Plain English with Derek Thompson - A Psychologist Explains Four Reasons the Internet Feels So Broken

Episode Date: April 9, 2024

Jay Van Bavel is a professor of psychology and neural science at New York University. His lab has published papers on how the internet became a fun-house mirror of extreme political opinions, why the news media has a strong negativity bias, why certain emotions go viral online, why tribalism is inflamed by online activity, and how the internet can make us seem like the worst versions of ourselves. At the same time, Van Bavel emphasizes that many of the group psychology dynamics that can make social media seem like a dumpster fire are also core to what makes humankind such a special and ingenious species. We discuss the four dark laws of online engagement and the basics of group psychology. If you have questions, observations, or ideas for future episodes, email us at [e-mail]. Host: Derek Thompson Guest: Jay Van Bavel Producer: Devon Baroldi Learn more about your ad choices. Visit podcastchoices.com/adchoices

Nathan Hubbard, spring has sprung, the birds are chirping, and the pop girls are pop-girling. Oh, and you know what that means, Nora Prenziotti. Every single album is back. This spring is packed with new releases from some of the biggest pop stars in the world, including our girl Taylor Swift, and we'll be covering it all. We'll, of course, break down every angle on the tortured poets department, and we'll also cover new music from Beyonce, Dula Lipa, Maggie Rogers, Casey Musgraves, and Ariana Grande. It's Pop Girl Spring on every single album.

New episodes starting March 28th. On Spotify or wherever you get your podcasts. Here's a pop quiz. Imagine the typical social media user in America printed out all of the content she read in a given day as if on one continuous scroll. So it's like every day, all the posts and the TikToks and the tweets that she read or skimmed by

on her way to reading something else were all printed out and measured, like one long biblical papyrus. You took this daily scroll and you measured from the top of the first morning TikTok to the bottom of the last evening reel. Measured in feet, how long would that scroll of content be?

How long is the typical user's daily feed? The answer, according to one paper published by today's guest, is 300 feet. Every day, we casually scroll through and estimated 300 feet of newsfeed on social media. That's roughly the height of the Statue of Liberty. It's as tall as the tallest building in Washington, D.C. Our attention is our life.

What we pay attention to shapes our sense of who we are, what is true in the world. So what are the... 30 stories of content actually doing to our sense of identity and reality. That is the question at the heart of today's episode. And it's the question at the heart of Jay Van Bavel's research. Professor Van Bavel is a professor of psychology and neuroscience at New York University.

His lab studies the rules of engagement in digital spaces. Why do we look at, listen to, click on certain parts of that daily 300 feet of content and not on other parts. His conclusions are not all positive. In fact, most of them show that the rules of engagement on the internet lead directly to many of the features that we associate with its badness, negativity, extremism, tribalism. If the internet is slowly becoming a kind of hell for many people,

Van Bavel's team is a bit like Virgil, guiding us through the inferno, explaining to us how hell works. They've published research on why the news seems so negative, why Twitter feels so toxic, why Facebook is so full of emotional crap. And when you line up the work of Van Bavel, it accumulates to a kind of devil's playbook for getting attention on the web. A set of nauseous rules that explain why the internet is the way that it is. Why it seems to be a cesspool of negativity and extremism. why people I know who are wonderful and warm in the physical meat space are transformed in a monsters when they disappear behind a keyboard.

In today's episode, we share that devil's playbook, the four rules of online engagement and why they're driving all of us a little bit crazy. I'm Derek Thompson. This is plain English. Jay Van Bavel, welcome in the show. Thanks for having me, Derek. Jay, you and I are going to spend most of this episode talking

about your empirical research on how the internet and social media in particular holds a fun house mirror to the human condition and why I think the incentives and engagement structures of social media are truly driving many people slowly insane. But first, how would you describe your work? What's the big question that you and your lab are trying to answer? Yeah, I think the big question that's animated me for last 20 years in my research is about the nature of groups and identity and how they work. And I started originally looking at the brain, how our brain responds to being in a group or seeing members of other groups. And then we've been kind of our brains, which are very evolutionarily ancient, have been dropped in this modern technological environment with social media

and artificial intelligence. And now I'm fascinated by studying how that all plays out. How is that manipulated? How does that, how do we get triggered in various ways and so forth? So obviously, the brain is a pretty ancient piece of machinery. Group psychology has to be a relatively ancient concept. We've been living in groups for hundreds of thousands of years before the internet came along. Is there a general psych 101 principle of like, this is your brain on groups?

