[Auren Hoffman] Welcome to World of DaaS, a show for data enthusiasts. I'm your host, Auren Hoffman, CEO of SafeGraph. For more conversations, videos and transcripts, visit SafeGraph.com/podcast.
Hello, fellow data nerds. My guest today is Sinan Aral. Sinan is the Professor of Management, Marketing, IT and Data Science at MIT. And he's the director of the MIT Initiative on the Digital Economy and founding partner of Manifest Capital. Sinan, welcome to World of DaaS.
[Sinan Aral] Great to be here. Thanks for having me.
[Auren Hoffman] Okay, now, your book “Hype Machine” is really cool. It details how friend recommendations and feed algorithms create polarization and filter bubbles of information. But human networks, we're used to being clustered with people who are similar to us. Maybe the social networks are like a bit more extreme, with making like friend recommendations etc., but doesn't it just mimic real life? Or where's the change from real life?
[Sinan Aral] I think that it turbocharges the human desire to associate with people like ourselves. So human social networks are well known for being homophilous, which just means that birds of a feather flock together. We tend to make friends with people who are like ourselves, but at least two things are different with algorithms mediating our social relationships. The first is that we are more likely to run into people that are not like ourselves out there in normal, everyday life. So in the workplace, out in our communities, even though we tend to be homophilous, even where we live, if we are just going about our daily lives in our cities, or we're going to various events or socializing with the families of our kids or things like that, we're more likely to run into diversity, and in our workplaces. But an algorithm if it is trained to maximize the likelihood that you accept a friend recommendation or make a friend recommendation, it's going to really hone in on recommending people that are even more like you than you would run into in normal everyday life. So it's sort of turbocharges our desire to be homophilous.
[Auren Hoffman] It does seem like society in the last, let's say, 40/50 years is really moving in this direction, even without social networks, right? I mean, people are moving to areas that were once let’s say, 60% of one party or now 80-90% of one party, people are subscribing to very niche news that they want to hear. So they're kind of voting with their feet, they're voting with their wallet, and now they're voting with the social media algorithm as well.
[Sinan Aral] Yes, and you're right, there have been a lot of books written on this realignment, in terms of voting and politics where people have moved and districts have become either all Republican or all Democrat in the United States. And there's been a lot written about polarization as a result as well. The second part of the social media world is that you can really be in a bubble that's completely blocked off from other perspectives. It's not just that you socialize with people like yourselves, but your entire stream of information becomes tailored to this type of worldview. And that's where you get into some of the dangers that are discussed in the book and elsewhere.
[Auren Hoffman] Okay, now, there are a lot of people who are sounding the alarm about social media, somebody besides yourself, so many other famous people are like Eli Paris, or Justin Harris, there's many, many others. Where do you differ from their thoughts?
[Sinan Aral] Well, so I very much respect those people, and many of them are close friends of mine. And similar to them, I have experience in building these kinds of systems as an entrepreneur, as an investor, but slightly different from them, is my role at MIT, where we do large-scale peer-reviewed science, analyzing this data for a living. And I mean, you mentioned Eli and Tristan, Tristan having worked at Google, Eli having built multiple companies, Move On, Upworthy, and so on, they have experience as doers, but neither of them are scientists, in the sense of doing peer reviewed research.
[Auren Hoffman] Are there certain conclusions that you believe that they don't believe? Or they believe you don't believe?
[Sinan Aral] Yeah, I mean, we have differences in some of our opinions here or there. But, I think that the main difference is that I have a requirement of my job, which is that I have to be able to add anything I say, has to get past three anonymous reviewers. And if I say something that contradicts the known science in the matter that I'm going to be held to task for it at the next conference, or the next time I'm at the watercooler with my fellow scientists. And so that forces me to hold my tongue. In my book, in the introductory chapter, I say “Listen, I'm going to come to some places where I'm going to stop short of making a bold claim”. And that's because, with my scientist hat on, I have to be very careful about the rigor and the robustness of the things that I'm saying and the conclusions that I can draw. And I hold myself to that standard because that's part of my job as a scientist. I think that that's part of the difference.
