Listen: Psychologist Stella Christie on Why We Must Fail

Associate Professor of Psychology and Cognitive Science Stella Christie argued that failure can be a logical necessity in psychology during her talk for the Aydelotte Foundation's Second Tuesday Social Sciences Cafe series.

"You need failure. It's not just something you can learn from, you actually really need that failure," she said during her talk in the Scheurer Room.

Christie studies cognitive development and her research examines how the mind acquires knowledge through analogy and comparison, the effects of language on cognition in development, and cross-linguistics and multilingual acquisition. She runs the Cognition and Development Lab at Swarthmore, where she and her students research how infants and children acquire concepts and language and how language development shapes children's learning about the world.

Audio Transcript

Jamie Thomas: Hi, everybody. Welcome to lunch. Welcome to our next in the series of Aydelotte Tuesday cafes.

Thank you for joining us on this second Tuesday. Thank you for joining us in this kind of grayish weather. It's kind of bizarre, but that's why we've got soup for you.

I want to say a quick thank you.

I'm Jamie Thomas, in Linguistics, by the way.

I want to say a quick thank you to all the folks who have been helping out behind the scenes, to make sure we've got the lunch for you, that we have this room, we have our AV support, from all of our friends in Media Services, that we have the vision for this series, which is focused on failures by researchers, and what the process of research is. That this up and down that happens actually leads us to processes of discovery, of maybe, something about ourselves as researchers, or something about the world around us, which we really try to celebrate here, through the process of liberal arts, learning and education.

Today, our talk will be from Dr. Stella Christie, in Psychology and Cognitive Science, to tell us why it is that we actually must fail. Hm.

Next time we get together, in December, we will hear from Krista Thompson in Philosophy, telling us how she fell in love with really old dead guys. And you know what, between why we need to fail, who these old dead guys are, there's a lot to hang out with this on. There's a lot to come back for, and in the spring, I should say, we've already got some things cooking for you. I don't know if the soup will be on, per se, but we're planning on having a bit of a blitz in February, where we'll kind of have a fast round of different researchers sharing, very quickly, anecdotes from their research, talking, again, about this process of discovery of the up and down of doing research.

So, please do keep chowing down, and grabbing more stuff from the back, but we're going to go ahead and get started so that we can learn why it is that we must fail.

Stella Christie, everyone.

Thank you.

Stella Christie: Thank you, everyone, for coming on a Tuesday. Must be really crazy busy Tuesday for all of you.

So, I'm going to take us down a memory lane of this building that is no longer here.

That used to be my window. Used to be my playground. So, this building is no longer here, and I'm going to tell you a story that this is the only picture of Papazian, not the only picture, but the picture that I found, but it is not to my liking, because I actually want to show you. Usually, in the summer, there's a hug air conditioner up right there, and this air conditioner is really loud, so I actually don't like using air conditioner in the summer, and so I usually just leave my door open in the summer.

And, one hot summer day, there came a student wandering around the building, and asked me, Professor, do you have time? And I said, sure, I have time. It was a hot summer day, I was not really focused on what I was reading, anyway. It turned out to be a high school student who was really interested in coming to Swarthmore, and the student said, what do you do? And, I said, oh, well let me tell you what I do. I always love telling students what I do.

So, I spent some minutes, I tell the student that what I do is I try to understand how humans acquire knowledge through analogical reasoning and language learning. So, I'm really curious about how do we get to reason about the world, and how is our process of reasoning is through the process of acquiring language, and through the process of analogical reasoning.

I spent quite a lot of time telling her about what I do exactly, in order to answer these questions, and she was listening attentively, and after I finished, she said, so, you don't help people.

And I said, no. You're right. I don't. I don't help people. I guess, I don't do that.

And she was so disappointed, and she said, thank you for your time. And I said, you're welcome.

I don't know if she went to Swarthmore in the end, or not. But, you know, I tried my best.

So, you may think, coming to this talk, because the theme is failure, that I would actually tell you that because of that failure of convincing the student that I actually help people, I change my researched outcome, and I become a better psychologist, and I actually help people. Well, no. That's not what I'm going to acknowledge. I still don't help people. The only thing I study is how people think.

And, in fact, I'm not going to tell you about the kind of failure that you actually know very well, which is, we often think about, you analyze the failure in one domain, and because you analyzed your failures, then you think you will be, at the very least you know how to obtain success in the same domain. So now that you say to yourself, I've failed, I know what to do to be successful next time.

