John Markoff sat down with Nate Persily, a 2017-18 CASBS fellow and the James B. McClatchy Professor of Law at Stanford Law School. They discussed the challenge of ethically using social media data in both social science and politics, election security and law, and whether democracy can survive the internet.
Nate Persily’s homepage with links to publications and key works
Social Science One builds industry-academic partnerships to advance the goals of social science in understanding and solving society's greatest challenges.
CASBS Symposium: The Consequences of Technological Developments for Politics and Government
Center for Advanced Study in the Behavioral Sciences
@casbsstanford on twitter
John Markoff: From the Center for Advanced Study in the Behavioral Sciences at Stanford University, this is Human Centered. I'm John Markoff. Nate Priscilli is a former CASBS fellow and currently the James B. McClatchy Professor of Law at Stanford Law School. Where he has taught since 2013. He is known as a constitutional scholar, and he has served as the senior research director for the Presidential Election Commission. He has been a court-appointed special master overseeing significant redistricting cases. A prolific author, Persily has been cited frequently by the Supreme Court on issues of election law. We spoke about the question of whether democracy can survive the internet.
Nate Persily: So I'm Nate Persily. I'm the James B. McClatchy Professor of Law at Stanford Law School, but I'm also co-director of the Stanford Cyber Policy Center and the Project on Democracy and the Internet at Stanford. And a lot of my work these days is at the intersection of technology and democracy, particularly how social media and new technologies are affecting democracy around the world, what can be done about it, in everything from polarization and disinformation to political advertising and foreign election interference.
John Markoff: Let me begin by asking— I mean, let's go backwards first— your time at CASBS and what you focused on while you were here. And when you were— you were a CASBS fellow?
Nate Persily: I was a CASBS fellow from 2017 to 2018, and I also I am a faculty affiliate and was for the last 2 years from 2018 to what will be 2020. So during my time at CASBS, I began sort of the serious research on democracy and the internet and dedicated that year to not only the scholarship on that, which has been published in various forms, which I can talk about, but also thinking institutionally about how to do analysis of social media data to discern the impact of Facebook, Twitter, and other platforms on democracy around the world. So it was here that I started the work on what's now known as Social Science One, which is the effort to make Facebook data available to the world's scientific community.
John Markoff: And, well, tell me the backstory about that. You did it here. Was CASBS the backdrop, or did some of the conversations you were having at CASBS spark that?
Nate Persily: Well, part of it— there's sort of a long run-up to what has now become Social Science One. I mean, for several years Facebook had been having meetings with academics to talk about data access, and they had always been trying to figure out how to make this happen, and then Cambridge Analytica hit. And so Social Science One, to some extent, grew out of the ashes of Cambridge Analytica because it was the possibility that Facebook would go the way many platforms have gone, which is to shut down access for outsiders. And instead they took a different route, which was to try to figure out a safe, privacy-protected way for academics to get access to the data. Since I had been working on those issues here at CASBS as well as in my other sort of parts of my life, I was known to them as well as to Gary King, who really is the intellectual force behind this project. He's a Harvard professor who had developed the model of industry-academic collaboration that is Social Science One and that we hope to repeat outside of Facebook and other companies. And my— let's—
John Markoff: maybe I could do a brief update on where you are with Social Science One. How close have you gotten to where you started to go?
Nate Persily: So the interesting thing about the progress with Social Science One is that many of the excuses that I think Facebook had historically made for not sharing data, and which I sort of drummed up as mere pretext for their hoardiness with data, I now have become convinced are genuine legal objections. And so Facebook promised data to us that they simply have not delivered and couldn't deliver at the time that they promised it to us. And so we have been waiting for almost a year to figure out a safe, secure, legal, privacy-protected way to make the data that they promised to us available to us. And the main obstacle is privacy, particularly the GDPR, the General Data Protection Rule out of Europe. And so, if we were to do this differently, we would not have announced an RFP even yet, but we announced an RFP a year ago based on assurances that we had from Facebook that, all right, we would have the data available in a few months. And so we figured we didn't wanna like, we wanted to get the data in the hands of researchers as quickly as possible. So we issued the RFP based on what we thought was the model. Since then, they have spent, well over $10 million, maybe $20 million on this project trying to figure out how to take Facebook data and make it available to the world scientific community in a privacy-protected way with new technology like differential privacy added to it and ways to protect the identity of individual Facebook users. The problem is that every time we solve one technical problem, we confront a new legal problem. And so what we need is clarity from the governments that what we are trying to do is kosher under, under their laws. And, you know, what we— this research has to get done at some point. This, this model will work because we cannot be in a situation where most of the world's relevant social scientific data is locked up in for-profit companies where no independent analysis can be, can be done. And so as much as the privacy community and the regulators are right to be concerned about leakage of the data out of these companies, it's not as if this data is not being analyzed, right? The question is who's going to analyze it? It's being analyzed by people, the data scientists inside the firms, right? But no one trusts them to pronounce as to what their effect on democracy around the world has been. But so if you want to figure out the critical questions about how much disinformation there is on Facebook, You've got to be able to have somebody outside Facebook who can analyze the data, right? And you have to have the normal scientific process of people arguing with each other and questioning each other's models do that. And so, in order to figure out how much disinformation is on Facebook, you need to have access to the individual-level data. You need to figure out how many people are seeing certain links from Russian bots or anybody. Now, when I say you have to have access to the individual-level data, that doesn't mean you have to have access to your Facebook account or something that you personally have seen. We have to have access to aggregate statistics about what individuals on Facebook have been seeing. And so that's where the rubber meets the road, and that's the problem we've been trying to solve.