Like if I went to like group psychology 101, what would I learn on day one about how groups change my own identity and my sense of what is true and false in the world? Great question. So maybe I'll give you, a couple principles. The first is all it takes is a flip of a coin for us to feel part of a group and start to identify with it. And so we study this in the lab. It's called a minimal group where we

just literally flip coins and put people on the blue team or the red team. But if anybody has ever grown up playing like recreational sports, you go to the gym and you randomly get picked onto a side to play a sport for an hour or two. And that automatically triggers a different psychological mindset. And even people who were strangers a few minutes before are people we suddenly like more, pay more attention to, have an affinity for, and were able and willing to cooperate with them in ways we wouldn't have if they were strangers. And this happens to us also if you go to like a professional sporting event. You're giving a high five to someone sitting next to you who has like a Yankees jersey if you're a fan, even though you might never know them before, might never see them again

in your life. And so that's the psychology that all of us walk around in. The other part of it that a lot of people don't get, but it's part of the examples I give, is that it shifts from minute to minute. So again, I can create groups in my lab. You walk into a stadium with a jersey on and see everybody else wearing those colors. You have a different identity.

You turn on the TV, and it's election night, and you're cheering for your party to win. Your country goes to war. You're under attack, and even broader, more inclusive identity gets triggered. Or let's say you turn around a movie and you're watching an alien

invasion. In that case, it's just all of humanity against this alien invasion. And so we can get triggered to think of ourselves as different types of members of groups, or as members of different types of groups, depending on the situation we're in. It's almost as if the desire to belong to any group is pre-downloaded in us, and therefore, thinking outside of groups seems to be the really key challenge. If it is so easy for us to adopt a group psychology, then being online where it's all just one big mass of group psychology overlapping group psychology overlapping group psychology, the hardest thing to do is to try to quote unquote think for yourself when we so want to think as if we belong to one of these groups. So having spent several

hours reading your papers from the last few years this morning and yesterday and thinking about them as a collective, I came up with what I am calling for the moment the four bad laws of internet and social media engagement. Because I feel like you and your lab have been in researching how group psychology and dynamics and social media and the internet affect us. I've sort of found your way to these four laws of what it is that drives bad engagement on the internet.

And I'm just going to read them out now and then we're going to walk through them. I'll repeat them at the end and then we can analyze them. But here are the four laws that I've taken from your research. Number one is that negativity drives engagement. Number two, extremism drives engagement. engagement online. Number three, outgroup animosity drives engagement online. And number four,

moral emotional language drives engagement online. And for listeners at home, if you didn't understand what some of those terms meant, we will absolutely define them in just a few minutes. But what I want to do with you with our time together is to explain where these ideas come from, because they're all derived from your work in last few years and what they mean. I want to start with the paper you published last year on negativity bias in the news, in a randomized study of a 105,000 headlines and 370 million impressions from a data set of articles published by the online news dispensary, Upworthy. You found that so-called negative words increased the click-through rate by more than 2 percent, and the prevalence of positive words in a news headline decreased

the likelihood of any headline being clicked on. In your own words, what did this paper show that was so important? Yeah, I think one of the things that psychologists are very familiar with is the idea that bad is stronger than good, that people pay more attention to or are more motivated by bad things than we are good things. And I'll go back to a bit of an evolutionary explanation for this, again, is that, you know, you can imagine your ancestors walking around, you know, the African Savannah, and they're looking for food to eat. And so there's an approach, motivation, to try to find food. But if they potentially see something and it could kill them, that's dramatically more significant to their survival. And so they have to err on the side of avoiding things.

that are negative and risky. And our ancestors who did that over generation after generation after generation were more likely to survive. And so we have brains that are wired to detect threats and super attuned to them more than we are to detect rewards. And so that gets manipulated classically by the news. The famous quote from journalism is, if it bleeds, it leads. But of course, social media and internet websites have hijacked that as well. And And the data we had, I'll just say briefly, came from, you know, the classic website that engineered virality online. It was from Upworthy.

So we got access to the Upworthy data archives. And what Upworthy was famous for in the early days of the internet was doing A-B testing. So they would take a single news story and they would try one headline on their website and see how many people clicked and then randomly assign half of the other people who came to their website to see a different headline. And so that way they would test different headlines, look at their data, and then pick the one that was most popular and keep it up. And they were the pioneers in this.

And in fact, they were so viral at one point, they had more traffic than the New York Times. They were so viral that Facebook had to change its news algorithm because people were getting really frustrated by all the clickbait from us upper the headlines and leaving Facebook to go get it. And so they really engineered this through data. So when we got that data set, we thought, well, we can see not what their theory is about what works.

We can see with their experimental AB testing, what actually gets people to click through an article. And, of course, they're more likely to click to negative content, negative headlines, for the exact same story. I think a lot of people are going to think, Upworthy, isn't that the site that had headlines that were like, this baby panda learned to break dance, he won't believe what happened next? Yes, it is that same website.

But one of the really useful things, as you said about Upworthy, is that they, A, B, tested headlines for the same stories. You can test, you can control for the story, right? Like a story, when I wrote about this article last year, I said, a story about, say, Harry Stiles' breakup is automatically going to get more clicks than a story about a Vermont pension accounting change, like every single time. But if you can test on each story positive and negative headlines, you can control for the underlying substance and essentially say, controlling for that substance, are people more likely to click

on the positive headline or the negative headline? And you guys essentially found that negative headlines tended to get people's attention. I totally agree for something that's very deeply ingrained in us. We alight to five alarm fires in our lives, and we are very good at habituating to the non-fires in our lives. So we're looking at a news feed, we're scrolling through the newsfeed. When we see the five-alarm fire, that's where we stop and say, ooh, I need to click on this.

