The Brain in the Social World

Thank you so much for that incredibly kind introduction. It's such a treat to get to visit such a unique place with so many creative and stimulating people with such diverse backgrounds, working on such fascinating questions. And I've had a really great time so far getting to hear about some of the really interesting work that some of you here are working on. And I'm looking forward to more of that as the day

goes on. And I'm excited to have the chance now in the next 45 minutes

goes on. And I'm excited to have the chance now in the next 45 minutes to an hour or so to share with you some of what my collaborators and I have been working on. And before I get started, I wanna be sure to thank the many people involved in the research that I'll be telling you about today, including members of my lab at UCLA and our collaborators on this work at UCLA

and also at other institutions, because everything that I'll be telling you about today is really a team effort. And the work that I'll be focusing on today, as Isabel alluded to, is all related to questions regarding how we represent and navigate and how we shape and how we're shaped by the real world social networks that surround

us. So in other words, the interaction between our cognition as individuals and the webs

us. So in other words, the interaction between our cognition as individuals and the webs of social relationships that we all inhabit. And I think that Understanding this can be really critically important to a lot of aspects of just everyday human social thought and behavior. Because you could take, for example, what might be some of the

simplest possible kinds of social encounters you could have in your day-to-day life. Like for

example, when you're walking down the street or across campus, for example, and you bump into someone and how you respond to this person. So how

much attention you might pay to her if you stop to say hello, the kinds of things you might say where you just stopped to say hello, even things like how much you might trust this individual, these things depend in part on aspects of her as an individual, right? So there are things like the knowledge you have about her and her preferences and her traits, your memories about her, but how you perceive

and interact with her is also going to depend on your relationship with her Right.

So one aspect of this is how close you guys are. Is this a friendly stranger, a loose acquaintance, a close friend? And then there's also the question of your respective social status and roles. So is this your boss? Is this a friend? And so on. But so, of course, it's not just characteristics of individuals in

friend? And so on. But so, of course, it's not just characteristics of individuals in isolation that are going to shape how we perceive and respond to each other. There's

also the question of our direct relationships with them. But it also gets more complicated than that, as I'm sure many in this room would appreciate, since we each have our own sets of social connections, who in turn are connected to each other and to others. And so we inhabit, of course, these networks of social relationship. And information

to others. And so we inhabit, of course, these networks of social relationship. And information

about someone's position in our social networks can matter for our behavior. So for example, some people might be better connected than others, and we might have to modulate our behavior accordingly, because now there might be more pronounced consequences for doing things like divulging sensitive information or for wronging them. And so

some scholars have even prominently argued that the cognitive demands that come with tracking and encoding and managing all the relationships that make up typical human social networks played an exceptionally significant role in the evolution of the human brain and much of what's distinctive about it. So this is arguably something that is

really important for us as a species. It's something that we seem to be concerned with a lot of the time just to we go about our day-to-day lives. And

relatively speaking, it's still something that we know pretty little about. So how

do we understand, how do we shape, and how are we shaped by the patterns of social relationships that surround us? And so to begin to try and get at these kinds of questions regarding the interaction between individual cognition and what processing within individual brains and people's social networks, we generally start by taking an approach that looks

roughly like the following schematic. So we get everyone in some bounded community. So sometimes

that's a school, sometimes that's a dorm, sometimes that's an isolated village. And we ask everyone we can in that community who their friends are typically using more precise criteria than just straight out saying who are your friends, but this is just shorthand that fits easily on a slide. And then using this information, we can map out the

friendship network of this set of people. And then this allows us to get a rich set of information about these people without having to ask them much beyond essentially who their friends are and without having to necessarily clue them in regarding what the rest of our studies are going to be about. And then we can glean things

from this network data, of course, like who's really well connected and how we look at that in a lot of these studies more in a few minutes. And

in human social networks, this is thought to be associated with things like protection from maltreatment and protection from scapegoating and social status more generally, as I'll touch on later. And of course, we can also tell who tends to bridge or broker between

later. And of course, we can also tell who tends to bridge or broker between otherwise disparate sets of people. So people who might be those sort of social chameleons who might participate in lots of distinct social circles. And we can also discern the distance between people in the network. And these are just a few of the kinds

of pieces of information that we can easily extract from this data that might be behaviorally relevant to the people within the network when they're interacting with each other.

And then we can have some or all of these people come into the lab for behavioral or neuroimaging studies where we can ask questions like, how much do we know about other people's positions in our social networks? And do our brains retrieve this knowledge when we encounter each other to help us sort of gear up for appropriate and beneficial interactions so that this information can, in other words, beneficially shape how we're

going to think and how we're going to act? Also is how we process the world related to how our social networks are organized. And we can also examine shared understanding in social networks and how this might relate to people's overall levels of isolation or social connectedness within the communities that they inhabit.

