the impact of social media in spreading misinformation and disinformation. Sinan is a global authority on business analytics, award-winning risk researcher, entrepreneur and venture capitalist. He is the David Austin professor of Management, Marketing IT and Data Science at MIT, where he also directs MIT’s initiative on the Digital Economy. He’s also a founding partner at Manifest Capital. Sinan has won numerous awards, including the Microsoft Faculty Fellowship, the National Science Foundation Career Award, and the Fulbright Fellowship. In 2018, his article on the spread of false news online was published in Science. It went on to become the second most influential scientific publication of the year in any discipline. And in 2020, Sinan published his first book, the Hype Machine, which received  a best book on AI award from Wired Magazine. Welcome Sinan.

Sinan Aral: How you doing Nosh? Great to see you.

Noshir Contractor: Thank you so much for joining us on this podcast. You’ve had quite a year. I want to first start by asking you what got you interested i looking at information on the web?

Sinan Aral: Well, I was a PhD student at MIT and I knew I wanted to understand technology, I didn’t know exactly how I wanted to get into it. As I was studying at MIT, I was taking statistics classes that assumed that all of our observations in the data were independent. And I was taking sociology classes with pictures of network diagrams in the research articles that uncovered the tremendous interdependence of our world. And I thought, a lot of the models of society could really be explained by the ebb and flow of information between us in our complex interdependence. The major thing that’s different today in that web of interconnections and the flow of information, even though the human species has been interdependent for a very long time, is the digital flow of information. And so I thought that’s where a lot of the answers to unexplained phenomenon in society were going to come from. And I’ve been researching it and studying it ever since.

Noshir Contractor: You talked about this web of connection in the digital economy, which is exactly the focus of why your work is so central to web science. Can you give an example of what you mean by an unexplained phenomena in society that can be well studied through these means.

Sinan Aral: So for instance, businesses try to understand demand patterns, for example, you know, it’s thought about in terms of the products that are sold, it’s thought about in terms of consumer preferences. And for many, many years, the thinking was that there’s some distribution of product characteristics. different products exhibit different characteristics. There’s some distribution of consumer tastes — different people have different tastes. And you know, there’s a match between these products’ characteristics and consumers’ tastes, and that can explain changing patterns of demand for different products over time. But one thing that has become clear over time is that people talk to each other. And they share their own opinions about products, and now we have digitized that communication in the form of social media likes, comments, and so on. And in the form of reviews and ratings where we are constantly describing to each other in digitized recorded mass scaled format, what we think about — and not just products, political candidates, you know, bills before Congress, you name it. And it turns out that the patterns of choice that human beings make, whether it’s voting, or whether to get a vaccination, or whether to buy a certain product, are heavily influenced by the patterns of communication and sharing of information about those political candidates or public health behaviors like vaccination or products. And so a significant fraction of the variance can be explained by the ebb and flow of information online, whether it’s through person to person communication on WhatsApp, or through microblogging services, like Twitter, or whether it is reviews and ratings from the crowd. The scaling of public opinion is changing the nature of the way we decide and act.

Noshir Contractor: I want to take you back to a decade ago, where you published one of the first articles that looked at doing natural experiments in the field. So to speak, I’m thinking back of your article in Management Science, titled Creating Social Contagion Through Viral Product Design. That article was one of the first attempts in my opinion that tried to look at how word of mouth campaigns could be digitally transmitted. And you did this experiment on Facebook that I would love for you to describe. And in particular, tell us about what got you to think about the question, but also the strategies that you used to answer those questions, which was quite novel at the time.

Sinan Aral: Our experiment was an experiment in the wild among 1.5 million Facebook users. So it was perhaps the first very large scale online experiment about the causal effects of information flow in networks on behavior. And the reason I became interested in it is to basically scratch an itch, which is how a lot of research starts. And that is to understand really the causal effects of networks on outcomes and behaviors in society. What we did was we developed an app for Facebook, along with a movie studio. And it was a movie app, you could friend celebrities, buy movie tickets, rate movies. We wanted to understand this concept of what we call viral design — Can we design products that are more likely to be shared amongst friends? So we built several features, a invite your friends button, and then a passive awareness campaign feature. The first feature allows you to press a button, there were buttons throughout the app that invite your friends, that showed you a list of your Facebook friends, you can pick who you wanted to invite and invite them. The passive awareness whenever you took a key action in the app, like rated a movie, it would send a message to all of your Facebook friends that said, :Hey, Sinan just rated this movie four out of five stars, you might be interested in the app, here’s a link to download it.” We created three versions of the app, a control group with neither of these features, and two other experimental versions of the app with these features. And then as people downloaded the app, we randomly assigned them to one of these three versions of the app. And then we just simply observed the sharing and diffusion of the app through the Facebook network. And any differences in the speed, breadth or depth of the diffusion of any of these versions of the app has to be causally driven by the existence of these features, because we randomly vary just toggling on and off those features. And that was it. And we found very significant differences. And then we were able to study, what is the power of a personal invitation? What’s the power of a passive awareness campaign? And as well, what are the actual differences in the rates at which apps spread with or without these features?