And the reason I think this is so important in terms of understanding the makeup of the news and the makeup of social media, is that one of my crusades, the journalist, totally self-serving, is to help people understand two things about negativity bias. The first is that negativity bias is, I believe, the most fundamental bias in news media. Some people say, the news is bias to the left, the news media is biased to the right, the news media is corporatist or the news media's anti-tech. Okay, some media sites might be all of those things or none of them.

The most fundamental bias in news is a bad news bias. And second, and I want to know if you disagree with me here, the bad news bias does not just come from us journalists. I think, as you said, negativity bias arises from an audience preference to click on things that they think represent a kind of danger, a kind of lack of safety. And that, I think, clicks into, no pun intended, this evolutionary instinct to pay closer attention to threat in their environment. Anything that I said in there that you think is unfair?

Yeah. I mean, when we see this happen with influencers on social media, where they're they produce more and more negative content because that's what gets engagement from their audience. And this is the same incentive structure that drives journalists, that drive major media outlets. And so they're going where the eyeballs are. And where the eyeballs are is in part what is appealing to our human psychology. And so that's actually like we can think of these, you know, the media is manipulating us. But partly what they're doing is giving us what we want.

And what we want is and how they measure that is what grabs our attention. And so everything is starting to be conditioned on what captures our attention. Upworthy was incredibly good at optimizing its content for virality. And you can sort of flip that coin and think that in a way, in order to tell the truth about the world, a truth that is not colored by this negativity bias, you have to embrace an editorial strategy that is suboptimal, right? You essentially have to say, I know that if I put the most negative treatment of the story as a headline, it's going to get more attention, it's going to be shared, it's going to create more

virality. If I create the more honest headline that is less negative, it'll be a better reflection of the truth, but it'll be suboptimal. But I think it's really important for people listening, I guess especially in news media, that the truth actually is often suboptimally phrased because the optimal phrasing of it is so negative. All right, the second rule that I want to move to, rule number one is negativity drives engagement. Rule number two is extremism drives engagement. And this is from your most recent working paper with Claire Robertson and Karina del Rosario. The name of that paper is inside the Fun House Mirror Factory. You claim that moderate opinions go missing on social media.

That is, for any given issue, abortion gun control, moderate positions that exist in the actual electorate are often missing in the online conversation that we believe, mistakenly, represents that electorate. To tell me more about this finding, why do you think the internet attracts more extreme positions than are reflected in reality? Yeah, so I guess first I'll just tell you a story. So I work, my office is right on Broadway, and I would like log into social media, and I would have this sense that like people were way more extreme than they were on the street. And I live in New York, you know, there's no shortage of loud extreme people.

But I would be like on Twitter and then I would like log off, go to get a coffee or grab lunch. and everybody just seemed way more chill. And then I also noticed this, even with my academic colleagues, these are professors, you know, you think of people are pretty reasonable or rational. And they would just be like fire breathers online. And often they weren't, the average professor that I would encounter at conferences was nowhere like that.

And I started to realize that the opinions that were being presented online on almost every issue, and then we're not just talking about politics here, we're talking about scientific debates and so forth, were very extreme and exaggerated. And I think part of it is from a few things I've learned. One is that very few people drive the conversation on social media or really anywhere online. So one of the statistics we have is that 97% of posts about politics on Twitter or X came from just 10% of the most active users.

And of those 10%, those tend to be people who are very ideologically extreme. And then I think we're starting to see that you see this across lots of domains. means. If you tune into like a debate about sports, you're going to see the hardcore fans are driving most of the conversation. They're commenting the most and they're commenting nonstop. Some of them are producing four or five, seven, ten posts a day. And other people who are like more, you know, on the fence or not hardcore fans, maybe tune in and check it out, but they're not commenting. So one of it is like who's driving the conversation. The other thing is the algorithms are conditioned on things like comments. So for example, the algorithm on Twitter, you

get more, you get bumped in other people's news feeds if you generate a lot of comments on your post. Well, what generates a lot of comments on your post is you taking a very extreme position because other people will share it who see it and are also extreme. And other people who don't like it will start commenting and criticizing you. And so those two things generate attention for the most extreme posts. The most nuanced posts just die. They never pop up in anybody's news feeds. And so what that does also is it provides a reinforcement loop. So if I post something really nuanced and complex, and I don't get very much attention, and the next day I post something on a similar topic, but have a really extreme take on it, a hot take. And I generate way more

get way more feedback, get way more comments, go more likes and shares and followers, then I'm more likely to post something like that in the future. And so, and then the third part of it is that I see other people doing it. It creates a norm that that's what people are how they're thinking. And then I'm more likely to kind of generate opinions that align up with the norm so I can fit in and be a good group member with whatever I'm talking about, whether it's politics or sports or work stuff. And so those are the dynamics that incentivize this extreme people or people with extreme views commenting all the time. I was going to ask you if the extremism bias that you found in your work is about the users or about the algorithm. And you're saying it's both and. It's both that hardcore users are more likely to comment and that the algorithm preferences the engagement that the hardcore comments have.

