Lindsey Powell: Developmental Social Perception
June 11, 2014
June 11, 2014
All Captioned Videos Brains, Minds and Machines Summer Course 2014
Topics: Infant studies reveal non-verbal building blocks for social perception; looking time methods: preferential looking (what/who is more interesting/pleasing), (de)habituation (what/who is similar), violation of expected looking (what is surprising), anticipatory looking (what will happen next); infants’ ability to recognize faces and social categories (e.g. gender, race, attractiveness, language) and visual cues used to make these distinctions; infants’ expectations of imitation among groups (members of social groups act alike); active preference measures, such as whether infants prefer helpers or hinderers, and what visual cues contribute to this preference; with regard to individual actions, infants appear to use an inverse planning model to interpret actions, and a similar strategy may be used to reason about social interaction
LINDSEY POWELL: Hi. I'm Lindsey, Lindsey Powell. I'm a post-doc in Dr. Saxe's Lab at MIT, although most of the work that I'm going to talk about today is work I did while a graduate student at Harvard with [INAUDIBLE] doing mostly behavioral research with infants. So today, I'm going to talk about the developmental side of what Ken introduced, the topic of the behavioral study of social perception.
So just to recap, because I know we're still also [INAUDIBLE] connecting to these central themes-- what I've been communicated to the Thrust 4 goal is that we want to understand the content of the systems that support social perception. So people go around interacting with one another very intelligently all the time, but what's the material on which they're basing those interactions and how is that material structured and used to support social interaction?
So some of the things that we're trying to figure out how to answer are who is there? What are they viewing? Who's doing what to whom? And what does one person think of another person?
I think you guys got a lot on the developmental side of those first two questions from Laura Schulz while she was here. She talked a lot about infant's perception of agents and basic actions they might be engaged in on their own. So throughout the talk today, I'm going to pull on examples from the [INAUDIBLE].
I'm mostly going to try to talk about social interactions and relationships between people who they are in terms of their relationship to other people out there as well. All right. And also, what will happen next?
So clearly a lot of adults' ability to answer those questions and to act on the answers to those question comes from social experience. So you need to know how telephones work to understand what a person is doing when they're talking into this inanimate object. It's not going to be immediately clear to an infant what kind of social interaction is going on there.
But there are also a lot of building blocks in place in infancy, and we can study-- especially if the goal is to understand nonverbal social cognition-- their ideal population is to sort of understand what are the early building blocks of nonverbal social cognition in terms of understanding who people are, what their actions mean, and what they're going to do. Because really, that's all they can do anyway, right?
But it's not as simple as you might think. So you could say, OK, you could see some really simple examples of tasks that adults do. So for example, Melanie [INAUDIBLE] research where you show people still faces and have them make snap judgments of that person's sexual orientation. So maybe babies can tell all kinds of things from faces, but you can't put a face up there and ask them to respond in the dimension of do you think this person is gay or straight, right.
So while they're really good at perceiving nonverbal social stimuli, they're uncoordinated research participants that you can't instruct or extensively train the way that you could train essentially nonhuman animals [INAUDIBLE]. But the challenge is to figure out what kind of nonverbal social perception they're doing with the kinds of things that you can have them do in an experimental setting. And so a lot of what I'm going to do today is talk about the kinds of methods that people apply to understanding social perception in infancy, and then I'll give a few examples of the contents that's come out of those [INAUDIBLE] methods to illustrate each one.
So most of the methods that people use with infants under a year are looking time methods. So can you put up [INAUDIBLE] videos of animals engaging in social interaction? And you could potentially tell a lot by watching how this animals behave. When it comes to four month olds, you're not going to be able to tell that much by watching how they behave because they don't really do a lot yet.
Not only can you not give them instructions, even if you could, they couldn't necessarily carry out those actions. All they really do is sit there and look at the world. And so researchers should capitalize on that tendency for them to just sit and observe, looking for patterns in the way that they choose to observe the world.
So there's a few different ways that people use [INAUDIBLE] methods. One is preferential looking. So just put two things side by side and see which of the two the babies choose to look at. And there's sort of two things you're actually trying to assess here. One is whether they find one of things more interesting or more pleasing or affectively valued than the other, drawing their attention to it.
But implicitly, there's also something in there, which is the question of whether or not they can distinguish those two things at all. Write In different cases, you might be testing one of the sessions more than the other. So for example, if you show a baby two shapes that are very different in terms of their contour and color, it's highly likely that they can distinguish the two, but they very well might not have a preference for one versus another, so you won't get any difference in where they choose to look.
Alternatively, you could imagine-- it's hard to think of the right example for a baby. But for an adult, you could put two diamonds in front of them and know for sure-- a diamond and a fake in front of them-- and know for sure that they prefer to have the higher value diamond. Maybe the problem there is just that they can't tell the two apart. So preferential looking tests are alternatively testing either what is more interesting or whether or not the two things are discriminable for a baby.
A method that came up a little bit, I think, in Laura's lecture, although maybe not quite in the context that I'll talk about it today-- is habituation. So basically, what things are similar to one another in the world? When you show infants repeated stimuli or to even changing stimuli, you can see the extent to which they consider those things to be the same by how quickly their interest declines. Then you can also see, once their interest has declined, if you show them something new, whether it rebounds, suggesting that they see that new thing as being dissimilar from what they've been watching up to that point.
It's really a little bit to the kind of violation of expectation looking that we're going to talk about some time. So for example, I think she talks about a paradigm where there's an actor reaching for an object over and over again, and I'll go over this again in just a minute. But the idea is that once you've seen an actor reach for one of two objects, if you switch the location of the object, you can ask what infants expect that actor to do by measuring their looks to both of those two kinds [INAUDIBLE] that you show them next to see if they look longer at one than the other, the idea being whichever one they look longer at is [INAUDIBLE] unexpected given the introduction.
Other times, the things that are surprising are surprising by default. So she talks about an experiment with objects where there's a rotating screen going back and forth. And babies see that for a while, but really, this is just an introduction to what's going on in the scene.
What comes next is you put a box behind half of the rotation of the objects. The baby's over here, and you see whether they expect the screen to stop when it hits the box. Does it keep going flat to the ground? You can imagine doing that experiment without doing any kind of introduction at all. It might just be the case that from the get go, they have an expectation about what that screen should do when it hits [INAUDIBLE].
And based on that sort of baseline theory about the world, they're going to be surprised when something happens that doesn't match it, and that's going to cause them to look longer at something. So a violation of expectation can either just see what's surprising by default given your prior [INAUDIBLE] world, or what's surprising given what you've just seen an actor or person doing in a given situation.
A final looking time method that people use is anticipatory looking, and you can see how for intelligence, this is actually particularly one you might be interested in whether or not infants are capable of because what's really helpful in the world is to know what someone's going to do next so that you can anticipate it and interact with them in a way that's going to [INAUDIBLE] their actions so that you can coordinate behavior with other people. In retrospect, you have to coordinate by thinking about what people are going to do [INAUDIBLE].
Finally, infants do have some motor skills. They get better, especially as they get older. So there are a bunch of infant studies especially looking at social cognition given that it's an area of cognition that's all about interacting with other people and not just observing them. So there are some infant studies that use active preference measures, things like reaching for, calling for one individual over another, choosing a toy offered by one person versus another, choosing who help if you have to make a choice between who you'd like to help. And these measures are taken as kind of who infants like or trust.
So that's just the range of methods that I'm going talk about today. And as I go through each one, I'm going to give examples of research on social perception that has used each of those methods. I've tried to sort of keep themes throughout the different methods so that I'm talking a little bit about the same kind of social perception as it's been studied [INAUDIBLE], but it's also going to hop around a little bit. So feel free to stop me at any point if you have questions about anything. Here we go.
All right. So preferential looking. As I mentioned, this is just putting two things side by side and seeing which one babies look at more.
It's a relatively ambiguous measure in terms of what's driving the preference for one thing over another, but it can tell you whether or not infants can distinguish two things and what they find most interesting in the world. And so when it comes to social perception, one of the earliest challenges is identifying people or agents out there in the world that you should be paying attention to so that you can learn about them and interact with them socially. And people monitor newborn infants.
Even newborn infants tend to orient toward faces, but people started to use preferential looking as a way to figure out just exactly how are these brand new babies who haven't really seen people or agents and their faces very much at that point, how are they recognizing those faces and [INAUDIBLE] them out in the world? So what perceptual schema might there be that from two minutes after-- or probably more like 10 or 12 or 20 or 30 minutes after they're born given how [INAUDIBLE] cooperative their parents are-- what kinds of perceptual schemas do you have that are helping you pick out faces in the world that you can pay attention to? And so the general observation that people had started with is that if you make a schema of a face where that are sort of top heavy, that are more dark spots or points of interest in the top half of the face than in the bottom half of the face, babies will look longer at the one that's top heavy over the one that's bottom heavy, which roughly corresponds to a right side up face where you have two eyes over a mouth as opposed to an upside down face, where you have the reverse.
