 Google wants to hack our election. Google could well elect the next president. They have a monopoly on information on what you know and what I know. Fight against this threat to our democracy. As the 2020 presidential election heats up, alarmism over big tech's influence on voters is once again reaching a fever pitch. Democrats fear the impact of so-called fake news and Russian bots, while Republicans complain of supposed anti-conservative bias and censorship on platforms like Google, Twitter, and Facebook. They all agree on one thing. The federal government needs more power over online speech. We will hold social media platforms accountable. Russia was running all those bots about me. They treat conservatives and Republicans very unfairly. It's not right. It's not fair. It may not be legal. Republicans have put a spotlight on the claims of one noted researcher and self-proclaimed proud Clinton supporter to bolster their allegations of liberal activism by tech companies. Psychologist Robert Epstein. Dr. Robert Epstein, a Harvard-trained psychologist. Dr. Epstein, I found your testimony incredibly powerful and incredibly concerning. Dr. Robert Epstein produced a series of studies that he claims show that search engines and social media companies manipulate their users to help Democrats. A former editor-in-chief of Psychology Today, Epstein has called Google's search engine the most powerful mind control device ever devised. In 2016, Google's search algorithm likely impacted undecided voters in a way that shifted at least 2.6 million votes to Hillary Clinton, whom I supported. If any headline comes out of this hearing, that should be it. While Google has called his work nothing more than a poorly constructed conspiracy theory, Donald Trump cited Epstein's research to explain why he lost the popular vote by nearly 3 million votes, prompting Hillary Clinton to call the work debunked. Last year, Epstein was invited to testify before Congress by Texas Senator Ted Cruz to warn of the existential threat that big tech poses to democracy. Democracy, as originally conceived, cannot survive big tech as currently empowered. In 2020, you can bet that all of these companies are going to go all out. They will have the power to shift 15 million votes. He's been sounding the alarm about what he alleges could be even more pervasive interference in favor of Democrats in the current election. You have hard evidence that Google is sensory and you can back that up. Oh, absolutely. So I explained this. I told this to Congress. The study that I referred to was irrefutable and no one has been able to disprove him. But a careful review of Epstein's work reveals that though he has conducted scientifically sound experiments, his findings don't fully support the claims that he's been making about the influence of tech giants on real-world elections. Election by election, I can actually calculate how many votes these companies can shift. A big part of that is what you call the search engine manipulation effect or SIEM. So can you just briefly talk about what that is and how you arrived at your conclusions? 50% of all clicks go to the top two search items. 95% of all clicks go on that first page of search results. So I asked a simple question, which is, if people trust those high-ranking search results that much, could I use them in an experiment to shift people's opinions about candidates perhaps, maybe even shift how they would vote? I predicted I could probably shift people one way or the other by 2% or 3%. That was my guess. The first experiment I ran, I got a shift of 48%, which I didn't believe. I redid the experiment with a new group. I got a shift of 63%, which is enormous. It's one of the largest behavioral effects I think ever discovered. In Epstein's initial lab-based and online experiments, participants spent up to 15 minutes researching details on a foreign election they knew nothing about. They used a fake search engine similar to Google that reordered results to favor a particular candidate. Then Epstein asked them how much their opinions of the candidate had shifted. He called the effect he was measuring the search engine manipulation effect, or SEEM. Bias search results can easily produce shifts in the opinions and voting preferences of undecided voters by up to 80% in some demographic groups. I think when a lot of people hear things like you can shift 80% of people to a different candidate, that seems on its face kind of ridiculous, right? So you observe this in the laboratory and then you're saying that we'll see that same shift in the real world. And what gives you confidence that that's the case? I want to be very, very clear on one issue here. I'm my biggest skeptic. So when I get a huge number in an experiment showing a lot of people's opinions just shifted, believe me, I'm very, very, very skeptical. That's why I repeat experiments over and over and over again. I do them in different ways. They are unbelievable numbers, but they're real. That's the problem. What I'm asking about is what applicability that really has to a real world election in which you have millions of other variables that you're not measuring, that could be confounding it, and you have people that are bombarded by all sorts of news sources that don't just rely on search ranking to decide their political