And so I'm going to begin with some research that falls within this first family of questions. And here we were interested in asking, as I just alluded to,

of questions. And here we were interested in asking, as I just alluded to, how much people know that where others sit in their social networks and if their brains retrieve this knowledge spontaneously when encountering one another, seeing one another to help them gear up to behave in an appropriate and beneficial way. And there's

reason to think that this might be the case. So there are other aspects of people's knowledge about others that are familiar to them. Like for example, knowledge about others personality traits that are thought to be spontaneously activated whenever people think about or encounter those individuals and the purpose of doing this is thought to be

helping people sort of gear up to think and act in an appropriate and helpful way to them. So this allows us, it's thought, to kind of often seemingly effortlessly shift gears depending on who we happen to find ourselves around or who we happen to encounter in day-to-day life. So we can act appropriately, retrieve relevant information, and

so on. What is spontaneous? Spontaneous just means you're not asked explicitly or

so on. What is spontaneous? Spontaneous just means you're not asked explicitly or instructed to retrieve this information, but just When you're presented with, for example, someone's face or you see them in real life, this just happens without explicit instruction. Yeah,

it's a great question. Yes. How fine and grained is that in coding with personalities, for example? Is there a big category? What's known about that? Yeah, so a lot

for example? Is there a big category? What's known about that? Yeah, so a lot of the work has been a little bit coarse because some of it's sort of older work and social psych doing things like looking at characters that people learn about through things like vignettes. or sometimes people who are like famous figures, and then looking at reaction time to different kinds of semantic

content when that person is presented and looking at if the presentation of or in the encounter with particular people sort of makes that relevant information more readily accessible. Yeah. I'm not sure more granular research into that, but there might be more recent work that I'm not aware of. It's a good question.

And so we know information about personality seems to be sort of activated by the brain when we encounter or think about familiar others. But also

information about social ties, like whether somebody is a friend or a friend of a friend or characteristics of their position in a broader network, like how well connected they are. These things also seem to inform our thinking and our behavior.

are. These things also seem to inform our thinking and our behavior.

But while there's been a fair amount of work contrasting things like how we respond to friends versus strangers or people who are familiar to us in some sense versus people who are totally unfamiliar to us, there really isn't much known, relatively speaking, about how we encode and are affected by more specific social relationship information about

people within that spectrum of familiar others. And this is despite the fact that these are the people like friends, coworkers, neighbors, family members, and so on, with whom a lot of our significant everyday interactions take place. And so to begin to get at this, in this first study, we took all the first year students in

an MBA program at a small institution in a rural area. So the cohort was about 300 people. And we characterize their social network simply by asking them to answer this survey question about who they socialize with most often as part of an activity they didn't know would be related to an FMI study they'd be recruited

for a few weeks later. So the question was, consider the people with whom you'd like to spend your free time. Since you arrived at this institution, who are the classmates you've been with most often for informal social activities such as going out to lunch dinner, drinks, films, visiting one another's homes and so on. And it will spend a little more time unpacking some of the methods for this first study, since a

lot of the same methods and concepts come up in the other studies I'll tell you about later. So here people essentially just name their friends using this way that we operationalize friendship. And then using this information, we could map out the network of the cohort. So all the nodes here are people, the lines are friendships between them.

the cohort. So all the nodes here are people, the lines are friendships between them.

Of course, the properties of people's positions in such a densely connected network are pretty hard to see meaningfully when visualizing them in this way, especially showing all the nodes and all the edges. It just looks like a big ugly hairball graph. So when

visualizing this here, it's really just meant to illustrate the procedure of the study rather than anything about the network itself or the characteristics of people's network positions that we actually care about. And so, What's the size of the null signifying? Oh, this

is proportional to their eigenvector centrality. Yeah. And what's the mean degree of this? The

mean degree, I don't remember offhand. I think it was something. It

looks like a page or something like that. I don't want to misstate it, but I think it's something in that vein. Yeah. I think it was something around that.

I can, yeah. And so in order to actually highlight some of the aspects of social network position that we're interested in, I'm just gonna use a simple toy network of about 13 people. And I'm gonna talk about things very generally, since my understanding is that there's people here with a broad range of backgrounds and interests, but I'm happy to chat more detail afterwards with anyone interested. So one important consideration that

you might have as someone in this network is other people's social distance from yourself, right? So this is pretty easy to get at in this unweighted network. So it's

right? So this is pretty easy to get at in this unweighted network. So it's

just like playing the six trees of separation game, of course. So for any participant like Mary, we can see who she's one degree away from like Tim, she's two degrees away from like Luke and play this game for everyone in the network and map out how close or far they are for Mary. And you can imagine how this information might be relevant to Mary's behavior, right? So the actions and mental states

and emotions of people who are closer to her might be more relevant to Mary than those of people who are not directly connected to her or to her circle of friends. And so she might be more likely to tune into them. And we

of friends. And so she might be more likely to tune into them. And we

can also extract other important aspects of sociality that perceivers might care about from the pattern of ties in this network too, like how well connected someone is, well connected to others. And this might be especially important for humans because for us, our power

to others. And this might be especially important for humans because for us, our power to influence other people And things like reserves allocation and conflict resolution, these things don't just depend on who's most physically formidable or strongest or dominant, or who is most prestigious or knows the most. If you have a lot of people who support you, who in turn have a lot of people who support them, you can

gain a kind of social power or status. So here we operationalize one dimension of social status in terms of eigenvector centrality. And in this context, the people who are high in this facet of social status or sort of big fish and big pond socially like Bob here. So even though Bob and Tim both have four friends, Tim's

status is lower in the sense because his friends aren't as well connected.