Well, I'm actually not going to tell you about this kind of failure, because I still don't know. A lot of the times, I still don't know how to be successful in that very one domain.

And, also, this is something that you know very well, so I don't think I can add much to this kind of failure. Instead, what I'm going to tell is about a different kind of failure. What I'm going to discuss with you, today, is about looking at failure, analyzing failure in one domain, actually helps you obtain success in a different domain.

So, it's not about that time you failed, and next time you will succeed, but about that time you fail, and what you actually take from that failure to actually inform you.

So, in fact, this kind of failure, I would say that, the failure itself is not necessary, because if you succeeded in the first place, you don't need that failure in the first place, right? So that failure, it will be useful to learn from that kind of failure, but it is not a necessary failure.

Today, I'm actually going to tell you, and try to convince you, that in empirical research, as a scientist, the failure is logically necessary. You need that failure. It's not just something you can learn from, you actually really need that failure.

So, I'm going to use a couple of examples from my work to show you why the failure is logically necessary.

So, to do that, I'm going to start with something that you're very familiar with. Just, I'm going to coach it in terms of this logical necessity. So, we must fail, as empirical scientists, because it's a logical necessity.

So, I'm going to start with a tale of two pills, and here is a pill, let's call this pill Zyrtec, and in order for you to say, confidently, that Zyrtec's really improved your allergy, what do you need to know? To draw the conclusion, very confidently, that this pill works, that Zyrtec is really improving your allergy, what do you need?

Audience Member: Some other intervention.

Stella Christie: Great.

Audience Member: That did not improve your symptoms.

Stella Christie: Excellent. Perfect.

Some other intervention that does not improve your symptom.

And, this is not just a nice thing to have, this is necessary. It's logically necessary to have it.

So, usually you hear about placebo, right? You have to be able to say that the placebo fails to work, only, and only if the placebo fails to work you can say that the pill actually works. So, this is usually R-D-A 101 in psychology, but this is very important. It's logically necessary. In order for you complete anything works, it's logically necessary that the placebo failed, right?

So, this is the basic premise, and in many of the works, scientists and psychologists, we actually heavily rest on this assumption, okay?

So, let me illustrate that with a number of examples.

So, as I said to the student, my work is looking at the fundamental aspect of human reasoning, and that is analogical reasoning. I am actually not going to focus on analogy perse, because the topic is failure today, but I can tell you, briefly, why I am so, until today, many years thinking about analogy. It's still something that keeps me awake at night. It's something I question that's exciting.

Analogy is, not only fundamental for everyday reasoning, so every time you use a metaphor, you say that this morning I killed two birds with one stone. What you're doing there is actually analogical reasoning, because you are thinking about a pattern that can apply to many different kinds of situations, and that's what analogical reasoning is. And because you can think about the same pattern that applies for many different situations, including novel situations, it breeds creativity.

This breeds scientific discovery. Most of scientific discoveries come from analogical thinking and analogical reasoning.

In many ways, people who believe, and there's a construct of intelligence, think that intelligence rests on analogy.

So, in fact, if you look at IQ tests, and the Raven's matrices tests, if you're familiar with Raven's matrices tests, those are analogical reasoning ability. So, analogy is a very important process that leads to many good things in human reasoning, and that's what keeps me awake at night.

So, I think a lot about how you get to analogical reasoning, because we are not born, from evidence that I actually collected, we're not born with analogical reasoning. So, how do you do that to ask this question? Well, I would, with this kind of populations a lot, so I ask this kind of population, this is my son, I ask this kind of population a lot about questions on how do you reason for analogical reasoning?

You probably can guess, that when you work with these kinds of populations, you will be faced with a lot of failures, okay? You cannot just ask them simple questions and then they will answer you simply. Failures are abound, and that actually are important to understand the human reasonings overall.

So, here's one example. Here's exactly a simile that I use to ask 2, 3, and 4 year olds, and I want you to answer the question, okay?

So, see this card? Which of these is the match for the top card? Left card or right card?

Audience: Left.

Stella Christie: Thank you. All right.

I doubt that any of you need extra time to answer this question, right?

And you probably can see this is analogical reasoning. In a very, very basic fundamental thing, which is that the way you're able to say that the left card is the match, it's because it contains the same relation.

So, here is a relation of identity or a relation of sameness. Here is also a relation of identity or relation of sameness, so you match these two relations together, and bingo, you've got the match. Right?