preferences. In one of those experiments in India, those were actually real voters, right in the middle of the voting period because voting goes on for three weeks in India. These are people being bombarded, as you said, bombarded by all kinds of sources of influence, but we got an enormous shift with these real voters. We had, in some demographic groups, shifts of 60% or more. That's a very significant result as far as I'm concerned in how easy it is to change what people say, how it changes whether or not they vote. I think that's a totally different question. I asked Aaron Brown, a statistician and former chief risk manager at the hedge fund AQR Capital Management to take a look at Epstein's research. He says that in experiments of this sort, it's not hard to manipulate what participants say and that Epstein is over extrapolating from his results. When Epstein ran his online experiment in India in 2014 on real voters, he exposed them to biased search engine results for an average of five minutes and then asked them how much their opinions had shifted. Whether that translates to actual changes in the vote people ultimately cast, I think there's zero evidence for that. Brown says that participants in this type of behavioral research often answer in the way they think the experimenter wants them to and even if their opinions really do change, by the time they show up at an actual polling site, it's likely that all sorts of other factors have changed their opinions yet again. The people whose votes you're going to shift with something as trivial as five minutes on a biased search engine are going to be the people whose votes are easiest to shift. They're probably the people who aren't voting for one thing. So these people are going to shift back and forth a hundred times before election day if they ever get to the polls at all. And so when he says that he got 60% shifts among some demographics in that study, do you think that that's a meaningful result that we can sort of map on to real elections? Not at all. We can't assume that that's the only thing changing people's votes. I mean, that's an absurd assumption. Their algorithm all by itself, it is already determining the outcomes of close elections around the world all by itself. Epstein has used the findings from his search engine manipulation effect studies to make bold claims about real election results. For example, he asserts that as of 2015, bias in Google search alone has been changing upwards of 25% of the outcomes of national elections worldwide. Can you just explain how you arrived at those numbers from the experiments that you just talked about? That's actually pretty simple. We simply looked at the margins by which national elections around the world were being won. And it turns out in national elections especially, the wind margin is often extremely, extremely small. And then we looked at, again, the SEEM research which tells us what proportion of undecided voters can be shifted in any given election and just kind of put the two together. And that's where we came up with this estimate of upwards of 25% of the national elections in the world. As of 2015, we're being decided by a Google search engine. The logic he just described is just wrong. There's lots of other things that influence people's votes. How many people changed their vote because of what their spouse thinks, what their friend thinks? An ad they saw, a book they read, a movie they saw, an experience they had. What you would have to show to show that they were actually influencing elections is you need a controlled experiment. You need actually to see how many actual votes shifted, not how many people reported a shift in opinion. As evidence that tech companies can affect voting behavior, Epstein points to a study run on 61 million Facebook users on election day in 2010 that tested the impact of get-out-the-vote messages. They actually measured how many more people got off their sofas and went and voted. So they sent it to 60 million people that caused an additional 340,000 more people to go vote. But Brown says this fits Epstein's pattern of emphasizing seemingly large numbers without giving their proper context. The Facebook study found that only one out of five people who said they voted after seeing the get-out-the-vote message actually did and the most effective message tested increased real voting by just 0.4%. It makes Epstein's higher estimates of his effect seem implausible. So of whatever percentage, you know, 15% or something who said they changed their mind in this experiment, maybe only 3% of them voted and maybe, you know, only 0.3% of them actually changed their vote on the basis of this. We just have no idea. But it's going to be much smaller. It's going to be much smaller than the effect you observed. That's a very far cry from the numbers that you're talking about, right? So why do you think that when we look at this 60 million person study, that's happening? Is it the actual turn out-to-vote that's so much different? Or what's your take on that? Actually, what that means is that if on Election Day in 2016, Mark Zuckerberg had decided to send out a go-vote reminder just to Democrats, number one, no one would have known and that would have given two Democrats that day at least 450,000 additional votes. If you can shift 15 million