John Markoff: I want to come back to Facebook, but a little bit more on CASBS in the context that that you were working in.
Nate Persily: Did you not come here—
John Markoff: were you not working on a book-length project? You had a book-length project. Was that the original? I'm sort of asking this from a CASBS point of view because traditionally CASBS had been about individual sabbaticals and working on books, and it's moved more in a project direction.
Nate Persily: Well, so that— no, but that— so that has— there are two parts of that that have been published now. And it's interesting because it has also ended up being kind of a project as well. So the three components of my proposal to CASBS were thinking about how technology is affecting democracy in general, political campaigns in particular. And so the three sides of that were thinking about how the impact of technology and how we vote. This has to do with online voting and electronic poll books, and I can talk a lot about that as well, election administration, which I have a background in. The second is How it affects political parties, in particular in the era of big data, how does the data that is being made available to certain firms, or these new kind of firms. The argument was gonna be that that is actually what a political party is right now, it's a database, right? And to think about what that means over time. And then the third is the stuff we're talking about now, how it affects campaigns, whether it's microtargeting or use of social media and small donor fundraising. And so I've actually, there have been 3 things that I did accomplish here at CASBS, CASBS, that, that have gone forward. The, the, on the election administration front, what it ended up being is the report of the National Academy of Sciences on election technology, which, um, has had a huge, uh, impact, at least in, in the way that we are talking about things like voting machines right now. And, um, even just, uh, this week, as we're talking, right here in late or mid-August, there's a court decision that just came out yesterday out of Georgia saying that they cannot use their electronic voting machines, showing and pointing to the National Academy of Sciences report.
John Markoff: Does that seem to dial or push the dial in the direction of paper trail? Yeah, yeah. So that's a good outcome, basically.
Nate Persily: Yeah, and that was what we recommended. And I talk a lot about how I had this National Academy of Sciences panel that I was sort of involved with at the front end, which was led by Lee Bollinger, the president of Columbia University, and Michael McRobbie out of Indiana University, and had all kinds of scientists that were on it. But that— so that, that is it, one book, you know, that, that was produced. The second, um, is the, the main project that I ended up doing here at CASBS, which was on the, the effect of social media on democracy. Then I've released as a monograph with, for the Kofi Annan Commission. So one thing that also happened during my time at CASBS toward the end, I actually met Kofi Annan. He came here and he was very interested in the effect of technology on democracy around the world. And we met here with various folks in Silicon Valley, and that led to the creation of what is the, the Kofi Annan Commission on Elections and Democracy in the Digital Age, which we are running through Stanford, but includes the former presidents of Costa Rica, Mexico, Estonia, former prime minister of Belgium, Alex Stamos from Facebook, Megan Smith from formerly Google, but now former CTO of the US, and so a whole lot of other folks as well. And so that's the framing paper for that. That is the book that I wanted to, but I wanted to get that out quickly. Updating it into a book as well. And then there is another book that will be coming out by Cambridge University Press that I'm editing this year on social media and democracy, the state of the field and prospects for reform, which should be out in 2020, which is all— it's on basically every sort of topic at the intersection of social media and democracy, disinformation, polarization, political advertising, all those other topics that I was mentioning before.
John Markoff: One more general CASBS question before we leave it. The value, you know, look back on your year and was there a value to serendipity? You know, this notion of taking these disparate individuals and their disparate disciplines and putting them together. Now that you've been here and been through it, what were your sense of the value of that?
Nate Persily: I think that's absolutely the case. I mean, in some ways, you know, CASBS is an expression of the Silicon Valley ethic in that regard, which is that, you know, If you get a lot of smart people and throw them in a room, they will bounce off each other and create new kinds of energy. And so that was, I mean, look, meeting you here, and I remember sort of in detail your talk that you gave here because it's something that I had, I hadn't quite appreciated though I'd read the books on it was sort of the genealogy of the Silicon Valley ideology, right? Or, and—
John Markoff: Still struggling.