So, you are not matching based on physical properties. You are not matching based on how the color looks like, how the shapes looks like. You are matching based on the relation, based on the pattern, and that is the fundamental core of analogical reasoning, you perceive pattern similarity. Not about the objects, not about the physical things, but about the patterns, right?

So, I ask this to 2, 3, 4 year olds, and I actually started this work thinking that they will do this, and I will actually test my other hypothesis with more complex kind of things, right? Notice that the relation I'm testing here is not relations like symmetry, or justice, or any kind of relation that's complex. This is relation that is so banally mundane. Same. I mean, how could you not know what same is? Well, it turns out that 2 and 3 year olds don't know.

So, when you ask this question, they're like, uh... They chose at random, and it was so mind blowing to me that I had to repeat this experiment like twice, three times, and then I actually gave these as talks, and I ask people, can you please reproduce these things, and like make sure that I'm not just gathering some not so smart kids.

This is done in my graduate class, and it turns that I wasn't. My kids were actually smart in normal ways, and this was replicated many times, and all labs actually got the same results, that 2 and 3 year olds answered this at chance. So, they were random in this. They couldn't tell you which one is the match. It is not because they don't know the match, by the way, it is not because they don't understand the task.

So, if I actually change this task into, you know, like flower in a vase, and another flower over here, and a car over there. They would actually tell you the match. So, if it's about how it looks like, then they can tell matches. Right?

So, this failure is really meaningful, because it gives me an understanding of a limitation of the mind to reason analogically.

All right. So, I cannot just rest my conclusions on one failure alone, it has to be combined with a pattern of success.

So, what on the questions... So this is like a [inaudible 00:15:03], you can see that the chance is 50 percent, and the 2 and 3 year olds just choosing at chance, if you look at the relation.

So, what I want to test is my contribution about [inaudible 00:15:18] thinking, that the way that you can so easily do the task, that simple analogical task is partly, not all, but partly because of language, and as that language fosters analogical reasoning.

And so, to do this, I actually, to buttress this theory, to give evidence for this theory, I actually have to do empirical tests, and my empirical test is just like the placebo. Notice that the placebo looks identical to the pill, except that very one thing, which is that Zyrtec component of it. So, that's how I want to do it in my experiment, right? I have a failure from that very test, and I should actually look at a very same test, except, I want to test this hypothesis that language is going to give you the ability.

So, what I have to do, I am going to give them exactly the same kinds of stimuli, and I also test the same age group of children, preferably around the same time, with the same kind of populations, and that's what I did in this very first experiment, but except now, I'm actually inserting a language.

So, there's a couple of very specific theories that I have presented and I have proposed, which is how exactly the language works.

So, today I'm just going to tell you about one of them, and I'm going to tell you that the language kind of works in a very interesting way, that even if you give them novel labels, it actually works.

So, the 2 and 3 year olds now see, now in a new group of 2 and 3 year olds, see the same kind of stimuli, the same task, which of them match, except the way I ask the match is not true match, but true label. So, I say first, see this? This is called traffic, and there's no training whatsoever. They right away see this card, and they hear the word traffic. They don't know what traffic means. The reason why I gave them the word traffic, and not the word apple or the word square, even, is because I do not want them to rely on concepts that they have coming in, because I actually want to test the hypothesis that novel label, just having a label, in your representation allows you to perceive similarity of patterns.

So, they don't know what traffic is. I never tell them what traffic is. Sometimes I ask, at the end of the experiment, what do you think traffic means? And, sometimes they do actually give me answers. The 3 year olds, in particular. Three year olds, well, they're just cute when just sit there and look at me.

Okay.

So, when you actually do this with a traffic, a very simple difference, and you ask which of these is also a traffic, lo and behold, the 2 and 3 year olds are actually now able to say, this is traffic, and they can do that.

Remember, these are the same age of children. So, it's not about that they have developed more. It is particularly about the language.

Now, I only are able to say that the language is necessary, that the language plays a very important role in your analogical reasoning because I have these two comparisons.

So, this is the baseline with a non-label, it's hovering around 50 percent, and this is with a novel label. The improvement is not much, but it is significant. Only because I compare the failure, without label, and the success in the novel label, I'm able to pinpoint to particular theory, that language plays a very important role in fostering your analogical reasoning.

So, that's an example how failure is necessity. Without the failure there, these results would be mundane. I'm not going to actually give any kind of insights about the workings of language and the workings of the human mind's representations in terms of analogical reasoning.