And a third important aspect of sociality that we can extract from the pattern of tides in this network that people might care about as participants of the network is who spans gaps between otherwise unconnected parts of the network. So people who occupy brokerage positions. So people who connect, connect

brokerage positions. So people who connect, connect people whose social circle don't otherwise overlap. So you could think of a person maybe like Tim here, who perhaps in high school might have been friends with those and lots of different cliques that weren't themselves connected to each other. So he'd be in

a minimally constrained network position since his immediate contact contacts aren't really tied to one another. In contrast, Lisa's friends are all already friends with one another, so she's in

another. In contrast, Lisa's friends are all already friends with one another, so she's in a more constrained position. And so compared to Lisa, Tim's more likely to be exposed, of course, to diverse information. He's more likely to be a kind of social chameleon who adjusts his behavior to fit different social contexts. So a high self-monitor in the

way that people often discuss it in psychology circles. And so we can characterize how constrained everyone is in the network. We all vary, of course, in how well connected and constrained we are, and we're closer to some people than we are to others.

And so together, these features describe aspects of our positions in a network. And so

we can take everybody in this network, characterize them in terms of these three attributes.

And as I said before, this allows us to get a rich set of information about these people without having to ask them all that much beforehand beyond essentially who are your friends, without having to clue them into what the rest of the studies are gonna be about. So here we took some people from this network. So the

nodes that you see in Peachier, have them come in for an fMRI study. And

for each fMRI study participant, we can look at the network from their perspective and see who's one, two and three degrees away from them in the network. And we

tailor made a stimulus set for each participant that had the two highest and lowest status as we're defining it here, individuals at each of three distances from them in the network. So the two smallest and largest nodes at each distance represented by the

the network. So the two smallest and largest nodes at each distance represented by the silhouettes on the left here. And for everyone in each stimulus set, we had a video of them saying essentially the same thing, which we just showed participants in the scanner as they watched videos like this with a fixation cross in between, which allows us to separate out the neural response to teach person. And so this is kind

of meant to approximate the kind of sensory experience you might have when you encounter someone out in the world. And participants just had to press a button whenever they saw the same person twice in a row to make sure that they're staying awake because one issue that sometimes comes up in FMRI studies is that they have to be pretty long since you need many repetitions of each condition given the noisiness of

the data. And it's not always the most exciting paradigm and you're laying down, you

the data. And it's not always the most exciting paradigm and you're laying down, you have a blanket, there's like a soothing rhythmic noise, people fall asleep, so we need to task to make sure that people are awake. So they're just monitoring, am I seeing the same person twice in a row? And then after scanning, we had a questionnaire that probed how they perceived those aspects of others' network positions that I went

over earlier. And quickly looking at that data, so here each line is a participant,

over earlier. And quickly looking at that data, so here each line is a participant, perceptions are on the y-axis and reality is on the x-axis. And in all three graphs, you can see that perceptions and reality are quite positively related. So as

you can see, people seem to know a fair bit about where others sit in their networks, but it is knowledge retrieved when you encounter these people.

And so to look at this, we use what's called a searchlight representational similarity analysis approach. And all that means is that each point in the brain, we look in

approach. And all that means is that each point in the brain, we look in the local neighborhood of a sphere centered on that point, and we get the local response pattern to each person in your stimulus set. And that's what's represented by these rows here, sort of the multi-voxel patterns or the patterns across space in

each little region of the brain. And then we compare those patterns to each other to make a neural similarity structure that captures the distinction that this brain region makes.

So essentially how it organizes information as some people describe it, but really what it treats is the same, what it treats is different and by how much. And then

we can go back to the network data and look at the 12 people in each participant stimulus set and just construct similarity structures based on those three measures of sociality that I went over earlier And because now we've summarized the patterns from the brain data and the social network data in a way that's abstracted away from how we've acquired it. So they're just these 12 by 12 matrices indexed by person in

your stimulus set. We can do things like take the neural similarity structure at any given point in the brain and just model it as a weighted combination of the three similarity structures made from the network data. And so we can do things like look in the brain to see what kinds of social network information about the people you're seeing is reliably encoded in different brain regions. And so here

the colors signify what seems to be reliably encoded in a given area. And without

getting into too much like really wildly speculative reverse inferences about what kinds of mental processes must have been evoked given the location of the neural responses, Some of these results point to testable hypotheses about the underlying processes that might be differentially triggered or influenced by social network knowledge. So, for

example, information about others' eigenvector centrality seems to be carried in brain regions that, among other things, are associated with mentalizing or keeping track of others' apparent mental states. And in some newer work led by my former grad student, Miriam Schwick, we see a similar set of results when people are viewing the

faces of members of fictive networks that they learn about in studies before going into the scanner. And I just mentioned that because although it's nice to get the naturalism

the scanner. And I just mentioned that because although it's nice to get the naturalism that we can get from real world networks, it's helpful to kind of complement that with more controlled studies where we can separate out people's network knowledge from things that might co-vary within the real world, like for example, other kinds of person knowledge, trait inferences based on what people look like, personal familiarity, visual familiarity,

visual characteristics, and so on. Yeah, I think here you heard it first, and then...