Nate Persily: Well, but it's still, you know, that Fred Turner here obviously has written at Stanford, has written quite a bit on this. But, but seeing that and then seeing how it plays out, for example, or that you still see the shadow of that in decisions about like, you know, free speech and democracy, right? How can we, you know, how should we think about the utopianism of the early internet entrepreneurs in light of everything? The dystopia that we are now living in. Um, and so, uh, that I remember, um, when, uh, there were folks, uh, like that we had here who were, uh, very big on Wikipedia, right? And I became convinced, and it was through my conversations here, that Wikipedia is an absolutely essential part of the web architecture right now, and that people don't quite appreciate how central it is to all of these debates dealing with things like fake news or polarization, and more importantly, that the model that's being used through Wikipedia is really in some ways the foil for all kinds of reforms that we're thinking about throughout the ecosystem. I can talk a little bit about a project I'm working on, on crowdsourced fact-checking, that is trying to figure out whether average people can do can scale up the fact-checking regime that is really dealing with a very small slice of the disinformation problem. And it's only by meeting the folks here who were studying Wikipedia that we— I could talk, let alone, you know, I've actually traveled to Taiwan as a result of my time here at CASBS, meeting some of the folks who were from Taiwan here, so giving some talks there. It opened me up for me, relationships with people around the world who are working on the stuff that I'm interested in. Yeah.
John Markoff: I wonder, the original goal was to have a social science equivalent for the Institute for Advanced Study in the physics community. Do you think that that's still a good goal for a CASBS-like institution, to create that kind of center?
Nate Persily: Yes. And I think right now, we're at a— We seem to have an inflection point every 2 or 3 years, right? So, but let's say that we're at the newest inflection point, right? Of trying to think about what the mission of the social sciences is in this new age, particularly in Silicon Valley, and thinking about how applied the social scientific inquiry should be, how much we should be related to government work, right? I mean, that's one of the real innovations here at CASBS is that we have work that's going on in thinking about experimentation and social use of social science and big data with government. What is the boundaries of social science in this era of big data? One thing here at Stanford that we're really trying to do is to get the scientists to start playing nicely with the social scientists more so that we can exploit their talents Whether we should start thinking about us as particular siloed fields in social science or more just like, "All right, well, here's a problem, and let's throw all the intellectual energy at it to solve it." That raises one of the things that I've noticed at Stanford in the last 2 or 3 years that sort of plays on this.
John Markoff: Stanford is littered with courses on ethics all of a sudden. I mean, just for a variety of reasons, but I have only looked at it from the engineering side. What does it look like from the social science side? Does that ethics discussion come up? In those academic disciplines, and has it affected you?
Nate Persily: I think it does. We've had our share of controversies at Stanford, actually, on this. Well, historically, you go back to the Stanford Prison Experiment, we've been at the forefront of—
John Markoff: I was in summer school, just on passing, I was in summer school that year, and I saw that ad, and I thought about applying for it, I was looking for it, but I was too lazy.
Nate Persily: Well, I guess you would have ended up being a prisoner then, not a guard in that experiment. But actually here at CASBS, before I was a fellow, I actually was engaged with some folks here on questions about the ethics of experimentation in elections. I mean, this has been a big issue on the Stanford campus because of some mishaps that have happened there. And so the ethics of research regarding elections in the era of big data, as well as thinking about field experiments. right? And so we had some really good— and these are continuing discussions that have happened here at CASBS about, like, for example, can you do a control— or not controls— a field experiment where you randomly mobilize people to vote, right? Can you try to— Facebook did that. Yes, Facebook did. Well, yeah, Facebook's a company, right? And there is controversial. But for example, if we If we say, you know, what we're trying to do is trying to figure out what would motivate African Americans to vote, right? And you say, all right, well, we're going to take this community and we're going to send one out of five African Americans a message that says, and this is a classic field experiment that was done by Don Green, not with African Americans, but just generally, um, um, you, you know, all of your, uh, it is publicly known whether you vote or not. And so your neighbors will know whether you chose to vote. That's all we do, right? All right. Does that lead to an uptick in voting? And that was actually used by the Ted Cruz campaign, as it turns out, in Iowa, you know, last time around for when he was running for president. And so it has real purchase. But should social scientists be potentially affecting elections, right, by doing these kinds of experiments? And at the same time, though, if you want to understand voter turnout, and what actually works, you kind of need to, to be in the field. Now, those are tame kinds of experiments. There are all kinds of experiments that are done in the international, um, side of things that really, you know, matter of life and death, right? Where you, you randomize, you know, who gets mosquito nets, right? Or something like that. And to figure out whether the malaria—
John Markoff: well, this is interesting because one of the things you said recently in describing your work is that you compared yourself to an anthropologist studying cannibalism, which of course brings us right— I was just I mean, I guess you just did, actually.
Nate Persily: Well, but no, no, it sort of focuses the mind, right? So I often say that those who study elections in this day and age, and particularly election administration and the law of democracy, are like anthropologists studying human cannibalism. And by that, I mean you have to suppress normal human impulses of disgust and revulsion and replace them with fascination and curiosity. Because it's really a coping mechanism at that point, because we are living in a time right now where it is extremely easy to get dispirited and despairing about the state of democracy in the United States and around the world. And whether it's the stuff that I do outside the tech area in dealing with, you know, voting rights and redistricting, campaign finance and the like, or seeing what's happening in the tech area, right? Where, whether you're— I'm looking at genocide in Myanmar or bots that are manipulating elections, foreign interference, right? Or issues of life and death. It's a really difficult time to be studying these issues. And so that's what I mean is that you have to have some kind of academic detachment in order to cope with this.