So the researchers said, OK, fine. Is that all it is? Is it just about top heavy images that have more stuff at the top than they do at the bottom? So what they did is they reversed the contrast of the images in a bunch of interesting ways to see how specific is this schema of a face that babies are drawn to.
So they did this whole series of experiments. But like I said, the first thing they did was just reverse the contrast. So there, they changed the white face to black eyes and mouth to a black face with white eyes and a white mouth and found that infants no longer showed a preference to look at that right side up face essentially versus the upside down face.
And then they waned to know, OK, well, how much of it-- because if you need several, at least two dark spots at the top half of the image on a white background, but how large does that white background have to be? Are they basically looking for eyes?
And so they touched the black overall [INAUDIBLE]. But they added-- and you can see in experiment 1C, they added pupils to the white eyes. And now they recovered the preference that they had before for babies to look at the right side up face versus the upside down face.
Similarly, they took real faces and they just reversed the polarity of the contrast from a normally [INAUDIBLE] face to a reverse contrast face and found that indeed, what infants are essentially looking for are eyes, sort of dark pupils on white backgrounds. And you can see these across a variety of versions of these face stimuli, that newborn infants tend to orient towards faces as defined by eyes on the top half of the image over upside down pictures that have the same visual features in general but in reverse.
So this is an example of how researches use preferential looking to figure out what infants, what social [INAUDIBLE] information in the world [INAUDIBLE] immediately. But it's also something they can tell you what they learn through experience. So we can't build everything in.
So one of the things about being born into very variable social environments is that you're going to have to learn a lot about the particular people in your environment, the language in your environment, all those kinds of things after birth and through experience. And so babies do start to see a lot of faces as soon as they're born, and researchers wanted to know, are they extracting anything about what a typical face in their environment is by looking at all those faces of people that they see? And so they wanted to see does the experience of looking at faces all the time alter face preferences. And in addition here, there's-- yup.
AUDIENCE: So it's impossible to extract [INAUDIBLE] variables. Can you draw the conclusion that babies are looking for eyes, but if you actually track their eye movement, did they actually [INAUDIBLE] focus on the eyes?
LINDSEY POWELL: So in those particular experiments, the researches didn't use eye trackers. There are eye trackers. They probably don't work particularly well with newborns who-- you have to orient a person's relationship to the eye tracking system relatively well. But certainly, with older children, people do often look at how face perception works in terms of where in the faces the participant focused.
Did they focus on eyes versus the mouth? Did they scan the face the way an adult does or not? Those particular studies were all just looking at one versus looking at the other and not tracking where in the face they were looking.
But I think the idea is that given the relationship of the contrast pattern that they prefer to actual human faces and eyes, the idea is that that's the best explanation for why that would be the pattern that they would prefer, that they would prefer dark spots on a white background. So it would be interesting to try to track and see if they focus in on the eyes with that particular study. Yeah.
AUDIENCE: Can you go back to the graph? [INAUDIBLE].
LINDSEY POWELL: Yes, that's the master.
AUDIENCE: So is that [INAUDIBLE]?
LINDSEY POWELL: So typically, what they'll do is have some set window of time. So I don't know exactly what the total maximum amount of time they presented the faces for was for these infants. And then they'll say out of that amount of time, how long did they look at one, how long did they look at the other, and how long did they look right? It's not a 90 second look in constant duration. It's just given some window of time to look at those two things, how long did they choose to look at one versus the other within that window of time.
All right. So does experience alter the kind of face preference that babies have early on? So people have contrasted, for example, how long infants looked at genders of face. And so here, I was going to point out that there's actually two questions that you really want to ask here given these infants are so young.
One of them is do they have a preference for one face over the other. The other is, can they even tell those two faces apart? Given what we know about infant face perception, you might be entirely sure that those two faces are going to be discriminable to a baby.
But what it turns out is not only do infants have gender preferences, they actually are experience based. So infants tend to look longer at the face that matches the face of their primary caregivers. If they spend all day long looking at a female face, they prefer to look at the novel female face. If they're taken care of by a man and all day looking at a male face, they tend to look longer at the novel male face.
So basically, it seems like they're drawn to whatever face matches the average of their face perception to some extent. This is supported by the fact that you can find preferential looking to own race faces over other race faces, but that is actually only takes place in infants who are raised in racially homogeneous environments. So it's not based on, say, who looks like Mom or who looks like Dad in particular and sort of identifying social [INAUDIBLE] groups.
If you have Ethiopian babies raised in Ethiopia versus in Israel, you find that the babies in the racially homogenous environment tend to prefer own race faces, but the babies in the heterogeneous environment actually don't have a preference for the race of their family over another race that they see frequently out [INAUDIBLE]. So again, the kind of faces that you're seeing on a daily basis by the time you're four months sort of tend to make you gravitate toward one of the faces over another.
Ken mentioned that attractive faces are basically average faces, the sort of center of what you see, and it turns out that babies have a preference for attractive faces by six months of age as well [INAUDIBLE] an attractive person next to an unattractive person. I love that these papers never ever give examples of their stimuli to protect the identities, I assume, of the unattractive faces that they use. But yeah, babies will prefer to look at the attractive person over the unattractive person by six months, suggesting that they're at least picking up on the variability in the faces that they see in the environment and forming some kind of perceptual schema of a real face as opposed to, say, that schematic little face with dots that they're using to guide where they look in the world.
AUDIENCE: I wanted to ask you, [INAUDIBLE] unattractive faces or really unattractive?
LINDSEY POWELL: Really.
AUDIENCE: So, I mean, we take it with a grain of salt. I mean, [INAUDIBLE] almost malformed. [INAUDIBLE] really weird looking [INAUDIBLE]?
LINDSEY POWELL: Got you.
AUDIENCE: Just a little bit of a caveat here.
LINDSEY POWELL: They may not even pass the test, the [INAUDIBLE] test, the beginning of having [INAUDIBLE].
All right. So it turns out that visual-- I mean, we're mostly talking about social perception and vision today, but I just want to point out that it turns out that preferences for faces at this age aren't limited to familiar looking people. They extend to familiar sounding people.
So [INAUDIBLE] actually the [INAUDIBLE] lab, where I did my graduate work, contrasted babies' preferences for people who had previously been speaking their native language versus a foreign language and found that infants preferred to look at the face of a person who had been paired with their native language over one who had paired with their foreign language.
So one possibility is that these kind of preferences, being drawn to people who have familiar perceptual characteristics, are the basis of the kinds of in group biases that we see as kids grow and become adults who have all kinds of preferences for their own in groups, but it's not clear at all that these preferences are actually social in that way, that they're going to guide who infants want to actually interact with. They guide who they think is part of the same social group in the world regardless of whether they are or whether these are just perceptual preferences. Basically, ease of processing for faces that match your schema better might lead you to look longer at one face over another without actually being about a social preference that's going to guide the way you interact with other people.
So it's hard to tease that apart with preferential looking because it's so vague in terms of what's actually drawing infants to one face or another. I think the active preference measures that I'm going to talk about a little bit later are people's best attempt to figure out do these preferences reflect real social preferences, or are they just perceptual familiarity affect, something like scientist mere-exposure effect, where the things that you see in the world more often become things that you like to look at, but they don't necessarily mean anything about inferences about the social structure of the world.
I just want to point out one way that people have tried to prove that these preferences are really social over perceptual, and that's to show that, at least in this language case for what babies are preferring, is the native language speaker. They wanted to ask whether you only prefer a person who's paired with a familiar language as opposed to any object at all that's paired with that familiar in a [INAUDIBLE] perceptually appealing language.
So they played English and French both with the speakers, but also in the presence of geometric objects that were not moving like agents. I know that [INAUDIBLE] showed you [INAUDIBLE] and showed that you can interpret agents [INAUDIBLE] as agents are people, but these were just shapes bouncing around the screen or rotating in random ways that were there while either the native language or foreign language is being played.
And it turned out that after that kind of exposure, infants had no preference for the shape that was paired with the native language speech. It was particularly the case that when you thought a person, or an agent shown with a face that you could recognize as a potential social partner have impaired with this familiar language that you then had a preference to look at them over the foreign language [INAUDIBLE].
So moving on to habituation paradigms, but sticking with faces for a second, you could ask do infants-- are they processing those faces as individuals in the world, or are they lumping them together any kind of way? And this would relate to the idea of are those preferences that they're looking at going to become preferences for categories of people, in group or out group members later on. Do infants perceive those faces as part of social categories, or at least categories, of people?
And the way to look at that using a habituation paradigm is to take advantage of the fact that infant looking time often looks like this. So if you show them a habituation face, where you're getting them used to one kind of thing over and over again, you can then do a pair of trials to the end where you look for dehabituation. Basically, you show them yet another thing that's similar to what they were watching during habituation and you show them something that you think they might distinguish from what they saw during habituation and see if they recover attention to a greater extent.