So you ask people about other individuals that are those three steps thrown into the network. Yeah. The assumption is that I actually know something about the eigenvalues, the eigenvalues is probably what appropriate levels of people that does.

Yeah, we do see some level of accuracy when we ask people to estimate. We

give them like a lay definition and then we ask them just to estimate for each of these 12 people, relatively speaking, how high or low. And

are you, in measuring the correlations to brain activity, are you control for how far away they are? Is there one or two or three? Yeah, so we're using, you mean the distance from the perceiver? Yeah, so we remove that, the variance accounted by that from the other, from the similarity

structures for eigenvexterous neutrality and brokerage before entering it into the model. And so the results that you see for example, in orange,

the model. And so the results that you see for example, in orange, our eigenvector, where people who are more similar in terms of their eigenvector centrality, evoke more similar neural response patterns over and above what could be explained by their similarity in distance, since people who are more central

or might tend to be closer to each perceiver, for example. Okay, I just, I think, good. So, you see, you're, controlling for the fact

think, good. So, you see, you're, controlling for the fact that people with higher eigenvector centrality do tend to have higher degree, and people who have higher degree are more likely to be within one step of that. Yeah, so we try and control for that statistically. To an extent, we

that. Yeah, so we try and control for that statistically. To an extent, we try to, as best we can, get examples of higher eigenvector centrality individuals that are three steps out, but as you imagine, like, even the people with the highest, for a given perceiver, who have the highest eigenvector centrality distance from

them still tend not to have as large of values as the highest eigenvector centrality individuals, like one degree from them. And so, or one distance from them, so. Okay, and I guess how we just, I mean, a little string going a

so. Okay, and I guess how we just, I mean, a little string going a lot of, they're all examples where you can show that you do, most of the things that Yeah. But in a lot of typical networks, all of these things are strong. Sure. Yes. You know,

I would worry a little bit that people know that somebody has a lot of friends and that that's the main thing they know. And then that's correlated with all of these things. Yeah. So the reason that we use eigenvector centrality here rather than degree was that we weren't

sure if people would necessarily know the degree centrality people who are far from them in the network, but maybe they would have a better sense of like who's well connected to well connected others versus like who isn't. So like maybe you'd have a better chance of

knowing this aspect of centrality for people who are far from you than information that's just based on someone's local ties. who are really far from you. But as

you suggest, it is the case that obviously is really correlated with Indigree. And so it's, I don't,

Indigree. And so it's, I don't, if I remember correctly, it's been a few years.

We get more extensive results using eigenvectors centrally, because I think a reviewer asked about it, but I don't want to misstate that, so I can double check your question. Yeah. Just a clarification. For your training data greater

your question. Yeah. Just a clarification. For your training data greater than zero, then you're only looking for increases in blood flow? It's not increases in blood flow because we're not looking at where does seeing someone, for example, higher and higher vector centrality evoke a greater response, but rather where are two people

who are really different in terms of their eigenvector centrality, evoking really different neural patterns. So because we're the...

neural patterns. So because we're the...

How different are the neural patterns? Because we want to capture the information carried in the spatially distributed patterns of activity and not just like the mean activity within a region. So this is not showing you where you get like greater

region. So this is not showing you where you get like greater responses as beta goes up, but rather we're having more similar evoked patterns is related to having more similar values for each of these three variables.

Yeah. I'll go very short, two questions about this. Apparently I can't see any kind of overlapping regions in their table. So like, did you, just chose the highest one that did happen in any overlapping regions or could've treated to like it? Oh yeah, you can't see, there is some overlap, but because of the way

it? Oh yeah, you can't see, there is some overlap, but because of the way the opacity is appearing here, you can't see it. But then also because, so because we remove the variance accounted for, or we like orthogonalize the predictors before putting them into the model, sort of like to reduce the, overlap, but if you just look at the pairwise correlations between the neural similarity structures and the

social network-based similarity structures, you see kind of similar results, although you do see more overlap between the orange and the purple. So you see more extensive purple, essentially, if you were to just look at the raw social distance similarity structure. And among all of these coefficients, were there any

strong, high magnitude negative position, like a dad is super high magnitude. I see that also implies something.

Yeah, we didn't see that. I wonder, it could be because, like, for example, I could imagine you might, like, so you could have two faces in the stimulus set that have, um, two pairs of faces in the stimulus set that both have a difference in social distance

relative to the perceiver of zero, like two distance one faces and two distance three faces. And for, you might see, if you're thinking about where would you

faces. And for, you might see, if you're thinking about where would you see like more different patterns for more similar social network information, perhaps it's like you individuate more amongst people who are closer to you, you might see more, more specific, stimulus specific patterns for people who are one degree away from you

compared to who are both one degree away from you compared to people who are like one versus two, one versus three and so on, or people, but within three that might be different, right? So it could be that this is just that there are other relationships that we're not capturing here because some of the

distinctions aren't, that are important for the perceiver aren't fully captured by the model we're using. Yeah, that's a good point. Okay.

