174 - Amit Goldenberg: Collective Emotions and Social Media (REAIR)

0:00Welcome back to the Stanford Psychology Podcast, where we ask leading psychologists about their most recent work. This week, I had the exciting opportunity of chatting with Amit Goldenberg, Assistant Professor of Business Administration at Harvard Business School. Amit studies emotions and social interactions, for example, in political contexts and on social media. He was a journalist and author before becoming an academic. In this episode, I ask Amit to talk about how emotions operate in groups. Do crowds easily go mad? Would emotions spread faster in groups? Why are we drawn to people more politically extreme than us?

0:36How is social media shaping our emotions and political behavior? Finally, Amit shares his journey from being a journalist to being a psychologist at a business school. Hope you enjoy.

1:09this week on the stanford psychology podcast i am very excited to be talking with amit goldenberg about collective emotions and social media and why we are trying to get closer to people who are more politically, morally extreme than us. Many interesting things to discuss and only limited time. So let's jump right in. Thank you so much for making the time. Thank you so much for having me. I'm a huge fan of this podcast. I really appreciate it. Let's talk about collective emotions. All of us know what emotions are. We all feel emotions. But collective emotions, what are those? When I first learned about the term, I had some idea of what they are, but it turns out I was a little bit wrong.

1:53So what are collective emotions and what are they not? I started doing my PhD in a department that is focused on studying emotions. And I was always interested in emotions as they are expressed in group context. But I always felt like when we're looking at emotions at the individual level, we're not capturing what I was really interested in looking at. So when I thought about emotions in group context, I thought about situations in which groups get heated up so much that they tend to do things that are completely beyond what you would expect them to do. I started my PhD in 2013, and during that time, there were so many social movements and conflicts in the world.

2:45And yet our theory of emotions was so much individual focus. And after my first year of grad school, I got a chance to spend some time at the Santa Fe Institute. And the Santa Fe Institute is a center that focuses on complex systems, collective behavior. And I started to realize, and the idea is that as a result of multiple emotional interactions, there's these macro phenomena that have slightly different attributes than the individual parts in which they're comprised of. And I thought, wow, that captures really nicely what I want to think about. and I started looking for people who were thinking about emotions from a complex system perspective, it turns out that there were a few of them already existing, mostly computer science, physics, and we started to form a little bit of a community that is starting to think about this situation.

3:49And the idea is basically that when people get together, their emotions tend to influence each other in ways that lead to some emergent phenomena at the macro level that is interesting in of itself. And although you can measure individual emotional experiences and expressions, there is something unique that you can see at the collective level that is hard to reduce to the individual level. That's the idea, basically. So one last example, when you think about a forest fire, you can look at the forest fire and think about the individual trees and each tree has a trajectory. But there is something important when looking at the fire from a forest perspective.

4:38It's progress, it's intensity. How is it influenced by the network structure, by the connection between the trees, by wind? And this is another angle to look at this phenomenon in a really interesting way and I became really interested in that. The first example that comes to my mind when I think about collective emotions are the emotions I feel when I'm at a concert, right? I pay a lot of money to listen to a band I can listen to for free on the internet. What am I doing? Why do I have to commute there? I'm in uncomfortable seats. I cannot wear what I want. It's the pain. And yet I love going there because of this feeling that all of us somehow have together.

5:20And it's not the same when I listen to a recording of the concert online, even with good audio quality, it's just not the same thing. So I think part of this is maybe some certain amplification as you're alluding to. And also that in crowds, there's certain emotional dynamics that you cannot repeat when you're just by yourself. Yeah. So you're saying a few really important things. The insight that when people get together, they become more emotional is one of the foundations of social sciences. We think of Hegelian philosophy that have conceptualized this idea, and Le Bon and Durkheim's kind of ideas of rituals and collective gatherings.

6:01And the idea is that not only when we get together, our emotions tend to amplify, but also it's really useful for our well-being to be with other people and experience emotion with other people. It also gets to the philosophical question of what is the necessary condition for collective emotion to occur? And this is something that there's a lot of argument about. So for example, if you and I are watching the same concert from our home, but we know that we exist, so we know that we're not the only ones, is it enough to constitute a collective emotion? Or do we need to physically interact with each other for this to be considered a collective emotion?