[Auren Hoffman] I’m really fascinated with the spread and popularity on social media posts, both the misinformation I think that's what you studied, but just like how things spread in general. And it seems like from what you came up with, the core belief of how things spread fast is that they’re quote-unquote, “novel”? Is that right? And what does that mean? Like, if I want to just create a viral social media post, what should I do? How do you break it down?
[Sinan Aral] The study that you're talking about is a 10 year study of the spread of false news that we conducted with Twitter and published in Science in 2018. And, in fact, we also thought, “Hey, you know, maybe the reason false news spreads, is because it's spread by people with more followers or people who follow more people and so on”. And we looked at all of this, and the exact opposite was true: the people spreading false news had fewer followers, followed fewer people, were less often verified, had been on Twitter for less time, and so on. We had to seek an alternative explanation for why false news was spreading. And what we found was that it was more novel in that it was different from what people had seen in the last 60 days. But when you say “what is novelty?”, that's actually a very crudely defined concept in the science. We're writing a paper now called “Unpacking Novelty”, which is about how do you define novelty.
[Auren Hoffman] Is it like the classic “man bites dog” type of thing?
[Sinan Aral] Yes, in a sense. What we found was that in the false new study, the stories that were salacious, blood boiling, anger-inducing and shocking, surprising, shock and awe is associated with sharing: people are 70% more likely to retweet false news. And false news is significantly more novel, hadn't been seen recently, shocking, salacious, blood boiling, anger-inducing, and so on. Shock and awe works in a sense on social media, which is disturbing. But there aren't very good definitions of novelty and there's a lot of ways you can look at novelty. Is it the degree to which how far away it is from topics that you already know about? Is it all of the dimensions of the topic are different from what you know about? There's a lot of different sort of technical ways to define novelty and those specifics matter, because it turns out that it could be that if something is novel, but too novel, then it's too far afield from what you know, and not interesting. There could be a sweet spot of novelty. We don't know yet, because we haven't defined novelty rigorously enough and we haven't studied it enough yet.
[Auren Hoffman] Do you think there could be a point where a tweet comes in or whatever, and it's sub second, so it hasn't yet been seen by everyone? And you think there could be some sort of prediction algorithm based on that to know how viral it might become?
[Sinan Aral] There's always been this interest among researchers about can they predict virality. And the way I look at this is that hindsight is 2020. Using models, we can look back at things that went viral and we can describe and characterize them, but it's hard to predict a priori, which things are going to go viral in advance. I think there are a couple of reasons for that. The first is that they're very rare events. And so rare events are harder to predict. But that's not enough to prevent prediction. There are lots of rare events that we can build models of predicting rare events. But I think that while things that go viral tend to have certain characteristics, they also tend to have other things that aren't part of that stack of things that are associated with going viral, that are sort of the unknown quantity of them that is part of why they go viral. And each one has a different unknown quantity. So it sort of escapes us to be able to build reliable models predicting virality, but it's easy for us to build models that describe, in historical data that pick out the viral things from the non-viral things.
[Auren Hoffman] I mean, there's so many tweets, you could make the argument that maybe it's like the monkeys and the typewriter, one of them creates Shakespeare or something like that. Is there some sort of sense of ok, some are just going to go viral? And we just got to throw a lot of stuff out there?
[Sinan Aral] I think that there have been people who have been better able to create virality than others. But that doesn't mean that you can predict virality reliably. Now, you know, what is it? I think that there's a combination of trial and error of throwing spaghetti against the wall, and just survivor bias picking out that this person is reliably creating viral hits, but they make 1000 things, one of them goes viral, they make another 1000 things, another one goes viral and say, “Wow, this guy just made two viral hits”. That's pretty amazing, but you don't see the 2000 things that didn't go viral. I think there's a lot of survivor bias in the descriptions of things or people that are successful and making things go viral. But I think that once you land on something, you try it again, you tweak it, and so on. I don't think that virality is a science, I think that we can describe it in hindsight, we're not very good at predicting it in the future.
[Auren Hoffman] If you think of like social media news versus TV news, my unscientific study of TV news is that it's probably worse for your health than cigarettes. And there's just a lot wrong with TV news, at least with social media news, the vast majority of news is about my friend's babies and stuff, which is somewhat kind of interesting to the average user. How do you see those two? And have you studied the effects of TV news versus social media news?