And I think one level further or one more dimension to that, the same users learn over time how to make best use of the algorithm. And if the algorithm is going to privilege engagement, and engagement is going to track extremism, then users might become more extreme over time. I want to ask a question that pushes again. That makes it sound like every single group and every single conversation in every single social media ecosystem is on a rocket ship toward extreme chaos nonsense no matter what.

surely there has to be, as there exists in every single culture, some opportunity for a break system, for backlash, for people to say, wait, wait, wait, wait, things have gotten too crazy here. Have you done, is there any work on how those systems of cultural breaks and backlashes to extremism work? Like, when is it too much for people? Yeah, I mean, that's not something that I have any data on yet, certainly here. But normally people have a latitude of what they're going to accept, and then they start pushing back on it. So people will accept more and more extreme things from members of their group. They don't accept it from the other group. And so what happens is either they try to clamp down or ramp down or, you know, constrain the extreme

perspectives of the other group by criticizing them or silencing them or, you know, we have like terms from cancel culture. And the research, this is from a colleague of mine down the hall at NYU, Ashwini, Ashikumar, who finds that people who are more extreme in terms of their identities are more censorious, that they want to censor the positions of outgroups that disagree with them. And so that's a little bit of how it happens. The other way that it can happen is creating different systems. So I'll give you an example of a couple systems that seem to have better ways of uprating quality content.

My favorite is when I go to the New York Times and I read the comments, you can sort to the ones that are reader favorites that have been upvoted by a lot of other readers. And often I'll find those are incredibly nuanced and thoughtful, sometimes way more thoughtful and nuanced than the article. The same logic works on some platforms like Reddit, where you could upvote quality content. The best example of this, I think, online is Wikipedia, where they have actually like a complex moderator system.

And Wikipedia has created this, like, incredible repository of knowledge from this massive volunteer army, and they do it through, like, moderation rules. And so having healthy, having people who can moderate, based on like certain criteria, say truth value and accuracy and making sure it's referenced, which is what happens on Wikipedia. You can have, again, online on the wild west of the internet, the exact opposite type of outcome, which is just something that's an incredible resource for free for humanity. Well, you said the key word, which is moderators. The New York

Times has moderators that are likely to appeal to readers of the New York Times. If you're a reader of Fox News, you probably aren't going to like the most favoriteed reader comments than New York Times, but that goes to taste. The New York Times understands its own audience. The best subreddits have the best moderators. Wikipedia famously has excellent moderators. You could say that the quality of any discourse rises to the level of its moderation, but also maybe falls to the level of its moderation. I want to give a few more specific examples here with extremism, because you talk in your paper about how the prevalence of extremism on the internet drives, both pluralistic ignorance and false polarization.

Now, I think I know what pluralistic ignorance is. I wrote an article about it a few years ago. That's when everyone pretends to think X but actually thinks Y. So the Hans Christian Anderson story, the Emperor's New Clothes is sort of the canonical example of pluralistic ignorance. Everyone pretends to think the Emperor, the naked emperor is wearing clothes, but he is not.

In politics, you know, let's say I go online. I see a lot of people saying on the left, maybe abolish the police, on the right, maybe stop the steel. And if you ask me privately, at dinner, do I think we should abolish all of the police, or do I think Donald Trump actually won the 2020 election? I might whisper to you. No, of course not. But I'll publicly falsify my beliefs in order to avoid censure. And if you multiply that effect by 100,000, you get the phenomenon

where everyone at a rally is screaming, stop the steel, because they think everyone else believes it, even though they're all privately ambivalent about it. I'm not predicting that that's the reality in Trump world. I've no idea what the actual reality is. But I'm saying that would be the pluralistic ignorance explanation of the Trumpist Stop the Steel phenomenon. So that's pluralistic ignorance. Can you give me an example of extremism creating false polarization? Like, pick any issue you want, but give me a concrete example of how extremism on the internet would create false polarization. Okay, so where we're finding a lot of evidence of false polarization in the research literature right now is that the average Republican and the average Democrat in the United States

thinks the other group hates them way more than they do. So there is real polarization. The average Democrat does not like the average Republican. The average Republican does not like the average Democrat, but they walk around with kind of a caricature of the other average member of the other party in their mind that that party really despises them and wants them gone. And so once you reveal to people, there's been several studies that have done this, once you reveal to people that the average person,

