Leor Hackel studies how we learn about other people and how we make decisions about them. He’s an Assistant Professor of Psychology at the University of Southern California, and he uses neuroscience, economic games, and computational models to sort out what’s going on in our heads as we’re getting information about other people.
This conversation explores how we learn about other people through interactive experiences, not just passive observation. Leor Hackel, a psychologist at the University of Southern California, discusses his research on the distinct processes involved in learning from the outcomes of our interactions (reward learning) versus forming impressions of someone’s character traits (trait learning). He explains how these two learning systems can sometimes come into conflict, such as when we like someone more because they gave us a large reward, even if they were less generous overall. The discussion also covers how we may simplify our impressions of others over time, moving from detailed memories to more categorical judgments. Overall, the research suggests that learning about people involves a complex integration of affective and cognitive processes that can shape our social decisions and relationships.
Things that we mention in this episode
- Dolf Zillmann’s disposition theory (Zillmann & Cantor, 1972; 1996; also see affective disposition theory [Wiki])
- The difference between “reward associations” and “trait impressions” in how we learn about other people (Hackel et al., 2020; 2022), including differences in brain processes (Hackel et al., 2015)
- People will give more to someone who gave them more, even if that person is just as “generous” a person as someone who gave less (Hackel et al., 2018)
- We can form impressions of others is various sorts of “gist” memories (Hackel et al., in press)
Transcript
Andy Luttrell:
Last night we were watching Ted Lasso because why wouldn’t we be? And there was a moment when the main character said something and I thought, “Oh, that is so Ted.” But how do I know? Ted’s not real. I’ve never actually met him. I’m also not like a TV critic, so I’m not deliberately reflecting on every nuance and decision he makes. And yet, even as just a casual fan of the show, I’ve developed a sense of who this guy is, how he acts. I can anticipate how he’ll respond to different situations, all without ever asking myself to paint some character portrait of him. It just happens. We do this all the time.
There are scholars in communication and media studies with formal theories of how this stuff plays out. One of the notable ones goes back to a guy named Dolf Zillmann. As a kid in Poland during World War II, he didn’t have a simple childhood. His home province was being yanked between Polish, German and Soviet forces. His dad went off to war and died. His schools were supremely underfunded. After the war, he ended up teaching him himself most of what he learned as a kid. But eventually he ends up in the United States to study communication and psychology, and by the ’70s, he’s really interested in humor of all things. When are we okay with entertainment that pokes fun at people? That gave way to a broader theory of how we engage with entertainment in general. Disposition theory. The idea is basically that we don’t just passively observe characters in stories. We root for the good guys and relish when the bad guys get their comeuppance. Later, Zillmann wrote that viewers of media are untiring moral monitors. In other words, we’re always on the lookout for who’s doing good and who’s doing wrong. This idea has blossomed into a bunch of interesting ideas about how we connect with and think about characters in movies, books and TV shows.
But for me, whenever I think about these theories of entertainments, I always wonder, is this just how we think about people in our real lives? Sure, I watch Finding Nemo and have a clear sense of who Marlon the clown fish is and why I’m rooting for him. But isn’t that just the app my brain uses to understand real people getting co-opted to understand this fish character? Is there a difference? I think there is a difference, and it’s one that makes me appreciate how much more is at stake when I’m sussing out the kind of person my neighbor is as opposed to an animated fish. I interact with my neighbor, my neighbor’s behavior directly affects me. I’m not passively observing a story I have no part in. I’m personally instigating some of the things my neighbor does that give me a sense of who he really is. So how do we do this? How do we learn about other people through the entanglement of our interactions?
You are listening to Opinion Science, a show about our opinions, where they come from and how they change. I’m Andy Luttrell, and this week I’m excited to share my conversation with an old buddy, Leor Hackel. He’s an assistant professor of psychology at the University of Southern California and he studies how we learn about other people and how we make decisions about them. For me, this falls well within the walls of opinion science because it seems we’re constantly sizing up other people. Do we like them? Do we trust them? What’s our take on John? His research uses neuroscience, economic games and computational models to sort out what’s going on in our heads as we’re getting information about other people. It was great to catch up with him and get his thoughts on all of this, so let’s get right into it.
As I read it, a lot of the work that you do or the way that you frame it in terms of how we learn about others is that we learn about them through interacting with them. I get to know the kind of guy you are by virtue of the fact that I have experience getting to know you. And some people might hear that and go like, “Yeah, right. That’s how we would get to know someone.” And so I’m curious, do you think that psychology has generally been good about acknowledging that as the mechanism through which we learn about people or is it actually a new thing to introduce this as like, no, actually we really ought to get back to basics and look at how we really learn about people in the world?