6:53There's a little bit of a disagreement on what is the necessary condition for that to occur. And a lot of arguments about that question. I think one point we have to address is this myth that I see in articles, pseudoscientific articles all the time about the madness of crowds, right? People get together and all these horrible things happen. Never get close to any groups, never be part of a group because that's how all these horrible things in the world are caused, which is such a really dark take on the psychology of groups. So just quickly, if we could clarify, you're not saying collective emotions lead to only negative things.

7:36Let's shy away from groups. No, not at all. I think that one of the things that we miss in our modern life is more groups. I'm a huge group supporter. And there is an interesting contradiction in the literature. On the one hand, crowds are very emotional. They tend to overreact. They tend to be more aggressive than separate individuals. On the other hand, there's a whole literature on collective intelligence suggesting that crowds are also much more intelligent than an array of separated individuals. And the question is, what gives? And the answer is obviously it very much depends on the situation.

8:16There are very sophisticated and interesting ways in which collectives find ways to regulate themselves, not only to increase intensity, but also to decrease intensity. One really good example is collective rituals. Cultures have been developing rituals since the beginning of history, not only to amplify emotions, which also is the case, but also to reduce emotions and also to reduce variance in emotions. And so I think the notion that crowds, emotionality is always destructive is a misrepresentation of human nature. And I also think that the notion that crowds are always more emotional when they get together is also misses out on a few really important aspects of human interaction that involves regulation.

9:09And while I know you write a lot about contagion and how emotions can be contagious in groups, there's also differentiation going on. It's not like people are always just easily copying all the emotions around them in some complicated dynamics. Sometimes people want to feel quite the opposite emotion or want to not identify with an emotion that a certain group of people is feeling. Yeah, this is how this insight is really what got me hooked into this whole space. I remember before I started grad school, I read a lot about contagion. And the general assumption at the time, and this is based on kind of Hathfield's, Elena Hathfield's kind of modern ideas on emotion contagion, was that this is somewhat of an automatic process.

9:56And I was very much inspired by more recent ideas by people like Maya Tamir and even Jamil Zaki that you know very well, who are thinking of emotional processes as motivated processes. And I had the same intuition for emotion contagion. I thought there are many different situations in which people are motivated not to feel similar emotions to other people for a variety of goals. I think the easiest example is actually to think about this in the context of parenting. You have two parents who are, we have a child who is doing something that makes them angry. Now, one parent is responding with a very strong emotional response.

10:44And the other parent feels like this is not appropriate. This is an inappropriate response. And what we find is that when parents are in that situation, they actually would experience much lower emotions than they would have naturally to offset this overreaction of their partner response and average the collective emotion up. Although this is a very small collective, in this case, two parents, there's still a a process to go against contagion. We also show it in much larger groups, situations in which, for example, liberals in the U.S. see that other people in the U.S. are observing a case of racial inequality and don't feel any guilt.

11:33And we find that when they learn that people don't feel any guilt, they actually amplify their guilt in order to bring in more people into the process and change other people's views. So in other words, there are many situations in which people use their emotions as a tool to influence the collective emotion. With all these complicated dynamics going on of how emotions spread, do we know anything about what kinds of emotions spread more easily? Is it negative emotions that spread more easily than positive ones or particularly emotions that have a very easy time spreading, whereas other emotions, they're really hard to scale up in a collective?

12:18Yeah, so the intuitive answer is that negative emotions spread more than positive emotions. And the reason for that is because we're tuned in to pay more attention to negative emotions because they provide us historically from an evolutionary perspective, they provide us with more information about the world, and it's more likely that it would have impact on our attention and therefore will lead to contagion. One way to test this is to look at content that people produce on social media, measure its degree of emotionality, whether it's positive or negative, and its intensity, and see the likelihood of it being retweeted, being shared by other people, which is a form of an indirect measure of emotion contagion.

