0:00

The

0:00

first step in taking care of your body

0:02

and your mind begins with eating and drinking

0:05

right every day. And what better way

0:07

to do that than with delicious palm wonderful

0:09

one hundred percent pomegranate juice.

0:12

Palm wonderful pomegranate juice is

0:14

whole pressed with no added sugar,

0:16

no preservatives, and no fillers. It

0:18

comes with seven hundred milligrams of polyphenol

0:21

antioxidants per a down serving and

0:23

has as much potassium as a banana.

0:25

Palm wonderful tastes great right out of the

0:27

bottle or as an ingredient in all sorts

0:30

of delicious drinks. Drink it daily,

0:32

feel it forever. To find delicious recipes

0:34

and learn more about palm, visit

0:36

palm wonderful dot com.

0:40

Hello, everyone, and welcome to the Mindscape podcast.

0:42

I'm your host, Sean Carroll's. very

0:44

natural when things go wrong, when

0:46

when something is very different than what you were

0:49

led to expect it would be, to

0:51

demand an explanation. To

0:53

ask why is it like this?

0:55

What is the reason why the explanation

0:58

for the state of affairs that you didn't expect?

1:01

But what does that mean? Exactly. Does

1:03

an explanation mean some connection

1:06

between one event or another? Or is it

1:08

a way that the world is. If

1:11

you remember back when we talked to Judea Pearl

1:13

about cause and effect, and

1:15

Pearl of course is one of the world's experts

1:17

in teasing out what

1:19

causes lead to what effects in all

1:21

sorts of messy situations in the

1:23

real world. He claimed

1:26

that babies spend their

1:28

time making causal maps of the world

1:30

saying if I poke this, it reacts

1:32

in this other way. So what

1:34

exactly is going on? Not just in babies,

1:37

but in grown ups also when we human

1:39

beings construct this

1:41

image of the world or this model

1:43

of the world, which says this

1:45

is an explanation for this other

1:47

thing over there. Well, today's guest is

1:49

Tania She's one of the world's experts

1:51

in exactly this question. She's

1:54

a psychologist at Princeton University

1:56

and her lab at Princeton is called

1:58

the Concepts and Cognition Lab,

2:00

which I love that as a name. I would love to

2:02

work in the concepts and cognition lab

2:05

where she studies what an explanation

2:08

is, what do we mean when

2:10

we say, here's the explanation for

2:12

that. What do we what should we mean?

2:15

What are we talking about philosophically as

2:17

well as psychologically? Do

2:19

people agree? On what explanations

2:21

are Why do we want them? What

2:23

what is it psychologically that

2:25

moves us to burst into

2:27

the room and demand an explanation? For

2:30

something. I like it because

2:32

it's not only a psychology topic,

2:34

but of course, it has something to do with

2:36

the structure of the world out there.

2:39

The fact that there are things in the world

2:41

that we accept as explanations.

2:43

This is an interesting fact all by itself and

2:45

there's other things that someone might say you

2:47

demand an explanation, which most of us would go,

2:50

no, that doesn't really work. Going

2:53

down to very, you detailed

2:55

questions at their, for example,

2:57

at the intersection of science and religion.

3:00

When you say, why is there something rather

3:02

than nothing? And someone says, well, God

3:04

made it that way. Is that a

3:06

good explanation? Does that satisfy

3:08

you? Should it satisfy you? And I'm not

3:10

telling you what the answer is. I'm saying these are very

3:12

good questions. We're gonna get into it,

3:14

so let's go. Donie

3:33

welcome to the Mindscape Podcast. Thanks for

3:36

having, Nishan. So you work on

3:38

explanations, which is

3:40

something that to me sounds like very natural

3:42

for a psychologist to work on. You know,

3:44

I it's always funny to me when I read

3:46

psychology papers and they say, like,

3:48

this is actually an understudied area of

3:51

psychology. And I I think that yours is

3:53

one. Is it true that explanation is an under

3:55

studied area of cognitive psychology? I

3:57

think it depends which types of explanations

4:00

you're thinking about. So two communities

4:02

have thought about explanation for a long time. One

4:04

is, social psychologists. They've been really interested in

4:06

how we explain our own behavior and other people's behavior.

4:08

And there's decades of research specifically

4:11

thinking about explanation in that context. And

4:13

another community that's been really focused on

4:15

explanation is educators, people thinking about

4:17

educational psychology and learning. An

4:19

explanation in those context. I think

4:21

what's newer is people

4:24

appreciating how fundamental

4:26

explanation might be to our everyday cognitive

4:28

lives. Not just in the social

4:30

domain, but more generally. So

4:32

I and many other people as well, I

4:34

think over the last, I think, two, three

4:36

decades have really focused

4:39

on new questions related to explanation,

4:42

but drawing upon what we know from social psychology and

4:44

educational psychology. And so but do you

4:46

sure work then in the in the domain of cognitive

4:48

psychology? I consider myself a cognitive

4:50

psychologist, although, but, you know, as with any

4:52

demarcations between disciplines, the boundaries

4:55

get a little bit fuzzy and I -- Right. -- I think they

4:57

should not be very well sharply

4:59

drawn. And I think what's most distinctive

5:01

about the approach that I've taken to thinking

5:03

about explanation is that many of the questions

5:05

I've interested in are the ones that arise in the context

5:07

of philosophers of science have

5:09

been interested in explanation for a very long time.