So imagine, for example, that you showed an infant a whole bunch of different versions of the letter a. And then at test, you showed them yet another new version of the letter a versus a letter b. So if they weren't extracting anything that was common about all of those different a's, then the new letter a isn't going to be any less new than the letter b.
If there was no commonality between those elements for the baby, then both of those things at test are new to them. On the other hand, if they could extract what was common about all those letters, and they were getting sort of habituated to that common core of the category of the letter, now when they see a novel stimulus, but within the same category they've been looking at versus a totally novel stimulus category and [INAUDIBLE], then you can get a distinction in looking time to those, too.
So you can imagine doing the same thing with faces. You could show babies a whole bunch of novel male faces, and then show them another novel male face versus a novel female face. You can do the same thing with race. You can show them a whole bunch of different faces of the same race, and then show them yet another new face from that same racial category, and then [INAUDIBLE], see if they are categorizing faces along those dimensions that you're showing them.
And so with both of these examples, for gender and for race, it turns out that infants at the age where they're showing preferences also show this sort of categorization phenomenon. So I mentioned that infants prefer the gender of their primary caregiver. That means most infants have a preference for female faces.
But given that baseline preference also competes with this kind of preference for novelty and this habituation phenomenon, you can show an infant a whole bunch of novel female faces and then at test prevent another novel female face and a novel male face side by side and now see that their preference for who to look at has flipped, or you can present them singularly one at a time and see how long do they look at each new [INAUDIBLE]. They actually now look longer at the male face than the female face.
So visually at least, they're also-- in addition to being able to discriminate these faces and showing preferences for one over the other-- they're also showing the ability to categorize underlying similarities between faces of different sorts that we recognize [INAUDIBLE]. They also categorize along this dimension of attractiveness.
So this is not necessarily saying that they've picked up in some way on the way that adults categorize the world. We often might apply labels or segregate by gender or by race, but maybe less so-- maybe, maybe not-- but maybe less so by attractiveness. All right.
But this is not the only way that you can use habituation to try to understand categorisation in infants, and I'm going to demonstrate a different way that you can use habituation to do the same thing in contrast with a different question, which is OK, so we've shown that infants can perceive faces.
They can perceive relatively subtle differences between the way that faces looked. They can categorize along those dimensions, but that's not really what social groups are about for adults all that much. Perceptual information can be telling or can help you in some cases identify using [INAUDIBLE] group.
But what social groups are to us really are groups that are defined by affiliation between individuals. And so who has some sort of social bond with another person? And a lot of times, those bonds are evident not in the way that people look, but in the way that they interact with one another.
And so we wanted to ask-- since I'm going to show you a brief study of my own here, we wanted to ask whether infants could use any kind of interaction cues to figure out what social categories are in the world, even when you hold perceptual features of an individual constant. And the way that we wanted to do that is rather than showing infants a whole bunch of instances in [INAUDIBLE] one category and then seeing if they distinguish between another [INAUDIBLE] category and a new [INAUDIBLE], what we wanted to do is look at the rate of habituation. So you can imagine that this is like going back and forth between a, b, a, b, a, b and a, c, a, c, a, c, a, c and seeing whether or not infants see a and b to be more similar to one another than a and c based on how quickly they habituate.
So say, for example, that a and b are two different members of the same social group whereas a and c are members of different social groups, you might then expect infants to perceive a and b as more similar to one another because they share some category membership whereas a and c are less familiar to one another because not only are they different individuals, but they don't even share a category membership with one another. And so in that case, what you're going to expect is more rapid habituation in the a b case than in the a c case.
Does that make sense? Yeah. All right.
So when it came to what kinds of properties of the groups, we wanted to see whether or not infants [INAUDIBLE] to categorize them. We turned to the adult literature. So there's a whole literature out there called entitativity which is possibly the most tongue twisting word you can try to say correctly during a talk [INAUDIBLE] times.
But what it does is it looks at what kind of cues adults use to group people together into social categories that they [INAUDIBLE]. What kind of Gestalt principles of social groups are there out there for adults. And not surprisingly, one of the things that adults use is proximity. Who is near one another?
Sometimes this is a good cue, and then there's a bad cue. [INAUDIBLE], for example, are probably a cue for proximity is a very bad cue if you're part the same social group. Everyone's there for an instrumental purpose rather than wanting to be near one another. And in a lot of cases, people orient themselves towards others, and approach other people, and cluster together in ways that give you information about their social relationships. So proximity is a perceptual cue that we thought might help.
In addition to that, we know from some research on, for example, minimal groups in adults and children that perceptual cues alone failed to delineate people into social groups a lot of the time and that instead, how those perceptual cues are used is actually important. So a study that I like is one by Rebecca Bigler. She took a classroom full of preschoolers and she had half of them wear red shirts and half of them wear blue shirts, an extremely salient perceptual cue that split the kids in half.
But she either had the teachers use this distinction to have the kids do different things. Like for example, OK, red group, go eat your snack. Blue group, do your work. And [INAUDIBLE], she had them just [INAUDIBLE] the way that the classroom was run and the way that the kids were sort of structured to interact with one another.
And what she found is that only the kids in the group where the shirt color was actually used to determine who was doing what way did the kids end up showing an ingroup preference based on shirt color at the end of the experiment. So it's not just about really visually salient perceptual cues to who is different from who. It's actually about how those cues map onto the way that people interact together.
So we used proximity and common action as a way to delineate our social groups for infants. So here's how we did that. So Laura talked a lot about how there are different cues for infants, maybe that things are agents, or have minds, and are capable of self propelled behavior. And in fact, they're sort of in some ways circularly defined.
So the way that you recognize an agent is by seeing whether or not it propels itself, whether or not it has a face, whether or not it seems to have perceptual abilities that are sounding contingent [INAUDIBLE], something like that, and that is essentially what we use to define an agent. So we gave these guys little faces, and they're going to control their own motion. And we either had all four of these little characters face one another together in the center of the screen and move together in a big circle or we put one pair over on one side and another pair over on the other side, and we have them take turns moving in separate pairs or separate groups.
[INAUDIBLE] one social category based on those interaction cues or refer to two different social categories. After that, we ran a habituation paradigm where infants were just looking at one of the characters at a time. So they were looking at four different characters, and you could imagine that it's sort of either like a, b, a, b or a, a, a, a in terms of category membership depending on which of the two conditions you were in. So obviously, they all look different from one another, and there's a little bit of a two by two visual distinction between the characters. But how similar they are might also depend upon which of the two ways they were grouped together socially at the beginning of the experiment.
So I'm going to show you guys [INAUDIBLE] it's a good example. It's the best example I have of this kind of method in a social context. The data is still in progress. Here's what it looks like so far.
So over the course of the experiment, you can see that, especially toward the end, all of those red dots, which are the looking time to individuals from the one group condition are ending up lower than all of those dots to looking times of individuals [INAUDIBLE]. Now--
AUDIENCE: Has the habituation [INAUDIBLE], or was there a whole dynamic thing, or--
LINDSEY POWELL: So there's a dynamic aspect to it. Sorry. They do a little-- whoops. They do a little dancing together. I didn't put the demos here, but basically, they move in these curlicue patterns. They go around in a circle, and so they're going either around in a circle with everyone all together, or each pair is going around in its own [INAUDIBLE], where there's proximity and common action. Right. Right.
So then when we show them the individuals [INAUDIBLE], they're habituating more quickly. When we've shown them dynamic cues, the idea that these guys are part of the same social group as opposed to are in different social groups, even though their perceptual information is held constant.
And so one thing I want to point out here, so part of this idea is getting back to that question of are those preferences that infants had for gender, faces of a particular gender or faces of a particular race, are they social or they perceptual? And part of the idea here is that we're going to be able to perceive people visually in the world, but the way that we interact with people depends upon construing them not just as agents out there with certain perceptual properties, but what their social relationships to us is. And so just like you can't tell for sure whether a category is a social category for instance in the sense that a test-- it's a category that's made up of people versus it's a category that means something about the affiliation between the members, you also can't tell whether a preference for a person is a perceptual preference for a familiar looking stimulus or a preference for a person who you think is going to be a better social partner for you.
And one idea by being able to define groups in different ways is to say, OK, well can we figure out whether or not the group is defined by people who look the same way is the same kind of group for an infant as a group that's defined this way by interaction? And one way that it might be able to do that given this kind of phenomenon is to take advantage of the fact that habituation actually has a neural correlate.
So I know you guys have been talking a lot about brain imaging methods this week, and the idea of repetition suppression might've come up, the idea that there are regions that are responsible for processing a particular kind of stimulus, and they respond less when you repeat the same stimulus than when you're altering the nature of the particular stimulus you're showing across trials. And so that affect is basically what's the minimal correlate of habituation given what a particular region cares about. If you show the person something similar within that category, you're going to get both less looking and less neural response to successive presentations from within that category.