Okay. And so just the fact that the eigenvector centrality or also the degree centrality of someone who you're encountering seems to be encoded with regions that are involved in things like making sense of other people's thoughts suggests the possibility that being well connected to well connected others has consequences for how much you pay attention to what other people seem to be thinking. And I'll just quickly tell

you about some work where we're testing this notion that was suggested to us from the neuroimaging results that I just showed you. And so one simple way to examine how much people are paying attention to others' mental states is with what are called gaze-curing paradigms. And these paradigms are very simple. The participant's job is simply to watch the center of the screen and indicate whether a target shown by this T here

appears on the right or the left side of the screen. And the only catch is that right before the target appears, a face appears gazing towards the right or the left side of the screen. And even though the direction of the face's gaze doesn't have any predictive value for where the target later appears, and participants can even be made aware of this beforehand, people are consistently faster to correctly

respond about the target's eventual location when the cueing face had just been looking in that same direction So on the valid trial shown up top here compared to when the face had just been looking in the opposite direction. So an invalid trial shown in the bottom here. And the gaze-cuing effect, so you being faster to respond correctly about the target's location on valid compared to invalid trials, arises because people

automatically track and shift their attention towards where others seem to be attending. So

what other people seem to be thinking about. And so the strength of this effect reflects how much you're attending to the apparent mental state of the cueing phase.

So if we do attend differentially strongly, the apparent mental states of really well-connected people in our social networks, as those FMI results that I just showed you might suggest, then those people's faces should exert stronger gaze-cuing effects. And this is exactly what we found so far in some work where we characterized the social network of a different group, had as many of them as possible come in to get

photographed to be stimulated for a gaze-cuing study where the participants were their fellow group members shown in peach here. And I'll just quickly show you some of the key results. So the gaze queuing effect is on the y-axis. The centrality of the queuing

results. So the gaze queuing effect is on the y-axis. The centrality of the queuing face is on the x-axis. And as you can see, the faces of better connected people are exerting stronger gaze queuing effects, which fits with past findings that have found that the extent to which you pay attention to others' mental states or social

attention, in other words, is modulated by other ways that people often operationalize social status, like the apparent masculinity or physical formidability of a queuing face, arguably because people who are high in those sorts of traits are often more behaviorally relevant for perceivers, which makes their mental states and their behaviors more important to predict and

monitor. And so here, oh, yes. I

monitor. And so here, oh, yes. I

interrupted you. That's fine. When I think about paying attention to the mental state of others, I think of things like they're being angry or preoccupied or something like that. Now, if I were doing your test, I would say, I'm trying to find where the T is, and the other person's looking over there, so I'll look over there. And that's what you've shown, that it

begates effect. Am I really attending for mental state, the person whose

begates effect. Am I really attending for mental state, the person whose face I'm looking at? I'm just trying to find the T. I don't care about her. Well, the sense is that you are told beforehand that it's not going to

her. Well, the sense is that you are told beforehand that it's not going to help you to look at the face and they're not even allowed to look over there. So they're fixing the whole time. It's just their covert attention is transiently shifted

there. So they're fixing the whole time. It's just their covert attention is transiently shifted towards where the face had just been looking. And so the idea is that people and also other primates just reflexively track where others seem to be attending and often pair that, as you're saying, with tracking others' emotional states and to see like, oh, somebody's scared about something that's over there. I better monitor what's over

there or this person keeps looking over there and glaring. Or maybe there's something really important in this direction, for example. So it's just a very, very, it's like the simplest possible way to look at attention to others' mental states as just looking at others' mental states without bringing in other information

where there might be more complex, like motivational phenomena entering the mix um so it's very pared down yeah in terms of what aspects of people's mental states were capturing it's really the idea is like someone's paying attention to something over there maybe they're thinking about something over there i care about what they're thinking about i'm gonna look there or divert my attention there even if

not looking there i don't care about it i don't know and I noticed that a member of my species is looking somewhere. My own visitors, I should probably look there because it would probably decrease my visitors to be watching out with everybody looking over there. And that has nothing to do with the animal's interest in the mental

over there. And that has nothing to do with the animal's interest in the mental state of the animal. It's just that where other people are looking at. Well, so

the idea- Probably worth looking. Just like our eyes pick up edges for a while.

Yeah. We tend to visualize motion rather than stationary things.

That's just a business trip. Well, so most animals actually don't track eye direction. Dogs

do, for example, and it's like visible in some animals, like where the iris starts to be able to have a cue to others sort of gaze direction. And it

often correlates with social cognitive abilities in animals because there is, there only is, I think, relatively small number of species that show these kinds of effects. Um,

And I guess the idea is that, yes, this is, I guess, maybe we're using the term mental state differently because I guess in a lot of psychology circles, just knowing where others attending, that is knowing about their mental state. So it's like their internal state, like where they're directing their attention. And so there is like obviously richer psychological information as well that we often use when we use the term psychological state

in other contexts. But I guess here, that what you're describing counts as someone's mental state. And I guess, yes, there is sort of like, just there are very basic cues in like someone's eyes that provide this information. But the idea is even if you keep the cues constant, for some

this information. But the idea is even if you keep the cues constant, for some people, some cueing faces, you're gonna show this to a greater extent because you care more about where those people are attending. So for example, If you show these pictures to strangers or people who don't know these individuals, they won't show the same modulation of the gaze-cum effect by centrality.