5:11

Overwhelmingly thinking about the role of explanation

5:13

and science, but not exclusively. And

5:15

I think if you take

5:18

seriously the idea that there is some important

5:20

connections between everyday human cognition

5:22

and what we do is everyday research trying to make

5:24

sense of the world and what scientists are

5:26

doing in their quest to make sense of the world

5:28

then it's very natural to think that the kinds of questions

5:30

that arise about explanation in the context of

5:32

science and philosophers of science have been interested in

5:34

also have analogues in the context of just

5:36

everyday human cognition. I did wanna ask you about

5:38

that because I I can imagine the

5:40

philosophers want to know what

5:42

explanation is and, you know,

5:44

what's a good explanation, whatever, whereas I

5:46

might imagine psychologists are also interested

5:48

in how we do

5:51

explanations, how we start doing explanations

5:53

as children and things like that? So do you see

5:55

quite a bit of overlap and intersection

5:57

there? I think that's right. So

6:00

I think philosophers have asked largely

6:02

what explanation is and also what

6:04

explanations should be. So asking a more

6:06

normative question. I think those questions

6:08

are also relevant for psychology. But

6:10

on top of that, we might want to think about

6:12

where explanations come from developmentally. Mhmm.

6:14

But also one of the things that I've asked a

6:16

lot in my own homework is, what do

6:18

explanations do? And I

6:20

think that might be a more useful

6:23

entry point for starting to think about this. Right? So rather

6:25

first defining what an explanation is, and

6:27

then maybe thinking about its consequences, we

6:29

can ask, what do explanations do for

6:31

us? Why are we the sorts of creatures who explain?

6:33

Right. What's the function of this activity?

6:35

And perhaps by getting a grip on that,

6:37

on the role of explanation and what explanations do, we

6:39

can then work backwards to thinking about what explanations

6:42

are. Right? So rather than starting from a

6:44

definition sort of starting from a functional

6:46

role? Howard Bauchner:

6:48

And as if I recall correctly, there's

6:51

even the possibility that much

6:54

is in physics, you could get

6:56

feedback from the science end

6:58

into philosophy, and I think that in one of your

7:00

papers, you're saying that this philosopher

7:02

made the following claims about our intuitions, about

7:04

explanations. So we tested them, and

7:06

it turns out those not our intuitions. Yeah,

7:09

that's right. And we've done that across

7:11

a variety of projects and sometimes we find that our

7:13

results accord very closely

7:15

to what further said, and sometimes we find interesting

7:18

departures. And, of course, whenever that happens, there's an

7:20

interesting question. Is it just that humans

7:22

are wrong or the philosophers are

7:24

wrong? I think it varies

7:26

case by case. Do the philosophers then

7:28

listen? Are they are they do

7:30

you is it an active in practice

7:33

back and forth between your work and those of the

7:35

philosophers? That's an interesting question. I think it

7:37

it really varies by what question

7:39

philosophers are asking and, of course,

7:41

switch philosophers a huge variability

7:43

in the field as I'm sure you know better

7:45

than I do in the extent to which

7:47

people engage with empirical work of various kinds

7:50

and philosophical work. But I think I think it does

7:52

vary. I can give you an

7:54

example if you'd like an thesis where I think

7:56

there is interesting cross

7:58

talk back and forth. So one

8:01

area where isymologists in

8:03

particular have been really interested in explanation is

8:05

in the context of what's called inference the best

8:07

explanation. And this is something that I

8:09

think most people kind of intuitively are familiar

8:11

with, you know, Sherlock Holmes claimed that

8:13

he engaged in deduction, but a lot of the time when

8:15

he was used to really inference the best explanation. He

8:17

was looking at a bunch of evidence

8:19

and trying to come up with it would best explain

8:22

that body of evidence and on the

8:24

basis of that concluding that explanation might

8:26

be true. So that's a pattern of inference that

8:28

we see in science, but we see all the time just

8:30

everyday cognition. And one

8:33

question might be, when

8:35

we to engage in this kind of reasoning.

8:37

Right? Is it ever normative

8:39

warranted? Is it a good kind of reasoning or is

8:41

it a mistake? And I think there's

8:43

truth to both of those things. There's conditions under which

8:45

it's good and conditions under which it's bad. But there's an

8:47

interesting dialogue that's happened back and forth

8:49

between some of the empirical evidence and some of the

8:51

kinds of normative claims that small just might make,

8:53

but have to do with thinking about what the

8:55

goals of that kind of inference might

8:57

be. So you might think our goal

8:59

in engaging in

9:01

certain kinds of inference like it's like this might just

9:03

be to be, like, as accurate as possible --

9:05

Yep. -- meaning that we want to sort

9:07

of, like, minimize our long term inaccurate

9:09

see. We just want to, like, in the long run, get things right,

9:11

and we Tania have practices for updating our

9:14

beliefs. That mean that we're gonna in the long range

9:16

be least raw. And if you have

9:18

that view, you might think that what we ought to do is

9:20

something like play what's called base rule, base

9:22

in inference from statistics. It's sort of a

9:24

rule that tells us how to combine our prior

9:26

beliefs. With the evidence we have in order to arrive at what they

9:28

call a posterior probability. Now it

9:30

turns out that if you look at human

9:33

explanation, evaluation, and so on, and these

9:35

processes adverts the best explanation, there's

9:37

some systematic departures from what bayes rule

9:39

tells us that we should do. No. Yeah.