So the idea is we could look both for regions that respond when you show, say, one female face after another and regions that respond when you show one social group member after another defines this more dynamic and interactive way. And see whether or not there's really just some general purpose [INAUDIBLE] of processing categories of agents, or whether there are different kinds of categories of people when they're defined socially based on the way that they interact with one another or the way that they're affiliated with one another as opposed to just based on perceptual similarity between [INAUDIBLE]. And these methods sometimes do have ways to apply the general principle to other populations in case not all of you out there are planning on designing infant studies in the near future.
All right. Any questions about habituation [INAUDIBLE]?
All right. So the next kind of build input method is pretty similar to habituation, and it's one that I think Laura talks about a lot, one I sort of mentioned that she'd gone over at the beginning of the talk, and it's what we call violation of expectation looking. And again, what you're doing is showing an infant a series of events or images a lot like habituation.
The idea is that it doesn't always have to be the case that what you see during habituation though is quite so analogous to what you're watching [INAUDIBLE]. So for example, in the classic study by Amanda Woodward. She showed infants and agent reaching for an object. And the first time that agent reached for, say, the teddy bear, the babies looked for quite a long time up there on the left.
And each successive time the person reached for the teddy bear, the baby looks less and less until they've habituated, or she's shown them some number of familiarization trials. And then you swap the location of the objects. And so now, you're not going to be showing them anything that's exactly similar to what you showed them before, right.
But the question is, which of these things is going to be unexpected given what babies have seen up to this point? The person making exactly the same motion, reaching for the new objects in the old location, or the person making a new motion by reaching for the same objects in that new location? And the question here is in terms of infant social perception, what are they extracting from patterns of motion that people are engaged in the world? Are they just tracking where the people move, or are they tracking, say, what the endpoint or the goals date of the action is, when it completes?
So of course, I think you guys know this already, but what she found is the babies are surprised by exactly the same motion that goes to a new goal object. They look longer at that, especially relative to habituation, but [INAUDIBLE] relative to reach to the old object now in a new place. The thing about violation of expectation, though, is it doesn't always have to be based on this kind of set up and then test kind of situation.
Whoops. [INAUDIBLE]. I've animated [INAUDIBLE]. There we go. Expected and unexpected. OK.
So it doesn't always have to be based on setting up some information about the world. So babies come in with all kinds of expectations to our experiments, otherwise they wouldn't work. And sometimes, you can just test those right off the bat without showing this kind of introduction, this slow introduction of events that lead up to an expectation. So take, for example, studies of infants perception of other people's beliefs.
So I actually don't know if you guys have this what you talked about theory of mind at all, but people have looked at whether or not infants use what a person knows about the world to predict what they're going to do, and this is exactly that kind of experiment. So there's a person there. She has a visor on, but the idea is that she can see that watermelon as this slides from the green box to the yellow box in the tops of the panels, but then she can't see in the bottoms of the panels, because those doors are closed, that the watermelon actually slides back from the yellow box to the green box.
So for adults, we would now think that she has a false belief about where that object is located. The question is, do infants similarly think that that perception of her information state play any role in what they predict she's going to do next? And in this case, the researchers show this introductory theme, but that's all they do. There's no long process of familiarizing [INAUDIBLE] to what that person is going to do next, where they're going to reach, that kind of thing. She actually just then shows them one of two things.
Either an expected event, where the person-- or for an adult, an expected event-- where the person then reaches into the yellow box, the last place they saw the watermelon go, or the unexpected event where the person reaches into the green box where the object actually is but not where she should think it is given what she's seen previously. And just like in violation of expectation events where you set up this gradual regularity of pattern, you find that infants look longer at the reaching into the green box even than in the reaching into the yellow box event.
So you can violate some expectation. That can be a measure of [INAUDIBLE] that the baby comes to the experiment with or ones that you set up by showing them some body of evidence before you test them. So just to stick with the theme of are social categories differentiated from just basically categories of people, of potentially social things.
For infants, I'm going to show you another violation of expectation study to ask whether we treat affiliated social categories of the sort that are defined by interaction differently than just the perceptual categories, either of inanimate things, or even of animate things. And the way in which we wanted to know whether they would treat them differently is whether they would expect different similarities for social groups than they would expect for non-social or inanimate groups. So one thing to point out is that the common feature of affiliated groups of people is that they tend to act like one another.
So for example, if people all dress alike, they all use a particular mode of transportation, maybe they don't use electricity, they're a different special [INAUDIBLE] social group and they have a different set of choices about some of these things. But these people can live in really close proximity to one another.
So it's not just that they've observed only one set of actions or possible ways to behave and so that's the way that they behave. It's actually that they choose to act like their social group affiliates. And given the range of clothes you could decide to where, or the range of places you could decide to live, or the ways you get around the world, they've chosen to do things the same way as the people that they have social bonds with.
So we wanted to see whether or not infants would have this same expectation about social groups. They expect social group members to act like one another. So we took eight-month-old infants. Then we showed them two groups defined the same way that I just showed you in the other experiments.
So now that little line's on there depicting how they move together. But there's the three right guys. They're facing one another. They move along those little black lines. They look fairly patterned together. The yellow guys do their own kind of dance together.
And after infants have been introduced to those two groups, what they see is that two of the right guys do one thing. In this particular experiment, it was hop on that purple box. In other experiments, we use other kinds of actions. Sorry.
And they also see the two yellow guys do the other action. So they would go and dump in a similar manner, but on a blue box rather than a purple box. And then we show them what the final red and yellow characters do, and they're both going to do the same thing.
So there, you see the right guy jumping on the purple box. That means the yellow guy also does. And so given a violation of expectation paradigm, if infants are expecting social group members to do the same thing, they might be surprised when that finally the yellow guy jumps on the purple box, but less so [INAUDIBLE] given that that's what his group was doing all along. Does that make sense? Remember the [INAUDIBLE] paradigm?
So there's lots of reasons why infants could expect this. They could've just gotten visually used to seeing a red thing on top of a purple box. What we wanted to ask with the study is is this a particular expectation they have for social groups as opposed to other kinds of categories?
And so [INAUDIBLE] ask that selectivity question, that domain specificity question, and to rule out some [INAUDIBLE] accounts for [INAUDIBLE] expectation, we asked whether this kind of expectation would generalize to inanimate groups. So Laura talked about a lot of the cues the babies used to animacy. So one of them is faces.
Rearrange the faces so that they didn't look like faces anymore. And another one is sort of self-propelled non-Newtonian motion. So instead of having the guys do dancing at the beginning, we still have them move synchronously together as two separate groups, but they were sliding along this triangular frame together instead of dancing in a sort of self-initiated kind of way.
But then we showed them exactly the same familiarization and tests in terms of two of the red objects landing on one of the boxes and two of the yellow objects landing on the other, and then asking whether they'd expect the same landing locations for the final object to reach [INAUDIBLE] with one matching its category and one not matching.
And finally, we also tested pretty much the same situation, keeping the cues to animacy but removing those proximity and common action cues to social group membership. So now, they see the same six agent like characters that they saw in the beginning. But rather than being grouped together based on those perceptual features and rather than dancing in groups, they're sort of intermingled with one another and each of them dances alone in an unsynchronized fashion. And then we show infants that two of the yellow characters land on one box, two of the red characters land on the other box. We ask if they expect the same thing out of block two.
As I was mentioning, the violation of expectations studies can take a bunch of different forms. We either habituated infants to what was in the middle column right there, showed it to them over and over and over again until they were extremely bored, or we just familiarized them to show them a few times what these guys did, what these guys did, or these objects did and these objects did before moving on to test trials.
So it turns out it doesn't matter what kind of setup [INAUDIBLE], it's a little bit or a lot. When it comes to the social groups, babies expect the characters to do what they're groups were doing. When they act like the other group instead, babies look longer. So this is a percentage of how much time they spent looking during the test phase in terms of what trial were looking at. And in the inanimate case, so these are the two conditions that we contracted as both familiarization and habituation. In the [INAUDIBLE] case, they no longer expected that third object to land on the same box as the first two, and the interaction between the two positions [INAUDIBLE] significant.
As far as the non-social group went though, same yield. This was a familiarization study only. So again, the [INAUDIBLE] was longer, and the social condition, when the last guy in the group doesn't do what the first two did, but in the non-social conditions, they don't care if that [INAUDIBLE] triangle hops in the same box as the first two did.
And so the idea here is-- oh wait, sorry. I'm missing a slide that gives you the summary message. Basically, the summary message is yeah, so the babies are perceiving categories in the world. It's not the case that every category supports the same kind of inferences about the world. So you could imagine that babies just make perceptual categories about the world.
Maybe they're based on things like proximity or common action, but in the end, all it boils down to is they group some objects together versus others, and those categories all serve as the basis for the same kinds of inferences. But what this suggests is the way that you create the category matters for what babies do with it, right. And so in a particular-- when it comes to social categories, it seems like action, interaction, and the relationship between individuals matters more than their perceptual similarity in deciding whether or not you should make particular inferences about a kind of thing.