So it's like the knowledge of someone's centrality that's causing you to pay more attention to what those people seem to be attending to. Yeah. How much individual difference is there in this effect? You mean at the level of the perceiver? Yeah,

that's a great question. That's something we're collecting data on now because we need a greater sample size in order to look at individual differences. So we're actually collecting data right now to get at that because that's something we're really interested in. There's some

really interesting work in primates from, in racist macaques from Michael Platts group that shows, for example, so they show that the dominant the position the dominance hierarchy of the cueing face modulates the gaze cueing effect for macaques and interestingly for the lower ranking perceivers there

it works for all cueing faces or I mean their their social attention is guided by all cueing faces more so by high status macaques but for the high status perceivers, they don't actually follow the gaze of the low status queuing faces. They

only seem to depend to the parametral states of the other high status individuals. And

so it'd be interesting to see if we see parallel effects in humans with this other facet of social status. And so, yeah, here we're seeing kind of parallel effects to some of those effects have been observed with other facets of social status in humans and in other species. And I think this makes sense

given that well-connected people are especially impactful when it comes to things like spreading information, setting norms, shaping reputations. And so especially for people who for better or for worse, we as psychologists and neuroscientists often study, so often it's students on college campuses, And people who have relatively peaceful everyday lives, so for

whom physical formidability, for example, might be less behaviorally relevant on a day-to-day basis, how well connected others are might be a facet of social status that might be more chronically behaviorally relevant in everyday life. So this is just one example of how social network knowledge, once retrieved by the brain, might shape downstream processing and behavior. Yeah. Well,

we should probably hold up mostly present to get along. Yeah, we should let a little apart from the doctor. Sure, okay. Yes, so

along. Yeah, we should let a little apart from the doctor. Sure, okay. Yes, so

now I'm gonna pivot a little bit. We can also use this kind of general approach that I outlined at the start of my talk where we characterize the social network of abounded communities and recruit a subset of people to come in for in-lab studies ask other questions regarding how individual cognition relates to the social networks that people inhabit. So for example, if how we process the world is related to where

people inhabit. So for example, if how we process the world is related to where we sit in our social networks. So of course, many of us have that sense that, you know, we're friends with others who are like us, or birds with feathers flock together, as the saying goes. And as you guys, I'm sure, are well aware,

sociologists have confirmed this tendency time and again across diverse contexts, for example, when it comes to things like age, gender, and so on. But most of us have sense that our friends aren't just people who match us on sort of a surface level. It's not just, for example, demographic information that we match our friends on, like same age check, same gender check, and so on. But then demographic groups,

we of course resonate with some people more than others. Hopefully you have friends who fall into different demographic categories on your own. But until relatively recently, a lot of the evidence for homophily, of course, not all of it, but a lot of it had concentrated on some of these coarser variables, despite the perception that a lot of us have that our friends are people who we sort of click with. So who

respond to things like we do, you see the world from a similar perspective. And

so here we want to test this notion. and one window into how we respond to the world, so into our thoughts, our emotions, how we allocate our attention and so on, as these processes unfold, it's provided by neural response time series to dynamic naturalistic stimuli. So in some really elegant work done by Yuri Hassan's group and others,

naturalistic stimuli. So in some really elegant work done by Yuri Hassan's group and others, it's been shown that if you take a set of people and show them some engaging dynamic stimulus, like a movie, for example, and extract responses from corresponding brain regions as they watch it, they become sort of locked together in time, which is what this cartoon schematic is meant to show. And the similarity of these responses,

depending on where in the brain you're looking, of course, captures things like similarities in interpretations, emotional reactions, what people focus on, remember, and so on. And so if we are exceptionally similar to our friends and people who are closer to us in how we attend to interpret and respond to the world, we'd expect neural responses to rich real world stimuli to be exceptionally similar among friends. And so to get at this,

we characterize again the full friendship network of another MBA cohort, had them come in for an fMRI study where they just watched a set of videos while we scanned them in the MRI scanner. We had a few criteria for choosing the videos we showed them. So we wanted things that participants hadn't likely seen before to

showed them. So we wanted things that participants hadn't likely seen before to avoid similarities among friends just being inflated just because they'd watch the same things together before perhaps. We also wanted things that would be engaging. So hopefully they'd recruit a

before perhaps. We also wanted things that would be engaging. So hopefully they'd recruit a good portion of the cognitive and emotional processes that characterize our everyday mental life. And

also to minimize mind wandering, which could be a source of noise since different people could mind wander about totally different sorts of things when they're disengaging from something they find boring. So the hope was that this would minimize variability between people due to

find boring. So the hope was that this would minimize variability between people due to thoughts are just totally unrelated to the study. But we don't want to squelch variability between people completely. We also wanted videos that would evoke more meaningful variability between people because different people might interpret them differently or have different emotional responses to them or

attend to different aspects of them. So we wanted to sort of minimize that more uninformative aspect of intersubject variability by being novel and engaging, but promote more meaningful variability between people. So you had some things that some people might find really sweet and touching and adorable. Others might find a kind of cloying or sappy

or overly sentimental aspects of things that might appeal to people with a certain sense of humor, but not others might feel to appreciate or even detect the humor. In

the examples we showed them, we also had things that might be understood or parsed differently depending on one's existing knowledge set or interests. And we also had examples of arguments that might really resonate with some people, but evoke skepticism or even contempt in others. And just to give you a sense of the kinds of things that people saw in the scanner, I'll show you some brief excerpts of some

of the stimuli. And in the interest of time, these have been trimmed down substantially to just be a few seconds long each. But in the study, each clip was a minute or a few minutes long. When I was outside on my first spacewalk, I was on the dark side of the world over the Indian Ocean. By having

workers who only had to do one thing, they could pay them a low wage.