[Sinan Aral] Yeah, and it's getting worse, right? Because I think that the cable news channels, at least the United States, have sort of hit on this business model of catering to one point of view and going to the extreme and becoming more extreme in their rhetoric at that extremity, if you will. And that's engaging and we see this engagement model being prevalent in the attention economy, the same thing happens on Facebook.
[Auren Hoffman] And the need to create news, that's what their business is so they constantly have to sensationalize things that aren't really that sensational.
[Sinan Aral] That's right, I challenge you to turn on one of these channels and not see the words “breaking news” at the bottom of the screen. Breaking News is always breaking. And the reason for that is because similar to what we found in the Twitter study, if I can impress upon you that “wow, this is shocking. It's novel, right now it's happening. It's important now”, that gets your attention. But, as a result, breaking news is always breaking 24 hours of the day now on these cable news channels. What I've thought about it is that then people say to me, isn't the cable news just as bad or worse, or the spread of misinformation and the hyping us up of society and so on? And my answer is that there are feedback loops between all these communication channels. What is on social media then gets rebroadcasted by cable news, and amplified, what gets seen on cable news then gets shared on social media. And it creates an ecosystem of reverberating engagement and escalation, which is in and of itself, a complex system that generates some of these dynamics that we see.
[Auren Hoffman] Even the New York Times, which historically has been one of the most venerated news organizations, they have a business model now of catering to a core group of subscribers. A lot of their stories are just quoting things from Twitter. And it seems it does seem like we're as a society, is there any way to break out of this thing in society, where we’re moving into this more polarized news sources?
[Sinan Aral] Yeah, in my book I described four levers that we have to break out of this. And I think this has to do with the larger attention economy that we've built for ourselves, and the engine of that attention economy being this sort of engagement model that hyper engages us or tries to hype us up, that's why the book is called “The Hype Machine”. And I think that there are four levers which I call money, code, norms and laws where money is super important and is the business model, you have to create the incentives for all the actors and stakeholders in this in this engagement economy. Code is the design of the platforms and the algorithms and the systems. Norms is how we adopt and use the technology. What we say whether we're paying attention to how we get hyped up or engaged. And then laws is obviously regulation, we've got a lot of work to do in terms of ensuring competition. Right now, there's no competition in the social media economy. And we'll get into that perhaps, without competition these platforms make money hand over fist. Same with the cable news channels, they have no incentive to change or fix some of the things that we all know are wrong with them. We also have to worry about the balance between free speech and harmful speech and so on. We don't have a handle on much of this at the current moment. But we need all four of those levers, money, code, norms, and laws rowing in the same direction if we hope to break out of this.
[Auren Hoffman] Let's get into the competition. That's very newsworthy right now as well. I know that you mentioned that breaking up companies won't solve the market failures in social media. Tell me if I'm wrong, but you believe that instead, we need reforms like interoperability data portability to create competition. Why will that work better than breaking these companies up?
[Sinan Aral] Well, breaking a company up is very sensational and newsworthy. And when you mentioned words like interoperability and data portability, it's very nerdy, technical, everybody goes to sleep. But I'm not against breaking a company up. But what we need is, that's a band aid on a tumor. We need much more structural solutions if we hope to have sustainable competition in the long term. So let me try to explain interoperability in a way that makes sense, which is that the social media economy runs on network effects, which means that each of these platforms, its power, and its ability to attract users and retain users is a function of its size. The way that network effects work is that I can keep people on my platform if access to the people on that platform is only on my platform. If you want to talk to your Facebook friends, you have to be on Facebook. That's how Facebook keeps users there. If you want to talk to your friends and family, you got to be on Facebook, or iMessage, any of these closed walled gardens. And the reason why they're closed and walled gardens, is because these companies know that if they can close them off, they get a lot of value from the fact that they already have a lot of users and people will want to join and will not leave because they want access to all those people that they've got locked in this walled garden. Interoperability breaks that and makes the new innovative company that comes online, able to connect with those 3 billion people on Facebook. And when you legislate and force interoperability, it enables innovation and competition sustainably for the long term, because new entrants can come in and plug right into that network effect.