And we found evidence of network effects that when you use the app with your close friends, you were less likely to give the app up. If you use the app with acquaintances, you were more likely to give it up, which indicated that network effects varied depending on the closeness of the relationship. Now we’re doing massive scale experiments to measure that across all the platforms, Instagram, Snapchat, Facebook, Twitter, and so on. Because we think that’s such an important part of the economics of social media today.

Noshir Contractor: And one of the things that I think distinguishes that kind of work you do, this carefully controlled experiment where you put people in different conditions, helps address an issue that has plagued prior research in this area. I’ve heard you talk about this before — talk about that analogy you’ve used here.

Sinan Aral: When we, as network scientists or web scientists study patterns of behavior in a population, sometimes we tend to foreground the explanations that we are investigating, and tend to ignore that there could be many other confounding effects. So when we study the diffusion of a behavior through the network, and we see that Wow, people who are connected to one another, tend to exhibit this behavior in succession, one after the other rapidly, we say, “Wow, there must be some sort of network effect here that that this behavior is passing from person to person to person, because it only shows up in the network in succession between people who are linked together, rather than more randomly in the network of people.” And that seems like a logical conclusion. But it’s not always a logical conclusion. In fact, many times it’s not a logical conclusion. 

So the analogy is to a crowd of people in a field watching a political rally or a concert, and you see the first umbrella go up in the bottom left hand corner of the field, and then immediately the umbrella next to it opens. And then the umbrella next to it opens. And then the umbrella next to it opens. And you see this crowd of umbrellas, opening one right after the other from the bottom left hand corner of the field to the top right hand corner of the field. And one explanation for that is that the first person, open their umbrella and then nudge the person next to them and said, “Hey, open your umbrella, you know, it’s the cool thing to do, everybody’s doing it.” That’s one explanation, the social influence explanation. But another explanation that’s much more likely, is that there’s a passing shower that is moving dynamically from the bottom left hand corner of the crowd to the top right hand corner, hitting raindrops on the heads of the people. And that is what’s causing them to open their umbrella. And so I use that analogy, because sometimes as web scientists and network scientists, we assume social influence, where it’s really probably some third factor that’s causing that pattern of behavior and the data. So we have to be careful about correlation and causation, if we intend to be rigorous about network effects.

Noshir Contractor: Sinan, when you began working in this area, I got the sense that a lot of your interest was to see how products could be made viral — how they could be designed so that we could have social contagion, and do it in a way that was scientifically grounded and effective. Somewhere along the way, I got the sense that what you thought was going to be a positive set of strategies began to trouble you where information was being sent, or misinformation was being sent. How did that happen? 

Sinan Aral: I mean, I think that technology is agnostic. And you know, one of the themes of my book Hype Machine is that if we intend to solve the social media crisis that we find ourselves in, we have to get past this debate about whether social media is good or evil? Because the answer is yes. And we need to understand how it can promote positive outcomes in society and how it can be dramatically negative for our democracies and our economies and our public health. And so I’m interested in both the good and the bad, in part to promote the good and counterpart to contain the bad or to work to build systems that reduce the negative effects of social media.

Noshir Contractor: Let’s talk about the book, the hype machine. I love the title. I’m going to first ask you, how did you decide on the title of the book, which I noticed is not only the title of the book, but also the title of chapter three in the book: The Hype Machine.

Sinan Aral: I considered a lot of things. I thought carefully about the chapter titles. And it’s super interesting to have that conversation with you. Because I’ve done many interviews over the last year. And nobody’s asked me that question, I’ve been waiting for someone to ask me. You know, I titled the book, the Hype Machine, because social media is built on an engagement model, a business model on engagement. And the way it works, obviously, is that social media companies sell attention as a precursor to persuasion. And they sell that as ad inventory. So the way that they maximize the opportunities to sell advertising, is to engage people, and to keep them engaged. And so the machine is designed to hype us up. And that’s where the title of the book comes from.

Noshir Contractor: You have a chapter that is titled “The End of Reality.” And that sounds very depressing, Sinan. Tell us about how that title came about, and what’s the thesis of that chapter.