John Markoff: Yeah. Where are we on the question of one person, one vote these days after the last wave of— could you speak?
Nate Persily: So in redistricting generally, we were delivered two gifts by the Supreme Court in their set of decisions in June 2019. One dealt with partisan gerrymandering, the other one dealt with the census. As it turns out, so one other thing that kind of distracted from my time here at at CASBS, and I felt guilty about in some respects, but felt I had to do it, was that I was appointed as special master for— to draw state legislative lines in North Carolina, and then as the court-appointed expert to help with the court plan for the Pennsylvania congressional districts. So that did suck away some of my time here at CASBS. I asked our director, Margaret Levy, for permission, but it was It was a pretty important set of cases. Not to get too into the details of those, but in Pennsylvania, they had struck down the plan as being a Republican gerrymander. And so the court's plan was remedying that. In North Carolina, they had struck down some of the districts there for the state legislature as a racial gerrymander. And so I was instructed to correct that. This is something I've done for the last 20 years or so in various states. And, and then what the Supreme Court did this year is that the Supreme Court handed down a ruling that said that under federal constitutional law, there is no bar to partisan gerrymandering, and specifically that the courts will not entertain these cases because there's no administrable standard. So since partisanship is always some part of the redistricting process, The question has always been, how much is too much? And the court says, we can't figure out how much is too much. There's no, like, there's no rule like one person, one vote, which is 6th grade arithmetic, right? And so if we can't come up with a simple rule as to define what is too much partisanship, we're gonna sort of say the federal court should retreat from this, and that's what they've done, and that was the cases out of Maryland and North Carolina. That doesn't mean it's the end of the line. You could still have these state court cases. You could still have referenda and other kinds of reforms, but at least the federal courthouse door has been shut. Related to that, in some ways maybe more significant, depending on what happens in the next few years, is the Supreme Court saying, with regard to the census, that the administration's attempt to add a citizenship question on the census was illegal, or at least that they hadn't justified it sufficiently under the Administrative Procedure Act. Now, everyone had complained that this decision to add a citizenship question would, wreck the census in a particular way, which is that Latinos and other immigrant groups would be less likely to fill out the census form. And as a result, their communities would be underrepresented in redistricting and reapportionment of Congress and the like, as well as some federal funding. So the Trump administration has retreated from adding it as a question so that you will not see a census form that has a citizenship question on it, but They've said they're going to try to gather citizenship data from all other federal records and then to release a redistricting data set that will include that, which might make it possible for jurisdictions to draw districts on the basis of equal numbers of citizens as opposed to equal numbers of people, which is the way that we've done it up till now. Right. Right.
John Markoff: You know, we've had the Mueller report, and we've had endless reporting on Cambridge Analytica, and I just was wondering You today, do you think it was established that the GRU employed Cambridge Analytica data in attempting to target Americans?
Nate Persily: Is it— No, I don't think there's evidence that the GRU used Cambridge Analytica data, in part because they didn't need it. I mean, because the Russian incursion into the U.S. election was targeted in certain ways, right, but not targeted in a way that you needed, like, psychographic profiling provided by Cambridge Analytica. They developed audiences themselves, right? They used Facebook in the way that Facebook is intended to be used, right? And they did everything that, you know, if you were a grassroots organization that you would have done, you figured out who was interested in your product or ideology, right? You advertised in a way to develop those audiences. You could see, you know, who clicked on, what types of people were clicking on your links and the like. They had media companies like Sputnik and RT that were very popular on not just Facebook but on YouTube and other platforms. And they didn't need to be surgical in their precision because a lot of what they were doing— I mean, it's clear that they had a candidate preference in the election. But a lot of what they were doing was also fostering division and mimicking the polarization that was already happening in the political ecosystem here so that the— as is now famously known and revealed in the Mueller report— that the Black Lives Matter fake group that the Russians had started rivaled, if not exceeded, in popularity the authentic Black Lives Matter page that was on Facebook. And so we've published through the Cyber Policy Center our first publication called Securing American Elections, which has a series of chapters on the different aspects of the Russian incursion and then what to do about it. And so I wrote the chapters on political advertising and on dealing with foreign media organizations like RT and Sputnik, as well as I did some work on the voting machine chapter as well. And so I think, you know, what's happened is that Facebook was completely unprepared for what the Russians were trying to do, because when they thought about security, they were thinking about account takeovers and the kind of typical cybersecurity stuff. They weren't thinking about, you know, hacking people's brains. They were thinking about hacking people's systems, right? And so what the Russians did is they hacked people, right? They didn't hack the actual systems. We still don't know enough about what was actually the kind of— some of the details of the Russian campaign, but we certainly know a lot of it from the Mueller report.
John Markoff: So now you have these new mechanisms to provide advertising transparency.
Nate Persily: Yeah.