So just because all three of those yellow triangles looks the same doesn't mean they're all going to act the same way. But when they hang out together and when they dance together, all of a sudden, infants now have this expectation of a phenomenon we find in the social world amongst adults. So there's more evidence that basically, you can make perceptual categories out of people in the world, and really young infants can do that, but it doesn't play the same kind of role in our model of how other people are going to act and how we should interact with them as information about the relationship between individuals does for babies. Sorry, [INAUDIBLE].
All right. So that just goes to support the idea that social cognition is not about just perceiving people in the world. It's about interacting with them and figuring out what the best way to do that is. So maybe the best way to measure social perception is to actually look at infants' attempts at interacting with other people.
I mentioned that they're relatively uncoordinated and they owed you a lot, but one thing they can do is two things, especially from about six months of age. So sometimes, infants will be presented with studies like the ones I just showed you that had agents with faces. And babies will get to choose one of the agents over another because you can make them small and hand sized as opposed to a person, which might be difficult for a baby to figure out in a kind of way.
So you do have people as your stimuli. You can also give them, say, the opportunity to choose one of two identical toys presented by two different actors. So it's not about they're choosing one of two things that look the same [INAUDIBLE] one another. It's that who offered it to them as opposed to a preference for one of the particular objects over the other.
And then finally, by the time infants are about a year to 18 months, they start to get more coordinated at doing additional things in the world, and they actually start to help. So one of the interesting things in the literature and the ways that babies' behavior has come out recently is that by about 14 months, babies spontaneously help other people, things like if I dropped a laser pointer off the edge of the table and reached out and there was a 14-month-old down there, he might grab the laser pointer and give it back to me.
And this actually interestingly enough is not something that they're trained to do through instruction by their parents. In fact, feedback from their parents or reward for helping has the kind of affect that rewards for good grades have on kids a lot of the time which is that they become extrinsically motivated rather than intrinsically motivated, and they actually help us. So it seems like infants are sort of motivated intrinsically to help other people, but they can also choose who to help.
So this particular image, it comes from a study where she was trying to get that box into the cabinet. She needed someone to open the doors for her. And the question was, would babies be more likely to open the doors for her based on the way that she'd acted toward them in the past? So you can use all of these kind of active measures as ways to look at how infants are evaluating people and how they're evaluating interactions in the world.
So some examples of studies of active preference measures. And this is one that Laura might've mentioned, at least in passing, and that some of you might be familiar if you are in tune with the developmental social literature at all is that [INAUDIBLE] people have asked is do infants prefer helpers to hinderers? So like I mentioned just a minute ago, they're intrinsically motivated to help by the time they're about a year old. But is this reflecting essentially even earlier developing preference for other people to engage in helping and evaluation of that as a positive behavior?
So a classic version of this study is something like this, a classic from 2007. There's a scenario where that little red guy has been trying to climb that very steep hill and failing. He goes up almost to the top and up almost to the top and doesn't quite make it each time.
And then after this introduction to his goal to get to the top of the hill, babies see two different things. They see a helper come along and push him to the top of the hill and they see a hinderer come and push him down to the bottom of the hill. And then infants are given a choice to reach for either the helper triangle there on the left, although obviously they counterbalanced across babies which one was helpful and which one was not, or that hindering square there. And what they find is that infants robustly preferred the individuals who helped that circle guy to the top of the hill over the ones who pushed him down to the bottom.
But what was it about this pattern of events that infants found made the helper more appealing? So for example, when you're judging someone's actions, do you just judge the actions on the basis of what that person is doing, or does the consequences of that action and the target of the action play any role in how you perceive and evaluate that action? [INAUDIBLE] a question you could ask about babies' perception. [INAUDIBLE] behaviors that they find positive and behaviors they find negative but that they can't really sort of think about them in a broader social context. Or maybe the social consequences of the actions are actually what make them good or bad for babies.
So a way that you can ask this question-- again, trading on the fact that babies are so easy to push in terms of seeing something as an agent or not an agent-- is to replace the little red circle guy climbing the hill, which is a ball with no eyes. So now when you push the ball up to the top of the hill, you're not helping an individual who has a goal to get to the top of the hill.
You're engaging in exactly the same motion, but its consequences in social meaning is different, and same deal with the hindering. He just fell down to the bottom, but now he's not preventing anyone from reaching their goal. And the question is do the babies do not like that downward pushing motion. For example, does something sound unappealing about that, or does it matter that it was being done to an agent who wanted to be doing the opposite thing?
And again, if you are given a chance to reach for one person to the other, and now the preference goes away. So now they're no more likely to reach for the guy who went up the hill than the guy who went down the hill. And so I think this reveals-- yeah.
AUDIENCE: [INAUDIBLE] surprising because they think that that ball [INAUDIBLE], they think of that as an agent.
LINDSEY POWELL: Right. So I have. So actually, they skipped that part. So with the ball, it never moves up the hill. In addition to taking the face away, they take away the part where it tries to get up the hill. Yeah.
So it's mostly just a control for the actual action, the actual movement of the helper itself. That might not be that surprising. In this case, you might have thought, OK, going up versus down a hill is not going to be intuitively appealing to a baby regardless. But there are actions where you might think the actions themselves might be intuitively appealing or unappealing to infants.
So for example, [INAUDIBLE] et al looked at sort of battery actions, bashing things versus crushing things. And they had two actors, one who bashed an object and crushed a person and one who did the reverse. And even with those actions, where you might think the actions themselves are the things the babies really care about, what they found again is that it matters who you were doing the actions to. So they preferred the person who bashed the object and liked the person [INAUDIBLE] wanted to do the reverse.
So one way that infants seem to identify social interactions is not just based on the pattern of the behavior you're engaging in, but whether or not there are two agents involved in the situation at all. All right. So I mentioned at the beginning when I was talking about face preferences that it wasn't clear what kind of preferences those were.
Are they just preferences that are driven by the perceptual familiarity of certain categories of face, or are they actually going to reflect those social Interactions [INAUDIBLE] want to engage in with those people later on. And here were two cases where we got a preferential looking result with young infants. Four and five year old infants prefer to look at own race faces or native language speakers.
And so to give us an impression of whether or not these preferences are perceptual or social, people try now moving to this realm of active preference measures where babies are really interacting with people. And they did, in fact, these toy choice kind of studies that I just mentioned. So now they had a native language speaker and a foreign language speaker.
Babies were introduced to each of them separately, and then they were put side by side, now silent. So they're not speaking the languages at the time, but they each held out a toy and offered it to the infant. There's this lovely magic trick that Katie Kinzler invented where the people are on the screen. They reach their arms forward, and there's these PVC pipe arms that come up and then present objects in front of the baby like a person reached out of the screen and put the object right in front of you.
Babies seem quite convinced by this magic trick, and in fact, they were more likely to reach for the toy that was presented by the native language speaker than the foreign language speaker. It turned out that when they [INAUDIBLE] looked at race preferences using the same method, there was no choice based preference to take one of two animal toys from an actor who shared the baby's race with an actor who was of a different race.
And so I want to point out that again, this seems to go back to this central idea that what social groups are for babies is inherently linked to behavior and action and interaction. So language is a way that you interact with other people, and so there's possibly a good reason why you prefer someone who acts in a familiar way, acts in a predictable way over this kind of perceptual preference the babies seem not to treat as a social preference unless everybody else is treating it as a social preference, and it actually influences who you're interacting with.
So basically it's saying, if these are like the two t-shirt colors, and sure, if babies had mostly seen blue t-shirts in their life, they might prefer to look at someone wearing a blue t-shirt over a red t-shirt in a visual preference test. But if they've never seen that dimension having behavioral consequences in the social realm for who's going to interact with who, who's going to coordinate their behavior together, then they're not going to care about it when it comes to social preferences about who they want to interact with. And instead, what might be more meaningful for them are these behaviors that people are going to engage in together.
AUDIENCE: OK, so we're looking at newborns, and then we're jumping to eight months. So has anybody done any longitudinal work between the newborns and the eight months to see if this change happens suddenly, or whether or not it's a gradual shift, or--
LINDSEY POWELL: Yeah. That's a great question. So for a lot of these studies, the answer is going to be no.
People have just, say, looked at a particular age as their sort of representative of pre-verbal infants, and part of what a lot of the studies are trying to claim is that whether or not newborns do this, they didn't learn it through language. So for example, with the groups acting alike, it turns out the generic language has a big influence on our expectation of category familiarity.
When parents make generic statements, children learn that the groups in the statement are going to share a common property, and it's actually a big influence of how they think about stereotyping and social categories. But if they're showing that at eight months, that can't be how they initially learned that expectation, right. And so part of it is just ruling out that is this about language and broad social interaction.
But it's really interesting to ask that question of what's the developmental trajectory going to be like for these different preferences. And so people have looked a lot at language preferences and perceptual narrate for your ability to perceive different language, contrast in different language. When this sort of language preference comes online, they have looked younger and younger, and they find it at roughly five months.