other group isn't that different from you and doesn't dislike you as much as you think. People become way more open-minded and willing to interact with those individuals. Part of it is because on social media, we find the most extreme members of a group that we don't like, and then we hold them up, screen capture them,

quote, tweet them, post how absurd they are, and we constantly hammer home that this is what this group thinks. But of course, if you take the most extreme example, which of course, as I've already pointed out, are the people posting the most, you hold them up as if they're representative of that entire group that is going to misrepresent that group. And it creates an illusion that there's more of a gap between the groups in terms of what they believe on many issues, but also how they feel about each other. And so that's why you have, and that seems to be correlated to how much time you spend on social media. So it seems like certainly the news, you know, I think many news sources, Fox News used to have this section called Campus Crazies, where they would find something absurd happening on a campus and then presented on like the evening news. And of course, it would often be something that I'd never seen. I've spent my whole life on university campuses, never seen anything remotely like that. But they hold it up as if this is like taking over campuses.

And so that's an example, I think, of how the mainstream media would do this. I think there's also been some of this in the left where they go to some rural small town diner to look for Trump supporters and they might quote and edit together some piece that makes them look absurd, but not including the quotes and the conversations with people

who seem more nuanced or reasonable. And so that creates also an illusion in that and left-wing media. And so this is something that we do all the time in the mainstream media, but I think it really is on overdrive, on social media, again, because the most extreme people post the most,

it's a great way to generate attention and content for yourself, can make you look like a virtuous person by constantly dunking on the craziest person from the other group. And then it's incentivized, it seems to be, by algorithms because it generates a lot of engagement. Is it fair to say that pluralistic ignorance deludes us about our own side and false polarization deludes us about the other side? Yeah, that's reasonable. Although there's some evidence that say like the average Democrat thinks that the average Democrat hates Republicans more than they actually do, too. Right.

So we're a little bit diluted about our own. Yeah, but you might say that falls under false, floristic ignorance. Yeah, I was saying, you know, we have some listeners who are more included into politics, some other listeners who are more clued into sports. And in sports, pluralistic ignorance would be everybody pretending to hate the coach online and posting terrible things about the coach, but privately everyone thinks the coach is just fine. And false polarization is like everybody in the Boston subreddit screams at Philadelphia and every in Philadelphia screams about the Boston subreddit.

And each side thinks each other like hates them, but actually, actually, like, they'd be totally fine going to the bar with someone from, like, the 76ers, someone from the Celtics. They'd be totally fine having a drink with each other. So that's, that's more false polarization. Is that a quick off-the-cuff sports analogy, but is that fair to say? I'd say that works. Okay, excellent. Let's move on number three. And number three is the principle. We've all, we've done negativity, drives engagement, extremism, drives engagement, and we've checked the boxes of pluralistic ignorance and false polarization. Rule number three is outgroup animosity drives engagement.

Here you looked at several million posts from Facebook and Twitter and found that, quote, each individual term referring to the political outgroup increased the odds of a social media post being shared by 67 percent precisely, end quote. Help me understand what outgroup animosity is. Yeah, in this study, this was led by Steve Rathgi. We looked at news stories and what was getting shared not only on Twitter, but also on Facebook, which is actually the biggest social media platform in the world. And we looked at what political elites, like political leaders, you know, members of Congress, the Senate were sharing and what drove engagement for them as well as media websites. And what we found is when people shared a story that shared something really negative about the other

party, that's what got engagement. So imagine Trump sharing a news story that made Joe Biden look really bad or Joe Biden sharing a new story about something that make Trump look really bad. Those are the things that are going to drive by far the most engagement for them. In fact, the outgroup animosity effect was bigger than any of the other effects we've talked about today. That seemed to be like rocket fuel, especially for political leaders. And so it really incentivizes them to share a lot of content like that because that's what really generates a lot of engagement for them.

The first principle you shared is that people so want to belong to teams that they are willing to develop this really, really life or death sense of team-based identity, even when something as frivolous as a flip of a coin determines what color jersey they're going to wear. It seems to me like the flip side of that is, well, if we really deep down want to belong to a team, then coinciding with that principle is that deep down we want to root against some opponent. And so it seems like on social media, where these feedback loops are so, so tough. A lot of people have figured out that the best way to drive easy engagement is to unite your own team by identifying the outgroup. Is that the idea?

Yeah. And I'll use a sports analogy. Like, there's a lot of Yankees fans, but there's probably just as many people who hate the Yankees, you know, a big part of, I used, I grew up in Canada. My favorite hockey team was the Emmington Oilers. And I remember it was common in that part of the world to see license plates and stickers that said, my favorite teams are the Eminton Oilers or who's ever playing the Calgary Flames, which is like their arch nemesis. And so that's a huge part of what drives us or when we have a nemesis in any context that we're excited to see them lose or something bad happened to them.