9:42

I know. I know. I'm sorry. I'm sorry to be the bearer

9:45

of bad news for somebody who

9:47

wants to add things. But

9:50

and I'm happy to say more about what those look

9:52

like, but the interesting thing is, you know, so

9:54

philosophers could just say, well, I

9:56

guess, you know, lay people are are just wrong

9:58

about this. They're doing poorly. But I think

10:00

a really interesting idea that's that's what

10:02

proposed is maybe maybe the thing humans

10:04

are trying to optimize isn't long term

10:06

inaccuracy, maybe it's something else.

10:08

So what might add something else be? What might something

10:10

that has I guess that's just what I was gonna

10:12

ask, you know, human

10:14

beings evolve under evolution, natural

10:16

selection, you know, reproductive fitness and

10:18

things like that. So in some sense,

10:20

the answer is always reproductive

10:22

fitness. Right? But I could

10:24

imagine that having an accurate

10:27

and predictive model of reality

10:29

helps with my reproductive fitness. Right? Yeah,

10:32

that's right. So the one

10:34

of the proposals and there's a handful of people

10:36

who've argued for this under the banner of

10:38

exploitationism, is that what we want to do is

10:40

get things mostly right in the short

10:42

term rather than being least

10:45

wrong in the long term. And

10:47

that's something that I think you can imagine trying to

10:49

motivate in terms of reproductive fitness.

10:51

Yeah. Right? We don't we don't know if we're gonna be

10:53

around long term. But if

10:55

we can get things mostly right in the short

10:57

term, that might be that might be the standard

10:59

that actually matters. And it turns out

11:01

that if that's what you're going for, you

11:03

shouldn't always use space roll. Sometimes you should do

11:05

something a little bit different. You might

11:07

also think that what we ought to do

11:09

depends on our cognitive limitations is humans. Right? If

11:11

we just don't have the the cognitive

11:13

capacities, that allow us to

11:15

engage in certain kinds of statistical

11:17

inferences or very complicated kinds of

11:19

mental computations. You might think that we ought

11:21

to employ the kinds of shortcuts that

11:23

are gonna be good enough given the

11:25

cognitive machinery that we have. And so that's another

11:27

idea that I think again makes sense within an

11:29

evolutionary context. But gives

11:31

us goals for what we're trying to do that look a

11:33

little bit different from, for

11:35

example, just applying basic inference. Yeah, I

11:37

mean, thinking and being cognitive Carroll's. Both

11:39

takes a lot of energy and a lot

11:41

of space, I guess, a lot of neurons, but

11:43

also time. And maybe you don't have time when

11:45

the lion is bearing down on you to think too

11:47

hard about what to do next. That's right. That's right.

11:50

And so being fast, being efficient, having

11:52

sort of heuristics or mental shortcuts arriving

11:54

at conclusions might be very beneficial. And

11:56

that's some of what we In terms of the explanations

11:58

we prefer. Right? So for example, one of the things

12:00

that we've looked at in my lab is people's

12:02

preference for similar explanations. And

12:05

you might imagine simpler explanations

12:07

are easier to process, easier to

12:09

remember, easier to arrive at and so and

12:11

so you might see a host of these benefits. So

12:13

I I was going to ask, and maybe that's

12:15

the answer. What what are the systematic

12:17

deviations from sort of perfect bayesian reasoning

12:19

that we might expect in real human Right? So there've

12:21

been a handful, and I'll tell you just a couple for

12:23

which we have the most evidence. Mhmm. So

12:25

one of them is that there is some evidence

12:27

that people seem to be more

12:29

sensitive to the

12:32

evidence that they ought to

12:34

compared to patient inference, meaning that

12:36

they will converge to an explanation that

12:38

fits the data well more,

12:40

you know, more quickly than they should and

12:42

consistent with this idea that they're trying to get to the it's

12:44

probably right fast. Interesting.