votes, you can, I mean, you can create a 20% win margin. In response to Brown's criticisms, Epstein reiterated in an email to Reason that his estimate that tech companies could shift 15 million votes this year is quote, conservative based on what I know we can do in our experiments. Though his studies measured shifts in opinion and not voting behavior directly, Epstein says that research on polling suggests that people's stated intentions are excellent predictors of their votes. Because Facebook had a way of at least estimating who actually voted, it's much more stronger result is when we can have a lot more phases. It's not a theoretical thing based on a sort of chain of logic. After his Seam experiments, Epstein had to answer one more critical question before he could arrive at the number of real votes Google was changing. Is the company actually biasing the information it shows to users? The Google whistleblower has come out of the shadows. Epstein says the answer is yes, pointing to leaked documents published by the conservative activist group Project Veritas and interviews with ex-employees that alleged Google sometimes does intervene to re-rank search results or remove content it deems objectionable. We're pushing down the fake news. We're demoting it and we are increasing the authoritative news and promoting it. I felt that our entire election system was going to be compromised forever. These companies are absolutely determined to control the outcome of the upcoming election. The term re-ranking is used in the documents. They use Blacklist to either remove content or demote it. There's a lot of material on what they call within Google, algorithmic fairness. Algorithmic fairness. So simple example would be most CEOs in the U.S. are men, back white men. Now if you type CEO into Google and you're looking at images, you're going to get a mix. You're going to get some more minorities, some more women thrown in. Whether that's representative or not, you see they have ways of balancing from their perspective, balancing content any way they want to balance it just like I do in my experiments. Tech platforms have often been inconsistent in explaining how they apply their community guidelines to remove or limit certain content. But experts in online search argue the technology is necessarily complex and always evolving and several studies have concluded that there's no clear evidence Google biases results for political purposes. Company executives also deny ever doing so. A Dr. Epstein study we have investigated. We don't agree with the methodology. Epstein needed proof they were doing just that. So in 2016, he ran a study in which he secretly monitored the search results of 95 field agents in 24 states after they typed neutral terms such as Hillary's health plan into Google, Bing and Yahoo. We had the searches so we could see whether they were biased in any way. We found a pro Hillary Clinton bias and I was a very strong Hillary Clinton supporter in all 10 search positions on the first page of Google search results but not on Bing or Yahoo. That's extremely important. The fact that we're seeing that bias on Google but not the other search engines that says that again for whatever reason their algorithms are actually favoring Hillary Clinton. After measuring the bias, Epstein combined these results with the shifts in opinion he was able to generate in his earlier seem research to calculate how many votes Google had supposedly shifted in real national elections. If there is any kind of favoritism in search results we know from the seem research without any doubt that that will in a big election shift millions of votes. In the weeks leading up to the 2018 election bias in Google search results may have shifted upwards of 78.2 million votes spread across many races. What we're seeing here is actual rigging. I mean, this is not paranoia. This is fact. This is rock solid fact. But Epstein wasn't able to determine the source of the bias in his field agent's search results. And his monitoring work has been criticized by other social scientists in part for using a small number of participants who are unlikely to be representative of the U.S. electorate. It's shifting the people who are vulnerable who are undecided and the shifts are simply enormous. And though his claims focus on undecided voters his data on the bias they're exposed to is especially limited of his 95 field agents in 2016 only 21 identified themselves as undecided and he didn't determine the political preferences of others. Isn't that a confounding variable if you don't know what the political preferences of people are? Because like you're saying one of the big differences between your study and what Google actually does is that Google has basically a dossier on every person that has all of these behaviors and they customize search results and things to the individual. So isn't it possible that the people in your group they were seeing bias search based on their own pre-existing political preferences? We would have ways of detecting that for example if that if that were true then we'd expect to see you know a pro-democrat results in blue states and pro-republican results in red states. If we're seeing pro-democrat results in both blue and red states but so you had 95 people right so you couldn't have had more