Leor Hackel:
Yeah. This is such a great question. I think this is something that’s been ignored at least more than it should have been in the world of social cognition. It sounds intuitive. A lot of the work, the vast majority of work in social cognition over the last several decades looking at how we learn about people has really looked at really passive kinds of learning. So you see on a computer screen that John helped an elderly person across the street and you form an impression of John. We’ve learned obviously a ton from that work about how we form impressions and which kinds of impressions we form and how they cohere. The perspective that I take is that there are a number of different systems for learning and memory in the human brain and they contribute in different ways to behavior. And I think one of the things that’s really different in interaction is that yes, we form impressions, but we’re also experiencing things.
So if John helps us cross the street directly, there’s an affective experience there and in particular there’s an experience of reward. We have an outcome that feels positive or that feels negative. In theories of learning this corresponds to instrumental learning, which is a kind of learning from rewards where we actually perform actions and then get some reward feedback. Either we get a reward or we don’t, and then that’s going to shape our future actions down the line based on those outcomes. This is thought to be a somewhat distinct kind of learning from more passive kinds of learning through observation. It is thought to have some distinct behavioral consequences. So for example, instrumental learning can give rise to habits where we keep on doing the same thing even when it no longer really fulfills our goal. And so we started looking at social learning through interaction in order to say, well, all right, the social cognition literature has really been focused on impression formation, but that’s been through passive kinds of learning. Meanwhile, in non-social work, there’s been a lot of focus on how we learn through actual reward feedback, but that’s largely been in non-social settings. So how do these things actually come together in social interactions? To what extent are we learning from these affective experiences of good and bad outcomes, things that are rewarding or not versus these abstract impressions that were forming of people?
Andy Luttrell:
Does it seem like it’s something that’s qualitatively different, like this interactive learning versus, oh, I just get more information, I get some information by just being a passive observer and then I get some extra by being the recipient of some of these actions or being closer up to what’s happening? Or is it like, no, I’m just in a new mode of learning about you now that I’m involved?
Leor Hackel:
Yeah. So based on work on learning and cognitive neuroscience, the idea would be that there really is this additional mode of learning that’s coming into play. We have some data that we have yet to publish or we’ve looked at this about differences when people merely observe versus actually interact and get rewards. And we see the reward effects when people actually interact more so or entirely as opposed to when they’re just observing. That’s not published yet, so I don’t want to make too much of it. We still want to replicate those findings. But at the very least I would say the prediction from the learning literature should be that this kind of learning really comes into play when you’re interacting,
Andy Luttrell:
Meaning that it’s also different not just from other forms of learning about people, but other kinds of learning in general. For me to get to know you requires something different of than me learning physics or whatever.
Leor Hackel:
That’s right. I think what really this gets at is the idea of social action. So social psychologists have long said that thinking is for doing. But I think really studying how we actually learn which social actions to take brings in these other layers. So one way I can learn what social action to take is to learn Andy’s a really trustworthy guy and so I should take the action of giving him this loan or telling him my secret. And that is one way I can learn what action to take. Another way I can learn what action to take is I can tell you a secret, experience that you keep it and learn, okay, I should do that again in the future based on that rewarding outcome. And both of these things can be happening at once. One of the things that we’ve explored is that these things can also diverge. So there can be situations where you form a positive impression of someone even though you didn’t have a very rewarding interaction or you could have a really rewarding interaction with them even though you think they’re a jerk. And so these things can also point in different directions at times and lead people into different paths.
Andy Luttrell:
Is this a program that’s always running? Because it could be where you go, I don’t have any secrets, so I don’t really care. I don’t need to know this about this person. And so you could imagine that I boot up the program of learning about people only when it’s really important as opposed to maybe another perspective which goes like, well, no, to be a person in the world, it just means you’re always running this program. That’s the one thing we’re always doing. That’s probably too big a question to really answer, but what’s your sense of that distinction?
Leor Hackel:
I would lean more towards the latter. I think that at least for most of our social interactions … And I’m happy to say more about this. The perspective I come from is that one of the things that’s really unique about social interaction is just how much knowledge and expertise and expectation we come into it with. And so the vast majority of the time there’s some relevant meaning that we can make out of someone’s behavior. Even if I’m not going to need to tell them a secret again, maybe I have a general impression that they’re a good person and that might be relevant for me. I can make sense of what’s happening through social knowledge that I have. I lean towards the latter. I think both of these things are just always happening as we engage in interaction. I think if you were in a social scenario that is completely novel to you and you had no way to make sense of it given your pre-existing knowledge and you would never be in that situation again in the future, and so you’d have to work really hard to think about what this means, yeah, maybe then you’re not going to do that. But I think most of the time we do.
