John Markoff chats with Sandra González-Bailón, Associate Professor of Communication at the University of Pennsylvania and 2019-20 CASBS fellow, about the influence of social media platforms on news and political activism.
“Bots are Less Central than Verified Accounts during Contentious Political Events”
Her book Decoding the Social World
Facebook 2020 Election Research portal
Narrator: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered. Today on Human Centered, We'll hear John Markoff in conversation with CASBS Fellow Sandra González Bailón, Associate Professor of Communication at the Annenberg School for Communication at the University of Pennsylvania. Her work focuses on how information flows through social networks, particularly in the context of news and political protest. The two discuss recent developments in the framing of social media technologies with respect to political information and election influence. We'll hear about the use and abuse of bots, how activists use social media, and the new Facebook 2020 research project she's involved with, which allows researchers to use Facebook data collected during the 2020 U.S. election.
John Markoff: I wanted to ask first if you've watched the documentary that's now on, uh, Netflix called The Social Dilemma.
Sandra González-Bailón: I haven't. It's on my list of things to do, but I've heard— I know it's controversial, and I don't want to make up my mind before I actually see it, but it seems to me that they left out a lot of interesting— it's very— what I've heard is that it's very sensationalistic, and they didn't really talk in fairness about a lot of the real problems that they, that the private sector is facing in terms of social media companies and whatnot. They didn't echo a lot of the research that they should have echoed, but I haven't seen it myself, so.
John Markoff: Well, it really, I mean, you've heard everything that they say. And they have this kind of narrative plot device where they have a family that they, a fictional family. So it's interesting, but it's basically Tristan Harris and Jaron Lanier who have been saying these things about social networks for what, 3 or 4 years, and Roger McNamee who also. And so, But it's clearly inflammatory and Facebook responded. I don't know if you saw Facebook had a formal response to, so. So, and I know that you're working now on a project to explore the impact that Facebook might have on the 2020 election. I know that we don't know yet, but I wondered if you, Two things I wanted to ask. One is your perspective and then if there is a consensus because I've struggled with all of discussion around the impact of the Cambridge Analytica data. I've read everything I can. I've seen the back and forth and I wonder two things. If you have a perspective and do you feel there's a consensus on whether or not they played a real role in the outcome of either Brexit or the election?
Sandra González-Bailón: As a researcher and scientist, I have to say that I haven't seen persuasive evidence that there was an exposure, sort of exposure to information on Facebook or other social networks, and in particular, that these information campaigns that were seeded and that targeted specific groups of people on the basis of their psychological profiles I haven't seen robust empirical evidence that that had an effect on behavior or opinions. And by behavior, I mean off-platform, right? So in terms of then you voting for a candidate or another, or you changing your perception of what matters and about candidates and whatnot. That doesn't mean that there wasn't an effect, but I have to say that the evidence is very weak. I don't think anybody has proven that that happened.
John Markoff: Did you see that it continues to come up? There was a recent Channel 4 English documentary that purported to, I thought it was very revealing. I mean, they actually went into relevant wards in, was it in Milwaukee, and asked people specifically who were influenced, who didn't vote, if they had actually seen any of this material. And they weren't able to come up with anybody who actually said, yeah, my vote was turned by seeing this material on Facebook, which I, I thought was interesting.
Sandra González-Bailón: Yeah, I mean, the truth of the matter is that it's very difficult to find— it's very difficult to measure these effects. So measuring media effects is in general a challenging task, right? Like, so even before the era of social networks and social media, you know, you have campaigns investing millions of dollars in trying to persuade people one way or the other. But then when it came to actually measuring the impact of those interventions, like, it's not necessarily, you know, easy. And it never was, but particularly now it's even more difficult because of the way that audiences are being fragmented, the way in which these campaigns are trying to target individuals. And then, you know, when you take the more interpretive account, when you go and ask people, whatever they say, they're just like specific individuals that would tell you very little about the overall impact of these information dynamics, right? And so I think it's— there's a lot of very suggestive stories in the press about, you know, how Russians interfere in the elections and how that changed things or how YouTube is contributing to dynamics of radicalization. But those stories are usually based on specific cases that were sort of cherry-picked to illustrate a story, a human story, but that doesn't really tell you much about the overall picture. And from the side of academic research, I think we're struggling with is how do we get access to the right data to be able to produce the right sort of evidence so that we can then have an informed discussion of whether or not we should regulate these companies or whether or not we should take measures. And that is what some of us are trying to do by forging collaborations with social media companies, for example, right? So data scientists at Facebook.