[Auren Hoffman] So like when Gmail started, obviously. Email is completely interoperable, and so they could just plug on this already existing email system, everyone already used email to communicate. Gmail was a significantly better system than what already existed, and then everyone can move there. Whereas like, if you already had this internal network on Yahoo Mail, you will be very reluctant to move over to Gmail, even if the product was better.
[Sinan Aral] Perfect example, could you imagine if you as a user of Gmail could only email other Gmail users, you’d tear your hair out and go crazy thinking “Wow, that is insane”. Imagine if you can only SMS text message, from Verizon to other Verizon customers or from Sprint to other Sprint customers. Yes, you’d think that's crazy. But that's exactly what we have in social media today, which I think is crazy. So as a condition of the AOL Time Warner merger, we forced the Aim Instant Messenger to become interoperable with Yahoo Messenger and MSN Messenger. Before we did it had 65% market share, one year later it had 59%, two years later it had 55%, and three years later, it seeded the entire market to new entrants. That's what interoperability does to competition in the market.
[Auren Hoffman] We have this new chair of the FTC, Lena Khan, and, and she's written a lot about monopolies. And she seems to be a little bit more on the breakup side of things. Some of them meet some of these big tech companies, maybe they're clear lines on how you break them up. Amazon has AWS and has the store, and maybe it has Twitch or something. Other big companies like Apple, I don't even know how you would break it up if you wanted to. There doesn't seem like any other recourse except interoperability. And by the way, does interoperability also include allowing more people access to app stores, and some sort of pricing controls, or how do you think about some of these other types of things?
[Sinan Aral] Well, I think the other major issue right now is, as Elizabeth Warren has said, this idea that you can't be the umpire and a team in the same game. And that's the idea of Amazon controlling the marketplace for products and then selling its own basic Amazon basic products in that marketplace and then privileging those products in the search results or and collecting data on all the choices that are made to privilege its own products. That's a separate sort of regulatory issue. But in my mind, all of these issues need structural reforms to the economy itself, a breakup of a company doesn't prevent the next company from doing that same thing, you need a law that says you can't be the owner/operator of the marketplace and have your own products there, if that is the type of competition that you want to create. Europe is ahead of the US on the structural reforms that generate competition with the Digital Services Act and the Digital Markets Act.
[Auren Hoffman] One of these I know, they're really forcing banks to make sure that your financial data is interoperable. You can move your data between one to the other, you can aggregate your data for new services, etcetera. Is that what you mean by interoperability?
[Sinan Aral] Yes. So the European regulations specify a lot of restrictions on what they call gatekeepers, which are these sort of big companies in a space that either dominate the marketplace or set the rules of the marketplace. And there are a number of different elements to those two new forthcoming regulations that are about limiting or sort of controlling the way that these gatekeepers operate in a marketplace. That is the type of thing that leads to sustainable competition because it applies not just to the current market leader, but anybody that comes after that market leader, and the rules of the road for how business is done in that ecosystem. Again, I'm not against the breakup of any particular company, I just don't think that that is going to create sustainable competition. I also think, by the way, with the case of Facebook, that that antitrust case, the Break Up Facebook case faces a very uphill battle, it's going to be 10 years, it's going to be slow and laborious.
[Auren Hoffman] These are just the lines WhatsApp, Instagram and Facebook, right? You have limited lines. Now, Google has a higher market share in Europe than it does in the US. Facebook has a higher market share in many European countries than it does in the US. It does seem like there's still a lot of concentration of power in Europe as well.
[Sinan Aral] Well, Europe is just beginning down this road with Digital Services Act and the Digital Markets Act. What is to come? We'll see what happens. I just think that that's the better approach is to have structural reforms of the economy itself, rather than dismantling one company and thinking that that's just going to change the market economics of the market itself.
[Auren Hoffman] Okay, now, going back to the spread of misinformation, and part of the reason it seems is that there’s just this general distrust of experts globally. Some of that mistrust, maybe quite a lot of that mistrust is quite justified. What is the expert class, broadly? I don't know how you define that. What does the expert class need to do to better engender that trust?