Andy Luttrell:
It reminds me a little bit of the … I hate the name that we give this, but spontaneous trait inference if we’re to use the fussy label for it. But their whole premise is like we see people do things and we cannot help but interpret that action as saying something about them and they do all these things to try and stack the deck against it and they go, no, you still come away with an impression of this person just by virtue of having seen it. And that seems like that would only ramp up if it’s I’m involved in this. This is happening to me or I am navigating a relationship. I can speak for myself. I can read into a lot of things that people do and say.
Leor Hackel:
Yeah. Yeah. Exactly. I think that’s exactly how I think about it. I think the same thing is true about the reward side of things where just in the background we’re constantly experiencing things that feel good or bad and we are more complicated than that. Humans are not going to just thoughtlessly repeat actions because something good happened or avoid actions thoughtlessly because something bad happened. It’s a simplistic form of learning. But I think it’s also a mistake to think therefore that it has no influence. There still is this contribution I think of this more affectively driven form of learning that it influences us, at least to some degree when we do experience these good or bad things happen.
Andy Luttrell:
It’s reminding me just like a week or so ago I came across this paper, which I haven’t actually read so all of what I’m saying is tentative. But my impression is that the finding is something like with people who have severe memory impairment and can’t actually remember episodic events, if you treat this person poorly later on, they’re going to go, I don’t think I’ve ever met you, but I don’t like you. Which to me highlights that this program is churning. Even when we can’t reflect and go, I don’t know why I don’t trust you, but at the moment my brain was like, let’s hang on to this because this person may not treat you well in the future. And similarly, I’m really speaking at the edge of what I really know how to say. But in the neuroscience area, I have heard and read that the processes in the default mode network, this set of brain areas that seem to just be persistently doing stuff are also the same kinds of brain areas that are involved in when we try and figure out what other people’s deal is. Is that a finding that’s stayed consistently true or have I butchered it in some way?
Leor Hackel:
No. No. That’s right. So this work from Megan Meyer, Bob Spun and Matt Lieberman showing that when people show more activation in these regions at rest, they are then faster to respond if you ask them to make some social attribution about other people or themselves. And so that this is the idea they’ve put forth is that it’s the default mode network that’s active even when we’re not doing a particular task might be getting us ready to engage in social inference.
Andy Luttrell:
Okay. So let’s talk about some of what you’ve actually done. So you’ve raised this distinction between general affective rewardy reactions we have to other people and maybe more nuanced impressions that we form. So just to make sure that everyone’s on the same page, what exactly is the difference there and why is that a meaningful difference?
Leor Hackel:
Well, I can answer that in part by talking about how we’ve studied it. A lot of how we’ve studied this has been through economic games where you are playing some kind of social game with another person and you have to learn who you want to interact with over time. So trying to model the process of finding out which people in the world do I want to form some kind of tie with or partner with. What we’ve done though is manipulate two things. So first of all, what proportion of money available does someone share with you in this game? So you are going to be recipient and someone else has made a decision to share some amount of money with you. And so the first question is, what proportion are they going to share with you? And this is going to get at their more abstract generosity or fairness, a more social impression.
But different people can have different amounts of resources available. So imagine that one person starts with a dollar and they share half of what they have. So you’ll walk away with 50 cents. Imagine another person starts with $2 and they share half of what they have. You walk away with a dollar. So they’ve been equally proportionally generous. They’re showing identical willingness to fairly split their resources, but you walk away with a really different material reward. And so we have people play games like this where they choose a partner and then find out both the amount they’re getting from the partner and the proportion they’re getting. And then what we can do is ask, all right, who do you go back to over time? Are you gravitating back towards the partners who shared a large amount with you and who were rewarding? Or are you gravitating back towards the partners who were being proportionally generous? This more abstract quality where regardless of the amount of money at stake, they’re splitting it fairly.
What we find is that the answer is both. In the studies we’ve run, typically people primarily go back to the generous partners and you have this smaller but quite robust effect of reward. So people generally are going back to the generous partners. One caveat on this is that it really depends on the context. So if you run a study where there’s some people who have a thousand dollars and other people who have $1, people are going to choose the ones who have a thousand dollars, even if they’re equally generous. Another possible caveat here is culture, which I can say more about, but one of my collaborators has been exploring whether culture also shapes this distinction. That said, to come back to your question, so that’s one way of thinking about it. Reward is this concrete outcome. How much money are you walking away with from this interaction and how good or bad do you feel about that? Whereas an inference about someone’s generosity, again, it’s this more abstract inference about how fairly they’ll split resources regardless of what’s at stake and it leads you to expect they’ll probably do the same thing again in the future.