And it turns out that the point at which this language preference emerges in terms of the preferential looking to one or another is about the same time when babies start to lose the ability to perceive contrasts in the other language. It's about when they stop using foreign language words to create noun categories in the world. So it seems to relate with their sense that at the point at which they sense there's a particular language that applies to them and that they should attend to seems to correlate with the age at which they start to show social preferences based on language. So younger than that, you might expect a failure, but it might be because they're not actually telling the languages apart, not because they don't actually have the language preference yet. But some of these things you might expect to be more--
AUDIENCE: Well, OK. But, I mean, people have shown that there are language preferences even prenatal, right.
LINDSEY POWELL: Oh, right. Like [INAUDIBLE].
AUDIENCE: So the other question I had-- which is probably not true-- but has anybody done the language preference study on bilingual kids?
LINDSEY POWELL: You know, I think that they were starting to-- so their initial study was only from 2007. I think the answer is yes, that the results are sort of in progress and not publish it. But, yeah. And I think the answer is that kids prefer both of the languages to which they are familiar with, but not a foreign language that they're unfamiliar with. It's not like hearing multiple languages makes everything fair game. It's like they now recognize [INAUDIBLE].
AUDIENCE: I would be interested in knowing if there is any difference at all, because I have two kids that we're bringing up bilingual, and there's an overwhelming abundance of English in the household. But, I mean, there's also I would say about 30% to 40% of the language that they hear is Japanese. So it'll just be interesting to see if that affects their social perception in any sort of a weighted sort of a way.
LINDSEY POWELL: Any other questions about these [INAUDIBLE]? All right. OK. So I've covered almost all the potential ways to use these methods so far.
One last one is anticipatory looking. So not only is social cognition for interacting with other people, if you're going to do it really well, you're going to predict what they're going to do next so you can coordinate your actions with what they're going to do next. And it's important to point out that violation of expectation tests don't really quite get you there in terms of knowing that infants-- they're called violation of expectation tests. They can also be sort of violations of postdiction tests so that you can see an event and then decide afterward that it was surprising rather than having an expectation about what's going to come next and then having that expectation to be violated.
So think about two alternatives for the Woodward scenario. Either infants saw those objects move, and then they're like ready for the person's next reach, and they're predicting where that reach is going to go, and they see the reach to the previously preferred object.
And they think, OK, fine, that meets my expectations. Or they see the reach to the other object and they say, oh wait, that's not what I was expecting. Or alternatively, they're just hanging out, looking at the objects. All of a sudden, the actor reaches for one or the other and they consider post hoc which of those two things seems more surprising, or how well does that meet what I thought that person might do if they had reached.
So the question is can infants anticipate actions ahead of time? And here's where you really want to be able to do the eye tracking kind of thing that someone asked me about. At the beginning of the study, not just measure how long do infants look at a given event like we've been doing almost all the way through, but when there's a reason to expect that something's about to happen, where do their eyes go at that particular moment to predict where the action is going to be?
So imagine now you've got those two objects. The person comes back and the baby has some reason to expect that he's going to reach right away, you can then measure where do their eyes go in anticipation of the reach. Do they go to the object the person was reaching for all along, or it's the location that the person was reaching for all long before the reach even happened? So now there would be a better sense that there's actually an expectation about what that agent is going to do next.
So I want to talk about this one again in the context of helper versus hinderer studies. Babies, [INAUDIBLE], have that preference for helpers over hinderers themselves, but you could also ask what do they think about the red guy? Who do they think he's going to like? They have any sense beyond their own preferences that that character should like the prosocial agent over the antisocial agent?
And so what you can do is put the right character right in between the two and see where babies expect that guy to move, and that's exactly what some of the research has done. So for example, the character moving-- I think that's helper over there he was moving toward. But you can show that event, or you can show the character moving toward the hinderer.
This would be kind of a violation of expectation paradigm. You can measure, then, how long do infants look when he moves toward the helper and how long does he look when he moves toward the hinderer. Are they surprised by one or the other?
And in fact, by 10 to 12 months, infants are looking longer when that middle character moves toward the hinderer. When he moves toward the helper, they find it surprising for him to move toward the guy who didn't help him. And it turns out that at 12 months, you can also get anticipatory looking.
So if babies know that that character is going to move one way or the other off the top of the hill, and then you put him up there and you've got the helper hinderer on either side, you can then track whether these moves, before the guy, the middle guy, even moves. And the way that their eyes move is in anticipation of a movement toward the helper over the hinderer. So these are the kinds of anticipatory looking measures you might want.
I've gone through all the methods now, but I just want to take a couple minutes if I have time-- yeah-- to point out why we would want to do all these different kinds of things to look at these different kinds of scenarios and what you're going to learn from using multiple methods with babies, which is to point out they're not doing this anticipatorially [INAUDIBLE] of violations of expectation until they're 10 to 12 months old. The earliest data I showed you it was from the six month old for babies reach for the helper over the hinderer, and in fact they found similar results, even with three months old.
The question is, why are infants themselves preferring the helper to the hinderer when they're extremely young infants but not expecting the protagonist, who was the one who was helped or hindered, to do the same until they're much older? And is this telling us anything about the goals of the social perception that infants are engaging in or the kind of algorithms they're applying to understand social behavior? And point out that really, these methods are most powerful in combination.
You want to know not only what infants prefer themselves, or not just how they expect people to act, but how well those things fit together. And in other words, it's best to understand what kind of content infants are extracting [INAUDIBLE]. So for example, showing that there's not only a looking preference, but also a social interaction preference for language groups whereas you only see the perceptual preference for race groups. It gives you a richer understanding of that kind of social preference than just one of the measures [INAUDIBLE] on its own. And so I want to focus on this distinction between infants' own preferences and their expectations about other people's preferences to try to explore this a little more, and I'm going to show you one more example that's going to come [INAUDIBLE] of my own work.
But first, let's point out that when it comes to individual actions, we hypothesized-- at least we have in some cases [INAUDIBLE] at some point-- that infants use an inverse planning model to interpret other people's actions. So this is an example from an adult study that I think you guys maybe read as part of the course work you're all familiar with. But basically, you watch this guy acting. He's choosing a food truck. He goes right past that first one, looks around the corner, and then goes back to the first one after he looks around the corner.
And so the model that has been developed that explains how we then figure out what that character's preferences were basically takes the way that you would plan your own actions given a particular [INAUDIBLE] preferences and runs that program backwards and says, OK, what array of preferences best explains the plan that that agent chose. And the idea here is the potential third truck that could've been around the corner but happens not to be when the guy looks around the corner, and that's why he first goes to look, because that's his favorite truck. But then when it turns out it wasn't there, he goes back to his second favorite truck that he already walked past, right.
So this is an adult example of this study, but you could say the same thing for how we think infants are arriving at their expectations in that [INAUDIBLE] paradigm that I've used over and over again. A person reaches for the same object over and over and over again, and the idea is that what the infants are saying, OK, I would engage in that action. If my preference were for that object, my plan would be to reach for that one.
I see this person engaging in that action, and so I'm going to use my inverse planning model to decide if that's because they had a preference to reach out for that object. Similarly, there's things with characters jumping over walls, and you expect once the wall is gone that they're going to go in a straight line because you think the role of the jump was to avert an obstacle that was in the way of your goal state, not to engage in the jumping action itself.
So we tend to think that infants, when they're reasoning about the individual actions of agents, are using this kind of model to figure out what the people are doing and why. So the question is, when it comes to social actions, are infants using a similar rational planner model to figure out what people are doing and why? So we reason about social interaction in a similar way.
So to try to think about this question, I'm going to tell you about some experiments that I did really, really quickly that look at preferences for imitators in young infants. So how many of you guys have ever heard of the chameleon effect? Wow. Not enough in this social psychology class right now. How many of you guys have ever been having a conversation with someone and noticed that you're sitting with the same posture, or you're nodding when that person nods, or you're talking louder when they talk louder, or any of those kinds of things?
So you're sort of mirroring their behavior subtly, right. So this is the chameleon effect. You're changing yourself to mirror the other person. And it turns out you do this more when you like the person you're talking to, when the person you're talking to does [INAUDIBLE] like them more. There are all of these kinds of positive social causes and consequences of this kind of imitation of other people.
And so we wanted to see sort of with infants, does this imitation preference apply to really young infants who actually don't do a lot of imitating yet. They get imitated very frequently by their own parents and other adults. If you ever want to watch something funny, just watch an adult interact with a baby for a little while. And as soon as the baby does something interesting, the adult will undoubtedly repeat it back to the baby.
And so what we wanted to ask is do infants reason about these kinds of interactions? Will they prefer an imitator to a non-imitator themselves? And Like you might if you were watching people interact and mirror one other versus interacting and not mirror one another, will they make inferences about relationships between people based on who chooses to mirror?
So the first thing we found is that if you show infants something like this where there's a purple guy who's repeating the same action over and over again and a blue guy who imitates that action [INAUDIBLE] red guy who doesn't, you just give babies a choice between the imitator and the non-imitator in either a preferential looking or a preferential reaching test, young infants prefer the imitator. But it turns out that really young infants, four month olds, just like the four month olds in the helping studies, though they have this preference themselves for imitators, they don't have any expectation that the target is going to approach the imitator over the non-imitator. Even though the behavior was directed towards him, just like the helping and hindering behaviors were directed toward that protagonist, they're not actually reasoning about who that guy likes.