I've been other studies where the motion that people experience when an out-group member fails or suffers or is in pain is what's called schadenfreude. It's a great German word that means I take pleasure at the suffering of another group or another person. And we see that all the in study after study when these outgroups do engage in something bad or suffering or get caught with some corruption or fail at something. That drives a lot of engagement. In fact, as I said, that was the single best predictor of engagement we found online. I'm going to be honest, back when I had a really strong rooting interest for football, I was a huge Peyton Manning fan and obviously his nemesis was Tom Brady. I don't think that my happiest moment as a football

fan was watching Peyton Manning win his first or second Super Bowl. I think it was watching Eli Manning beat Tom Brady in the Super Bowl, especially during the 18 and O season. And I'm sorry to all the all the Boston fans out there. He is the goat, but it was an incredibly satisfying moment for me. And there was a way in which the purity of Schadenfreude is more delicious even than the taste of pure victory. I don't know why that would be true evolutionarily. You, you might have a hunter-gatherer story off the cuff that you can share to defend it. But in many categories, even outside of politics, there is something that's just so cleanly delicious about Chaudenfroyd.

And I'm not sure why. Yeah, I mean, I'm not sure why how far that one goes back, but it's something that we see. Like, actually, there's been studies, this was by a collaborative of my Mina Chikara. She put people an MRI scanner. And these were hardcore Yankees and Red Sox fans and showed them like little clips of bad things and good things happening to the Yankees and Red Sox. And she found that they are a reward center of their brain lights up when something bad happens to the other side.

So there's something very deep-rooted about this kind of pleasure center that gets triggered by this. Rule number four is that moral emotional language drives engagement. In a study of 563,312 tweets, you found that moral emotional language is the most shared. Before we go into this one, obviously we have to define the key term here. What is moral, emotional language? And how would you distinguish that from, let's say, non-moral, non-emotional language? Yeah, great question. So we started

by generating these dictionaries. We pulled from other studies. There were some dictionaries of emotional words. So let's say, like, sadness isn't a classic emotional word or happiness or joy. And then there are moral words. And they might be things like abstract concepts, like justice and being right or wrong. And then there's words that kind of like, imagine a Venn diagram of those emotional words and those moral words. And there's some words that kind of sit in that middle of that Venn diagram, like contempt or hate or disgusted. About outrage?

Yeah. Yeah. So they're often associated with words like outrage. They can also be words like love or something like that that has a moral connotation that's good. And I'll give an example. I'm sorry to interrupt there. Is this related to Jonah Berger's theory of high arousal language?

So Jonah Berger had, I think, studied the New York Times most emailed list. And his theory was that high arousal headlines were more likely to be shared than low arousal headlines. So like sad is low arousal, outrage is high arousal. Like happy is low arousal. Ecstatic or awe would be high arousal. So is there any relationship between your own? Yeah, there's some relationship that the moral emotional words also tend to be higher arousal.

We have run some studies where we try to control for that. but for the most part, I think in the real world, they're pretty overlapping. And so what we found in many, many studies now, and other labs have now around the world have found something similar, is that if you use a moral, emotional word in a message online, like I'll go to like a post on social media, it's about 10 to 20% more likely to be shared. And so you can take a message and you can load it up with those words

and put five or six or seven in a message, and all of a sudden it's like twice as likely to get shared. And so we found that that was really only happening for the moral emotional words online when people are talking about political topics like climate change or gun control. And for the most part, those are often moral outrage words that are negative, you know, in the topics I just mentioned. But we also had a data set where we looked at right after the Supreme Court ruled in favor of same-sex marriage in the United States.

And the hashtag love wins was one of the ones that was going viral. And so a lot of the messages at that time were positive. People were exciting and celebrating. And so it can be positive words or negative words. But once you put those moral emotional words in a message, it changes it. And we've now run experiments where we take the exact same message in the lab and just change one word to make it a moral emotion or not. And we find the same thing that even in very carefully controlled experiments, it's about a 10 to 15% increase. The other thing that happens is we say we have participants see these messages for social media and we just change that one word.

And when people see that type of language and a message, they see that person as a more hardcore partisan, so more identified with their group. That's the signal that it sends out to the world. And so if you're on their team, if you are part of the same party, you're more likely to want to contact them. If you're not on their team, it sends a signal that they're close-minded and people report that they don't want to engage with them. And in fact, when we looked at those 500,000 words that you met our posts, across all those topics we originally studied. Moral emotional language was making messages go viral. But when we looked at who is going viral,

it was almost like a perfect red-blue map, almost like a cell dividing, almost a perfect map of polarization because when you're using that language on these topics, suddenly no one is sharing anything from the other party. So liberal users are only sharing liberal content. Conservative users are only sharing conservative content. So these messages, when they have this language,

are kind of a double-edged sword. It's more likely to spread, and so therefore you're getting all the reinforcements, but it's closing you off to people that are different than you because they're seeing you with them and they're not engaging with you or sharing your content. And so there's kind of like you can think of it as if you go online, you might have two goals.