And it was very easy to find someone to replace them. The water squeezes out of the cloth. And then because of the surface tension of the water, it's... It's

insulting when a coach is making five to 10 to 15 times more than a college president. These babies are not strong enough to cling and they fall to the ground repeatedly. At times like that, you don't often think of funny things to say, but I did. And I turned to the rescue workers and I said, talk about a rough day at work. Talk... That was...

Referring to the spending the time up there being pretty rough.

Talk about a rough day at work. Okay, and so while participants watch these videos, we measured how activity across the brain rose and fell over time. On

top you see this activity summarized for two people who watch these clips. On the

bottom is an activity summarized for the circled region, been sped up a bit for visualization so I don't have to linger too long on this slide. But you can see that sometimes these people seem relatively in sync with one another. These two people also happen to be friends. And here we want to see if in general people who are friends and who are closer together in the network would have more similar

responses. And so for each pair of neuroimaging study participants, we just

responses. And so for each pair of neuroimaging study participants, we just divided their brains into constituent regions based on anatomy. So different regions are shown by different colors here. And then we measured the similarity of their response time series in corresponding regions as they watched the videos. If we can go back to the social network data and see how close these two people are in the network. And then

we can relate their social network proximity to their response similarity. And I'm just gonna show you some descriptive stats here to give you a sense of the overall pattern of results. And I should note that the results I'll show you here are all

of results. And I should note that the results I'll show you here are all controlling for all the demographic variables that we're able to get our hands on. So

here, warmer colors will mean more similar than average for a particular brain region and cooler colors will mean less similar. So for people, three degrees of separation in the network, we're seeing that are cooler throughout much of the brain, suggesting that these people are responding pretty dissimilarly to one another. The opposite seems to be true for distance

two dyads. So here we're seeing warmer colors suggesting that these people who aren't friends

two dyads. So here we're seeing warmer colors suggesting that these people who aren't friends with each other, but are friends with one another's friends are responding to the videos more similarly than pairs of people farther removed in the network. And then if we look within friends, we see warmer colors still in many regions indicating that they're processing what they're seeing in an exceptionally similar way And the story does get a little

more complicated. I'll note when we look at the, I think three or four dyads

more complicated. I'll note when we look at the, I think three or four dyads that were four or more degrees apart in this network. And I'm happy to chat more about that afterwards with anyone who's interested, but just in the interest of time here, I'm just honing in on the three categories of dyads that made up the bulk of our sample. And we see this pattern is especially pronounced in regions like

this peripietal lobule circled here, which determine things like how we allocate our attention in our environment, where we see friends are especially similar, and then that decreases with distance, and also regions like the angular gyrus just below it, where past work has shown that similar neural response time series are associated with similar high level construals

of situations. And in some ongoing behavioral work, we're starting to see converging

of situations. And in some ongoing behavioral work, we're starting to see converging evidence that people who are closer together in their networks interpret situations more similarly.

So for example, if I showed you a photo like this and you interpreted it positively like a man anticipating a visit from loved ones, it'd be more likely that your friends would too compared to if you looked at this and saw it as someone looking at the window because he's worried about what's out there. And you can see that sort of in this set of graphs from some work in progress

where we're looking at behavioral response similarity on the y-axis and social distance on the x-axis. And in some newer work that's also in progress, we're also looking at what

x-axis. And in some newer work that's also in progress, we're also looking at what kinds of similarities in mental processing might account for the relationships that we're seeing between neural similarity and social network proximity in different brain regions. And so to kind of return to the question that I started this section with, do we see the world like our friends do? things that we do, but it's important to note that

all the data I've shown you so far has just looked at people who already knew each other. So there's this important open question of, do we become friends with people who are already like us and how they process the world? Or do we grow similar to and sort of align ourselves with our friends over time? And what

ways does this happen? And so to start to address these sorts of questions, we've been gathering some longitudinal data where we can try and tease these things apart. And

so I'll tell you about that next. I'll try to move quickly because I realize the time. So in a recent study led by two really talented graduate students, Lisa Shen and Ryan Hyun, because we wanted to look at if preexisting similarities in how people respond to the world predict future friendship,

we recruited people who were just about to move to a new town to start an MBA program. And the vast majority of them were about to move into university provided housing where their roommates and their locations would be randomly assigned as with the study groups with whom they would do most of their coursework. And so for most of our participants, we were able to scan them within a couple of days of

their arrival. So before they had a chance to meet many people in what would

their arrival. So before they had a chance to meet many people in what would be their new home for the next year or two, much less get to know them and substantially influence or be influenced by them. And just like in that previous cross-sectional study, we had them come in to the scanner and essentially watch a variety of videos. And then we examined how their neural responses rose and fell over time.

of videos. And then we examined how their neural responses rose and fell over time.