Now, let me take you to another work that is highly related to this, and see, again, about failure. And that is, when cognitive psychologists, like myself, ask questions about the origins and the learning process of the human mind. We have to do this true comparison, okay? We often, developmental psychologists, then use young children and ask, what differs within young children and adults? What differs within 2 year olds and 7 year olds? What are those differences between 2 year olds and 7 year olds? We can also ask this across species. You can also ask what differs between us humans and other species.

And so, in my work, I actually have looked at, also, chimpanzees, bonobos, orangutans, and gorillas. These are the phylogenetic trees, so bonobos are actually our closest cousin, and chimpanzees next, orangutans next, gorillas next.

And by the way, as a trivia, you may not know this, but I'm from Indonesia, so orangutans come from Indonesian language. Orang- means people, and -utan means forest, so in Indonesian language, it means people from the forest. Just to tell you how the closeness and similarity of these.

So, with the same thing, this very fundamental, simple thing of our M-P-S, the task, the relational match and sample task, you can also ask the questions, how do other animals, especially the ones who are really close to us, right? It doesn't make much sense if I started this program by asking whether ducklings, or chicks, or ants can do this, because if they fail, it isn't going to tell me much. So, some failures, you have to actually have to couch the failure so it actually is a reasonable failure of comparison.

So, when you actually test this with chimpanzees, identical task with chimpanzees, even if they're hundreds and hundreds of trials, chimpanzees actually fail to do this. These are adult chimpanzees, so they behave like the 3 year olds, and furthermore, the chimpanzees are given hundreds and hundreds of trials.

Now, you may say, well, that's not a fair comparison. Maybe the task is not interesting. The task was actually interesting for chimpanzees. They were actually rewarded for this, so every time they chose correctly ... The way we test chimpanzees is you allow them to touch one of the two panels, and every time they touch the correct one, then you give them rewards. So, they're motivated to touch the correct panel. So, you see, after like hundreds of trials, whether they have reached that criteria.

So, the chimpanzees are very motivated to actually do this correctly, but studies from many labs actually showed, and these were started by Premack, David Premack, who's at UPENN, in the 1980s, and up to today, none of the chimpanzees every managed to do this, except in one situation, and that situation is if I first taught these chimpanzees, or the experimenter taught the chimpanzees with a symbol training.

So, the chimpanzees were taught, look, every time you see two that are the same things, then raise the symbol diamond, and not the symbol heart. So, if you see two buckets, raise the symbol diamond. If you see two cans, raise the symbol diamond, and so on. So, they were first trained until they got really, really good in knowing that when they see two of the same thing, they have to raise the symbol diamond, when they see two different things, they have to raise the symbol heart.

So, only after they managed to do this, if the chimpanzees managed to pass this symbolic training, really know the symbols, just these two symbols alone, then they are able to do a relational matching sample test.

You may now notice, again, there is [inaudible 00:23:31] here, there is a failure without the symbol system, and there is a success with the symbol system, and only this pairing, the failure without the symbol, and the success with the symbol allows me to say that in the cognitive constructs of the chimpanzees, symbol training is instrumental, is necessary, for them to perceive that relational sameness, to perceive that sameness in terms of pattern.

And you probably draw a connection between the chimpanzee's results, the two sets of failure and success that I have told you, and indeed, together, you can see how these sets of successes and failures tells us that language is very, very important, if not necessary for analogical reasoning.

The chimpanzees do not have language, but what they have is a rudimentary training of a symbolic system, and that is what language is. Language is a symbol system. You have symbols for things in the world. Cup is a symbol for that concept of a cup, and that symbol is arbitrary, just like the heart is an arbitrary, it can be heart, it can be ... Sorry. Just like that diamond, it can be heart, it can be diamond, it can be anything, it's arbitrary. So is our language system, but we have it as a system. They don't yet have it as a system, but even the rudimentary symbols are very useful for the primates, for the chimpanzees.

But, the highlight here is that you can only draw this conclusion if you look at the patterns of failures and success. The failures are necessary.

Now, let me move on to a second set of examples that shows how analyzing failure in one domain actually tells you a lot about a different domain, and it's still a logical necessity.

So, to do this, I'm going to give you an example from this work, so as when many psychologists talk, you, again, are going to have to respond here.

So, here's the experiment, all right? You are the participants.

There are these two sets of groups. There are two groups of animals, and if I want to find out about all animals, right? If I want to find out about all animals, you, as the scientists, and you only have a chance to look at one group, but I want to find out about all animals, and you have a chance to look at only one group, which group do you want to look at?

Audience: A.