As for the question about then, well, so why does this matter? We’ve looked at a couple different implications of the way people learn here. One implication we’ve looked at for example, has been for economic inequality. So typically people reciprocate with others who are generous. So if somebody does something kind for you, you become more likely to do something kind for them. And this is typically thought of in terms of those impressions of generosity and fairness. But if you also have this contribution of reward learning where when we have good outcomes in an interaction with somebody, we feel good, we feel more positively towards them, we’re more driven to interact with them well, we might end up reciprocating more with individuals who start out wealthier and therefore can provide us with more material resources even with equivalent levels of generosity.
And this is something that we found. So if you experience a scenario that I’ve talked about before, one person has a dollar and they give you 50 cents, another person has $2 and they give you a dollar, later on if we give people the opportunity to now reciprocate and say, “Hey, the last thing that’s going to happen in this study is we’re going to switch the roles. Now you can send money back to these people just out of the goodness of your heart. You won’t interact with them again. There’s no strategic concerns. Just who do you want to send more money to?” Now it turns out that there is this effective reward where people are likely to send more money back to the person who started out richer because that person evoked this kind of more positive disposition. Now people also of course reciprocate based on generosity, but above and beyond that you have this effect of reward. And so if this is happening in the background, you can get these effects that can propagate inequality, for example.
Andy Luttrell:
So just to summarize, you can rig a game like this where I’m learning about another person so that I get a lot of money from a stingy person, but I get very little money from a generous person and this poses this potential dilemma. Like I really like when I have this money and person A gave me more money, and so I should be inclined to like them. But at the same time I’m capable of understanding that person A was giving me relatively little money relative to what they had. And so they’re than person B. And yeah, interesting. So generally it’s true that I would still reward the generous person and get like, “Hey, thanks. You’re great.” I feel almost maybe a moral obligation to pay back some to you because you went about this game really honestly and virtuously. But I also go, but yeah, but also person A gave me a lot and so I feel indebted to them. And so still there’s this additional urge to reciprocate the amount that was given. Does that capture what you’re saying?
Leor Hackel:
That captures it really well. I think that the reciprocity part, I would just add two notes. One is that, yeah, if you had one person who is generous and one who is stingy, people will reciprocate more with the generous person. If you have two people who are equally generous, but one gave you more, that’s where I think you’ll really see these effects of that reward. The other thing I would say is that I think that that reward influence, the way I think about it at least is as something affective where we kept on experiencing these positive feelings and it builds this positive disposition towards this other person. Maybe we even end up thinking that they were more generous than in fact they were. And that is something we’ve seen in some studies that people misreport these individuals as having been more generous. I’m not sure people are explicitly trying to rematch the amount they got. I think it biases a bit your disposition towards this person to feel positively towards them.
Andy Luttrell:
Some of this stuff is stuff that you’ve looked at in the brain too, this reward versus trait inference stuff. So what is the distinction there? What do we gain by taking that level of analysis to it?
Leor Hackel:
Yeah. That’s a great question. I would say there are two things that I think we’ve gotten from doing that. The first thing is to try to really dissociate learning processes in ways that are hard to do in behavior. So behaviorally, we can say people are learning from these two signals, they learn from the amount they get, absolutely, and they learn from the proportion they get. Are these really qualitatively different kinds of learning or is there just one general learning system that attracts two different pieces of information and two different inputs? And so neuroimaging has been helpful to say, well, actually these things are dissociable in the brain and learning signals for each of these have different patterns of brain activation. And so it does seem like there’s a more meaningful distinction that these are different kinds of learning happening at once.
The other way that I think neuroimaging has been really useful for this is to serve as this common bridge to other literatures in other areas. So for example, in just a vast amount of work on non-social reward learning or reward processing, a region known as ventral striatum is active when people are learning about rewards. So when they get rewards that are better than expected and has been found in tons and tons and tons of studies. Well, in our work in social interactions when people are learning from rewards, we see that ventral striatum is responding to better than expected rewards. Now that suggests that there really is this potential overlap in process of this general reward learning process that’s been studied in a lot of other ways. And I think that’s useful for generating new hypotheses, for being able to say, well, maybe we can take what we know about the reward learning literature from non-social cases and test if that’s going to be true in social interactions as well. And so having that bridge that can suggest new ideas as well.
Andy Luttrell:
And nice also though, to be able to say that that’s not sufficient to completely account for social learning. It might feel like, oh, we can just look at all this ventral striatum stuff and be like, “Oh, this is going to tell us how we learn about people.” And your perspective is actually like, yes, but there’s this other process that people seem to really be leaning on that you’d totally miss if you were just thinking about it in those terms.