13:09the initial assumption in literature was that, in fact, the negative emotions are more likely to spread. And findings by Billy Brady and others suggest that is the case for political context. But here's an interesting thing. When you actually look at the way that people interact act in their daily lives, it doesn't seem like people are more likely to share negative emotions compared to positive. The only situation in which negative emotions seem to be shared more on social media, but only two situations are politics or intergroup situations. And interestingly, when the content is produced by a popular figure.

14:01We have a paper that is coming out with my graduate student, Jonas Shone, that shows that negativity is much more likely to be shared when it's produced by public figures. And there's many reasons for why that would be. One reason is that the connections between the public figure and its followers is less strong. And negative emotions seems to spread further in weaker ties. We already know this from other works. And this is kind of how this idea of weak ties spreading negative emotions further translate itself to whose negative emotions are more likely to be spread. The other context is obviously political context, as I said, in which inner group conflict or perception of the inner group situations are more likely to generate negative emotions, and these emotions are more likely to receive attention.

15:04One phenomenon that has received a lot of attention in psychology and sociology in different fields is the idea of homophily, people aggregating with people who are like them in many dimensions, including political beliefs, right? If I'm a moderate Republican, I like to hang around other moderate Republicans, extreme liberal, I want to hang out with extreme liberals, and that effect has been replicated again and again. But in the political domain, at least, it's not the only effect that is at play. And so you have recently co-authored a really wonderful paper introducing a separate phenomenon called acrophily, which I think is just absolutely key in understanding how collective emotions in the end can be shaped and explaining how people really aggregate towards people who are more extreme than them.

15:55So the insight for Crawford started from the emotional domain and has generalized to other domains. And the idea was, if we're more likely to attend to a stronger emotional expression of our group, we're more likely to prefer people who would express these views, and these people will tend to be more extreme than us. The general insight, as a social media user, I've always noticed that people around me seem to be more extreme than I am. And we tend to blame the algorithms for that. But the algorithms in some ways reflect something really inherent in our tendency. And so if you're a liberal who's a seven on a one to 10 scale, and I offer you someone who's an 8 or someone who's a 6, which one would you prefer?

16:52If we're looking at the literature on Thai formation, it suggests that the only thing drives the system is homophily. In that case, homophily, the distance between 7 and 8 and 7 and 6 is equal. So there's no prediction regarding who would you choose. But that seemed to me like not reflecting of how people actually behave. And so we created these series of studies in which we had people respond first emotionally and then in other providing their attitudes on a variety of issues. And they rated how they felt or how much they supported a specific issue. And then we gave them a few people to choose from who responded to the same issue.

17:38And we asked them, who do you be interested in interacting with? Who would you like to keep seeing in response to these stimuli? And what we found was that in addition to homophily, people do what we call acrophily, attraction to extreme, which is that people would much prefer the eight over the six if they're liberal, or if they're conservative and their five is neutral, so there are three and we offer them either a four or a two, they would much prefer the two, meaning the more extreme conservative. And this is a really basic phenomena that really accelerates the processes that we thought we're seeing in our day-to-day kind of political interactions of polarization and segregation.

18:27because if you're homophilizing, that's one degree of segregation. But if you're constantly attracted to people who are more extreme than you, the system is more likely to segregate much quicker, which presumably would lead it to be more polarized than we think it is. And we find a lot of evidence for this now in a variety of contexts, not only in LabX methods, but also on Twitter, on social media, And even in social networks, there are kind of real-life social networks in universities, for example. Looking at all this like a researcher almost gives us this bird's-eye view of what are we doing here?

19:10This is only making polarization worse. But from the perspective of a participant or a regular person or speaking for myself on social media, the question is why are we engaging in this effect? And it turns out it has a lot to do with who we see as representative. Right. So we had all sorts of ideas of what drives this tendency for Crawfordly. But the thing that seemed to be sticking the most in terms of the empirical evidence is this idea that we tend to think that the normative response of our political group is actually more extreme than it actually is. and this creates a gravitational force for us to prefer that view because we think this is representative of our group and so we're attracted to it and indeed we find that people who think that the view of their political group is more extreme are also those who are more likely to choose those more extreme people now that's interesting i could see why i would think my political extremes people to the same political meaning but more extreme than me i'm more representative there's almost a sense of they hold the same ideals but more strongly and i'm just too weak i don't really stand up for them as much as i should as they are standing up for them right so they are representing the ideals and i'm just like a watered down version But then there's also like another way in which extremes are not representative, in which I might say these people are literally too extreme, right?