Stella Christie: Thank you. All right.

Why do you want to look at A?

Audience: [inaudible 00:26:32]

Stella Christie: Different kinds?

Audience: [inaudible 00:26:39]

Stella Christie: All right, because sharks are ten different kinds. They are different kinds of animals.

So, what about different kinds that makes you want to look at them to find out about all different animals?

Audience: [inaudible 00:26:49]

Stella Christie: More categories means more information, yes.

So, this reasoning comes naturally to you, and this is what we call inductive reasoning. Generalization by inductive reasoning.

So, if I give you these kinds of two sets of evidence, it's natural for you to have your inductive reasoning turned on, and say, the diversity of the evidence will give me better generalizations. That I'm more confident to generalize based on a set of evidence that is more diverse, right?

This is pretty abstract, currently, in the ... This is the actual stimuli we give with young children, but these reasonings are very important in everyday life.

Why is that?

Because you are not faced with every single data from the world. That means, for you to draw conclusions about, let's say, what are the food that actually is served in shore, is good or not. You actually make generalizations, right? Now if you have actually tasted food from many different kinds of occasions, not only Science Tuesday Cafes, but also like faculty lecture, or board meeting that happens in shore, now your generalization is going to be a lot stronger. But, if it only comes from one series, then it is going to be less stronger. But, it's probably impossible for you to get the data, for every single data, for you to make generalizations. And yet, you have to operate on generalizations.

Every thing that you think about in the world, you are basing it, mostly, only generalizations, and this generalization has to come from somewhere, and the question is, where is that somewhere? It turns out that not all things are gained from generalizations, so that's the basis of inductive reasoning.

So, now, surprisingly, when we asked 5 year olds, they actually ... This is work that I did with Alexander, my former student, Alexander [inaudible 00:28:52], who is now at Yale pursuing a graduate study. When we actually did this, 5 year olds are, again, failing. So, they chose randomly at chance.

So, despite thinking like you do, that generalization is better from a diverse than from narrow evidence, the 5 year olds were choosing randomly within this diverse and narrow evidence. Our work was not the first one to do that. In fact, this is a very, very old problem. Starting in the 1990s, there's at least by now, there are almost like 30 studies that document, over and over again, not until around the age of 9, children spontaneously make use of this inductive rationale. Okay?

So, Alexander and I got to think about this failure a lot, and what we thought about like, why did they have this failure? All right. So, the very first thing that you may think about why did they have this failure, is because they simply don't have the inductive rationale. Right? They just don't possess this. They don't get what inductive rationale is, they haven't processed anything about it.

However, there is actually a possible second explanation. There are a couple of other explanations, but the second one is probably the strongest possible explanation, and it's not because they don't have inductive rationale. They actually do have inductive rationale, but they perceive both sets of evidence as diverse.

In other words, maybe what they're seeing is that a zebra and a tiger are two different animals, and so are a zebra and a horse. They also are two different animals. So, if you ask me, which set of evidence is more diverse, which set of evidence is more different from each other, then you're going to say, well, they are both groups are diverse. Both groups are different.

So, that is a possibility, and these two explanations, obviously, is important for us to determine what actually is going on in the mind. And it's the failure of this that make us think about these two possible explanations.

So, when we think about these two possible explanations, we are actually thinking, let's test this. Given this failure, how can we test which of these alternative hypotheses is actually true. So, to test this, then, we actually look at people. So we test this thing, they again show failure in our lab, so we replicate a prior work that shows they are unable to choose between these two sets of animals.

And, Alexander and I were the first ones to say, well, let's think about people. Let's ask the children about people. Why? Because there's no way that the children will say that a boy and a girl is just as similar as a boy and a boy. They probably know very well that a boy and a girl are very different entities, and a boy and a boy are less different than a boy and a girl, right?

So, this was our intuition, and if this intuition is true, then we should see ... And if they actually have the inductive rationale, and yet, what they did before is simply not differentiating between different kinds of animals, and lo and behold, they did.

So, when you actually, in the same test, those children who like were at chance at this one, they were actually successfully saying, if I want to find out about all children in the world, then I'm going to look at this group of children. They correctly say that the more diverse evidence is the one that's better for generalization, okay.

So, we know from this pairing of failure in a domain of animals, and a pairing of success in the domain of people, then one of these explanations is not true, and it's that one. So, it is not that they don't get the inductive rationale, but it's because they think about different sets of data, different entities, differently. And, if they actually can see the diversity, they're able to tag on, and explain this, using inductive rationale, and this is really powerful, because it tells us that inductive rationale is there from the very beginning, but the manifestations of those inductive rationales really depend on how you perceive the simile in the first place, how you perceive your data, how you perceive the entities in your environment.

