At a Senate hearing on social media algorithms Tuesday, lawmakers on both sides of the aisle sat rapt as former Googler Tristan Harris explained why Facebook, Twitter and YouTube are "attention vampires" steering people into "different rabbit hole[s] of reality" in what he described as a "psychological deranging process."
A little purple, perhaps, but the committee ate it up. Because at a time when Democrats and Republicans can't agree on much, they share a common view on this: Social media platforms and their shadowy, unexplainable algorithms are driving Americans to extremes.
For Chris Bail, director of Duke University's Polarization Lab, that explanation is too easy and ignores all of the deeply human reasons why people behave the way they do online.
In his new book, "Breaking the Social Media Prism," Bail argues that the primary failing of social media is not the way it radicalizes or confines people to echo chambers, but the way it distorts political discourse as it actually exists, rewarding status-seeking extremists, muting moderate voices and giving us all the impression that the other side of the aisle is more intolerable, and more different, than they really are.
Through a series of experiments at the Lab, which he outlines in the book, Bail has studied what actually works to make people behave more civilly online — and just as importantly, what does not. Bail spoke with Protocol about those experiments, what people misunderstand about political polarization and how tech giants can and should evolve.
This interview has been edited and condensed from a longer virtual conversation hosted by the Mechanics' Institute in San Francisco.
What do you think are some of the biggest misconceptions people have about the way social media contributes to polarization?
A good place to start is the idea of the echo chamber. I think all of us have experienced how social media can allow us to surround ourselves with like-minded people and insulate ourselves from people who don't share views. People are concerned that algorithms may reinforce that tendency, creating filter bubbles.
Four years ago, right after the election of Trump and right after the Brexit referendum in the U.K., these were shocking developments to a lot of people. I thought the echo chamber just explained this. As someone who's liberal, whose social media feeds leans left, I didn't see that Trump was generating excitement. We had the idea in 2017 to try to break people out of their echo chambers and see what would happen. If we did that, would they become more moderate?
So let's talk about this experiment. You paid people to essentially follow bots that were on the opposite side of the political spectrum, after you had already gotten some sense of where their baseline was. And you wanted to study how following those bots for a month would influence them. So what were you expecting to find from that experiment?
One idea going into this was if you break people out of their echo chambers, they should be able to understand each other and empathize with each other. The key question, though, of course, is: Is social media the right place for that to happen? A rival idea would be that, actually, this might make things worse.
So when we went to do the experiment, we surveyed about 1,220 Republicans and Democrats. This was all on Twitter, all in 2017. And we asked them a bunch of questions about things like climate change, racial discrimination, government, regulation, the economy, all sorts of stuff. Then a week later, we invited half of them to follow these bots that we created. We told people: Hey, we're gonna pay you up to $26 if you can correctly answer questions about an automated Twitter account that we'd like you to follow. They didn't know that the bots were going to tweet about politics, they were only told that they were going to tweet 24 times a day.
At the end of the month, we sent them the same survey, and we looked at how their views changed. We really wanted to see that people would become more moderate, because that would dovetail with the story of the echo chamber, right? Unfortunately, nobody in our study became more moderate. And most people became either a little bit more polarized, or a lot more polarized, especially Republicans.
What do you think explains that phenomenon?
We want to think about social media as this place where we go to get information. When you take the pulse of social media for real for a minute, you realize that it's really not about that at all. Most people use it for this more basic human purpose, which is to really figure out our identities. Social scientists, for a long time, have known that every day each one of us puts on a different presentation of ourselves. Neuroscientists will call this the social brain hypothesis — that we're constantly surveying our environment for cues about what's working with other people.
Chris Bail is director of Duke's Polarization Lab. Photo: Alex Boerner
We did this long before social media. But I think the really interesting question is: How does social media change this all too human thing that we all do? I think it happens in two ways. The first is that we have unprecedented flexibility to present different versions of ourselves. And then we also have all these new tools to monitor what other people are thinking about us: things like follower accounts, "like" buttons. They distort the social landscape. This strange feedback loop can fuel status-seeking extremists, but also make moderates just totally disengage.
Those are two twin problems that I think really explain why an attempt to take someone out of their echo chamber doesn't make them more moderate. It just puts them into this identity war that we're all fighting constantly on social media. We're trying to figure out ways to make our side look good, and the other side look bad. Exposing yourself more to the other side only brings you further into the war.