20:46Why can't they be more moderate like me? What is going on with them? And it turns out oftentimes people at the extreme can be very loud. And as you say, their social media behavior, for example, gets amplified more easily than the behavior of people who use less moral, emotional language, for example. And so it can almost create this idea of these people are everywhere because their content is everywhere. But actually, it's just a few people who are very loud. Yeah, exactly. And here's another angle to the same thing. And I should say, I do think there are other things that drive this process.

21:17For example, there's a really interesting paper by Federico Zimmerman. And they took 5 ,000 people and divided them randomly into pairs and asked them to talk about six political issues. And then after they had this discussion, a five-minute discussion on these six political issues, they ask each people to rate how much they like the other person. And what they found was that people were attracted to similarity, but they were also attracted to coherence, meaning people who represented a very coherent view of a political view, meaning they were just left on all six issues, were more likely, were more preferred than people who had a more mixed kind of message.

22:02And this directly relates to this idea that people who are at the extreme are also often more coherent in their ideas. They just have a very clear sense of the view, and their view is really hard to interpret. And it seems like that's something that people are also attracted to. Finally, there is the Thanksgiving effect, which is we all know that we're going to face an angry cousin on the other political group on Thanksgiving dinner, and we just want a sharper toolbox for that conversation to come. And so the person at the extreme provide us with that sharper toolbox. And we don't want to express these views, but we think that knowing them can help us a little bit in improving our toolbox for the future.

22:57We have no empirical evidence for that, but this is something that we're really interested in exploring. And it seems to be really hard to gauge on social media how much my in-group rolls their eyes at our political extremes, right? So if someone to my political extreme is posting something and it goes viral, I see all the likes, I see all the retweets, right? On Facebook, I see all the positive reactions, maybe some negative reactions. Oftentimes the negative reactions I might see are like toxic comments or quote tweets, which are not really a mild reaction as I would voice it, even if I disagree.

23:33There's no eye roll emoji. And it seems like such an emoji would do many bad things. So I'm not saying we should have it. But I think if we had it, it would give us like a little bit less of this almost pluralistic ignorance where we think, oh my God, everyone in this in-group loves this content. When actually a lot of people are like, oh God, this again. And now I have to care about this? Yeah. So first of all, if you're listening and you have some control over the world of emojis, please design an eye-rolling emoji. That's the best idea I've heard in a long time. But the second, your point is very true.

24:09There is the people who are silent and rolling their eyes are unheard. And that leads to a false representation of the norm, a false representation of the norm on the other hand or the other side. and an attraction and this misrepresentative of the norm leads to an attraction that is likely to amplify these dynamics even further. And the real question in my lab, we always have a big argument of what's the solution to the problem? Is the solution to the problem filtering? Meaning, do we need to design algorithms who will filter content in a way that would be healthier, less polarizing, and so on and so forth.

24:52But in some ways, we fiddle with the content that people see. We don't naturally represent the world in some way. Or is the right solution is to inform people and maybe providing them with information that the content that they see is not representative is the right solution. And we go back and forth. And I don't have a good answer of which one is better. One is obviously more aggressive, but it's an interesting question that I think should be kept open when we're thinking about how should we solve these destructive dynamics? There is no data behind this. This is just me as one sample point. But having learned about all this research, I enjoy social media a lot more because every time some hostile content goes viral, I don't think, oh God, now I have to pretend And I am this hostile because everyone is like this.