Leor Hackel:
Yes. Thank you for saying that for me. Absolutely. I only addressed one side of the equation, but yeah, I think I mentioned some of what I see as the insight for social cognition about how interactive learning and reward matters, but absolutely. One of the points we wanted to make for the more general learning literature in cognitive neuroscience was to say, yeah, look, when people are learning about in this task generosity learning signals, they also activate a ton of regions that have been associated with impression updating in past work. And so yeah, reward is really not the only thing or in our study, even the primary thing that’s going on. Definitely.
Andy Luttrell:
So the examples that we’ve been talking about so far are reward in terms of you decide to give me money and me learning about your generosity. That’s the trait. But you’ve looked at this for other of trait learning too. So pitting reward against other sorts of traits. The one that I could name is work that you’ve done on competence, maybe you’ve done others. What do some of those look like where you again are rigging the rules of the game so as to pit these two potential ways of learning about people against each other?
Leor Hackel:
Yeah. We see pretty consistent results. So we’ve done this with competence where you’re learning how much workers earned for you in a trivia game, both through the proportion they earned, but also the amount that you got, which is just dependent on the number of points that at stake for them in the game. More recently we’ve been doing work on how people learn from social acceptance and rejection. The idea being that somebody could really want to interact with you and they could think very highly of you, but for one reason or another it might not work out. Maybe they would rank you third out of a hundred to be on their team, but it’s a two person team and you don’t make the cut. Or maybe they would rank you 99th out of a hundred, but it’s a 100 person team, and so you do make the cut. So across these really different scenarios, we pretty consistently see that both of these kinds of learning take place. People learn from the social inference. This person is competent and they will be in the future, or this person likes me and they will try to include me in the future. But people also learn from the rewarding outcome. Hey, this person earned a lot of points for me in the game or hey, at the end of the day I got accepted by this person, I got to play this game with them.
Andy Luttrell:
It’s an optimistic view of the world actually, as I think about it. Isn’t it nice that we can not really gain much from a person but still learn about their strong qualities? I can still come to learn that you have a lot of positive qualities, even when I personally stood to gain not very much. And so from my world of thinking about what overall impression do I have of you, I guess the question is do people integrate? If you just ask people, how much do you like this? How good is this person? Do people compartmentalize the feeling that I got when I interacted with you from the kind of person I figured out that you are, or do I just mash them together and go like, well, you are generous, but I got not very much from you? And so I like you, but not as much as if I also got a lot from you and you’re generous.
Leor Hackel:
Both. So people will mainly be able to separate these. By and large, if you ask people, how much do you like this person, or how generous do you think this person was, or how competent do you think this person was, the biggest effect you’ll see is in fact, how generous were they proportionally, how competent were they in terms of their actual performance? How did they rank you? But you do also see this smaller, but again, really, really consistent bias where people think that you were a little more generous if you gave them more or earned them more. So that’s why I say it’s both. To some degree people can keep these separate, but there’s affective spillover is how I think of it, at least to some degree.
Andy Luttrell:
So it seems like maybe the primary thing we turn to when we gauge how good of a person you are these things that I’ve learned about you. Like what kind of person are you? Are you generous? Are you competent? Are you kind, virtuous? And then I disseminate some bonus points or I take some away based on the feelings I got when I interacted with you. I hate stage theories of things, but it feels like a two-stage process of first I anchor on how good you are as a person based on what I know to be true of you. And then I nudge that based on the feeling.
Leor Hackel:
Yeah. I’m not sure how sequential it is as opposed to just simultaneously integrating these different sources of information. But I would agree with the metaphor of it’s like getting some extra bonus points on your impression based on how you made someone feel.
Andy Luttrell:
Okay. So more recently, this stuff has been pushed in another direction, and what I really liked about this new stuff that came out is maybe the themes of my question so far is how are we reflecting specifically on what we’re learning versus just getting a general impression of who you are? So I think I’m mostly just teeing you up to talk about this project. It starts with that question of when I am in these interactions, am I writing in my journal everything that happens so that I can look back on it later and go, oh, you did these three nice things, you are a good person. Or am I just vaguely getting this intuitive sense that you’re a good person? And so what is that distinction and how can we play with it?
Leor Hackel:
Yeah. So this is recent work where we really wanted a bridge … It’s two lines of work, one in the field of memory and judgment, and on the other hand, one in social impression formation where we started thinking that maybe these are talking about the same thing without realizing it to some degree. So in the domain of memory and judgment, fuzzy trace theory is this theoretical framework that we drew on. And the idea of fuzzy trace theory is exactly what you’re describing, that when we experience really any event, we simultaneously form these two memories. First off, there’s detailed episodic memories, so exactly what happened in an event. But parallel with that, at the same time, we’re also forming what’s called a gist memory where we’re just extracting what’s the ultimate meaning here? What does this mean? Forget all the details. One common example is I could tell you that there’s a 30% chance of rain or a 70% chance of rain and you could remember those details or you could just extract from it, okay, there’s a low chance or a high chance. And you don’t have to remember if I said it was a 70% chance or an 80% chance. That detail isn’t super important for guiding your decisions that are to come. What really matters is it a low chance or a high chance? Should I bother taking an umbrella or do I not need to?