[Sinan Aral] That's a good question. I think that it's a shame because a lot of times these experts have spent years decades trying to understand the very specific problems that they're focused on. And to disregard them, I think we do that as a society at our peril. But I also think that it's understandable, in a sense, because of a couple of reasons. One is that I think that the experts need to be more transparent. Right now, I think that they just speak in a way that says “here's the answer, I am the expert and you should take it as the answer.” If you speak in accessible language, large swaths of the population are smart enough to understand what you're saying and can get it. So don't shroud yourself in the white lab coat and say “Hey, I'm the expert. And that's the answer, that's the path that we should follow”. So, transparency is one. The other thing is that I do think that sometimes experts, having spent so much time thinking about something or researching something or focusing on something, assume that the audience is not capable and so not only talks down to them, but also doesn't give them the benefit of the doubt to explain and assume that they can engage and consume and understand and react and participate in the conversation. I think that's a big miss on the part of the experts, because I think when you talk to more people, you learn yourself. If you are open to learning, you can learn in sometimes the most unexpected ways and in some of the most unexpected conversations, and openness, transparency and not sort of talking down or assuming that your counterpart is unable to understand what you're saying, I think that's where experts are going wrong today.
[Auren Hoffman] There always seems to be a lot of pressure on scientists to change their work at least a little bit to kind of conform to a particular narrative. Now, I don't know enough historically to know if this is a new thing, or this is an always type of thing. But I have many friends that are scientists that have had a lot of pressure. One of my friends is a cancer scientist and is very against smoking but did a study that in this particular study showed that people who smoked at older ages didn't learn a higher risk of cancer. And he didn't like the results of this study, but still wanted to publish the study. But he got a lot of pressure from his colleagues in this community not to publish the study because they said “well, okay, this is going to hurt the overall narrative of what we want to do”. How do you see these pressures scientists coming to bear?
[Sinan Aral] Well, it's interesting because I think scientists are also a communal species, in a sense. They live in a social environment, and they are very much subject to the opinions of others. But at the same time, I think scientists are genuinely skeptical people. And that cuts both ways. If you have an overwhelming set of results that go in one direction, and then you have a study that in a in a certain narrow niche area goes against that, they're going to be skeptical of it. If something is a new discovery, it's initially met with traditionally a lot of skepticism. I think that this skepticism is valid.
[Auren Hoffman] Like moving the whack, covering the wagons and everything like that, putting them together and getting together.
[Sinan Aral] No, I don't I don't think it's scientists circling the wagons and saying “we need to defend the narrative that exists currently” or anything like that. I think that if you're an individual scientist, and you've seen 1000 papers that say that smoking causes cancer, and then you see one that says it doesn't, you're naturally thinking “wait a minute, let me read that with more care. What do they do wrong?” Because I mean, these 1000 papers before it, say the opposite? How can this one be right if these 1000 are wrong? And it could be right, in a sense that well, like in this age group, or, if they only smoke menthol cigarettes, it could be right, but there's a natural skepticism that some paper that goes against the weight of 1000s of papers before it. A simpler explanation is that they did something wrong, then that they discovered something new. I think that that skepticism is really important to science, because it forces us to be rigorous, and really show things that are robust, and hang our hats on things that are robust, but it can also create knee jerk reactions to new things as well. Okay, there's a famous saying, by the way, that that is that science advances one funeral at a time.
[Auren Hoffman] That’s Niels Bohr, right? I love that thing. All right. Now you and I share what maybe possibly is an unhealthy obsession with Wall Street bets, right. I'd love to just jump into that route a little bit. Now there's all these new sub-Redditts that are gaining popularity, on stock picking, stock means, etcetera. Where do you see the Wall Street bets movement going in the future?
[Sinan Aral] I think Wall Street bets is old hat now. All the apes have left that forest and they've dispersed to other places, right? They call themselves apes, so that's not my own pejorative term. I think that what we saw is online collective crowds successfully exercise collective behavior coordination. And I think that that means once that was demonstrated to be effective, I don't think it's going anywhere. In other words, I don't see it fizzling out, in the sense that I think that it will continue, it will disperse into new channels, it's unclear to me where the new sub Reddits are going to be or which new social media platforms it's going to appear on, or where it's going to, in the same way that it's hard to predict virality, it's hard to predict where the next successful group talking about equity prices is going to go. But I think it's going to evolve. Part of the reason I think it's going to evolve is that I think that institutional investors are going to take it more seriously, insert themselves into the conversation, maybe surreptitiously, maybe not.