And, this only comes from us, together, thinking about the failure in one domain, and the success in another domain, like people versus animal, for example.

So, the failures are critical. Without the failure, these results, they're successful with the people, it's interesting, but it's sort of not that telling to us, it's just mundane, right? They can do it, right, just like adults, but that's not what we found.

Audience Member: [inaudible 00:33:56].

Stella Christie: Yeah?

Audience Member: [inaudible 00:33:58].

Stella Christie: Yes.

Audience Member: ... when they were looking at stripes.

Stella Christie: When they are looking at the stripes?

Audience Member: Stripes.

Stella Christie: Oh, okay.

We did, but we also know that if I test this ... There are a lot of similes in the study, so another possible simile is a zebra and a mouse, so that's even more striking. So, even with a zebra and a mouse, they still are failing, meaning they are still at chancing this, but that's a good questions, because I think you're thinking about the stripes maybe threw them off.

Yeah, so we tested with many different kinds of animals.

Another question?

Audience Member: The question is deeper than the stripes.

Stella Christie: Mm-hmm (affirmative).

Audience Member: We take it for granted that the more important similarity ...

Stella Christie: The kind.

Audience Member: The species ...

Stella Christie: That's right.

Audience Member: ... rather than the superficial similarity of stripes versus solid.

Stella Christie: Exactly.

Audience Member: But that's, to me, that seems that's knowledge of the world.

Stella Christie: Absolutely.

Audience Member: Not our reasoning ability.

Stella Christie: No, that's precisely what we're saying.

It is not about the reasoning ability, but it's about knowledge of the world, and what's really this pairing of things that tell us that the reasoning, the inductive reasoning, has to be paired with that knowledge of the world to predict how you actually draw a generalization.

Audience Member: So you think that could be applied, now, to 3 year olds? Because you could say that by using the world trumpet, you've signaled that you think of those two objects as a single thing, but you didn't say a pair of trumpets.

Stella Christie: That's right.

Audience Member: If you said, look at this pair of trumpets, which of these other cards is a pair of trumpets?

Stella Christie: Yeah.

Which, maybe... That's ...

Audience Member: And what's the correct answer?

Maybe trumpets, or things that have right angles.

Stella Christie: Examples.

Audience Member: And non-trumpets don't have right angles.

Stella Christie: Exactly.

Yeah. So there's a lot of these exciting questions that generated from the failure, right?

And, indeed, we have actually tested one of our qualified questions, which is, what does a language do, exactly? And I don't have time to go through everything, but one of the things that we tested, is we actually use, not language, but just like a connecting bar, and really tell them, look, they are connected to each other, and in this case, they were actually still at chance.

So, there's something above and beyond about the language that actually give us that connection.

But, absolutely, it's about how you perceive things.

My claim is, indeed, that language changed the way you perceive things in the world.

So, one last thing that is actually an illustration of this, some failures are truly logical in the children's mind, and I think this is ... We actually haven't done this with adults, but it's something, a set of interesting questions.

And that is, when we look at these abilities, so we now know that with people, the children are successful. They can perceive the differences, whether it's deeper, or through perceptual similarity, and then they actually applied inductive reasoning.

And then, my student, [inaudible 00:36:46], actually came in, and said that, well, you know this is all and good that they actually are able to perform this rationale of inductive logic, but we also know, from many prior literature, that children are sort of biased, preferring people from their own group, right? And this is not biases that are limited to young children, this is biases that are present in all of us, and I really mean all of us. We all prefer people from our own group. We may counter this with explicit effort to ourselves, like don't prefer people of our own group, but these biases are prevalent in the human mind. Okay.

So, what [inaudible 00:37:28] was testing is this. If now, I ask you to make a choice between people from your own group and people from diverse race, how are you going to choose? So, the task is this. Children were told that here is a far away school called the Jiffy School, and this is, again, done with 5 year olds, and then I need to buy a toy for the children in the Jiffy School, and I don't know how the toys look like in this bag, and in this bag, because we don't want to show them the toys, so they don't choose based on the toys alone.

But, I know that this group of children like the toy in this bag, and this group of children like the toy in this bag, okay? So, which toy do you think will be liked by most of the other Jiffy children? Which toy should I buy to make all the Jiffy children, or most of the Jiffy children happy?