John Markoff: Well, do you think that— Facebook has access to data that would, from that time period, that they're not making public that would be interesting for you to sort of illuminate some of these questions? I mean, there was a debate and it's kind of gone away, I think.
Sandra González-Bailón: So the short answer is I don't know, right? Like, so I don't know what they had access to in 2016. And for the most part, a lot of the data from that time period doesn't already, doesn't exist anymore, right? Like, so one of the first things you realize when you start talking with researchers. And then I would like to make a distinction between Facebook as an organ— as a company with a hierarchy of command and the business side of the company, and then the scientists, the research scientists, the data scientists who are doing the nitty-gritty work of measuring things and trying to inform the higher-level decisions. And most of my interactions, most of my information about the data that these companies have or have not is my view of that is based on my interactions with the data scientists, right? And so, and I do know that from the outside, we sometimes talk about these companies as if they had like omniprocess, like only sort of superpowers essentially to get data and about everything and transform that data into actual knowledge. And my sense having interacted in the rather frequently in the last few months, at least with some of these data scientists is that a lot of the data we think exists doesn't exist anymore. They deleted from their servers. Partly to comply with privacy demands, but also because it's literally impossible to keep everything, to store everything they keep track of. And so to go back to your question, I don't know what they had access to in 2016 or not. I do get a better sense of the kind of measures we can, you know, that could help us answer some of the questions we have with regard to the impact that Facebook has on the electoral process in terms of channeling information. And it's a lot and it's very sort of, it could potentially give us interesting answers to questions that are still open, but it's sort of not as much as sometimes you would say they have when you just hear some public discussions about the power these companies have. So, you know, I think there's a humility that comes with actual, with doing the actual research and trying to get access to data and then transform that data into useful information and that into knowledge. There's that humility that comes out of going through that process that I wish some of these kind of more journalistic accounts sometimes had more present, right? Yeah. You do have access to a lot of data. That's obviously, you know, that's a fact, right? Yeah.
John Markoff: One of your recent papers focuses on the impact of the shift to mobile access in terms of information ecosystems and ideology and all of these things. And, but before we talk about that specifically, I wanted to ask you generally, if you could, do you have the sense or could you describe how the information environment has changed between, you know, well, over what period? I mean, is it changing significantly?
Sandra González-Bailón: Yes. So the internet arrived to disrupt many, many things, including the media environment we inhabit. And so, you know, from, you know, there's been a radical transition from the broadcasting era where most people were exposed to a limited number of sources. And so there was a lot of common ground in terms of the information that citizens get exposed to. And that was mostly via TV or newspapers or radio, but there was a very limited set of sources that citizens could get exposed to. And with the emergence of cable TV first, but then more drastically with the emergence of internet technologies, the web, and everything that happened after, social media apps, then the information environment has decentralized, it has fragmented in terms of the number of sources available, and the costs of publishing content have been drastically reduced, which is not to say that everybody has the same voice, right? Like, so the fact that it doesn't cost anything or next to nothing to publish a blog or publish a website doesn't mean that you will have an audience, but it has as we all know, right? But it has multiplied the number of sources that are available, and on the other end, people have more choice, right? And so we as consumers have more choices as to what are we going to consume. And so with this transition, there were many claims around the impact that that would have on the public sphere. And most of those claims seem to suggest that just, you know, that people would self-select into like sort of into venues or sort of sources and outlets that would echo their, their, their, their, the ideas that they already agree with, right? And so that the cross-exposure to a diversity of ideas would go down because we all self-select into like-minded outlets. And so we would all be operating within these future bubbles reinforced by algorithmic filtering. And those ideas are very powerful, very catchy, very intuitive. But they were based, I would say, mostly on technological affordances and what technologies allow us to do, not necessarily what people actually do. And it's also true that technologies work both ways. And so what we show in this paper that you mentioned is that, so that, you know, that actually, you know, mobile access to news, so mobile on the go, right? Instead of incrementing this self-selecting process, is actually expanding the number of sources that people get exposed to. And the reason we hypothesize about is it's probably because of all these news aggregators and, you know, the information you see on your feed in social media comes from different places. So you don't actually go to a particular news source that, you know, it's going to give you the information that you're looking for. Technology has also created entry points to news that are actually more diverse than they used to be in the past. Or, you know, and so you're more likely to see news from a diversity of sources if you are consuming news from your phone than if you're sitting at your desktop computer and you just go to the New York Times domain and to, you know, to access news that way. And so when we wrote this paper, we had two audiences in mind. One were the set of, you know, public intellectuals, academics, opinion makers who have been, who have had a lot of influence in terms of painting this picture of increasing fragmentation, increasing balkanization of the public domain. That was one audience we showed that the best available evidence we have access to doesn't support that picture. But also we were thinking about our fellow academics who in the past were using behavioral data, tracking what people do online using desktop computers, which are easier to, you know, to track. You just install a little add-on on your browser and that allows you to track what participants in your study do. But if we only focus on that, we're also missing an important part of the picture, which is what people do when they are on the go, on their cell phones. And tracking mobile devices is trickier, is more difficult, but it's actually crucial because the vast majority of people uses mobile phones or tablets to access information. Yeah. So it's a really interesting point.