than like two or three people right in any given state? Again I need to emphasize that these systems I've set up so far are proof of concept projects. In 2020 I think there's more at stake here and I think that we want to we want to have numbers that aren't just statistically valid. We want to have numbers that are psychologically valid and that means having a much, much larger number of field agents. What we need are data literally millions of pieces of data coming in every day from a very diverse group of people whose demographics we know. Why are you comfortable giving these number estimates now? Why not just say this is a significant effect? We don't know what effect it will have in the real world. We think it'll shift it this way and we're going to wait to see why do you give these numbers where you're extrapolating directly from your data to the real world? I'm very careful when I make extrapolations I'm very careful to word things in a way that I think has integrity. So what I said in 2016 was that if the level of bias that we found in the sample we collected if that level of bias had been present nationwide that would have shifted somewhere between 2.6 and 10.4 million votes to Hillary Clinton. And you know, I stand by that. I've never said I know that level of bias was present nationwide. I've never said that. You testified before this committee that Google's manipulation of votes gave at least 2.6 million additional votes to Hillary Clinton in the year 2016. Is that correct? I must correct you. The 2.6 million is a rock bottom minimum. The range is between 2.6 and 10.4 million depending on how aggressively they used the techniques that I've been studying now for six and a half years. Again, I'm very careful in the way I word things because I'm very, very skeptical about my own work. I think that's why my work is pretty solid. His results are pretty solid and they're very scientific. Yeah, he overstates them but mostly what he does is present them artfully. Aaron Brown says the issue is not Epstein's experimental design but the fact that he continues to repeat numbers that would only be accurate if the effects he's measured were the sole factors influencing voters. And in the context of all the things that influence votes he has not convinced me that this is a significant one. But Epstein writes that Brown has no idea how powerful new forms of influence can be especially influence that people can't see. Seem is an especially dangerous form of influence because it is in effect subliminal. The political campaigns I mean they're they're doing all kinds of shenanigans on these platforms and they're buying ads and you know so doesn't that make a difference? And the answer is no, that makes no difference. All of that is competitive which is healthy. That's a good thing. But if the platform itself favors one candidate or party you cannot counteract that by any means. Unlike other forms of attempted online voter manipulation like political ads, biased news stories and even Russian troll farms Epstein argues that tech companies influence is more insidious since users assume computer generated content is neutral and they can be manipulated repeatedly over long periods of time without knowing it. Google can take a 50-50 split among undecided voters and change it into a 90-10 split with no one knowing that they have been manipulated. I mean my jaw is open. But Brown says the world is full of implicit messaging and unlike in Epstein's research studies have shown directly that other types of unconscious influence can have a large effect on real voting behavior. You know the top person on the ballot gets somewhere between 1 and 10% more votes and they would get if you reverse the ballot or there's plenty of subliminal and you know implicit message that people get all the time but you don't need an international conspiracy or a secret group of technocrats in Silicon Valley to influence elections. If anyone draws news out of this hearing I would encourage you to review very carefully Dr. Epstein's testimony. Epstein's work is once again gaining traction in this election cycle as politicians seek greater control over activity online. We as a country cannot tolerate political censorship blacklisting and rigged search results. My executive order calls for new regulations under section 230 of the Communications Decency Act. President Trump recently signed an executive order targeting social media and tasked the FCC with investigating tech platforms for violations of section 230 otherwise known as the Internet's First Amendment. I think this hearing has underscored the need for Congress to revisit that. Nobody wants to see a federal speech police deciding what is and isn't permissible. The goal here, Kim, is not to get less speech. The goal is to get more speech. Republicans in Congress have proposed legislation that would largely eliminate the speech protections in section 230 allowing the government to punish tech companies if officials were to deem a user's post unacceptable and the Department of Justice is reportedly planning to file a new antitrust case against Google before the election. And the step on the line beyond that is a Teddy Roosevelt step. And it's not just Republicans agitating for control over online speech. They behave like monopolists. They need to be broken