12:46

Okay. In my own lab,

12:48

one of the things that we have looked it as I mentioned to

12:50

simplicity. And the way we've

12:53

we've tried to get at that in psychology experiments

12:55

is by giving people scenarios where there

12:57

is a winner a couple of effects that they

12:59

observe and they're trying to come up with the best causal

13:01

explanation. So they're trying to come up with explanation that

13:03

sites one or more causes that

13:05

generate those effects. And under those

13:07

conditions, if you know the probability of

13:09

the causes and you know the probability of the effects,

13:11

given the causes, you could just

13:13

do the math. And so in some sense, we kind of

13:15

have the ground truth

13:17

for what the right pattern of reasoning

13:19

would be if people were just doing was most

13:21

likely. But what we find is that although people are

13:23

very sensitive to the probabilistic evidence,

13:25

they choose simpler explanations more often

13:27

than they should, where we define simpler explanations

13:29

as those involve fewer causes that

13:31

are themselves. So, you know, for example, if

13:33

you have two symptoms that you can explain by appeal

13:35

to one disease or by appeal to two

13:37

diseases that each does cause one symptom, you

13:39

find that people prefer the single disease explanation

13:42

more often than they ought

13:44

to given the probabilistic

13:46

evidence that they have. Physicists certainly act that way,

13:48

but maybe it works in physics in a way that it might

13:50

not work in medicine or everyday life. Yeah, it's

13:52

an interesting idea. You know,

13:54

in our in our lab cases, we

13:57

know what the ground truth is. Right? So we set

13:59

things up so that we we have some basis for

14:01

saying what the what's most likely or what's

14:03

not. But in most real world

14:05

cases, we don't have the

14:07

ability to cleanly say when people are

14:09

getting things right and when they're getting things

14:11

wrong. And so what we could be seeing in

14:13

the lab is the overgeneralization of

14:15

a strategy that actually maybe does make a

14:17

lot of sense in a lot of real world cases. I mean,

14:19

in the physics cases, and I'd be curious

14:21

if you have a a thought

14:23

on our physicists' right to be doing this. Have they

14:25

gotten to be doing this? Has it helped

14:27

scientific practice of theorizing for

14:29

them to show a preference for explanations that have this

14:31

kind of structure? I think

14:33

is a is a big ongoing debate to be

14:35

perfectly honest, especially in in my

14:37

little corner of fundamental physics, where

14:40

progress is a little stalled these days.

14:42

And so people are debating

14:44

different methodologies based on their

14:46

personal preferences. Right? And, you

14:48

know, some like like I have a

14:50

strategy which is to step back and think about

14:52

the foundations in a more philosophical way

14:54

to make sure that, you know, we're on a firm

14:56

ground where we where we do

14:58

our further down the road

15:00

reasoning. Other people are just

15:02

gonna, like, switch fields and, you

15:04

know, come to some area where they can

15:06

come up with a better explanation, test against the

15:08

data, and others are gonna argue about

15:10

beauty and math and all these things.

15:12

So I I that's why I was asking you. I real

15:14

I It's -- Yeah. -- physicists love

15:16

what works. Right? And when

15:18

some strategy works for one problem, they declare

15:21

victory, but the next problem is always

15:23

different, so we don't know. Yeah. Yeah. And I

15:25

think the everyday cognition cases are hard. If if I had

15:27

to guess, I would say that there

15:29

are many cases where preference for explanations

15:31

for the structure, in fact, does lead us so

15:33

sorry. Right. And That's that's the guess. I think

15:35

it's very hard to quantify that in in everyday

15:38

cases. Well, as a psychologist, is there

15:40

is there an understanding of psychologically

15:43

why we are driven to seek

15:45

explanations? You know, is

15:47

there is it just that there's a part of a

15:49

human brain that is naturally curious?

15:51

Or which sounds

15:53

kind of lofty almost

15:55

like we're naturally curious creatures or is

15:58

it is it more down to earth do you think? So,

16:00

we definitely have evidence that

16:02

basically from the moment kids

16:04

have the language capacities to ask

16:06

questions, they are asking

16:08

questions. And a lot of those

16:10

questions are what I would call explanation seeking

16:12

questions. So

16:15

it seems like it's an

16:17

early emerging and capacity, and we know it

16:19

plays an important role in human learning.

16:21

So what I've argued in some of my work, and I

16:23

think other people have argued for things along

16:25

similar is that it is a pretty basic human

16:27

capacity and it plays a very

16:29

important role in learning in particular.

16:32

And so I think there's one

16:34

version of that is very obvious because it's

16:36

so familiar to our everyday experience. But then

16:38

once you dig into it, it gets a little bit more

16:40

peculiar. So so one thing to

16:42

think about is that I very

16:44

straightforward why there would be some sort of adaptive

16:46

benefit for humans to be good predictors.

16:48

For example, right? If we could predict what's

16:50

gonna happen, that's clearly

16:52

very beneficial. We couldn't anticipate our

16:54

circumstances. We're going to know what's going to

16:56

happen if we intervene to find something about and so

16:58

on. But Explanation is different

17:00

from merely predicting. Explanations

17:02

are typically backwards looking, so we observe something

17:04

that's already happened, and then we wonder why did

17:06

that happen, and we to out what

17:08

happened in the past. So it's maybe not super

17:11

obvious why it is that we would have this

17:13

practice of trying to

17:15

explain things and wanting things with a particular

17:17

type of explanatory structure, if the thing

17:19

that's actually useful for us is just to be able to

17:21

predict. And so what I've argued is

17:23

that actually we can make sense

17:25

of why explanations have some of the capacities they

17:27

do, some of the characteristics they do

17:29

rather in this backwards

17:31

looking sense where the thought

17:33

is that Our explanatory practices

17:35

and looking for explanations actually helps us

17:37

construct the sorts of intuitive theories about the

17:39

world that are going to help us

17:41

predict. Down the line. Okay.

17:44

So to give you here's

17:46

here's one example of something

17:48

along these lines. So one of the phenomena that

17:50

I'm really interested is that we seem to learn better

17:52

by explaining to ourselves or to other

17:54

people. Right. Which is a little bit puzzling

17:56

because when you explain to yourself or to somebody else, you

17:58

don't get new information. Right? If you give me an explanation,

18:00

you've given me something new. But if I'm explaining to

18:02

myself, I'm just rearranging the pieces that are already in my

18:04

head. So

18:06

What's going on there? So we construct all of these lab studies

18:08

where we compare people who are learning

18:10

some tasks by explaining to themselves

18:12

versus doing something else. Like

18:14

thinking aloud or describing some sort of controller comparison

18:17

task. And we find that the people who

18:19

are explaining to themselves actually

18:22

learn better. They learn certain kinds

18:24

of regularities in the environment better than those who don't.