One of the stories that you tell is about a woman named Patty, and you describe her as basically an unenthusiastic Democrat and you said that she went into the experiment with not very forceful viewpoints. But after the month following the bots, she became increasingly more defensive and more likely to intervene. The hypothesis you make is that the more that she saw her side attacked, even though she wasn't necessarily a strong defender of her side, the more she went on the attack. That made me wonder: Is it really a matter of exposing people to other viewpoints that doesn't work? Or is it a matter of exposing people to extreme viewpoints that doesn't work?
I think you put your finger on it. What we see in the course of this month, and in talking to her after she followed the bot that she was never focusing on the moderate posts of, like, a David Brooks, a center-right Republican [and columnist for The New York Times] or a Steve Bullock, the centrist governor of Montana. She was focusing on [Sen.] Ted Cruz and the most polarizing parts of the continuum. Extremists are a huge part of the reason why people backfire, because what we're being exposed to leads us to get a dosage of identity, rather than a dosage of information.
In your experiment, you have people follow these bots for a month. But people's viewpoints are based on their entire lifetime of experiences. Do you think that we can really write off the potential that exposing people to differing viewpoints on social media would have a moderating effect over time? What further research can be done on that?
You're absolutely right. People are deeply complex, and they're encountering this information from different perspectives, and different media channels. One of the things that I wanted to do with this book was to really tell the story of the online versus the offline. So to do that, and to really figure out why people are having this counterintuitive reaction, we did interviews with about 80 Republicans and Democrats. The book presents the stories of people as they're stepping outside their echo chambers, and tries to really fill in how what you see online is such a small part of the story.
You interviewed internet trolls. What did you learn from those conversations?
One story that was shocking from the research in this book was the story of this guy, Ray. He claims to be a centrist Republican, and he's very polite, even deferential. He goes out of his way to say all these people on the internet are always getting in fights, and they're probably just losers who live with their mother, right? And then we go to look at his Twitter data, and what we discovered is this guy was actually probably the biggest political troll I've ever seen on the internet in 10 years of studying political extremism online. The question is: Why did they do that? What's this Dr. Jekyll and Mr. Hyde thing that he's doing every night?
What we discovered is that the reason this guy Ray does this is because he's a social outcast in his real life. He's recently divorced. He actually lives with his mother. He was literally talking about himself in our interview with him. He's literally created two separate realities for himself. The micro-celebrity that he's been able to achieve from sharing this unspeakable stuff on Twitter is really actually profoundly important to him. He checks nightly: How many followers do I have? How many likes did I get? And they're really fulfilling a sense of purpose for him.
If so much of what people are seeking is the feedback, the likes and the engagement and the retweets, how much of an impact do you think it would have for all that stuff to disappear?
To clarify, I'm not suggesting that all of us have the potential to become this guy, Ray. In fact, he's the outlier. The far more common story is relatively moderate people who are just completely invisible on social media. So the other story that really stuck with me from this research, was the story of this woman I called Sarah. We begin all of our interviews by asking people to tell us about the last time they used social media. She says, I was up late the other night, and the NRA posted something about how it's Americans' right to own guns. She posted something like: My husband owns a gun and he's responsible gun owner, something fairly innocuous in the landscape of America's gun debate.
And then she says, within minutes, someone looked at her Twitter feed, saw that she had kids and posted: I hope your kids find your gun and shoot you. Unfortunately, this is the story that is all too common. Experiences like this made her completely disengage. Unlike this guy, Ray, who is getting all this feedback that is vital for his sense of identity. For Sarah, this is actually a liability.
This is the much more common story that we see among this huge, but largely invisible, more moderate majority. It's not that we're stuck in an echo chamber and seeing the same type of people. It's that the people we're seeing are very different online and off. And that has terrible consequences for the rest of us and makes us all feel more polarized than I think we really are.
Instagram has begun hiding likes — is there any evidence it's working?
In the case of Instagram, one of the key problems is there's a lot of research happening inside social media companies that just never gets shared publicly. But we can look at some traces. A recent study, for example, looked at web tracking data, where someone agrees to share with researchers every single digital footprint that they make. And in an interesting study, some political scientists basically tracked how often people progress from a moderate view to a more extreme view. What they discovered was actually pretty surprising.