[Auren Hoffman] They’re almost certainly already in that conversation already.
[Sinan Aral] Absolutely, a 100%. That's true.
[Auren Hoffman] At the very least, they're monitoring that and trading on it and either trading against it or adding fuel to the fire, right? When these big hedge funds go down, it's not just the crowds. There are other hedge funds helping the hedge funds go down, they’re helping the crowds.
[Sinan Aral] That's exactly right. When one hedge fund goes down, there's a larger hedge fund that makes triple on that bet, going in the opposite direction, right? I think that we have to recognize this as an addition to the equity market landscape that will evolve and change over time, will be coopted in ways by institutional investors will not in other ways. It’s hard to know exactly where it's going to show up, but this is a part of our investing reality going forward. And I've been saying that since before GameStop. […] What I was saying was essentially that I wrote about this before GameStop. In fact, I talked about this in my book, which was published in September 2020. I use multiple examples of social media driving equity prices, and sort of laid out the research on it. I said that this is going to become more of a part of our investing ecosystem in the future. And then obviously, when GameStop hit, people turned to me and said “Wait, didn't you say this in your book?” And I said “Yes”, and they said “Well, what is going on here?” And I said, “Well, I don't consider myself an oracle. This is entirely predictable, if you study social media, because it enables this type of information flow coordination, and that's exactly what an equity market is”. I think it's part of our landscape, and it will evolve and it will be part of our landscape for many years to come.
[Auren Hoffman] Yeah, if I'm a hedge fund or something, there's all these different people trying to coordinate certain equities to go up. And what I want to do is figure out early on in the coordination, which ones are going to work, bet heavily on those and not bet on the ones that are not going to work, for every GameStop and AMC, etcetera, there are potentially hundreds of equities that people try to coordinate but didn't actually take off, right? How would you advise if you were an advisor to a hedge fund? How would you advise them going in and trying to understand this as early as possible?
[Sinan Aral] Well, it's interesting you say that, because I'm working on this exact problem right now. I find this fascinating and it's such an interesting problem of collective behavior. Part of what makes it interesting is that a lot is at stake. People behaving and making decisions in this particular sphere are putting their money on the line, their mortgages, their houses, and so on on the line. That changes the way decision-making happens in these collective situations. I think there are ways to be better at predicting these types of events. I don't want to reveal too much of the secret sauce, but I think that a number of things can help. For instance, diversification of your bets, obviously, is important. Building ensemble models, I think, is important, instead of relying on a single type of model, which is more fragile. And I think that data is always the key, as you know, in your line of business. In my mind, rough estimate, the success of a model is 80% or more due to the quality of the data and the way that you organize that data, and less than 20% is about the model itself. You can use a very rudimentary workforce model, but all of the work is in what we call feature engineering. And feature engineering doesn't get enough attention relative to the source of value that it creates in predictive modeling.
[Auren Hoffman] It’s the data munging, it's the data cleaning, it's the truthiness of the data. It's how easy it is to ingest into your systems.
[Sinan Aral] I would add one other thing, which is the identification and extraction of features from the data. Features being a definable data element from a large swath of data. I can take a data set of whatever sales of products, and I can slice and dice that an infinite number of ways. For instance, what's the variance of the sales? What's the second order variance of the sales? What's the temporal dynamics of the sales? What's the variance of the temporal dynamics of the sales? What are the correlations between this subset and that subset, and the subsets of subsets and so on, there are infinite ways to slice the data, which give you features that then can be predictive. And that's why I always say that data science and analytics is essentially a creative art. Because you have to be thoughtful and creative about new insights of what the data are, and how that data can generate value. That's where I think the best data scientists separate themselves from the pack is in that ability to ask the right questions, and an ability to be creative about the data. Those two things are undervalued, but essential in my mind.
[Auren Hoffman] Interesting. We had, we did a podcast with Hilary Mason and she said almost something verbatim to what you just said, really fascinating.