Sinan Aral: In 2018, we published a 10 year study of the spread of fake news online. We worked directly in collaboration with Twitter and had access to the entire Twitter historical archives. And we studied the spread of all of the verified true and false news stories that ever spread on Twitter over 10 years. And what we found was that false news traveled farther, faster, deeper and more broadly than the truth in every category of information that we studied, sometimes by an order of magnitude. While there has been misinformation and disinformation throughout human history, we’ve never had a technology that has essentially rewired the central nervous system of humanity within one decade, that accelerates the spread of information, as much as social media does, in an algorithmically controlled fashion. 

And so we’re at a particular moment of risk. And in this chapter, I discussed the potential impact of the spread of falsity on democracies, economies and our public health. I talk about coronavirus misinformation, I talk about meme stocks and how misinformation can affect the stock market. And I talk, of course, about democracy and elections: the 2016 US presidential election, the 2020 presidential election. I talk a lot about deep fakes and the science of fake news. What is it about human cognition that makes us susceptible to falsity? And what some of the solutions might be? In this book, I try very hard to remain rigorous throughout the book and allow the science to lead. And you don’t need to exaggerate it, for it to be dramatic, because there is dramatic, rigorous science out there that should be compelling enough to motivate policymakers and platform designers and leaders to the potential peril that we face with social media, as well as the tremendous promise.

Noshir Contractor: I’m aware of several people, perhaps Mark Twain, and you can correct me if I’m wrong here, who, a long time ago, said that lies can travel twice around the world before truth could put on its boots. So why is it that today, the web might be changing that particular phenomena? And is it simply a question of the lies traveling 100 times around the world before truth could even put on its socks, let alone its boots?

Sinan Aral: The phrase fake news was first mentioned in a Harper’s Magazine news article in the 1920s, so you’re right when you say that, you know, fake news is not new. And it’s interesting, the Mark Twain quote, is actually not a Mark Twain quote. I’ve heard that quote, attributed to so many different people incorrectly, which itself is ironic. To bring it back to current times, there are a couple of things that make the spread of falsity today particularly dangerous today. And that is speed and the algorithmic amplification of falsity, as well as the targeting that happens. So we don’t know who is seeing which information. The speed with which information travels today is nearly instantaneous. And it is much faster than even just a decade ago in terms of the spread of false news, but also true news. This favors falsity, because we’re 70% more likely to share a verified false news tweet than a verified true news tweet. 70% more likely, over a 10 year period of all the tweets on Twitter, that’s a big number. And false news travels about six times as fast as true news. The entire planet could be mistaken about something, about a consequential choice that needs to be made, and if the truth doesn’t catch up quickly enough, then we can make very significant errors. 

Elections are a dramatic example. The rise and fall of equity prices are another example. And then finally, whether or not you get vaccinated in time to stop the spread and whether you can achieve herd immunity against coronavirus. These are three dramatic examples. The hype machine was published in September 2020. And in the book I predicted all three of these things would happen. In the book I said that we were gonna see violence during the 2020 election. We had the Capitol Riot in January. I predicted the rise of meme stocks. We saw the GameStop stock price rise. I predicted that misinformation around vaccines would create protests around vaccines that would disrupt the vaccination process in the United States. In January 2021, we saw Dodger Stadium in Los Angeles shut down by anti -vaccine protests. And I don’t consider myself an oracle. These outcomes were entirely predictable. We’re at a particular moment in history, where the spread of falsity can have even more dramatic implications than it has in the past.

Noshir Contractor: Amazing. Now another chapter title: and Networks Gravity is Proportional to its Mass. 

Sinan Aral: I wanted to make a point about economics, but in a way that non-economists might understand. And you know, people who have taken grade school physics, understand mass and gravity in a sense. And really the point I wanted to get across in this chapter is how important the economics of the social media economy are to the outcomes we see regarding democracy, public health, our economy, stock market, business outcomes, and so on. And the main economic force that shapes the social media economy is what economists call network externalities or network effects. And that is to say that the value of a platform or a product is the function of the number of people who use the platform or product. And so the size of a Facebook network, the size of Twitter’s network, has implications for the attraction it has to new users and the stranglehold it has on current users. We’ve got a big conversation about whether we need to break up Facebook and whether we need to break up big tech, we’re worried about the rise of monopolies in the social economy. And what people I think are missing is that in order to really create competition in the social economy, we have to deal with the structural economic phenomenon that are creating market concentration. And that really is driven by network effects. And in an economy driven by network effects, big platforms have big power, they have power to attract new users, they have a stranglehold on current users. If you want to speak with your friends and family, you can’t leave Facebook, Twitter, and the major platforms today, because that’s where everybody is to talk to. And it’s hard for new entrants to get new users because of the network economics of the social economy. And so a lot of the outcomes we see, in terms of misinformation, in terms of effects on democracy, economy, public health, in terms of the very nature of competition in the social economy, stems from the simple fact that a network’s gravity is proportional to its mass, the amount of power that a social network platform like Facebook has, is proportional to its size. And if we want to address that, from a policymaking standpoint, or a business standpoint, we have to make structural reforms to the economy itself, that address network effects. And as I described in the last chapter of the book, that means instituting interoperability and social network and data portability, which is itself the main structural reform to the social media economy that will have the single biggest impact on the level of competition in that economy.