And there are a couple of points about fuzzy trace theory that we thought might be really relevant for thinking about social impressions. So the first was this idea that we formed these two kinds of memory and parallel because what this reminded us of was exactly what you mentioned before, a spontaneous trait inferences. So in really classic work on this, they had subjects read sentences like the barista tripped while carrying the drinks. And then later what they find out is that people formed two kinds of memories in parallel. People did form memories for the details of the sentence, and people could, to some degree, recapitulate the exact words in the sentence, simultaneously, people encoded in memory the gist of it via a trait word like clumsy. And so people also had a memory for this trait word like clumsy. And so that looks a lot like what fuzzy trace theory is talking about when it talks about gist.
The second thing that we thought was really interesting is that in the fuzzy trace literature, gists persist for longer than verbatim memories. So even after people have forgotten the details of an event, they might still hold onto that gist, which again gets rid of the details, but extracts the ultimate meaning of it in this simpler format. Well, that’s true of social impression formation. People can remember hundreds of impressions of whether they like someone or not based on their behavior, even when they’ve forgotten the behavior they read. So even long after you’ve forgotten that the barista spilled the drinks, you might still hold onto some impression of how much you like them or not.
And then the third thing that we thought was really relevant is that in fuzzy trace theory, there’s this idea that when people have more expertise in something, they really rely more on gist. So novices are going to get lost in the weeds and just attend to each individual tree. An expert can look at something and immediately see the forest and pull out the ultimate meaning here in a much simpler way. And as I started with today, I think humans have just immense social expertise. We have these concepts ready to go, trait concepts like generous and competent. Moral concepts like fair and unfair. Role concepts like helper or teacher. And so we have this expertise where we can really, really, really do a great job of extracting gist.
So starting with all of these ideas, we thought, well, okay, maybe we can draw further on the fuzzy trace literature to think about how this unfolds. So there’s two other things that I have to mention just as background here. I’ve talked so far about details versus gist, but fuzzy trace theory really says there’s a couple kinds of gist that push farther and farther in the direction of simplifying our memories. The idea is we really are going to just gravitate towards the simplest representation of something we can. So instead of 70% and 30% chance of rain, higher low chance of rain is simpler. But you could get to an even simpler representation, which is called a categorical gist.
And this would be, if I told you that there’s a 5% chance one day and a 60% chance another day, you might just encode this as basically no chance and some chance. And that’s a categorical distinction. This is even simpler because imagine you’re looking at the five-day forecast. You don’t have to keep in memory, okay, Monday has a higher chance, Tuesday has a slightly lower chance. Wednesday has an intermediate chance, Thursday’s the highest. All you have to do is for each day you know, yeah, Monday there’s some chance. Tuesday, no chance Wednesday, no chance. So it’s even simpler. And so there’s this pull towards the simplest representation that you can get to from details to relative goodness or badness or relative rankings. And finally to just black and white categories. Discrete ways of labeling something.
And so the insight of this new work that Peter Mende-Siedlecki and I just published is that maybe that’s true as well of social impressions. If social impressions look like just memories, they persist longer than our memories for details of social interactions we have a bunch of expertise in social cognition. Well, maybe we also form these two kinds of social memories, either remembering people in terms of relative rankings, this person was more generous, this person was less generous, and we can encode people categorically. This person was generous, this person was stingy. And again, maybe we have this pull towards simplicity that when we can, we’re going to go to the simplest representation we can get to. If we’re able to get to a relative representation, we’ll do that. If we’re able to go even farther and land on a categorical representation that maybe we’ll do that. That’s what we took a look at in the study as well as how this impacts our decisions that we then make about people.
Andy Luttrell:
And you’re able to pit these against each other. The metaphor I always use is you rig the rules of the game. I just feel like that’s the most intuitive way to talk about how we design studies like this. But what does this game look like and how is it set up such that we’d be able to tell at the end whether you were encoding this stuff strictly speaking as just like, I’m just regurgitating what you told me as opposed to having extracted some gist-like idea of who you are as a person.