18:26

So for example, they're more likely to

18:28

identify subtle patterns

18:30

that differentiate two categories that they're

18:32

trying to learn to differentiate. So

18:34

Sorry. Can you can you elaborate on

18:36

what the two choices are? Like, one person

18:38

is is seeking an explanation and

18:40

the other person is doing what?

18:42

That's right. So it varies across across

18:44

experiments because there's no perfect control condition.

18:47

But to make us all more concrete, Suppose

18:49

you come into the lab, then you're in one of these experiments. Your

18:51

task is going to be to learn to categorize

18:53

new types of robots. Okay? And I'm

18:55

gonna show you Eight examples of

18:57

robots four belong to. Gorb the Gorb

19:00

category and four belong to the Drent

19:02

category. So you're gonna have four labeled examples of

19:04

each type of robot. And you're gonna study

19:06

these in order to figure out how

19:08

to differentiate corpse and dents because I'm gonna show

19:10

you some new robots later and you're gonna have to tell me

19:12

if they're corpse or dents. So I'm looking for

19:14

the pattern. Exactly. So

19:16

everybody in our task is basically

19:18

trying to find what what the pattern is in these

19:20

stimuli that will allow them to

19:22

generalize to new cases. So here's what

19:24

we manipulate. Half the participants,

19:26

as they're studying the globes and the dents, get

19:28

asked an explanation question. So we ask them,

19:30

like, why do you think this one's a glob? Why do

19:32

you think this one's a drent? And they try to come up with an

19:34

explanation, and we don't tell them if they're right or wrong,

19:36

so they're not getting any feedback, but they're engaged

19:38

in explanation seeking. And

19:40

we compare what they learn to

19:42

participants in a control condition. And the

19:44

control condition we we varied across studies,

19:46

it could be describe the scope,

19:48

describe the strength, So they're

19:50

being forced to use language. They're being forced to

19:52

pay attention to the task. We could

19:54

ask them to think aloud. Just tell us

19:56

what you're thinking as you are studying this

19:58

corporate rent. We could give them no

20:00

instructions at all, but give them an opportunity to

20:02

study so that everybody basically has the same task,

20:04

has the same data. Now we we design

20:06

the stimuli, these corpse and drens or whatever whatever we're

20:09

studying that particular experiment, so

20:11

that there's some relatively

20:13

subtle pattern that

20:15

you might discover that differentiates the Korps

20:18

syndrome. There's other things going on too,

20:20

but maybe it only accounts for half of the Korps

20:22

syndrome or seventy five percent of the

20:24

Korps syndrome. The only thing that it counts for,

20:26

you know, that all it only the trends

20:28

have versus all it only the groups of this sort of

20:30

subtle pattern. And what we find is that the

20:32

participants who are prompted to explain are

20:34

significantly more likely than those in

20:36

these alternative conditions to discover

20:38

that. So what's going on there? Well, it looks

20:40

like by virtue of the fact that they were trying

20:42

to explain they learned

20:44

something real about the

20:46

structure of these stimuli. And they learned it

20:48

faster and better. Sorry? They

20:50

learned it faster and better than they would

20:52

have otherwise. That's right. That's right. That's right. And so

20:54

it seems like there's something about

20:56

human explanation seeking that at

20:58

least for particular kinds of structure in

21:00

the world, might be especially

21:02

good at making us formulate

21:04

useful hypotheses, test them

21:06

in effective ways and come

21:09

up with sort of a good way of

21:11

characterizing some real generalization in the And this

21:13

is a place where you might think simplicity is beneficial

21:15

even if the world is not simple. By

21:17

virtue of trying to find a simple pattern, we might

21:19

sort of look harder and try harder and

21:22

discover whatever structure is actually there

21:24

even if the structure that's there is not

21:26

itself simple. And in the

21:28

clips and dents are the

21:30

patterns that they will ultimately find,

21:32

ones that they can

21:34

corporate kind of functionally for the robot their

21:36

little robots so they have a preexisting idea

21:38

of what robots might do? Or is it

21:40

just like this one has stripes and this one

21:43

has spot? We've done the

21:45

studies both ways. So

21:47

sometimes they're totally arbitrary features that don't seem

21:49

very meaningful. So for example, some of

21:51

them have feet that are pointy at the bottom and some of them

21:53

have feet that are flat at the bottom. And you can

21:55

come up with reasons why that might matter for

21:57

being a corporate veteran, but I don't know a lot

21:59

about Galp syndrome. So probably not In other versions,

22:01

we actually give people more information that would allow

22:04

them to make that be meaningful. So for

22:06

example, we tell them that some of these are indoor robots

22:08

and some of them are outdoor robots. And

22:10

now all of a sudden, you can kinda come up with some

22:12

reasons why foot shape might matter. Maybe

22:14

they'll they'll, you know, scratch the wooden floor if they

22:16

have pointy feet and but that makes

22:18

sense on type of

22:20

material. And what we find there is that

22:22

explaining makes you more likely to

22:24

discover the sort of simple

22:26

pattern that accounts for all cases either

22:28

way but that when you have this rich background information,

22:31

you use that more when you're

22:33

trying to explain. And I think that

22:35

makes sense when about

22:37

explanation everyday cases. A lot of what explaining looks like

22:39

is trying to make sense of new observations in

22:41

the context of what we already know. We're trying

22:43

to sort of like fit it in to what

22:45

our existing beliefs and intuitive theories of the world