John Markoff: A specific example I wanted to ask you about. Do you get any sense that because of the rise of TikTok, has that been of the scale to be a measurable impact on Facebook's influence? I mean, has TikTok emerged on your radar as an important force to look at?
Sandra González-Bailón: No, and TikTok actually, and so this is also a generational thing, right? Like TikTok is really popular among younger users. And of course now it's becoming more mainstream. And TikTok presents a whole range of different issues when it comes to analyzing it because it's not textual data anymore. It's like it's all visual videos and that's a different format of communicating that will require us devising new tools to actually analyze what that content is about. Yeah. But I think Facebook is still quintessential to understand. And I wanted to note here that the research we were discussing before doesn't really tell us much about what happens within the platforms, right? So because social media are these walled gardens, right? So we know how many people go to Facebook, but we don't know what happens within Facebook. Same with YouTube. It is easy to see— I mean, YouTube is one of the most popular domains online, but it's very tricky to see what people do once they are in the platform. Google knows. Researchers have it more difficult to access that data. And the reason why I think that matters is not just sort of what happens on those platforms, It's not just because they create blind spots when we only— if we don't take that into account, we have a blind spot. We only know what happens online, on the web, through apps, but we don't know what happens on those platforms. And that's important because a lot of people who opt out of news consumption— so going online to consume political content and political news is not the most popular activity amongst online users. It's only a tiny fraction of all online community revolves around consuming news. And so what we are tracking are the individuals who are interested in politics, right? And so one of the things that we highlight in this research that we were discussing is that more than half of the online population doesn't voluntarily consume news at all, which begs the question, well, these people might still be exposed to political content on Facebook via their friends, via, you know, so the, the, the, how your own social network shapes exposure to that content is not voluntary in the sense that you are not proactively looking for political news or political information, but you see and you still encounter that content. And if you don't get exposed to legit political news anywhere else, you might be more vulnerable to believing those memes that are actually spreading disinformation. Or so, so there's a subset of the population that might be more susceptible to misinformation campaigns just because they don't have anything to compare what they see on social media with. They are just not as informed as they should be. And so we can't get that if we don't obtain access to the data, right? We can't determine if there are more susceptible populations on social media if we don't have access to what happens on platform, right? And same with YouTube, same with TikTok, Instagram, and so forth.
John Markoff: I was really quite disheartened as a retired member of the mainstream media to see your finding that less than half of the online users get their news, get their news from the news. That was a real, uh, statement about where we've gone. And the other thing that, you know, there was a remarkable event that happened just a couple of days ago on TikTok, um, where the 15-year-old daughter of Kellyanne Conway, Claudia Conway, outed her mother and criticize the president sort of in real time. And I got the sense that this is a window into the way things are moving in some, well, another thought about your research, you subtitled your book on data sciences, The Unintended Consequences of Communication. And that really brought that, what are we seeing there in terms of this real-time interaction that's happening that has nothing to do with traditional textual forms of communication.