Noshir Contractor: Let’s go to that final chapter, you title it, “Building a Better Hype Machine.” What you just described was, the larger the network, the more sticky it is, the more likely it’s to keep you down on it and not let you go away from it. How does interoperability solve the problem? And why would the large networks today have any incentive to participate in that?

Sinan Aral: That’s the main question. And I think the answer is obvious. They have no incentive to change, because they’re making money hand over fist, consumers have no choice but to use the platforms that have the greatest network externalities, because that’s where everyone is to talk to. Now, it’s interesting that you started with the 2011 paper that we wrote, where we first began to measure network externalities in the social economy. That was a decade ago. Now we’re doing a very similar experiment across all the platforms to measure, well, how big is the network effect for a Facebook or a Twitter or Snapchat? And what can we do to break this network effect? So if I was to hypothesize to you, how would you feel if you couldn’t send a text message from Verizon to Sprint? You think I was crazy. You think I was absolutely insane — What do you mean, I can’t send a text message from one mobile carrier to another, of course, I should be able to do that. And then I say, Well, why can’t you send a message from Facebook to Twitter, or from Instagram to Snapchat? And suddenly you think to yourself? Hmm, that’s interesting. Why can’t I do that? Well, the reason is because they have made themselves incompatible in order to retain their network externality and their network value.

And because of that, consumers have no choice. And when consumers have no choice, and they can’t switch from one platform to another, without incurring significant cost, then there’s no incentive for that platform to give consumers what they want, which is a clean internet that protects their privacy, that reduces bullying, that reduces misinformation, that does something about social media manipulation during elections, that reduces the amount of hate speech on the platforms and so on. Because all of that stuff is engaging and profit maximizing, it’s allowed to continue. But there’s a bill in front of Congress now called the Access Act, which would require any social media platform greater than 100 million users to become interoperable with other social media platforms. If that were enacted, and if I came up with a social network that said, I would protect your privacy, I will eliminate phishing on my platform, I will have tight security, and you can speak to anyone on Facebook, Instagram, Snapchat, and Twitter, then people would be able to choose that social network. And as more people chose that social network, the larger platforms would have to make reforms to provide similar levels of privacy and security in their platforms, and so forth. That is an example of how interoperability creates competition in networked industries.

Noshir Contractor: Now, for a long time, even before, before we got to social media platforms, the notion of having some kind of interoperability goes back to the very start of the internet. The brilliance of the web was in large part based upon HTTP, the hypertext transfer protocol and allowed interoperability. Do you see that there is any reason to believe that through legislation or technological innovation, we will see a breakthrough in this interoperability dilemma that we face right now?  

Sinan Aral: I’m sincerely hopeful that we do because I think this is just an obvious point — imagine a balkanized internet where you had not a global internet where you could share information, but you had many different privately controlled, intranets that people were part of and had to pay to be a part of, and to get access to information and so on. The level of innovation and collaboration and communication and life-saving health information, and everything that is shared, because the internet is interoperable and is free to build on is staggering. If you don’t create interoperability, you really stifle the innovation and creation of value tremendously.

Noshir Contractor: Well, And thank you again so much for joining us and giving us a chapter-by-chapter tour, in some ways, of the Hype Machine Book. It’s a wonderful read. And I would definitely recommend that to anyone in web science. I look at your scholarship as being one of the poster children of really excellent cutting edge web science. Thank you again Sinan, and good luck for all the additional work that we will be looking forward to seeing coming out of your research in the years ahead.

Sinan Aral: It’s a true honor and a pleasure to join you. I’m looking forward to seeing you in person again very soon and to give you a hug and a high five, because it’s been too long. We’ve all been through a lot and it’s great to connect virtually but I’m also looking forward to connecting with you physically in person as well soon.