John Markoff: Are you confident it will make some difference? in the next round, that the transparency actually will be used in some way to limit some of those—
Nate Persily: Yes, I think it does. So let's be clear about what these transparency measures are. So both Facebook and Google have done different things, but they— and Twitter, but they all have libraries or archives of the actual ads so that if you want to see what candidate-related ads are being put up on the platform, you can now see them through the library. Now, that is easier said than done in a lot of ways as to what is a political ad and what is a candidate ad. In addition to providing the advertising, if you are running a political ad, and they each have different definitions of what that is, then you have to go through a verification process. So they send a postcard to an actual address. You have to have the name of the person, right? And so it is still possible to game that system and to deceive. And that happens, but it's going to be a little bit more difficult for, you know, you're not going to be able to pay in rubles for ads like they did last time. Now, where things have gotten off the rails a little bit is to try to figure out how to regulate political advertising that's not candidate-related. So it's easy to figure out if someone uses words or images of Trump and Clinton, it's like, okay, That's political advertising. But if you look at the Russian ads, right, 80% of them were not candidate-specific, right, and they weren't advocating for the election of defeated candidates. They were on issues like guns or immigration or religious issues and the like. And so Facebook had— because Facebook's transparency regime is reverse-engineered in a lot of ways to deal with the Russian problem, they have said, OK, not only are candidate-related ads going to be part of this archive and required this kind of disclosure, but we are going to include advertising on issues of national legislative importance. It's like, oh, OK, so now who's going to decide what is an issue of national legislative importance? The answer is Facebook. And now they have stepped in it with this. I mean, it's— first of all, it is admirable that they do this. It is the right thing to do. And it is also an impossible problem. Anyone who— those of us who've been working in the campaign finance area for the last 20 years know that the so-called regulation of issue advocacy, is like the most— it's the thickest of political thickets, right? It is so difficult to figure out what is political because you can't figure it out ex ante. What's going— is football going to be a political issue? Well, only when Colin Kaepernick then takes a knee. And so therefore, when Nike runs an ad featuring Colin Kaepernick, well, are they engaging? Are they basically— should they be roped into the transparency regime? That is Facebook political advertising. Similarly, The New York Times, right? When it boosts a post on some political issue, okay, are they now a political advertiser, right? And you say Facebook has now originally said yes. They've now come around to a media exception. But who is the media in this day and age, right? And so, and then NGOs, like if you're doing, if you're, if you're the American Cancer Society and you're soliciting donations, —right?—at a time when healthcare is really a salient issue in people's thinking about politics. Are they now going to be deemed political by Facebook? They don't want to be. Now, all of this just means that they then have to enter the disclosure regime. It's not that they end up being banned from advertising or any of this kind of stuff. They have to go through the postcard verification process and then get their ads put in the ad library. But, you know, New York Times doesn't want to be called a political organization. These charities don't want to be called that. And then Facebook doesn't want to be in the position of figuring out like what they've done, which is that they've isolated 20 issues that are issues of national legislative importance. And one of those issues to start out was values. It's like, well, that seems pretty capacious, right? I mean, that could touch a lot of things, right?
John Markoff: And this was only for the advertising facet of the puzzle, not for the organic posts?
Nate Persily: Yeah, well, and that has to be the right way to do this, though. I mean, you can't possibly have like postcard verification for all political discussion on Facebook. It's, it's just going to be— and, and part of this is you might think that Facebook, that, you know, the rules for advertising should be pretty strict. I mean, basically, let's face it, Facebook and Google are advertising companies. That's what they are, right? And so this is central to the business model and who they are. But, um, political advertising is a tiny— however we define it— is a tiny slice of the overall advertising. And frankly, it's only significant in the United States in terms of money, right? And because of our political system and parties and consultants and the like, they have contemplated eliminating political advertising on Facebook if they could figure it out, because right now, given all the transparency rules that they're throwing on it, it ends up being more headache than it's worth. And economically, it's not significant. The problem is they still are going to run up against this question as to what is a political ad, right? They're going to have to figure that out even if they're going to be banning political political advertising. And, you know, they also— there are good arguments to say that, you know, if Facebook closed down political advertising, it would have particular effects on incumbents versus challengers. You know, Facebook is in some ways— it still is, like the internet generally— the ability of, you know, the small person to have an outsized voice.
John Markoff: If you took, let's say, presidential election advertising as a big lump these days, What fraction of it goes to social media?
Nate Persily: That's a good question. And actually, that also, I think that was, that was the, I did mention the third thing that I published when I was here at CASBS. So there's a report that we issued in December of the year that I was here at CASBS through the Bipartisan Policy Center on sort of campaign finance post-Citizens United. And there's a large section of that, which is on all the available data dealing with online and that, that, and I should say this was a coalition I put together of social scientists that was led by me and Bob Bauer, who was a White House counsel under Obama, Ben Ginsburg, who was counsel to the Romney and Bush campaigns, and we wanted to come out with a sort of statement about what, what had been happening in the campaign finance system. And so of the share of dollars that are spent on communications that are going social media versus other kinds of media, It ends up being somewhere around 20% or something like that, but it really varies and it depends what we're talking about. And, you know, but it's gone up, you know, gone like tenfold from previously, right?