Leor Hackel:
Yeah. So what we did was as follows. We told subjects that they’re going to be learning about other participants who had played a trivia game where they could win points worth money. They learned though about two groups in separate blocks or separate stages. One group, they learn about everyone’s pretty incompetent, so they get 25% of the points available on average except for one moderate target who does a bit better, they get 50%. In the other group, they learn about everyone’s pretty competent. On average, they get 75% on average except for one moderate target who gets 50% on average. Now, after they have learned about everyone, we give them a decision phase where we say, “Okay. You’re now going to hire people to play trivia questions for you. On each round, you can choose which of these people you want to play with.” And of course what we’re interested in is what do they do when these two moderate targets who won 50% are pitted against each other? Who will you choose to hire?
If you are remembering the exact details of what happened, you would remember that well, each of these people got 50% on average, they are identical, and there’s really no reason to have a preference for one or the other. I should just flip a coin in my head. On the other hand, if you are moving towards the simpler representation of relative gist, now you do something different. One person you learned about was a positive deviation from their group, so they got 50% when everyone else got 25%. So your simpler representation of them would just be they’re better or they’re relatively more competent than those in their group. The other 50% person you learned about was a negative deviation from their group. They were doing worse than the people you learned about with them. And so they would’ve gotten encoded as, oh, here’s the bad one. They’re worse than the group. And so now when you’re asked to choose who you want to hire, you should have this relative preference to choose the person who looked relatively good next to a bad group as opposed to the person who looked relatively good. I’m sorry, relatively bad next to a good group. And that’s exactly what we see. So when given the scenario in this way, subjects have this really, really robust preference to choose the person who looked relatively good over the one who looked relatively bad.
Now, so far, that only addresses the relative just component, how people relatively stack people against one another. But again, we thought, well, if people can even go further and categorize others in simpler categories, they should do that. So in a couple studies, we now added in the opportunity to do that. For example, in one study we told people, all right, look, anyone who gets zero to 40 points, that’s pretty poor performance. Anyone who gets 40 to 60, that’s moderate, anyone who gets 60 or above, that’s excellent. The idea is that now when subjects are seeing these players, they no longer have to rank them in their heads as this is the relatively good person, this is the relatively bad person. They can just categorize them as, oh, this person’s bad. Oh, this person’s moderate. Oh, this person’s good. If that’s the case at the end, we should now be able to completely eliminate this relative preference. So at the end, there’s now no reason to prefer one of the moderate targets over the other because they were both in that middle moderate category. And so you can categorize them the same way. And so that’s exactly what we see when we now give people the opportunity to move to that simpler representation, that’s what they do.
The last thing I’ll just mention in describing this is that’s heavy-handed. Explicitly telling people, Hey, use these categories of poor, moderate, excellent. So we also tried this with more naturalistic categories where they’re learning about investors who had invested in stocks. In one condition the investors gained small, moderate, or large amounts of money, and so again, you have to track their rankings. Who is winning more than somebody else. In the other condition though, they crossed category boundaries, they either lost money, broke even, or gained.
The idea is, again, that lets people move towards this simpler representation. Instead of tracking, oh, this person did better than that person. You can just label people. This person keeps losing money, this person keeps gaining money, this person keeps breaking even. And so using that scenario as well, we saw the exact same thing. When people can only use the relative ranking frame to simplify what’s happening, they show this relative preference for someone who is relatively good in their group versus someone who is relatively bad in their group. But when people can assign these categories instead, that relative preference completely disappears.
Andy Luttrell:
The main takeaway I’m getting from this is I should just surround myself with less competent people and just wear a badge that says, “Trust me. Put me in your highly competent group.” So one of the connections that came to mind was there’s this inspirational poster saying that’s like, people don’t remember what you do, they remember how you made them feel. Something like that. It strikes me as this, but with an important caveat based on the first round of things we were talking about, which is on the one hand we go, yeah, I’m not really paying that close attention to exactly what you’re doing. Like I am, but when it comes to the choices I end up making or what I ended up remembering later, I just remember this vague sense of like, oh, you were one of the good ones.
But it’s also a little bit different than purely how you made me feel because as you were talking about at the beginning, that’s only part of the equation. So you can make me feel a certain way, but I’m still on guard for who you are as a person. So I’m paying attention to the broader context, not just, oh, you made me feel nice inside, but I’m also sussing out what were the options you had in front of you, and who is the person that you seem to be in an enduring way? And that’s the trait level, which also is gisty. But I just think we need to make our own inspirational poster that just has a slight revision.
Leor Hackel:
Yeah. Just surround yourself with people who make you look good relative to them.
Andy Luttrell:
Yep. So why do you do this? You could study anything and it’s clear that it’s very cool stuff. But I wonder what is the thing about this that keeps these questions seeming interesting to you?