25:46I'm like, probably a lot of people who like this are just like, oh, God. Yeah, sure. I care about this. Everyone, please like me. Don't unfollow me. I'm one of you. And there's this test that I do where every time a moral issue goes viral and people seem to be debating it everywhere. I talk to my friends who have the same political leaning and ask them what they think. And almost every time it goes as follows. First, they say, oh, yeah, this is outrageous. And how could they? then I asked them by the way what is the issue about and they're like actually I have no idea I just liked it because it sounded really outrageous but really I don't know what's going on it's just like this performative outrage learning about the extent to that's true it's really useful so you fall in the information camp it sounds I think both camps can be true but certainly that one I think there's got to be a way that's true and it's also more empowering of course because you don't control people you just give them information and it's oh yeah that's useful.

26:42So there is a researcher, Chris Kelly, who did a really interesting intervention online. They told people that the content that this is to resolve the phenomena of what we call doom scrolling. So if you're in a bad mood, you will look more for negatively balanced information online. And what they try to do first was to tell people that the content that you're about to read has negative emotions in it. And that seemed to have no effect on people's choses. They modified the instructions by telling people that the content that they're about to see is going to make them feel worse. And that really made a big difference.

27:26They stopped looking at these negatively balanced information. So there is hope for the information camp. We just need to be very thoughtful, I think, Think about how we provide that information in a way that would help people, that would make the problem better rather than worse. On a methodological issue, how do we even know who is authentic on social media? So you mentioned as a somewhat recent tool, we can use social media information to study collective emotions, group-based emotions, individual emotions. but how do we know because people can behave in certain ways but maybe they're just pretending to go along versus do they actually mean it so here's the interesting thing about so one way to ask this question is do emotions on social media look like emotions outside of social media and so this answer it's a full circle because the answer to this question depends on whether you're answering this at the individual level or the collective level.

28:32At the collective level, work by David Garcia, what they did was they did sentiment analysis. They evaluated the emotions expressed in two and a half billion tweets produced in the UK. And they also looked at these really large scale surveys that were done on people's emotions every two weeks in the UK. and they correlated the emotional intensity of the emotions expressed on Twitter and the emotions that people rated in their surveys. And the correlation are crazy high. It's in some of their studies, it's 0.87, which is something that you would not imagine could be seen. So at least at the collective level, it seems like there is a really the emotion that people as a collective express online seem to track really well on their emotions in their life.

29:30This doesn't mean that the intensity is the same, meaning it could be that the emotions on social media is a phase up in terms of intensity, but very correlated in terms of the degree of positive and negative, but at least they're correlated. At the individual level, the results, it's a much harder task, because if a person is expressing emotion, think about how that study could be designed. You need to measure how people feel and then at the exact same time measure what they produce on social media. And these two things obviously influence each other in really interesting ways. But all sorts of elegant ways that have examined this found a really poor correlation between people's emotion rating and emotional expression on social media around R.1.

30:20So, it's either the methods that are inaccurate or that is something about social media that captures really well our collective emotions but not our individual idiosyncratic emotional experiences. I don't know the answer to that question. The point of accuracy is really interesting. So, you were talking about us researchers trying to be accurate about how individuals feel online versus not. But accuracy is a big issue in the study of emotions, of course. Can we read what other people are feeling? Usually on an individual-to-individual basis, we look at someone's face and they frown when we infer certain emotions.

31:02Are we right or not? And it's a whole mess. And I'm not even going to try to summarize what people are thinking because everyone's going to hate me and there's many camps and there's a lot of debate. What about collective emotions? I know that you've done some work finding that when people try to assess the emotions of a collective, they're not just aiming for accuracy. There are certain, I don't know what you would call them, biases, certain tendencies certainly to overweigh the extremes. Yeah, the question of how do people understand what other people around them feel was a really interesting question for me.

31:36And I could not find a really good way to test this in grad school. because it's really not clear. When I ask you, Eric, how do Americans feel about the current president? It's really unclear what people do to form that judgment. It's a really interesting problem. Do we think of one person that seems to be representative of the collective Do we sample from our memory a few people who are salient in our mind? It's really an interesting question that we don't have a good answer to. It seems like we're inaccurate. That's for sure. It seems like we tend to amplify. But one thing that we decided to do was, okay, let's try to take it to the simplest state.