John Markoff: One of the things you said recently in an interview is if $100,000 in Facebook spending is enough to swing an election, then we have a lot of political consultants who have not been behaving so smartly, which really hit me because I began to think about. What is this spend doing to the system, the way we elect candidates? I mean, sort of the big picture question: has it been completely transformed? But yes, well, it is.
Nate Persily: So there are a few things there. First is what I was trying to say with respect to the Russian advertising. Like, my view about the Russian incursion in the US election is that advertising is really a tiny part of what they did. All of these other aspects of what they did was much more significant. The organic social media posts, the use of Facebook groups, the use of broadcasters or websites like RT and Sputnik. The actual advertising was in some ways the most notorious thing that they did because it showed how the platforms could be abused and in ways that should have been preventable. And so they spent $100,000 on Facebook ads. So it kept— if that was enough to flip the election itself, then that, like I said, suggests that the campaign consultants have not been spending money wisely for a long time. Now, we still have no idea, particularly when it comes to presidential elections, what makes a difference in terms of vote choice from an advertising perspective. That was true with television advertising, right? The studies on television advertising are inconclusive as to whether campaigns even matter, um, because, you know, so much of it is determined by macro indicators, so much of it is determined by free media, Right. And part of the story, say, with the Trump campaign, it's like, yes, they were— they did very well on social media. They did some very smart things. But what they also did was get $2 billion worth of free media coverage, you know, which is sort of what money can't buy. So there is a culture shift among political consultants so that— and this is a generational shift, right, that you, you know, through the 2016 election, you still had the old hands who were, who were saying television, television, from television. And one of the reasons— there's a kind of industry capture side of this story, which is that the way political consultants often will get paid is as a share of the advertising buy on these, on television channels. They haven't quite figured out how to— what the payment system is on social media. It's complicated. Look, they're all getting paid fine, but, but it's like there'd been a typical way that the industry worked when it came to television. Disruptive. But with Brad Parscale, who is the social media director for the Trump campaign, you know, they— the innovation, I think, with them was to really leverage the expertise of the platforms themselves so that they were working with the tools that Facebook, Google, and Twitter were giving them, and the people who were embedded really with these campaigns that were helping them from Facebook, Google, and Twitter. And then the Clinton campaign, for the most part, had developed— because they had a lot of the Obama folks who were Jedi Masters, or at least seemingly so, on technology. So they had a huge sort of operation, and they spent as much, if not more, than the Trump campaign on social media, as it turns out. But a lot of it was spent outside of the actual platforms themselves on sort of developing the tools.
John Markoff: Looking forward, you know, what are you seeing on the horizon, the evolution of bots or conversational AIs? Are there things that already have appeared that seem to be issues for the next election cycle?
Nate Persily: Let me speak even more abstractly about that. Let me talk about what I think from the standpoint of the health of democracy, what are the kind of emerging issues that we're seeing, specifically specific reference to technology. So in general, we now recognize that there are certain features of online speech and communication that are in tension with democracy. In many ways, they were seen as the most pro-democratic aspects of the internet. Things like protection of anonymity online, which, you know, is, um, in some, in the US it's a First Amendment right to speak anonymously. Right? But that's what gives you the bot problem. That's what, what exacerbates the hate speech problem and the like. Similarly, the removal of gatekeepers like the major newspapers or the major networks to then determine the boundaries of political debate has now replaced— the internet has replaced them with virality as the coin of the political realm. So then what people see is what is popular and what's most viral. That means that, and we know as social scientists what types of communication are viral, those that appeal to emotion, those that appeal to outrage. And so those campaign strategies and communication techniques that are, that sort of work in a viral transmission atmosphere are the ones that are favored in this system. Similarly, so you've got sort of virality, you've got anonymity, you have the problem of homophily echo chambers, which is a debated idea how much really are online lives politically monochromatic as compared to, say, our offline lives, because the truth is you're probably more likely to run into someone on Facebook who's of a different political ideology than you are, than you are in your local neighborhood. Uh, but nevertheless, whether you're talking— if you're talking about like the deep dark corners of the internet, 4chan, 8chan, these subreddits where like Pizzagate conspiracy theories develop, that is something that is facilitated by the internet, which really didn't exist in the pre-internet age. Add to that problems of monopoly with Facebook and Google and problems of sovereignty where domestic governments are simply unable to wall themselves off from political communication and incursion into their democracies by whether it's Russian actors or foreign platforms and the like. So that's the basic structure of the internet, why I think it has challenges for democracy. Now what in a more sort of micro way are are things that we're seeing on the horizon. When you mentioned AI before, the issue du jour is thinking about deepfakes, right, and