[Sinan Aral] That's no surprise because we're good friends and we talk a lot.
[Auren Hoffman] Speaking of data, you're also well known for using lots of different types of data. I think you and I kind of came across each other when you used the SafeGraph data a year ago for COVID modeling. You used a lot of other types of data. What obligations do you think companies and organizations should have to making their data available for researchers and academics?
[Sinan Aral] As a researcher and an academic, I think they should really make all of their data available to researchers and academics for research purposes. I say that a little tongue in cheek, I think that there are some companies where we absolutely need more transparency, because they're having such an important effect on society. This is certainly true in social media, and I've made this point over and over again. I've spoken recently to Nick Clegg and had this conversation about how we have to become more transparent with social media.
[Auren Hoffman] Nick Clegg runs policy at Facebook.
[Sinan Aral] Exactly. And he agrees, but I think that there's also some sort of restraints on their ability to do that, and so on. I think a good example is, what they cite as well, we also are being pressured to focus on privacy and the security of the data of individual users of Facebook. And this is something I pointed out in an interview with MIT Tech Review in 2018, which I call the transparency paradox. We want these companies to be more transparent, but we're simultaneously asking them to be more private and secure with our data, which makes it hard for them to give out data. Remember, Cambridge Analytic, a scandal began with a release of data to a researcher from Facebook, and then used for purposes that it wasn't intended to be used for originally. We have to develop new methods for enabling transparency and privacy and security. Those types of methods will help us be more transparent and secure at the same time. Then we need more ability to analyze this kind of data. We need more transparency, we need more access to researchers, we can have data safe harbors. Social Science One is a good example, where industry academic consortia are trying to create safe harbors for data where scientists can analyze that data under certain you guidelines and restrictions. For instance, the Bureau of Labor Statistics and other types of national data holders, or the census will allow researchers to become in a sense certified by the census or the BLS and then go into their own facility, like a clean room and analyze data within that space but not be able to take data out.
[Auren Hoffman] There does seem like for researchers, a little bit of a habit, you're at MIT, so you're probably much more likely to get access to really good data sets, then maybe somebody who is at a less well-known institution. If you think of like Raj Chetty, he's got access to the IRS data, he has probably produced more interesting papers on that data set than almost anybody because the data is so valuable, but suddenly, it's a little bit unfair that he has this amazing proprietary access, but like the random economist somewhere else can't get access to something like that. How do we also democratize that access amongst the academia class?
[Sinan Aral] I think that's true. I think Raj is amazing and I've been speaking to him a lot over the last two years, I think that standardization is a good way to do that. When you have rules and guidelines for how data can be accessed, then any credentialed academics should be able to have access to that data. A lot of times there is a bit of a subjective evaluation of particular research projects for which data is released, it's not necessarily the case that the privileged institutions have access, it's just correlated with being in a privileged institution that the research proposal that's made is a good one. A good way to deal with that and the way that scientists typically deal with that is blinded peer review. So instead of knowing what the institution is, and who the person is, you judge the idea on its merits while blinded, and that might enable more access and democratization of access. But you also don't want research to be done on topics that aren't worthwhile and or topics that have themselves dangerous to them or used in nefarious ways. Credentialing is also important. There are a number of different elements working simultaneously there.
[Auren Hoffman] Okay, now a couple of personal questions. It's been like 20 years ago, now you're really one of the first people to pay like attention and study social media. What advice would you give to the upcoming academic, the upcoming person who really cares about studying technology? Is it just like, “Go with your gut?” I mean, you were doing things back then that now are super popular, but back then were not popular at all.
[Sinan Aral] I think that there is there's at least two ways to think about how to make contributions in science and I've done both. One is to sort of put bricks in walls that already exist. And another one is to put a brick down in the hopes that there will be a wall there, where the risk is that your brick will be the lone brick by itself. No wall being built around it. I do both. Because I think that incremental science is important, and making refinement, especially where you think that something is going in the wrong direction, or needs refinement, or the mistakes being made, or we don't know enough or something, incremental things are useful. But I also think that the advice I would give to the young scientist is: don't be afraid to try and start a new wall. Don't be afraid to try and ask a question that isn't being asked currently, because that's really where the big breakthroughs come from. That's really where the new insights come from.