Sandra González-Bailón: No, and we expect for a long time that digital technologies arrive to challenge the traditional gatekeeping rules of the media, for example. In the past, if you were involved in a political protest, your first goal was to attract the attention of the media so that they would echo your mobilizations. And for that, you needed some— you needed attention from journalists and editors that would decide to put the news of the protest on the newspapers or to cover it on TV. Now there's all this talk about hashtag activism, Twitter revolutions, whatever. Some of them are overblown, but it's true that these online networks, decentralized networks, have challenged these traditional gatekeeping rules, which on the positive side allow marginalized voices to have visibility that they wouldn't have been able to obtain so easily in the past. And Black Lives Matter is an example. And I would say that the example you mentioned of Marian Conway's daughter, Athena, would also be on the positive side in the sense that she has different— she's like a great example of gatekeeping. You know, who's the gatekeeper here? No one. And so that's interesting. On the flip side, of course, you know, that's also the reason why we have all these dubious sources of information having an impact on so many people, right? Because no one is doing the editorial work of filtering of interpreting the news. And so it has, you know, there's a good side and a bad side on this debate. But the subtitle, The Unintended Consequences of Communication, is true. Like all these technologies are designed with a purpose in mind. And then, you know, users ultimately can repurpose those technologies to do things that were never intended. And I think that that's, Yeah. Can't have creativity without progress.
John Markoff: That takes us right to William Gibson which your other paper focused on bots in part and I wanted to read one passage from your article which really jumped out at me and I wanted to ask you to expand on it a little bit. You said we show that verified accounts are significantly more visible than unverified bots in the coverage of— the events, but also that bots attract more attention than human accounts. And that last part of that sentence, I wanted to ask you what that meant.
Sandra González-Bailón: Yeah, that paper came out of, so an idea, the need to, to my mind, the need to clarify the role that bots play in social media platforms, because there's many different types of bots. And by bots, you know, when we talk about bots, we refer to automated accounts pieces of software essentially that are designed to automate certain tasks. And so those tasks can be to target certain users on Twitter to promote certain messages. And so every time that there's a hashtag, some bots, they just react automatically to messages that have a particular hashtag to try to send a message. But media outlets also use bots to break news. And so, and we wanted to make sure that we didn't conflate the different roles that bots play on social media. And so this is why in this paper, when we identify automated accounts, we also differentiate between those that are verified by the platform itself and those that are unverified bots, right? And so, and the main, you know, as you said, John, one of the main findings is that, again, like we are trying to counter some of the, in our opinion, exaggerated claims about how relevant bots are online. There are many bots and we kind of quantify them in this paper, But they, you know, we also show that when you actually look more carefully at those automated accounts, many of those belong to media organizations or to, you know, to public figures whose behavior on Twitter is bot-like behavior, but they are actually kind of contributing to spread news and not necessarily misinformation. And we find that in the context of contentious political events, which are in theory, particularly susceptible to information manipulation campaigns. So to increase antagonism, for example, we find that those verified media accounts are actually the reference, are retweeted way more frequently than the other accounts. But the average human being on Twitter, and again, at least in the context of this to episodes of contentious politics, we, you know, they receive less resonance than bots, unverified bots. And of course, you know, in terms of volume, there are still many humans than verified bots, and they just because of that, human accounts generate more content, but bots seem to get more traction in terms of diffusing information. You know, we don't tackle the question again of the effects of being exposed to bot-generated content in terms of opinion and behavior, right? Like what happens on Twitter, does that matter for what happens offline? And we don't know.
John Markoff: So this is a more speculative question, but I'm, you know, I've always been fascinated by bot technology and I wonder if you're seeing significant evolution in bot technologies. I mean, bot A bot as an automated broadcast tool, it seems to be fairly straightforward and understandable, but now as we have the development of these language models such as GPT-2 and related technologies, I wonder if you're seeing that technology be swept in. I'm asking it particularly in the context of information manipulation.
Sandra González-Bailón: Yeah, I mean, I think that the actors who are creating those bots, malevolent and non-malevolent actors, they're becoming more sophisticated in trying to emulate human behavior. There's also, as you probably know very well, like very high-profile failure stories like Tay. I think it was called Tay, Microsoft's bot, for example, that was launched to learn from human interactions and it became a racist bot in a matter of hours, not even days. And I do— so, you know, so as AI technologies evolve and I think it's going to get increasingly sophisticated. But I'm an optimist by default, and my point of view around this is that that is not necessarily a danger. Like, we can also use bots, and I discuss this with my students a lot because they have very imaginative ideas on how we can design bots to actually improve civic engagement in online interactions, right? Sort of the quality of civic conversations online and whatnot. We can think of bots also as policing devices. And policing is a word that has the wrong connotations. And I don't know, like traffic lights, right? So we can use bots to automate, to monitor and sort of encourage the right— nudge people to behave in a particular way when engaging in online discussions. And there's research about these, right? So Kevin Munger, who's a colleague at Pittsburgh University, he has this paper where he work with automated accounts to try to sanction Twitter users who were harassing other users. And so he had this very imaginative research design in which he used bot-like accounts to try to instill certain norms of behavior on those users. And when the bots were perceived to belong— you know, this in-group, out-group dynamics, right? So when the avatar, the photograph of the bot shared the same characteristics of the person being sanctioned, of the user being sanctioned, the user responded to that. And so the levels of harassment went down. And so if we can scale that up and we can use all these new technologies to actually think about ways of regulating or improving these online interactions that would reduce harassment, we reduce instability. That's a good use of the same technology that those feeding misinformation are using as well. And so—
John Markoff: There was another aspect of your research that really intrigued me, and I probably longer than we can pursue it in depth, but I did wanna ask you just a little bit about the structure you've seen in looking at the emergent sort of social networks that come out of activist movements. You talk about, in one talk that I saw you gave, a committed core of activists who basically engage with a periphery that is less engaged, but they use that periphery to amplify their message. I just, I guess the general question I wanted to ask is what is the sort of the meaning or the impact of those kinds of structures and are they, you know, is that the way the model always works?