Your goal is to spread the word or signal your team loyalty, that moral emotional language is what you should use. If you're signaling, if your goal is to persuade or reach across or convince people who don't necessarily already agree with you, it's alienating them. And so people don't see the alienation. It's not like you're getting together at a dinner party and you can see them all rolling their eyes

and saying, I'm going to go home and cutting out early. All you see is the cheers, mostly, because you're seeing like the likes and the shares and the followers go up. So that's one of the problems we don't have perfect feedback when we're sharing this information. It's kind of like preaching versus converting, right?

Preaching to the choir or converting people who aren't, you know, yet members of it. The way that I thought about it as I was reading the paper is you could almost say in a way that contempt or outrage, you know, some of these negative high arousal, moral emotional languages, they almost like they solidify the walls of the echo chamber. Like, because people who agree with you are very likely to engage. People who don't agree with you are more likely to just ignore the message. And that means in a way, and tell me if you think I'm taking this too far. It's suggesting, me that the more angry and morally emotional people get online, the more likely they are to distort

their own sense of how popular an issue is. Because if they're fortifying their own echo chamber with this language, then they are solidifying the sense that their extreme views have a large and enthusiastic audience while they're hearing crickets from the people who don't necessarily agree with them. Is there anything to that, the way in which the use of, I'm sort of combining now, the rules here, but the use of negative, extreme, moral emotional language can almost delude us into thinking that our ideas are more popular than they actually are. Yeah, yeah, it's just because you're not getting full feedback. You're only getting the like button. There's no dislike button on most of these platforms. And so you're deluding yourself about. And even if there were, that wouldn't be a

perfect mechanism either, because there's different ways in which these different things go viral. So what I find so interesting about these rules, again, what drives engagement online? It's like if you were going to write a blueprint for getting engagement, not for a politician, maybe just for like a company, right? Maybe just for like a marketing firm. You might still start this blueprint with like, well, look, negativity drives engagement, extremism drives engagement, identifying outgroups drives engagement, moral emotional language drives engagement.

These are not, you know, tools from the dark side. These are just laws of an ecosystem that we've all accidentally co-produced together in a way. To what extent do you think these rules represent human psychology versus corporate decisions to create algorithms, machines, to present a certain view of reality to us? Is this, am I looking in a mirror? Is this mirror or machine, I guess is what I'm saying? Is it just human nature or is this a corporate creation? Yeah, I think it's both.

So obviously, I think these are part of human nature. These, as I said, often have deep roots in our evolutionary psychology. These are also profit-driven. These are ways that our attention gets monetized by billion-dollar corporations to keep us online as long as possible each day to show as as as as many ads as possible to extract as much money from our attention as can. I would say that some of them, although it might be partially unintended, I talked to people

at Facebook, for example, and they tried to change their algorithm a few years ago to like amplify social interaction. And so more comments and things like that. And they wanted people online like engaging and they thought, well, that might be a positive thing for people. Social connection is something that's dwindling in a lot of corners of society these days. Loneliness is high. But what they found is it actually, you know, I think it was somewhat unintentional. it intensified and amplified content that drove kind of polarizing in extreme views because those generated the most comments. And so sometimes I think they're stumbling in to these things, but for the most part,

they're conditioning their technology design and decisions and algorithms and reward systems based on what gets people to stay online as long as possible. And so part of it is they might be accidentally stumbling in to these aspects of human nature that they then exploit at some. point it might be unwitting and then eventually I think they know what they're doing. I mean, that's a pretty fair answer. I don't think that these companies are blameless, but I also think it is sometimes underrated the degree to which these dynamics are emergent in online group dynamics rather than purposefully engineered into them.

Before I throw two big fat caveats at you about this entire episode, am I missing a rule, negativity, extremism, outgroup, moral emotional language? Is there a another big rule that you would put in this sort of, you know, dark magic playbook to hand to politicians and marketers about what drives, I think often for the worst, user engagement online. Yeah, so let me tell you maybe a fifth rule. And then I'll also share a caveat so that marketers and activists don't walk away with the wrong message here about the downsides of this. So one is the first thing, I think, as an extra rule, is that a lot of this is status-driven. And so there's some great studies, actually done of colleagues of mine in Denmark,

and they find that people who have the highest needs for social status and the highest desire for it are the most likely to share kind of this hostile set of content online. Is this Michael Bank Peterson? Is this Michael Bank Peterson? We just had him on the show a month ago, two months ago. Yeah, we talked about his need for chaos papers. Okay, so he has his other line of research, which is really well done. He's a great, his great scientist, great lab.

on need for social status, that people have a high need for social status are the ones who are most likely to share this type of content that we've been talking about, this hostile content. And so you have to understand that if people who are looking for status are dropped into a system where status and followers and likes and all these things are conferred to people who share this type of content, they're going to figure that out and share more of it. Just in the same way that if you drop them in another environment and like positive or trustworthy, or accurate information is the coin of the realm, they're going to do that.