artificial video and how that will affect democracy. We here at Stanford have lots of projects on that, right? The ability of people to manipulate videos so to create alternate reality where someone can be accused of saying that they said something that they didn't actually say, right? And then you have the video to prove it. And the effect of the rise of deepfakes is not so much that people will disbelieve or believe something individually that is false, but now they will lack trust in everything that they see, right? Which is in some ways the goal of a lot of the Russian incursion was that. If you talk to the folks in Silicon Valley about this, a lot of what they say is, look, shallowfakes are bad enough, right? It's like, we don't need, you don't need the fancy technology. You've edited pictures that were seeing all over the place. You could— the Nancy Pelosi altered video was not a deepfake at all, where they slowed down her speech to make it seem like she was intoxicated, right? So it's part of a larger problem of manipulated imagery and how this— but again, you overlay that onto those other problems of anonymity and echo chambers and how this stuff gets shot around. Second, a less popular or, I think, recognized innovation is the effect of the internet of things on the way that we are now going to receive political communication and information. And by that, I'm particularly thinking of the digital assistants like Google Home, Siri, Amazon Echo, and the like. The way that people interact with those devices, which are going to become more ubiquitous in our environment, is very different than the way that they interact with a website or a phone or the like. And if you see the way— I see the way my kids talk to these machines and they ask it questions and they expect an answer, right? And it was all well and good from a disinformation perspective if what you're getting is a Facebook feed and then there's a particular piece of disinformation or propaganda that's in that. Or if you do a Google search and you get 10 blue links, right, then there's, you know, it's one out of whatever. But if you ask Google Home right now, Where are 3 to 5 million illegal votes cast in the last election? Right? It has to give you an answer, right? It's one answer. This is where the importance of Wikipedia actually comes in because now what's been happening, all these platforms are now defaulting to Wikipedia as being the answer to some of these incredible questions. But you can think about how this will happen, you know, in different aspects of our environment where we start expecting technology to give us an answer to a particular question. And so the stakes are much higher. As to what that answer is going to be.
John Markoff: Have you been pulled into Monica Lam's project called Almond on campus? No. Because she's got— she's trying to build a virtual assistant that she sees as a Switzerland of virtual assistants because she's worried about the creation of these proprietary stovepipes. And it's an interesting—
Nate Persily: Let me say just two other things here, particularly from the dystopian election administration perspective, which is that while Cambridge Analytica was in many ways a bunch of snake oil salesmen, that they really couldn't deliver on what— there are now international consulting efforts that are good at that. And whether it's buying bots and troll armies to help manipulate elections or doing the psychographic profiling, that now it really is becoming a formalized industry. And so we've seen it already where governments are and incumbents are hiring out these firms to help them with this, this kind of election manipulation.
John Markoff: So I wanted to come back to identity, which you brought up, and the problem of nobody knowing that you're a dog on the internet. Have you looked closely at the Facebook transparency data related to fake accounts? Yeah. The 2.2 billion created in the first quarter of 2019? I look at that number and I look at their assertion that they get all but 5%. Yeah. And I've actually talked to ask their executives about this, they actually believe those numbers.
Nate Persily: Do you believe those numbers? I do. I mean, I think part of this is— I, I almost feel like releasing that data is so misleading because it's, it's, it's sort of like, you know, trying to describe all of these dogs that are not barking, right? Because it is, um, they've got filters that prevent the creation of fake accounts, and if you do a lot of obvious things, you're going to get prevented. And so it's just like your spam folder, right? That if we didn't have spam detection on your Gmail or whatever, your email would be completely overrun and most of what you would get would be spam, right? And almost all of the, those fake accounts are like economically motivated spam stuff. And so, so great, their filters pick up billions of attempts to do it. So they're a huge company, so everybody's trying to go after them. The interesting thing about Facebook is that it's actually easier to detect fake accounts on Facebook than it is on certain other platforms, and here's why, which is that people, when they go onto Facebook, follow a kind of predictable path, right? If you're going to be a new person on Facebook, you'll probably friend your family first, right? And so you'll be a part of that network. Then you'll go back to your, like, oh, your high school friends who happen to be on Facebook, and oh, you'll friend them. And then, oh, college and people in your work, right? And that's something through AI that they, you know, can, can predict. If you're a Russian bot, it's harder to, to do that. Right. And so, so I do believe that now that's not to say that sophisticated, that there's still, you know, millions, tens of millions, maybe 100 million accounts, right. That are fake. Right. But this is the problem in the internet age is that there's 100 million pieces of anything, right? And so, yeah, and it's a problem, but it's less of a problem on Facebook than it is on Twitter, where let's say 10% of the US accounts are fake. But that's because they have an anonymity policy, and Facebook has a real names policy.
John Markoff: Last question, just how do you answer the question you posed, can democracy survive the internet?