And then we characterized their social networks two months later. So once they had attended orientation, had a few weeks of classes and gotten to know and socialize with each other a bit. And then again, eight months into the school year. And so I'll just quickly highlight the things that we're interested in at each time point in the data set. So at time one, before people had much of a chance to meet

data set. So at time one, before people had much of a chance to meet and potentially impact each other yet, once you look at their pre-existing neural similarities. And

then at time two, So a couple months into the school year, we want to look at who initially became friends with each other. And then six months later, at time three, we want to measure the same thing, which once people have more of a chance to get to know one another, meet more people and maybe form more stable friend groups. And interestingly, we don't really see much of a relationship between preexisting

neural response similarity and friendship and social distance when we look at the initial friendships that form just a couple months into the school year. But when we look at the distances between people and the network, once everyone's had more of a chance to get to know one another, so eight months in, we see a pretty clear relationship that looks like this. where in many regions of the brain, particularly areas of the

default mode network, they're associated with subjective control and social cognition among other things. And also in regions of the brain that determine things like how we

things. And also in regions of the brain that determine things like how we allocate our attention in the world, as I alluded to before. And we see there's exceptional similarity among the people who later go on to become friends, but then less so among people who end up two degrees apart, and far less so among people who end up three degrees apart in the network, which was three was the diameter

of the network here. So here we're seeing that similarity in how our brains process the world around us, collected before we even meet new people is linked to whether or not we become friends and to how socially close we become in the future after several months have passed and we've had the chance to meet more people, get to know each other and maybe figure out who we're compatible with. And maybe,

is there just six minutes left? Okay, maybe I'll skip this. So essentially, we also can predict, or there's also, yeah, we can also predict which dyads will grow closer over time versus drift farther apart. And so people who initially show greater neural similarity are people who are more likely to

become closer over time in the network. And people who are very dissimilar are people who are later likely to drift farther apart, maybe because your initial relationships might be determined often by things like circumstance, like who you get sat next to or happen to start talking to your orientation, then maybe you find out your people as the

months go on. And then they'll just give you like the one minute version of the last section. So there's time for questions after, because I saw some hands up earlier. So everything I showed you in the second part of the talk is just

earlier. So everything I showed you in the second part of the talk is just all about the relationship between dyadic similarity and social distance, we can also look at the relationship between one's overall level of social connectedness and how normative one's responses are compared to one's peers.

And we find that people who are more central in the network process world in a way that's more reflective of group norms. And then interestingly, consistent with like the Anna Karenina principles of this idea that All happy families resemble each other.

Every unhappy family is unhappy in its own way. We see that the people who are central in a network tend to process worlds in a way that's very similar to one another, whereas the people who are less central each process the world in a more idiosyncratic way. So we receive really high similarity among pairs of people who are both central in a network, even if they're not close to one another.

We control for their... distance from one another, but the people who are less central in the network seem to be dissimilar, both from the central individuals and from the low centrality individuals. And then we see a parallel set of results. If instead

of looking at centrality, we look at subjective loneliness. And so lonely people tend to show responses that are quite idiosyncratic with respect to their peers and, um, I think this is interesting because a lot of times people who are lonely will report feeling misunderstood or like others just aren't on the same page as them. And

so this suggests that this isn't always a misperception or an exaggeration, but that lonely people often do see the world in a way that's different from those around them.

And that perhaps, you know, Being surrounded by people who see the world very differently from you yourself could be a risk factor for loneliness, potentially, even if you regularly socialize with many individuals and are friends with them. Because we see these effects even after controlling for how many people each participant socializes with. And

they're just more objective measures of social connectedness. And so just to zoom out and sum up, so hopefully what I've shown you today is that we seem to know a fair bit, but where others are. in our social networks. We retrieve this knowledge spontaneously when we encounter each other. And consistent with the notion that the reason we retrieve any kind of personal knowledge at all when we encounter someone is to help

us gear up to act and think in an appropriate and beneficial way. We see

that social network knowledge seems to gate social attention in humans. We pay more attention to what well-connected people seem to be thinking about. And we seem to be exceptionally similar to people who end up close to us in social ties with respect to how we process the world around us. And relatedly idiosyncratic responding compared to those around oneself is associated with social disconnection. And so I think that by combining these sorts

of often siloed approaches for looking at individual cognition from psychology and neuroscience on the one hand, with looking at people's social networks on the other, we can get into what I think are some really interesting questions, because if we really wanna understand social thought and behavior, we can't always treat individuals as just isolated units or even dyads

and expect that we're always gonna be able to tell the whole story, right? So

even if we're just looking at two people having a conversation or how they're perceiving or reacting to one another within a dyadic interaction, oftentimes what's going on under the hood seems to take into account the broader social networks that surround them. And so

with that, I just want to thank you for your time and attention and all the thoughtful questions people have raised so far. And I'd love to take any questions in the couple of minutes we have left or also in discussions throughout the day.