apart. Senate Democrats joined them in a bipartisan effort to subpoena tech CEOs in advance of the election and numerous prominent Democrats including Joe Biden and Kamala Harris have promised further crackdowns on Internet platforms for the content users post. Section 230 should be revoked immediately. Should be revoked. Number one for Zuckerberg and other platforms. That's a pretty foundational loves of the modern that's right. Internet. Exactly right. And it should be revoked. It should be revoked. You've also said that there needs to be strict regulation of companies like Google. What's your case for that? Yes. I used to call for strict regulation. I've stopped doing that because with Google there is actually a simple solution to the threats that that particular company poses and that is to make their index which is the database they use to generate search results to make the index into a public commons. So Google's dominance would disappear. They could still make lots of money. You know, big customers would have to pay maybe to access their index. But the point is that's a simple solution. That's very, very light touch regulation. You know, there are things we can do. The EU has been extremely aggressive. If the federal government does ever step in to regulate, you're not going to like it. Embattled Facebook CEO Mark Zuckerberg is calling for stricter regulation of the web, including his own company. The irony is that some big tech CEOs actually welcome regulation because they already have teams of attorneys and lobbyists at their disposal. But new legal requirements are costly barriers to entry for upstart competitors. I'm a big believer in the free market but we have to admit when the free market's not working and it hasn't worked here. And I think it's inevitable that there will be some level of regulation. So would you work with us in terms of what regulations you think are necessary in your industry? Absolutely. Okay. Would you submit to us some proposed regulations? Yes. We should be setting standards not unlike the Europeans are doing it relative to privacy. A study done one year after the EU passed sweeping regulations in 2018 in the name of protecting privacy found that Google and Facebook had in fact gained market share. It's a well-thought-out crafted piece of legislation. And then yes. You think the Europeans have it right? I think that they get things right. Recently we had Mark Zuckerberg testifying before Congress begging them to regulate him and the Google CEO doing the same exact thing which is what always happens with these large companies they can afford the compliance costs and it limits their upstart competition. So you know why are you confident that these regulations would not do what they've done with other telecommunications companies? Well I'm very much afraid that the regulatory approach will not work at all. Partly because the regulations actually get written by the companies typically you know breaking up Google breaking up Facebook they won't remove these threats. In fact they'll really just enrich the major stockholders. But you know there's a much larger issue here which is how could law and regulation possibly keep up with technology? That's impossible. Technology moves so rapidly that I just don't think laws and regulations are going to be able to protect people. So how do you protect people long term moving forward and the answer is monitoring systems because monitoring systems are tech. Tech can keep up with tech. I'm raising funds now to build a much more comprehensive system in 2020 one that will allow us to catch big tech in the act. This is going to cost more than 50 million dollars. Yeah. Epstein is convinced that the influence he's measuring is just the tip of the iceberg since tech companies have vast troves of user-specific data to mine and can use many other so-called ephemeral effects through search and video suggestions answer boxes news feeds and more to target and manipulate voters in the real world on an ongoing basis. You know I haven't tried this but what happens if we hit people with five effects at the same time heading into the 2020 vote Epstein says his only interest is in free and fair elections but his alarmist predictions have aided politicians opportunistic efforts to get Americans to hand them more power over the internet and to control the free exchange of ideas online. Any of our congress people could look at these numbers they would be stunned and they would take action. I guarantee it. Well I hope that video of this conversation goes everywhere on Facebook and Google. Despite the limitations of his research in measuring the power of tech companies to influence election outcomes or whether they're actually trying to do so Epstein continues to speculate on the number of votes they're shifting confident that his claims haven't gone nearly far enough. A lot of the social chaos that we see that's all around us is actually driven by the big tech companies. You know out of chaos comes order right. It's going to be their order that's coming. I think the problem is much bigger than I've been saying it is and I've been saying it's a big problem but I think it's actually much bigger. So what that tells me is that the numbers I've been getting over the years which are high their big numbers are all too small.