Nate Persily: Not in its current form is my answer, which is to say that we need to start thinking broadly about new types of democratic institutions, and particularly as it relates to these multinational global information monopolies and how you bring accountability to those firms. And so I think Facebook has actually been moving in this direction with this attempt to create what's known as a content review board, something I'm also involved in, which is— and that we issued a paper from my students in the Policy Lab at Stanford This is what's been often termed the Facebook Supreme Court, right? Can you appeal from a decision that Facebook has to take down your communication? But we need to start thinking about those kind of alternative democratic institutions or accountable institutions of accountability that can check these big firms. And it's extremely difficult because of their multinational nature, right? Because it's not enough for the US to be trying to curb their excesses. But just to give you a sense of why I believe this, why we have to start thinking very creatively about how democracy works in an information ecosystem which is dominated by a few very large monopolistic firms. For the European elections, 2019 European elections that had been you know, just concluded in May. In January or so, Nick Clegg, who formerly was of the British Parliament but is now one of the vice presidents at Facebook, went to Brussels to announce what would be the rules for political advertising for the European elections. Right. And so just digest that for a moment, right, that you have an official from an American firm, right, going to Europe to announce what the rules of the game are going to be for their elections. And that's where we are. Right? And that is not a sustainable equilibrium. We need to figure out some way to have, whether it's international organizations or agreements, to then sort of democratize these companies and to open them up to some kind of accountability. How do you manage to do all that you do? It is a question I ask myself frequently, and one that my family asks me. I have— one of the things I've tried to do recently is, particularly when it comes to travel, is to combine— basically have constant periods of concentrated pain where I will then, you know, travel to several countries in a short period of time. There was— this then led me in the spring of 2019, in a 2-week period, to go to Mexico, Brazil, Bangalore, Delhi, Kenya, Dubai, London, and then back home. And then most recently, I spent the summer, I was away for much of the summer, mostly in Europe, which was nice, but with these different organizations. But the truth is, look, we're sort of striking while the iron is hot here. I mean, if you're someone who's working in the area of democracy and the internet, I mean, these issues are probably not going away anytime soon, but now this is, on the front burner in so many different domains. And what's important is the decisions that are being made right now are going to have such profound long-term consequences, decisions that are being made at the companies as well as by governments, right? Decisions that are being made about how to balance privacy and security, how to balance free expression on the one hand with safety on the other. They're so important, and we're going to look— I mean, I think everybody recognizes that certain design decisions that were made with these companies early on have helped produce these problems, and now we're making another set of these decisions that are going to have long-term consequences. Why did you decide to pursue this field of study? I had started as an expert in election law, uh, and the sort of issues of voting rights, election administration, campaign finance, redistricting, and political parties. I had, you know, I had done my first redistricting when I was a wee professor at University of Pennsylvania, and I was appointed by a court to do that because I had no political background. And so this was one of these areas where being a novice was actually an advantage because you didn't have a track record of working one or the other. Political parties. So I got involved early on in doing work in the intersection of law and politics. I should say I have a PhD in political science. Sort of my intellectual progenitor there was Robert Dahl, who was one of my professors at Yale, and then Nelson Polsby and Bruce Cain, who were at Berkeley. And then here at Stanford, Pam Karlan, who is a sort of prominent civil rights lawyer, constitutional litigator, and the like. And so they were, they were sort of my earlier inspirations to go into the field of the law of democracy or election law. So my entrance into the field of democracy and the internet, you know, happened lo and behold when I moved from Columbia to Stanford. But, but one of the things that I'd been thinking about at the time was how campaign finance and the way we think about campaign finance is completely outdated in the era of the internet. And so I wrote a piece called, um, uh, uh, uh, The Campaign Revolution Will Not Be Televised. And it was all about how the rise of internet communication, uh, will— should change the way we think about, uh, campaign finance. And it was focused on things like how the platform's rules and terms of service are in many ways more important than informal law when it comes to political spending, right? Um, I held here a, a conference that Bob Bauer and Ben Ginsburg, with whom I had worked on the Presidential Commission on Election Administration, which was, uh, a, um, uh, the, the commission that Obama put together to deal with long lines on Election Day. Well, we were all, had all been interested in campaign finance also. And so we, um, held a meeting here at Stanford with members of the Federal Election Commission. Republican and Democratic campaign folks and folks from the platforms to talk about the future of political communication online. And that was one of the first things I did here when I came to Stanford. That then led to a project I had done with Google, not getting paid for it or anything, but related to Google on trying to promote disclosure in political advertising, something that was then shelved but then suddenly gained more interest after the 2016 election, right? But that was something we'd worked on before. And so my entry point into these questions of democracy and the internet was from the campaign finance side and political advertising side. And that was the way— and that was still at a time when we had this utopian view of what the internet's potential was for democracy. Everyone was talking about these sort of Jedi masters in the Obama campaign who had figured out how to use microtargeting and raise money on Facebook and the like, or Howard Dean and Meetup and and all of that. Then of course, 2016 election took a decidedly darker turn in the way people started thinking about the potential of the internet. And so all of these questions of bots and trolls and disinformation and polarization and echo chambers and foreign interference then basically became the thing that we study when we talk about democracy in and the internet. And so since I'd been working on this in the context of the political campaigning online, it was the next chapter in that. I'd been sort of expecting to write something on the 2016 election and how technology was affecting the campaigns. But once you started looking at how technology was affecting the election, then that sort of opened up the can of worms that we've been trying to deal with since them.
John Markoff: Thank you for spending time with us.
Nate Persily: Thanks for having me.
John Markoff: To learn more about the topics in this episode, check out the show notes. There you'll find links to works by our guest and relevant articles. Thanks for listening.