Okay.

Now, if you think that they don't understand most, they do, because they're not at chance. Instead, what happened is that they're preferring the diverse evidence, only 30 percent of the time. So, these children strongly prefer that say, I need to buy toy B that is liked by this group of children. The participants, at this point, are 100 percent Caucasian white, from Swarthmore environments. They grew up in majority white environments.

So, because what I pitted here is between your own group versus the diverse evidence. Even though we know that children are capable of reasoning true rationale logic, pick diversity for strongest generalizations, when that's actually contrasted, you have to make a choice between that rationale with your social environment, they pick for their social environment. Okay?

So, this failure of choosing the diversity is actually kind of logical, and tell us about the learning processes, and that's very important for us.

Then, we actually went on to say, how flexible is this? So this is an ongoing work. How flexible is it? So, we first tell them like, look, this is the Jiffy School children, okay? And that's it. We remove the Jiffy School children, just one second to show them the Jiffy School children, and now we ask them, all right, new group of 5 year olds, again, all white participants from Swarthmore area neighborhood, which toy do you think I should buy? And now they are actually, 60 percent of the time, choosing the diversity.

So, they change their rationale if they're given evidence to the contrary. This is good news. It means that A, environment affects your learning, that's why we see the failure somewhat before, and that failure is logical. And B, at least at this age, we know, that this is still malleable, that you can change your perceptions about generalizations, if you're given evidence to the contrary, if you're given evidence that the school comes from a mixed background.

Question.

Audience Member: What happens if you do the experiment the original way, but ask them before they give their answer, what they envision the population of the school to look like? Because what comes to my mind as an immediate explanation is that kids from the white background ...

Stella Christie: Good.

Audience Member: Are assuming, without thinking, that of course the school has white kids.

Stella Christie: Yeah, good.

And we actually ask this question at the very end, in our original experiment, and what we found is a little funny. There is a tendency, but not significant, yeah, that they do, we give them 4 pictures. What do you think the second graders look like, because we asked about first graders, what do you think second graders look like?

So, the 4 pictures, like all white girls, all white boys, and then mix boys and girls of all races, and then boys and girls of just whites, and we thought they were going to say, in the very first study, boys and girls of whites, but what we found is that they actually tend to pick their own gender. So, the girls tend to say the school is going to be white girls, and the boys said the school is going to be white boys.

So, it does tell us that what they hold in the representation, but who is the population depends on their environment. So, [inaudible 00:41:53], in her thesis, actually, right now, going out to Chester, a black majority neighborhood, as well as West Philadelphia, which is a very mixed neighborhood. There a black, African American, obviously black muslim immigrants, Chinese immigrants there, as well. To ask, how does this diversity versus ingroup-ness, play out when you are confronted by different learning environments?

And this is very important, because, as I said, even in adults, the ingroup bias is very strong, and even though you, and what you've probably seen, is that, similarly, like in kids, who know how to do rational logic of generalizing, using of diverse evidence, you may actually be really compelled to go for your own ingroup, and that can undermine many important things, like do you choose a policy that benefits people who are diverse, or a policy that benefits your own ingroup? And I think people haven't started asking that question, and that is a very important cognitive question that even adults have to face in every day life, and we see these roots in young children.

All those comes from understanding the failures, and I just want to close there and say, we must fail, because it's necessary for us to understand the structure of the mind, and when you love psychology, you're really, really keen on thinking about these failures, right?

Papazian was beautiful, but our new lab is even more beautiful, okay? So, I want to thank you all, and also have a plea, we need your support. All this data must come from somewhere, right? This is not made up, and it's very difficult to test young children.

If you, or you know colleagues that have ages 1 to 7 year olds, please join us at a kid's lab. These studies are really fun. You see that the studies are fun, right? They are, obviously, not intrusive, the children love doing these studies. The children get gifts, get a tee-shirt that says "kid scientist", with this logo, and the parent also gets a gift. You also get a bag, so, it's an incentive for you, and there's free parking, and you can park there. All right? Parking is an issue here.

We moved to Whittier Hall, by the way. All right, thank you. And I'll take questions.

Yes?

Audience Member: Is there any way to take that analogy, in [inaudible 00:44:26], if you know what I mean. And I think hearing about nonhuman animals, there must be ways that most organisms have a way of judging likeness between situations.

Stella Christie: Yes.

Audience Member: That isn't dis-embedded in this fashion that is lived and then maybe you couldn't test in the same way ...

Stella Christie: Absolutely.