22:48

are. And so when we prompt people to explain,

22:50

we see them doing that more. They're sort of trying to come

22:52

up with a story like,

22:54

why pointing versus flat feet would make

22:56

sense given that they're inboard versus out for

22:58

robots? Yeah. I mean, maybe this is a crazy

23:00

overgeneralization, so correct me if I'm

23:02

wrong, but It seems that over and over again, we see

23:04

examples where we have an

23:06

idea of what acting in a perfectly

23:08

rational way would be and

23:10

human beings come close to it but not really

23:12

because they're doing something completely different than

23:14

being rational for other completely other

23:16

reasons, but nevertheless, the reasons have

23:18

led them to sort of mimic rationality in some

23:21

kind of way. Howard Bauchner: Yeah, no, I don't think that's another

23:23

generalization at all, and I think that's

23:25

a really interesting way to think about this.

23:28

In this particular case, one

23:30

thing that I have come to think is

23:32

that there might be a story about why it's

23:34

rational to prefer simpler explanations

23:36

for example or explanations with other kinds of structures,

23:39

but it's not the one that we might have thought. So I

23:41

think the the intuitive idea

23:43

is the idea that I think Knewton

23:46

advocated. He has this lovely quote in the prepared, yeah,

23:48

which is something like, you know, when he said he

23:50

said we should basically prefer simpler

23:52

explanations because nature

23:54

affects not the pump of

23:57

superfluous causes. Right.

23:59

So so the thought is something like, if

24:01

nature itself is simple, then we

24:03

should prefer simple explanations because those

24:05

simple explanations are more likely to

24:07

accurately reflect nature. Right. And so that

24:09

would give you one reason why

24:11

preferring simple explanations would be rational.

24:14

And I'm skeptical of that one. I'm really

24:16

skeptical of that. But I

24:18

think what our data might suggest is

24:20

that having the practice

24:22

of preferring simpler explanations

24:24

might lead you to learn

24:26

about your environment and look for

24:29

particular types of structure in your

24:31

environment in a way that might have the positive

24:33

downstream consequence of leading you to

24:35

discover the structure that's actually

24:37

there. And so it's almost more like a methodological

24:41

strategy to get you to the

24:43

right place. But without the

24:45

assumption that the world is itself simple.

24:47

Right? It's rather that having

24:49

humans be sort of picky about explanations is

24:51

gonna be something that motivates go

24:53

out there and figure out what's really going on. Because the world

24:55

is a weird mixture of simple and complex.

24:58

Right? And we As a

25:00

psychologist, I'm gonna say more

25:02

complex than simple. But yeah. We

25:04

have to work hard to find the

25:06

simplicity, I guess, is is the plan I'm trying

25:08

to make. And and and we do, like like like

25:10

you just said. I guess I didn't

25:12

want to let go of the intriguing thing you

25:14

said about explanations being

25:17

mostly looking backward in

25:19

time rather than predictive in some

25:21

sense. I mean, it makes me think of mystery

25:23

novels. We love mystery novels. I love

25:25

mystery. And mostly

25:27

detective is trying to come up with an explanation for something that

25:29

happened in the past without necessarily

25:31

helping us predict the future on the

25:33

basis of that, but can

25:35

we conceptualize it as part of a

25:37

larger strategy that if we know why all

25:39

these murders are committed that will help us in the

25:42

future or is there something else going

25:44

on? I think there might be two things going on. So I think sometimes

25:46

we care about the backwards looking part

25:48

because we hold people causally

25:50

or morally responsible. Right? And there it

25:53

seems like that judgment is really

25:55

playing almost more of a social

25:57

role in how we regulate other people and

25:59

interact with other people. So that's

26:01

important. But I think also the practice of

26:04

explaining why particular events occurred

26:06

is going to be part of what allows us

26:08

to construct a causal model

26:10

that domain or just more generally sort of

26:12

a theory of that domain. And then it's

26:14

the theory of that domain that will allow us to

26:17

predict things in the future.

26:19

Right? So, you know, to give a

26:21

toy example, you might imagine that by

26:23

virtue of trying to figure out

26:25

a particular murder in this

26:27

detective story and so

26:29

on, somebody comes to learn something about human

26:31

motivations that they didn't appreciate before.

26:33

Mhmm. They come to learn something about

26:35

how particular poisons work and how you can mask their effects that

26:37

they didn't know before. Right? And so

26:39

they come to learn all of this particular stuff

26:42

that actually might be useful. In

26:44

the next case. You know, maybe not not narrowly for preventing another

26:47

murder, but but

26:49

it is contributing to your repository

26:51

of knowledge about the world. Yeah.

26:54

that is going to be useful in the future.

26:56

Yeah. I I guess I'm actually yeah. I'm

26:58

interested in this because I don't know

27:00

why the mystery

27:02

genre. I mean, in some sense, almost all

27:04

genres or subsets of mysteries. Right? Like,

27:06

things happen and we don't know why we wanna fix them

27:08

and It's a very it has a very powerful hold

27:10

on us psychologically and

27:13

I'm willing to buy that

27:15

it stems from the same impulse as

27:17

the impulse to, you know, understand and

27:19

predict the world going forward. May

27:21

maybe that is it or or maybe there's extra ingredients

27:24

being fed. Yeah. I think that's right. I I love

27:26

this example. The case I thought about a little bit more

27:28

is something like riddles. Mhmm. Right? We

27:30

find riddles extremely satisfying.