So one possibility is that this reflects the fact that imitation, and maybe social perception in general at this age, is serving to identify good partners for the baby rather than serving for them to make more inferences about how people are going to act in the world. So we kind of talked a little bit about what are the goals of social perception? What is it adapted and used for? And possibly one reason why you're not seeing reliable predictions about other people's actions based on social interaction is that what infants are trying to do is figure out who they should want to go to interact with, and they're not at that point not really concerned about what's going to happen next in someone else's social interaction. Yeah.
AUDIENCE: Older kids use these cues to reason about other people? I mean, when you asked the group, the idea was familiar to some, but not everybody realizes [INAUDIBLE].
LINDSEY POWELL: Oh, the imitation? Yeah. So yes. No. Actually, I don't know. Four year olds think that the imitator likes the target more than the non-imitator does, but they don't necessarily-- and I'm going to actually show you some evidence that infants do the same thing in just a second-- but I don't think anyone has actually asked the reverse question about whether the target likes one more than the other or not.
And this is a relatively understudied aspect of adult social cognition. There are two studies on what we think about people who are imitating other people based on characteristics of the target themselves. So for example, if someone who's imitating somebody who's then being quite rude or mean to them versus someone who's imitating someone who's being quite nice to them, you find the imitator to be more competent in the case where the person was nice to them than when the person was mean to them. And you can find myriad cases where these effects hold.
Like for example, waitresses who repeat a customer's order verbatim back to them get higher tips than waitresses who paraphrase orders. In fact, the helping study I showed you there with the 18 month olds where the person was trying to [INAUDIBLE] a cabinet, the manipulation there is imitation. So has the experimenter imitated the baby, the 18 month old, or not. And 18 month olds are more likely to help people who've imitated them.
So it's true that the interactions hold quite readily. I don't know how good older kids are at precisely that inference. I would expect they can do it, but I don't know for sure. Yeah.
AUDIENCE: Are people who imitate more likely to demonstrate prosocial behavior?
LINDSEY POWELL: Are people who imitate more likely to--
LINDSEY POWELL: So the question would be sort of are they more likely to demonstrate prosocial behavior toward that particular target, right? And Nancy, is your question about whether or not they know that the target [INAUDIBLE] imitator back? So that kind of imitative behavior is driven by the situation you're in to some extent and how much you like the person.
So how much you imitate a person is reliably predicted by how much you say you like them. I don't know about other prosocial actions toward them, but if you say you like an experimenter a lot, then you're more likely to imitate them. Also, if you've been ostracized or rejected recently, if you're in need of social affiliation, you imitate more, and that's true for both five-year-olds and for adults. So it seems like it's a behavior that you have at least an implicit expectation will lead to social affiliation for you.
AUDIENCE: [INAUDIBLE] ingratiation.
LINDSEY POWELL: Yeah. Exactly. You do it when you want someone to like you, whether it's because you like them or because you need friends at that particular moment. Right. So this is one possibility, but-- and this is sort of what I just mentioned to Nancy-- that idea [INAUDIBLE], they're not going to make any inferences about anyone in that situation. They just decide who their own preference is based on what they saw, and they're not reasoning about the rest of the agents and what's going to happen next for any of them.
But it turns out that if you reverse the situation, and now you have two different targets, and you have a guy who imitates one of them and not the other, really young infants do, in fact, expect that guy who imitated to approach the person he imitated rather than the one he didn't. So this has never been done in a helping context. It would be like, for example, having someone who helped one individual and hindered another than asking is he more likely to approach the guy he helps than the one he hindered.
So are you making any inferences about the agent's social preferences based on who he acts nicely toward and who he didn't act nicely toward. And it turns out that before they're making inferences about the targets, they are in fact making these inferences about the actors. And this is where I particularly wonder about the rational planning model, right, because the way that you figure out a person's preference for one object over another is to think about the affect that their actions have on the world and then go back from that and figure out, OK, why was that plan-- what kind of preference does that plan reflect given the consequences that it had?
Reaching for the laser pointer versus the bottle of water is something where you can figure out that I wanted the bottle of water because of the [INAUDIBLE] I obtained the bottle of water, right? It's essentially saying here that infants aren't really understanding the psychological effects that acting nicely toward someone is having on them. They don't get that when you're nice to someone, they're going to like you back.
So how are they using the fact that you were nice someone to figure out that you prefer that person to someone else? It sort of seems like you're missing the understanding of the effects of the action that you would need to have made that action your plan in the first place. And the interesting thing is that infants this age don't actually imitate themselves.
It's not until they get to about 10 months, which is where they start to do things like predict that the protagonist will approach helpers over hinderers. That's the age at which they really start to engage in a lot of prosocial activities themselves-- imitating other people, helping other people. And so it does seem like their own planning and actions depends upon understanding the [INAUDIBLE] of that action in terms of the recipient of your actions then feeling back toward you.
But at the same time, they're managing to figure out what actors of prosocial actions are like before they're understanding those causal effects on the psychological states of other people. So it's possible that we need a different kind of model that isn't an inverse planning model to understand the earliest of how infants interpret social actions for other people. The focus is on the idea that you do nice actions toward people that you like without necessarily including the piece that says you do that because you want them to like you back and because you're planning the right way to get them to like you back, which is to do something nice toward them.
So that's my general introduction to how we use different kinds of behavioral methods to think about social interaction in infancy. And hopefully, I've given a few that are actually going to help us figure out what the building blocks are [INAUDIBLE] answer all these questions. Thanks.
AUDIENCE: So has [INAUDIBLE] methods with some sort of measure of brain activity, say [INAUDIBLE]? And at some point you said something about--
LINDSEY POWELL: Repetition [INAUDIBLE].
AUDIENCE: --infants as not being very [INAUDIBLE]. And you can imagine that [INAUDIBLE]. Like, you can maybe [INAUDIBLE] according to what sort of [INAUDIBLE].
LINDSEY POWELL: Mm. So the idea is that the different types of looking time measures might track with different sorts of neural phenomena.
AUDIENCE: Yeah. So I think that the [INAUDIBLE] fMRIs, but infants, it's hard to do that. But if you can imagine [INAUDIBLE] pattern perhaps, it may be the pattern [INAUDIBLE] going to be different [INAUDIBLE].
LINDSEY POWELL: Habituation versus preference versus anticipatory looking, yeah. So I don't know how much do you expect different patterns based on what kind of method you were using. So for example, I wonder if a surprising action that's surprising by default, if the way you would distinguish that neurally would be different than, say, [INAUDIBLE] habituation, recognizing a novel stimulus in a category.
It's certainly true that you should be able to use neural methods with infants to see the parts of [INAUDIBLE] different hypotheses though, right. So like I was talking about with different kinds of social categories, you might expect different regions to be tracking, say, oh yeah.
That's another person that belongs in a perceptual category that all of those people did versus that's another character that belongs in the same social categories as other people did. And you certainly should be able to use similar methods to the ones we've used here to ask those questions in terms of neural responses rather than in terms of behavioral looking time responses. The question, right, is what kind of method do you use to do it?
And so you mentioned fMRI. People are trying to do fMRI with infants. They do a lot of sleeping studies with babies listening to language because it turns out the auditory cortex will still respond to properties of language while you're asleep. And they're even trying to do awake studies. I mean, Nancy and Rebecca are involved in a project trying to look at sort of early distinctions in very different types of [INAUDIBLE] process in the brain.
But there's also another method that might hopefully potentially help work a little bit better with infants because they can sit in their parents' laps in an upright fashion. They don't have to be laying down motionless in a scanner, and it's called near infrared spectroscopy, and it uses infrared light to measure changes in blood flow rather than magnet, the fMRI magnet.
And so that's actually something that we have as part of the [INAUDIBLE] project as well. So we recently obtained [INAUDIBLE] system, and we're trying to use it to study infant social perception at the moment. So hopefully, we will be able to take things like the habituation measures [INAUDIBLE] repetition suppression analogies of them that we can do imaging studies with [INAUDIBLE].
AUDIENCE: [INAUDIBLE]. Also, I didn't exactly hear the question, but you can also approach this kind of stuff with adult functional MRI. So I cut this lecture because there were too many lectures, but [INAUDIBLE] people [INAUDIBLE] in my lab are trying to understand a whole piece of the brain that's along the [INAUDIBLE]. It kind of runs from here all the way down to the temporal lobe down to there. It's about 12 centimeters long, and all on either sides of that region of the brain, it's just a rich microcosm of high level social perceptual [INAUDIBLE]. It's really fascinating.
[INAUDIBLE] a lot of the stuff that Lindsey talked about and that Ken talked about is computed in there. And so one of the things that we're trying to do is understand [INAUDIBLE] social perception by figuring out what its components are-- if they have any-- by mapping out the functional associations in that part of the brain.