Audience Member: [inaudible 00:44:46].

Stella Christie: Absolutely. Yes.

Audience Member: And now it's [crosstalk 00:44:48].

Stella Christie: Yes, yes, yes. Absolutely.

In fact, I would actually, I would not say that animals do not have rudimentary analogies, so to answer Tim's question, are there a kind of analogy that is more appropriate for the animals? I actually have tested that. So, I've tested using spacial paradigm. So, their knowledge of space. Can they see relations of space like top to top, middle to middle, bottom to bottom, in order to retrieve an object? And when I tested that with chimpanzees, orangutans, and bonobos, we see that the bonobos and chimpanzees are actually really good, but the orangutan is not so good.

So, there are some kinds of analogies, like spacial analogies that I have tested that they actually are successful in that. So, analogy, being able... But this is going to a deeper point, it's right, that being able to perceive sameness in pattern, analogy also used for problem solving, right? Because you can imagine, you can solve one problem, but if you think about the pattern, then you will be able to do that.

It's very much dependent upon domain knowledge. So, how well do you know a particular domain? So, it's not that the children are failing across the board, for some kinds of things that they know very well, like if you ask a 5 year old if a tree has a knee, where would it be? The 5 year old would, now, actually be able to map, and that's a simple analogy. So, the domain, the knowledge of the domain itself is very important to predict how likely it is that you can transfer between one pattern from known to unknown.

Yeah, thank you.

I see a question in the back. Yes?

Audience Member: When did you ... Have you done anything [inaudible 00:46:29].

Stella Christie: I haven't looked at that question, but tell me why you're interested in that question.

Audience Member: Well, because when I first looked at it, the first thing I thought was, [inaudible 00:46:39].

Stella Christie: Oh, okay.

Audience Member: Because they both have red, and some tone of red, and they both have [inaudible 00:46:47] and ankles, and yet you took it as completely obvious in some evidence that it's the left one, and yet my first instinct was, oh, it's the right one.

Stella Christie: Okay.

Yeah, well thank you.

Audience Member: [inaudible 00:47:03]

Stella Christie: Yeah, thank you for the question.

Yeah, in our study we actually did 8 trials, so we have to do what psychologists call counter balancing, so for half the trials, the correct answer is on the left, and for half the trials, the correct answer is on the right. It is exactly to make sure that if we have a failure, it doesn't come from simply looking at one side.

Audience Member: Oh, no. I didn't mean left and right. I mean either characteristics. You're thinking about just one characteristic.

Stella Christie: Mm-hmm (affirmative).

Audience Member: And I'm thinking about well, the right one, they match in color, and they match in type of shape.

Stella Christie: Yeah, so the type of shapes is also something that we play with, so we actually have different kinds of shape, making sure it's not about the edges, it's not about the angles, so nothing about those kinds that kind of, would simply tip them off.

So, we show many different kinds of shapes.

But, of course, like any given things, there's interpretations, always. The question is, what tips you over to the patterning interpretation. That's the question, yeah.

Yeah, Paul.

Paul: Just following up on that. This is sort of getting [inaudible 00:48:16] primitives. Sometimes the correct answer is socially defined.

Stella Christie: Yes.

There's some socially defined, but it also, like cognitively interesting, right? So, perceiving the pattern itself, right? Rather than perceiving other things that might matter.

Paul: But things like the squareness and roundness of the shape, that is also a cognitive thing.

Stella Christie: Yeah.

Paul: [inaudible 00:48:35] a value judgment, in a sense that the relationship is more important.

Stella Christie: Yeah.

Paul: [crosstalk 00:48:42] I'll just finish with this one comment. [inaudible 00:48:46] what's the next number in the sequence? And the parents are arguing, one of them is saying, oh, well the first 9 numbers are all palindromes, and the next one must be 11.

Stella Christie: Yes.

Paul: And the other parent comes up with some even more complicated thing that gives ...

Stella Christie: That's a great one.

Paul: The kid's in the background saying, the next one is 10, and the parents are just telling the kid to be quiet because the grownups are busy doing real math.

But again, that sort of, it's the issue that what you think is important is, to some extent, dependent on your context of social [inaudible 00:49:27].

Stella Christie: Yeah, absolutely. Absolutely.

And what we're trying to figure out, is because patterns are important in our every day life, and that's why I was interested about how you get to that pattern in the first place.

And, some things like, colors and shapes, children are very good without having to, you know, without having to train them in any way. Right.

Thank you, again.