Leor Hackel:
I think there’s two things. One, I’ve just really always been interested in integrating cognitive work and social work. Trying to integrate different domains. And I think trying to bring together literatures on memory and impression formation or reward learning and interaction. To me, I think there’s something nice about trying to bring these together to say, how can we get these integrative insights that put together things we know from different domains into some broader picture? That’s scientifically why it excites me. I think the more practical answer is we want to understand decision-making, which I think we do. I think the way that people make social decisions, it’s really important. We want to know when somebody hires someone, what is going into that decision? Or when thinking about social connection and loneliness, when somebody decides to approach someone else versus just withdraw and think, no, they’re not going to like me, what is going into those decisions? What computations go into that and what things should we be on guard against or what things might we want to encourage or promote? And so I think if we really want to think about decision-making in interactions, we really need to understand learning and memory and interactions because ultimately our decisions are going to come from some kind of memories that we formed during learning, whether they’re instrumental reward, episodic details of what happened, or some really abstract gist.
Andy Luttrell:
Do you think this perspective applies just as well to how we think about ourselves? So it strikes me that something like my self-esteem is a little more rewardy. Something like my self-concept is a little more tradie. I’m paying attention to how I act in the world, but I may not remember everything that I’ve done, but I still get the sense that there’s a me in here and it is this kind of person. Do you think the same thing applies if we just turn the lens on ourselves or is that a different process?
Leor Hackel:
I do. And I will say a PhD student in my lab, Jean Luo is working in part on this question, so stay tuned. But there’s this work from John Kihlstrom, Stan Klein, and then Jenny Beer as well, from I guess the 1990s and early 2000s that really made the argument that our knowledge of our own traits is distinct from our memories of specific episodes. So you can have a neuropsychological patient who has some degree of amnesia, and they cannot tell you about things that they have done in detail, but they can accurately tell you about traits they have. I’m punctual, I’m caring, and their ratings will correlate with the ratings of people who know them well. And so I think that coheres nicely with the idea that there’s this distinction between episodic memory and gist memory and that we might think about ourselves in terms of gists. And in some of the work that Jean is doing she’s interested as well in this question of reward and how that can influence our self-concept as well, through, again, those affective experience of doing well or not.
Andy Luttrell:
Does any of this change the way you interact with the world? Knowing this is how we do things do you see people differently or present yourself maybe a little more strategically?
Leor Hackel:
That’s a hard question. I try not to have it change how I think too much about the world because I think, yeah, I don’t know. I think that getting too caught up in thinking about this sometimes you don’t want to be analyzing social interaction that carefully and strategically. You want to engage with it as it is. I wouldn’t say directly. What I would say though is I am interested in work that moves towards thinking about some of the applications I was mentioning for what can we do in decision-making or social connection. Once we have some good answers from those kinds of intervention studies, I’ll be happy to apply some of that.
Andy Luttrell:
For me. I wonder as a perennial people pleaser, if some of this is like that’s not the only thing. Also, you can also be attentive to what is this saying about me? There’s the potential that by completely copying to what is going to make someone feel good in the moments, you may actually sacrifice some perceptions of competence. You might sacrifice some other kinds of things that might be down the line, like useful things for people to think of you. And obviously you want to balance it. I’m not saying like, oh, don’t try to make people feel nice. That’s irrelevant because it’s not irrelevant, but it’s not the only part of the story, which I find is not the model that I maybe would’ve intuitively come to.
Leor Hackel:
That’s a great point and I wish I had said that. Yeah. I think that’s a great point. And just to add, not that it’s how you make people feel isn’t the only part of the story, in many situations it’s not even the main part of the story, it’s the smaller side part of the story.
Andy Luttrell:
Well, I don’t know that that’s the most inspirational way to end things, but that’s how we’re doing it. Just want to say thanks for taking the time to talk about all this stuff, and I appreciated learning about it.
Leor Hackel:
Yeah. Thank you so much for inviting me on.
Andy Luttrell:
All righty. That’ll do it for another episode of Opinion Science. Thank you so much to Leor Hackel for taking the time to talk about his work. You can find a link to his website in the show notes to learn more about him and his lab, and as always links to the original research we talked about Are there too.
To learn more about me, you can head to opinionsciencepodcast.com and follow the show at Opinion Sci Pod on Twitter. Subscribe wherever you get podcasts. And don’t forget that the website has links and transcripts for all the past episodes. But this outro feels short. Do I normally say more things at the end? Oh, review the show online. I usually say that. But I don’t know. I think this is all I have to say. I’m mostly just basking in the glory of the academic year being over where I have more time to do stuff like this. I’m also in the throes of a bidding war on eBay, which is a thrill I haven’t had since 2004. So maybe I’m a little distracted now too. But I don’t know. Thanks for listening and I’ll see you in a couple of weeks for more Opinion Science. Bye-bye.