And then just one example, not my work but work from Kevin Pelphrey's lab, he's used pretty much exactly the Amanda Woodward paradigm that Lindsey talked about. So you have the person who reaches many times for object A over here, and you just switch the position of A and B as you expect them to go through the same object over here or the same action over here. And then Woodward shows that infants understand-- whatever age, 10 months? I can't remember.
AUDIENCE: Five. 5 and 1/2.
AUDIENCE: Understand, expect the person to reach for the same object, not to just rotely do the same action. Well, you can do that with functional MRI. You can just show videos of that.
And what Kevin Pelphrey finds is that a big chunk of the [INAUDIBLE] focus is on more to the unexpected action than the expected action. So there's lots of work of that kind trying to map out in the brain some of this high level social perceptual stuff that Lindsey and Ken talked about.
LINDSEY POWELL: Yeah.
AUDIENCE: So one of the assumptions of a lot of this work, as Josh talked about quite a bit, is the idea that there's sort of this process of hypotheses that are developed over time. Now, has anybody taken a look at when infants learn to do certain types of things, and then kind of looked maybe correspondingly [INAUDIBLE], looked if there are particular physical maturation properties? So kind of similar to what they do in like B1 in mice with respect to the development of early vision in critical periods in those areas that shows that there's a physical correlate, or is it purely sort of a learning process where they actually are specifically connecting categories as opposed to [INAUDIBLE] maturation of particular pars of the brain?
LINDSEY POWELL: Well, I think it's much harder to study the physical maturation of the brain in [INAUDIBLE] living infant.
AUDIENCE: [INAUDIBLE]. So it might be you could do a lot of these kinds of things with looking with a young macaque and actually--
LINDSEY POWELL: Maybe. It's been mildly successful and mildly unsuccessful. Preferential looking is OK. Violates their expectation looking, [INAUDIBLE] tried to do a little bit of but not as successfully.
AUDIENCE: I just have one example. There's a study [INAUDIBLE] a few years ago by a guy named [INAUDIBLE] who was asking about the role of experience in development of [INAUDIBLE] systems in macaques The reason to do this in macaques is you can control their experience in a way that's unethical with humans [INAUDIBLE].
In any case, three baby monkeys for 6, 12, or 24 months without ever letting them see a face, never a single face in their whole life. They were born in the dark. They were raised in the dark. They were cared for by human caregivers who had hoods over their heads. So this is a really precious opportunity-- and they didn't have horrible lives. They got lots of nurturing and cuddling, but just by humans with hoods over their heads.
And they had rich visual environments, just no faces. So that enables you to act, at least in that case, whether experience of faces is necessary for the development of facial perception [INAUDIBLE]. He then used the methods that Lindsey talked about [INAUDIBLE] habituation. So you habituate to face A and you switch to face B in order to ask can the monkey discriminate those two faces?
So they did that test on the very first day the monkeys ever got to see a face at any point in their life, and the monkeys were indistinguishable from normal ones, which is a result so powerful and astonishing that I can hardly believe it. I mean, [INAUDIBLE], but this is so extreme. This says that it's not just that you can get very good face discrimination abilities.
I mean, [INAUDIBLE] fairly similar to one another. [INAUDIBLE] was just [INAUDIBLE] very subtly different faces and healthy monkeys who had never seen faces at any point in their life were as good as normal adult monkeys who are damn good at this. So that's just one data point. [INAUDIBLE] extremely astonishing. I would've thought that some of the faces [INAUDIBLE] innate, but you would still [INAUDIBLE]. This experiment says no, [INAUDIBLE].
AUDIENCE: Were they adolescents, or what age did they--
AUDIENCE: [INAUDIBLE] 24 months. 24 months is-- what's that? A juvenile. Did anybody go to [INAUDIBLE]?
AUDIENCE: Eight years old.
AUDIENCE: Two to--
AUDIENCE: Eight years old. Oh.
Like eight years old. Right. So they're still young.
LINDSEY POWELL: I'd say partly the ability to do that depends on whether or not the particular measure that you're using is a good measure of reliably getting individual differences, right. So it was part of the problem with a lot of these studies is that they work in the aggregate with whole groups of infants, but looking at any one individual infant's behavior in a violation of expectation study with just two trials is going to be hard to interpret. So any one [INAUDIBLE] macaque.
But preferential looking is actually a particularly good-- and dishabituation is a particularly good kind of measure because you give an infant or macaque two different choices [INAUDIBLE]. When you're measuring them over a period of time, every fraction of a second is a chance for them to make another choice. There's actually a lot of data that you get in preferential looking between two options [INAUDIBLE].
So some of these things, like violation of expectations through teaching, are things that you might not really see in every baby on every [INAUDIBLE] trials. In the preferential test, I think you can get better [INAUDIBLE], like when you track an individual's experience, or with physical changes to the brain.
AUDIENCE: [INAUDIBLE] mentioned that these deprivation experiments are really sort of interesting, and in that case, [INAUDIBLE]. They are only one sided in the sense that if you define something [INAUDIBLE], you can always make the argument that system was deprived and it actually needed stimulation for its normal development. [INAUDIBLE] you have a neural system that's already there and it's not been stimulated over many months, [INAUDIBLE] away.
AUDIENCE: It would, but when it's resilient--
AUDIENCE: Right. It's just it's a one sided kind of experiment. [INAUDIBLE]. That's fine. They got the result. But in a sense, it's not a completely powerful experiment. I mean, it's very--
AUDIENCE: They got lucky.
AUDIENCE: They got lucky. And then the other thing you mentioned [INAUDIBLE], the thing that's interesting about it, didn't they show them [INAUDIBLE] and those monkeys actually couldn't do it [INAUDIBLE] one kind of face. [INAUDIBLE] monkey face or a human face?
AUDIENCE: So what they did [INAUDIBLE] experience? In the initial testing, they were adult like and equally good at human faces and monkey faces. They then took those monkeys and put them for a month in an environment where they only saw human faces or only saw monkey faces.
And at the end of that period, the effect of the experience was not to improve performance discrimination ability at all. It was to throw out the ability to discriminate the unexperienced species. So there's this [INAUDIBLE] situation where some of those monkeys were put in an environment with human faces [INAUDIBLE] two years.
After that month, they could then discriminate human faces, not monkey faces. And then for years later in the environment with only monkey face-- or monkey and human faces, they never regained the ability to discriminate monkey faces. So there is an effect of experience, but it's a perceptual narrowing effect of eliminating abilities, not construction. Amazing.
AUDIENCE: So I was wondering [INAUDIBLE] faces, did they also [INAUDIBLE] experiment in which they expose them to, say, another object which they also have not seen?
AUDIENCE: Wouldn't that be great? They didn't do that. They haven't reported that yet. But, yeah. That's a great-- [INAUDIBLE]. Sorry.
AUDIENCE: I think also [INAUDIBLE].
AUDIENCE: Well, the other experiment of [INAUDIBLE] patterns on those hoods of the caregivers. There's just a whole [INAUDIBLE] to be done there. [INAUDIBLE] should be a 2008 [INAUDIBLE].
AUDIENCE: So one of the things I was thinking about-- just a quick question-- so when you mentioned about the [INAUDIBLE] path in infants, it seems like infants are so smart. They understand the perspective of other people and engaging in social behavior.
But then like in age three, for example, they felt the typical, like, the other theory of mind [INAUDIBLE]. And then they seem to regain it in, like, age five or something. So what happened in between? Or is it just a [INAUDIBLE] that for example, one is more visual, the other is more verbal? Or is it, like, depending on visual working memory, or something like that?
LINDSEY POWELL: So that's not a short question, is it? [INAUDIBLE] people spend a lot of time talking to me about. Part of an answer is definitely [INAUDIBLE]. So you can do implicit theory of mind [INAUDIBLE] two year olds, or young three year olds who will [INAUDIBLE] the verbal questions that you can do things that-- not looking time, but other kinds of implicit measures.
Like, for example, say that there are several boxes [INAUDIBLE] an object and experimenters put one and one [INAUDIBLE] someone else moved it to a different box. When that [INAUDIBLE] comes back, you could ask those two to three year olds [INAUDIBLE]. [INAUDIBLE] ask, do they need to [INAUDIBLE] find the object?
And three year olds, [INAUDIBLE], will be more likely to point when the experimenter has [INAUDIBLE] about where the object is and when they watch the transfer, and so they [INAUDIBLE] about where the object is. Does that make sense? [INAUDIBLE] coming back into the room, either with knowledge of the object [INAUDIBLE], knowledge of where they [INAUDIBLE].
And three year olds who fail a verbal [INAUDIBLE] are more likely to point for the ignorant experimenter than the [INAUDIBLE]. So it's definitely not like three year olds have lost some kind of understanding of [INAUDIBLE]. Definitely, the different tasks seem to [INAUDIBLE] than other [INAUDIBLE]. [INAUDIBLE] is entirely about executive function [INAUDIBLE].