27:32

We're very curious about the answer to a

27:34

riddle. Very satisfied when we get the

27:36

answer to a riddle, but it's hard to

27:38

make the case that riddles play a fundamental role in

27:40

human learning. Right? It sort of seems like it's

27:42

the it's the candy of the system.

27:46

And I think the mystery novels might have

27:48

some of the same characteristics. Right? They

27:51

mimic some of these cases where we

27:53

really are prompted to find a next nation and where that's really central

27:55

to our ability to learn about the environment. Right.

27:57

And so they sort of like push all the right buttons

27:59

without necessarily giving us

28:01

the same adaptive

28:03

consequences that we may be seeing the explanation

28:05

case. Mhmm. But by

28:07

virtue of that structure, we we we do feel

28:09

drawn in. We are very curious. We do feel very

28:11

satisfied when there's a good resolution. Another

28:14

thing I don't let go of is

28:16

the relationship between explanations and

28:19

causes. You know, causality is

28:21

something that philosophers debate a lot

28:23

about. We had Judea Pearl on

28:25

the podcast. time ago. You know, it's

28:27

a whole subfield

28:29

of understanding causal influences on

28:31

things. Is it

28:33

okay to think about the search for explanation

28:36

as mostly a search for the cause

28:38

for why something happened? Or is there

28:40

a division between those sure

28:42

you know, there's a debate about this in philosophy,

28:44

and and I'd say probably in psychology

28:46

as well. My own view is that

28:48

a lot of explanation is causal, but

28:51

not all explanation is causal. I think one

28:53

of the clearest examples of non causal explanation

28:55

is mathematical explanation. Right? So you can give

28:57

an explanation for the Pythagorean theorem.

29:00

And it doesn't look like there's anything causal going

29:02

on there. And so, you

29:04

know, I think that raises a question about the

29:06

relationship between the causal cases and the non

29:09

causal cases. there really

29:11

fundamentally different kinds of explanations? I'm

29:13

more attracted to views that that think

29:15

that it's really very similar across these

29:17

kinds of cases that explanations in general

29:20

understanding that the nature of human

29:22

understanding involves something about

29:24

appreciating dependency relationships of a

29:26

particular Tania. might be multiple types of

29:28

dependency relationships, causal dependency

29:30

relationships are a really important type that

29:32

characterize huge swats of what we

29:34

care about, but not the only type. You

29:36

know, things can be can have a

29:38

deductive or an entailment relationship in the case

29:40

of mad, things can have a constitutive relationships.

29:42

So I think there's actually lots of other kinds of

29:44

relationships that can support explanations and explanatory

29:47

understanding. But by and large, the ones

29:49

that I've studied are the causal

29:51

cases. And I think those are the paradigm cases we

29:53

typically think about. It's interesting because, of course,

29:55

we don't agree on what

29:57

constitutes a causal case. You

29:59

know? And some people are gonna say,

30:01

well, pythagoras's theorem is

30:03

true because of

30:05

the postulates of Euclidean geometry.

30:08

And, you know, to me that's a

30:10

very different notion of the of

30:12

the word cause, then the

30:14

person died because they were stabbed by

30:16

their friend or whatever. I mean, we're using the same

30:18

word, but these are very different

30:20

philosophical concepts, I think. Howard Bauchner: I think that's

30:22

right. I mean, I think I think it's tricky to

30:24

interpret because Carroll's you

30:26

know, we use because in in all sorts of ways and at

30:28

least I would hesitate to to say that

30:30

they're all causal. You

30:33

know, we did some research in my lab that

30:36

tried to narrow in on a part of your question.

30:38

So we thought, what's the

30:40

most minimal contrast

30:42

between a cause claim and an

30:44

explanation claim. Okay. And so we had

30:46

cases where, for example, a causes

30:49

b, or be because a.

30:51

Right? And so the thought is they're, you know,

30:53

kind of as close as possible basically. But

30:55

but one's one's an explanation talk and one's

30:57

in causal talk. And we found that these

30:59

two behaved very, very similarly

31:02

in terms of what sorts of evidence

31:04

people thought was relevant for assessing whether or

31:06

not the claim was true, but we did find

31:08

some differences Right? Today's one identical. So here's

31:10

where maybe I'll give you a concrete case and get

31:12

an intuition or how these worked. So we

31:14

wanted cases where there were causal Carroll's. Where

31:17

people would not antecedent we think were related

31:19

at all. So one of our cover

31:21

stories involved, you

31:23

had to imagine that you've gone to work for a

31:25

museum, and one of your task is to tabulate

31:27

lots of data about the museum, you know, who

31:29

visits, which exhibits, and what they do, and

31:32

so on. And you just notice this

31:34

correlation in all of your data. There's a

31:36

correlation between having visited the portrait

31:38

gallery in the museum and

31:40

having made an optional donation when

31:42

you leave the museum. Good. Okay. So one thing we

31:44

can vary now is what's the strength of that correlation? Is

31:46

it just like a really weak association or

31:48

is it like pretty much a perfect

31:51

relationship. You know, every single all and only the

31:53

people who went to the portrait gallery, bit not foundation

31:55

and so on. And we ask people

31:57

claims like, To what extent do you agree that visiting the

31:59

portrait gallery caused this person to

32:01

make an optional donation as a left a

32:03

museum? Or why did this person make an optional

32:05

donation when a left a museum because they

32:07

is a project gallery. Right? So that we have the kind of matched cause of claims and

32:09

explanation claims. K? So the stronger the

32:11

statistical evidence that there is

32:13

a relationship, the more people are willing to

32:15

endorse these claims. But that

32:18

has a bigger effect for the causal

32:20

claim than for the explanation claim.