Sandra González-Bailón: Yeah, so when we talk about online networks in the context of activism, but in general about online networks, we have this model of decentralized structures with no unit of command and control, right? Like, so there's no hierarchy in the sense that no one is really in charge, but they are, from the point of view of the structure, they are very hierarchical, right? And so these networks emerge from the bottom up. And so, but very quickly, there's, you know, they evolve in such a way that a small number of accounts attain a disproportionate level of attention, or in network terms, a disproportionate level of centrality. And so these are the reference points, right? And so these are not necessarily leaders in the way that we usually think about leaders in social movements, because there's no formal organization, right? But they are very hierarchical. And most online networks are hierarchical in this sense, right? There's a small core of highly committed protesters in the context of these mobilizations. But there's usually a small core of highly visible users or accounts or nodes, and then a large periphery that are only loosely connected among them, but they are all connecting to the core. And this has implications for how information diffuses in networks, and it's one of the reasons why online networks are such effective tools for the diffusion, for the fast diffusion of information in real time, which again, in the context of protests, especially when there's confrontations with the police on the streets, is very relevant because you get information out before traditional journalists, for example, could in the past. And these networks are also international. And so in this paper you were referring to, we pay attention to the protests that arose in Gezi Park in Turkey. And literally, the national television was showing— I think it was a documentary on penguins— as protesters were confronting the police on the streets. And the only way in which the news— other people living in Turkey could follow the events was through online networks and Twitter in particular. And so, yeah, so to your question, this is a very universal feature of online networks. Like, if we get technical, you know, most of these networks get a very skewed degree distribution, which essentially means that a minority of nodes have a disproportionate number of connections and the vast majority of nodes are very peripheral. And this emerges from the bottom up, right? No one is really designing these networks to be this way, but most networks have these hierarchical structure And it's good for the diffusion of information. Now, we are very careful. So, you know, I'm a sociologist by training, although now I'm in a communication school and most of my research has, you know, focuses on communication dynamics. Whatever happened online doesn't necessarily mean that there's going to be a change offline, right? So which is, you know, what I just said is true from the point of view of how news and information circulate online. That doesn't necessarily mean there's going to be an impact in terms of regime change or that traditional social movements are not relevant. They definitely are.
John Markoff: Yeah, interesting, but it's interesting in its own right. You know, a final question. I think this is based on something that you wrote, I think in the wake, this was in 2012, an essay in the wake, I believe, of the Arab Spring. You posed the question, I just wondered if you have an answer now, it's been 7 years. You said, can social media transform bursts of political activism into stable forms of participation. So now it's 7, 8 years later. I wonder if you have a view of that question.
Sandra González-Bailón: Well, I mean, I guess that it depends on where social media operates. And that's another complaint I have in the state of the field right now, because a lot of attention is being paid on the US, on the role of social media in the US. And the technologies are the same all over the world, but depending on the political context in which they are being used, the consequences is changed, right? And so in Spain, the Indignados movement, which was, it happened just before the Occupy movement exploded in the US, is a precursor of the Occupy campaign. The mobilizations were very similar. The protesters camped on the main squares of the main cities in the country. And out of that movement came out several new political parties. And so those mobilizations transition into more institutionalized forms of politics. And they are represented, like, one of these parties is in government right now. And they are represented in Europe at the local levels, local administration and whatnot. And so in that sense, it was successful. But I wouldn't say that social media allowed those movements to become consolidated. Social media was just a sort of it facilitated part of the process of enacting social change, if by social change we mean in this case having new parties come up and voting for them. And so, you know, social media can only do so much, but I do think it definitely helped them spread their voice and their message and gather a critical mass of people that then did the work to do the rest. Yeah. So I'm still optimistic about, you know, at the end of the day, social media and digital technologies are just tools, and we can do good or evil. And, you know, it's how we use the tools that matters. So that would be my answer.