And so that's the other thing, is a lot of this is being fueled by people who are seeking out ways to gain social status in their environment. And so there's certain motivations that are also driving people who I think are trying to, in their own way, for their own needs,

gain the system. And that can, you know, so you have to think about, like, changing the incentive. So those same people might do something very different if you gave them a different pathway to social status and society. That's a great caveat.

Let me give you the two caveats or asterix that I wanted to present at the end of the show. The first is that I do think that sometimes we lob accusations at social media that we could just as easily lob at traditional media, but traditional media just isn't as interesting, and so we don't. So, for example, I know I just said like moral righteousness and anger can sort of solidify the walls or echo chambers. But obviously that's what's happening when three million people are watching primetime Fox News or even primetime MSNBC. Like not only are they walking into a kind of ideological chamber, but also it's a broadcast rather than any kind of conversation. Like no viewer has the opportunity to jump into the television set and say, actually, I think Jesse Waters is totally wrong about this issue and then get upvoted after they sort of Matrix jumped into the television set. Like that doesn't happen. And so I want to be clear that I think that a lot of the phenomena that we're talking about, even though,

though it might be hyperbolic in 21st century media, was already existent in 20th century media. Surely the negativity bias of news is a great example of that. The second thing that I want to throw in here is I have a little note to myself in my show notes here that says, say something nice about the internet. I do think that what we're talking about here, these sort of 21st century tendrils of group psychology, we need to keep in balance. the group psychology is really good. It's probably what made human civilization possible,

the ability to think abstractly about the minds of other people, to try to persuade them, to try to work with them, to try to understand them, to try to get them to understand us, to try to get our ideas to, our own software to go viral

on as many pieces of cerebral hardware as possible. This is the stuff of human civilization. And so I want to be clear about the fact that I do think, think social media and the internet have poisoned a lot of this. But there's a lot of good to the underlying dynamics of group psychology. And as our group psychologist here, I wanted to give you an opportunity to either say something nice about the internet or say something nice about the group psychology basics that we're discussing here. Yeah, thanks for giving me the opportunity to say this.

I'll try to say something nice about both things. First of all, I use social media because I think it has a lot of upsides, which is spreading information, connecting with people. I wouldn't be on this podcast, probably, unless you hadn't seen my paper, which would, you know, historically have just gone to a bunch of other academics and been read by 12 people and then died on a dusty shop at a library. I saw it on Twitter. Yep. Yeah. So I think those things, it's connecting people and creating conversations that couldn't otherwise happen. It helps to spread information really quickly. In some cases, and the data backs us up, it might be, it seems to be actually eroding trust in in institutions and fostering polarization in developed democracies,

but in underdeveloped democracies or autocracies, it's actually a vehicle for freedom and expression and collective action. And so I think that we need not paint it with a single terrible brush because, you know, I met someone who was critical at helping to overthrow Roberto Mugabe in Zimbabwe. And he told me the story of how he was a pastor, and he got his message out with talking directly into Facebook, and regular people could see it and finally had a voice and could organize around that.

And so when you hear stories like that and you see lots of evidence of it and it's being documented, I think there's definitely two sides of the coin. I will say one other thing that I'll defend group psychology. So I just wrote a book a couple years ago in group psychology. And part of it is the downside of group psychology, but part of it also is I think group psychology is, you know, the key to our evolutionary success as humans. It's why we rule the world for better or for worse. It's because we're not stronger than other species.

We don't have sharp teeth or camouflage. We can't fly. What we can do is cooperate and communicate at ways that no other primate can. And so that has allowed us really to take over the world and to fly to the moon and all these other incredible things. And so I think the way I study it in my research and way I think we should all think about it is not to avoid group psychology and just lean into individualism.

because that has all kinds of downsides. It's to think about how we can leverage it in ways that are healthier for society and for individuals that, you know, people feel connected. They'll feel less lonely if they're part of a group. They'll also be able to accomplish more. There's so many benefits of it if it's done well. It's just understanding when it's getting exploited, manipulated, monetized, and all of these things that can take us into some of these dark corners.

I think it's very well said. someone once said to me that the internet among other things is a machine for helping people find each other. And that is just as true for mothers in a pediatric cancer help group as it is for white nationalists. And that sounds very strange to say to put mothers of kids with cancer with white nationalists. But the truth is that Facebook is very, very effective at helping both like-minded people find each other. And this is why they're advertising machines. is so valuable, and it's also why it has proved so useful for extremists around the world to find each other. The internet is machine for helping people coordinate on action and ideology.

And that is a mercenary skill. There's ways in which that facility is neither good nor bad, but it is really important to understand. Jay Van Bavel, thank you so much. Thanks for having me, dear. Thank you for listening. Plain English is produced by Devin Biroldi. We've got new episodes every Tuesday and Friday. If you like what you're hearing, give us five stars and a nice review on Apple Podcasts or Spotify or wherever you get your podcast. For feedback and episode suggestions, email us at plain English at Spotify.com.