[Auren Hoffman] Is it like, 80% on the old wall and 20% on the new wall?
[Sinan Aral] I think that portfolio allocation changes over the course of a career. I do less incremental work now, because I just feel free or to sort of, if I put a brick down, and people say, “Oh, boy, that that's really boring sit on”, I can sort of move on from that. As an assistant professor trying to get tenure, maybe that wouldn't have been such a great idea. But now I prefer to write a paper and then hope that other people follow with more follow up research, and then go on to a whole new topic and write another paper in a whole new area, that's riskier. And you're more able to take risks when you have tenure, or when you've had a little bit of time under your belt doing research and so on. I think that portfolio changes over time. But my advice would be: don't be afraid. If your gut is telling you that something is going to be really important, and you've sort of double checked yourself, in a sense, you've done some reality check, you've talked to other people you've, and you have reasons for believing what you believe you applied a skeptical lens to your own belief that this is something important. And you still believe it, go for it, because there's a good chance that if it pans out that it could create a whole new field, it can create a whole new area of research or a whole new research question. And that's really important. We don't want all our scientists spending 100% of their time making incremental improvements. That would be devastating to the advance of science.
[Auren Hoffman] Our last question we asked all our guests: if you can go back in time, like what advice do you wish you could have given to your younger self?
[Sinan Aral] I'm finding that I'm much better at this now, at my age today than I was then. But my advice would be to be more present in the moment. And I have achieved that. I'm still not great at it, but I'm much better at it than I used to be. I meditate more now, I do other things that helped me do this, but wherever I am, whatever moment I'm in, I am present in that moment, in that conversation, I'm listening, my mind isn't off somewhere else thinking “why is this important to me? Is this not important to me?” No. The point is, you're in that moment. The reason I say that is because insights, friendships come from the least expected places. If you're not paying attention, you're definitely going to miss it. Being present helps you fully engage yourself with the thing that you're doing in that moment, and really leverages your own capabilities to their fullest in every moment. But it also leaves you open to insights, new friendships, relationships, or things you would have never imagined, coming from the most unexpected places.
[Auren Hoffman] Great. I love that. Tell us where we can find more information about you on the broader interwebs.
[Sinan Aral] You can find me at sinanaral.io on the web, @sinanaral on Twitter and @professorsinan on Instagram.
[Auren Hoffman] Okay, awesome. And I follow you on Twitter. I highly recommend everyone does as well. Thank you so much for joining us on a World of DaaS.
[Sinan Aral] pleasure. Thank you for inviting me. It was it was fun.
[Auren Hoffman] Thanks for listening. If you enjoyed this show, consider rating this podcast and leaving a review. For more World of DaaS (DaaS is D-A-A-S), you can subscribe on Spotify or Apple Podcasts. Also check out YouTube for the videos. You can find me on Twitter at @auren (A-U-R-E-N). I’d love to hear from you.
Sinan Aral, Professor of Management, Marketing, IT and Data Science at MIT, talks with World of DaaS host Auren Hoffman. Sinan is also the director of the MIT Initiative on the Digital Economy, founding partner of Manifest Capital, and the author of the book “Hype Machine”. Auren and Sinan discuss why social media networks are optimized to connect users with like-minded people, what creates viral content, and how to solve market failures in social media. They also discuss the future of Wall Street Bets and why it will likely remain part of our investment landscape.
Hayden Brown, CEO of Upwork, talks with World of DaaS host Auren Hoffman. Upwork is a $6 billion market cap workplace company. Auren and Hayden cover how the COVID-19 pandemic massively accelerated the freelance market, Upwork’s strategy to become the single destination for all work-related needs, and various other examples of successful marketplaces. They also discuss Upwork’s proprietary dataset that provides unique insight into the labor market and how Upwork is working to make more of this information publicly available.
Hilary Mason, co-founder of Hidden Door and data scientist in residence at Accel Partners, talks with World of DaaS host Auren Hoffman. Hilary previously co-founded Fast Forward Labs, which was acquired by Cloudera, and served as the Chief Scientist at bit.ly. Auren and Hilary explore how data science has progressed in the past decade, the role of data science in an organization, and data ethics.