If you take the pulse of the public debate, you'd say, it's the algorithms that are the problem. Companies are profiting. This is a really neat, tidy explanation. But when we look at the data, this only seems to happen to probably well under 2% of people, and maybe as few as one in 100,000 people.
Now, these are preliminary studies, we need a lot more research, I think, and above all, we need more transparency from the platforms to really get to the bottom of it. But thinking about the human drivers of polarization is something that we haven't done enough of. I'm worried that if we focus too much on these popular ideas that have very little evidence that we might wind up not really moving the needle very much.
Sens. Ben Sasse (left) and Chris Coons speaking at Senate hearing on algorithms and social media.Photo: Al Drago/Getty Images
In the book, you also write about a social media platform you built at the Polarization Lab to test whether there was a way to design a new social network that would encourage civility. The main difference is it's anonymous. For me, that was shocking — I've seen what happens on Reddit. Why did anonymity work for this platform, which you called DiscussIt?
I was shocked too. The theory was that, OK, on the one hand, anonymity allows us to escape social norms, and maybe we'll be uncivil to each other because there's no consequences. On the other hand, escaping identity and public pressure to conform to our side could in theory allow us to entertain ideas that we wouldn't really feel comfortable talking about on a public forum like Twitter.
So we asked people a bunch of questions, all sorts of stuff about their political views. And then we asked a bunch of questions about two very controversial topics, immigration and gun control. After that survey, we waited a little while, and we invited half of them to test this new social media platform called DiscussIt. They were told they were going to chat anonymously with someone else. What they didn't know is that the invite code that we gave them to log on to our platform was pairing them with a member of the other party to talk about either immigration or gun control. We didn't tell them, "Please be nice."
What we found — which was really, really surprising — is that people who used the anonymous chat app to talk about either gun control or immigration depolarized much more than people who didn't. That effect was even stronger for Republicans.
I struggle to see how that can be applied to existing large platforms. You add anonymity and you lose the other parts of social media that are already not toxic, like sharing baby and dog pictures. Can you really overlay this on any of the platforms that we are living with today?
I would never say, "Let's make Facebook anonymous," for example. I think what we're seeing in social media is a splintering of all kinds of social media. Why should we have our political discussions in the same place that we have our cute kid pictures and cat videos? Maybe what we need is to recognize that actually, most people don't want to talk about politics. Something like less than 6% of Facebook posts are about politics. We shouldn't expect these platforms that are really designed for sharing cute cat pictures to produce rational debate.
Instead, we need to create platforms where you're rewarded for producing content that lots of different types of people appreciate. I'm imagining something more like Stack Overflow, which is a site where software coders can answer each other's problems, and when you answer someone's problem correctly, you gain a status. What I think we need is a platform for people who can and will engage in productive conversations about politics.
Does it really do us any good as a society trying to become less polarized, if all the people who are open-minded about having discussions with people on the other side just go over there to have their conversations, and we keep battling each other in the main arena?
I think if we move the conversation to another arena, the other platforms are going to be a much less entertaining place for trolls to play. You can think of it as quarantining some of the extremism. I'm okay with the idea of quarantining as long as it eventually allows the people who are willing to have productive dialogues to have a place to play where trolls aren't incentivized to ruin it all.
I think one of the most surprising parts of the book, any book about polarization written in the year 2021, is that you write that actually, a lot of the polarization we see is false polarization and we're not as polarized as we think we are when looking at social media. Rates of partisanship are actually pretty stable.
When I read that I thought: Is that really what we mean, when we talk about polarization? We're not really talking about our split of partisanship. We're talking about our antipathy toward the other party. And it seems like by a lot of measures, that is growing. So let's end with that: Is polarization getting worse or not?
We make a distinction between issue-based polarization and affective polarization: what you think of the other side, independent of their ideas. And in the journal Science, two months ago, I published a piece with a lot of other social scientists where we identify that for the first time ever, out-party animus has replaced in-group love. So you're absolutely right to say that this is the trend, and the one that concerns a lot of us.
Now, the important question is: Is that because of social media or not? Unfortunately, we know that that trend began before large scale social media usage. So, tempting as it may be to say that social media explains it all the way, I don't think that's the case. I think the question we need to ask is not does social media polarize us or not, but how could social media be reconfigured to help counter polarization?