Narrator: So CDA Section 230 has been popping up in the news a lot more recently, and it seems people have all political persuasions, uh, want a crack at dealing with it. And I'm wondering, over the last year or more recently, have you seen any useful evolution of the discourse on Section 230? Have you seen any improvements in how we are framing our understanding of Section 230 and that will allow us to better use and regulate and understand these technologies?
Sandra González-Bailón: Not really. Not really. But I do, in general, worry that we are framing these discussions sometimes on the basis of misunderstandings of how these technologies really work. And so I think there's a lot of work to be done in terms of public engagement as well. And this is with regards to the project that John mentioned, the Facebook 2020 project, of which I can't say much because, you know, under embargo and we can't really say that we have the findings. I think I already know that we'd have to go through a lot of public engagement in terms of explaining exactly what is it that the research can accomplish and how we can fit that into rethinking how these platforms should operate. And so I think I take that as, you know, I think that's the great contribution that researchers can make is they try to offer the evidence that will inform some of these conversations. And I don't think that right now the public perception of how these technologies are impacting democracy or more generally how digital technologies are changing society, like, you know, bots, like bots seem to be like these Leviathan kind of monsters that can manipulate everything. I think we need more work at allowing the public or communicating to the public exactly what these technologies can accomplish and what they do. And then we can have a conversation of what they should look like. And there's a lot of empirical gaps yet to be filled to be able to have that conversation in an informative way. But I do worry that regulators— and there's so much pressure that the timings are not aligned in terms of what the political pressures such as we should do versus what's feasible given our knowledge right now, given the state of the art.
Narrator: Well, speaking of state of the art and our current knowledge, what would you be willing to share with us about what excites you about the Facebook 2020 project you've been working on?
Sandra González-Bailón: I think this Facebook 2020 project, it's super interesting for many reasons. But mostly because we are forging a model of accountability that is still evolving, right? Like, so how do we hold these companies accountable? And in terms of, you know, measuring really the real impact, and so, and then, you know, on the basis of that, revisiting how they operate, right? And so how can we believe their findings when they don't make their data publicly available? And so this collaboration between academics and data scientists, I think it's gonna— I'm very excited because it's, you know, it's creating those channels for accountability that we don't currently have.
Narrator: And when can we expect some of the research from these projects to become available?
Sandra González-Bailón: So I think in the announcement it said like summer, next, in the summer of next year, because we'll stop collecting the data after the election and then we'll need some time to actually process the data. But so yeah, I think that was in the announcement. I think that the date that was given was summer 2021. And so that's when we expect to be able to discuss the findings. And then the discussion will be based on actual evidence, right, as opposed to hunches or speculations or fears. It's a lot of fear in these discussions as well that I think is adding some noise because the stakes are very high. I mean, these are really important issues and we don't want anyone interfering in elections. So yeah, so that's the timeline, Joe.
Narrator: Thanks.
John Markoff: Yeah, well, that's wonderful. Thank you for this. I have to tell you that I'm looking forward to having a conversation with you after your research group responds to the— and the dust settles after this coming election. So, look forward to another conversation.
Sandra González-Bailón: Thank you so much, and hopefully we can meet again soon.
John Markoff: Talk soon.
Sandra González-Bailón: Bye-bye.
John Markoff: Bye-bye.
Narrator: That was John Markoff interviewing Sandra González Bailón. To learn more about the topics in their conversation, be sure to check out this episode's notes. We've got links to Sandra's research papers and her 2017 book Decoding the Social World: Data Science and the Unintended Consequences of Communication from MIT Press. We've got more interviews ahead for the human-centered feed, and we'll be continuing to publish more episodes from the CASBS web series Social Science for a World in Crisis. So be sure to subscribe in your podcast app of choice And while you're there, you can always rate and review us. We've been loving your comments and feedback so far. It's really helping us to tune the show to something we think you're gonna love. You can always visit the CASBS website at casbs.stanford.edu or follow us on Twitter @casbsstanford for up-to-date postings on the topics and events in the behavioral sciences community. Until next time, from everyone at CASBS, thanks for listening.