32:22

Here's here's here's the on the flip side what we found.

32:24

Yeah. Now we give some of

32:26

our participants a mechanism linking these two

32:28

things. Right? So if we constructed our

32:30

stimuli correctly for what we're going for, it should

32:32

hopefully not be at all obvious why there would

32:34

be a correlation here and

32:36

what might that? we tell half of our participants that actually there's a

32:38

lot of research in social psychology showing that

32:41

if you're surrounded by watchful others,

32:44

like in a portrait gallery with eyes and faces.

32:46

That triggers these mechanisms where

32:48

you're concerned with your reputation and leads

32:50

you to act more socially and so on.

32:52

And so you're more likely to do something like make an

32:55

optional donation. Okay. So when you give

32:57

people the mechanism that makes

32:59

them more likely to accept

33:01

the cause it more likely to accept the explanation claim that the reason

33:03

the person for example donated was

33:05

because they were in the portrait gallery. But

33:07

that has a bigger impact

33:09

on the explanation claim than not in

33:11

the causation claim. So, you know,

33:13

what is What is the probability of this? This is

33:15

tricky. Yeah. Okay. Good. So so that

33:17

that is a bigger impact on

33:19

the claim that they donated because

33:22

they went to the portrait gallery than it

33:24

does on the claim that the

33:26

going to the portrait gallery cause them

33:28

to donate. Exactly. Very fine

33:30

distinction. We're drawn here. But, okay, good. Yes.

33:32

That's right. But, I mean, part of the reason I maybe we wanted to be

33:34

a fine distinction really trying to drive a wedge

33:36

between these otherwise extremely similar

33:39

claims. Right? The kind of like bear calls will claim,

33:41

the bear explanation claim. So that's just

33:43

the empirical finding. But of course, I think the

33:45

more interesting question is, like, why? Right? What does this tell

33:47

us about the nature of explanation and causation and

33:50

so on? So what we think might be going on is that

33:52

this might give us some hints towards

33:54

what the functional role

33:56

of explaining is for people. Right? What do we want

33:58

our explanations to do for us? Such that we'd

34:01

be more satisfied when we have, say, the

34:03

mechanism information in this case. And

34:05

what we argue in this work is that one of the

34:07

reasons why mechanism information might be so

34:09

central to explanation is because mechanism information is what

34:11

allows us to generalize to new

34:13

cases. So suppose I now ask you to imagine

34:15

a case where some of the patrons of

34:17

the museum visit the

34:20

sculpture garden. And you need to predict whether or not those who visit garden

34:22

are also more or less likely to make

34:24

an optional donation. Well, if you

34:27

have the mechanism, you have a

34:30

basis for making that prediction. You might want to know,

34:32

well, where the sculptures

34:34

figured it? Did they involve faces?

34:36

If so, I'm going to predict that you're more make

34:38

a donation. If these were all

34:40

abstracts, sculptures, then no. You have no

34:42

reason to think that you're more likely to make a donation.

34:44

So by virtue of knowing the mechanism,

34:47

you are able to generalize from the original

34:49

case outside of the data that

34:51

you'd already observed. By contrast, if you

34:53

only have the correlation,

34:55

or the statistical evidence there. That tells you how

34:58

strong the relationship is in the

35:00

population that you studied already. But

35:02

it doesn't give

35:04

you guidance for how to generalize from that population to a novel

35:06

population. And so part of what we think

35:08

is so key about explanation is that it's

35:10

going to

35:12

direct you make you look for the sorts of things that support generalization.

35:14

And this ties back to our earlier discussion about

35:16

how explanation might support prediction.

35:18

Right? If what explaining is making you

35:20

do is

35:22

not just on any kind of structure in your environment, but

35:24

specifically the kind of structure that's likely to

35:26

be useful for generalizing to new

35:28

cases for predicting the future, then that would

35:30

make a lot

35:32

of sense. It does. So in other words, what what

35:34

looking for a good explanation is

35:36

about is more than just finding

35:38

a pattern but finding a

35:40

pattern that sort of fits into the rest of our

35:42

knowledge of the world in such a way that it has

35:44

some implications for other things we might

35:46

say going forward. That's right. That's right. And I think

35:48

that it helps explain by merely

35:50

predicting something accurately doesn't give

35:52

us a sense of understanding and explanatory sense of

35:54

satisfaction. If you had a black box

35:56

that allowed you to predict lots of

35:59

things very accurately. no idea how to

36:01

use it to get

36:03

to new cases. That's not

36:05

going to give you what we want in explanations even though

36:07

you're getting some predictive leverage there. Howard

36:10

Bauchner: It is weird because

36:12

in certain corners of

36:14

modern physics. I'm I'm bringing up physics

36:16

more than average in this philosophy

36:18

in this psychology But there's a

36:20

movement be precisely because some

36:23

modern theories of physics invoke

36:25

things we can't observe. Right?

36:27

Like the multiverse or string theory or whatever,

36:30

there's a sort of countervailing argument

36:32

that says, all I care about is

36:34

making predictions

36:36

for observations. And, you know, we have to stand up for the idea that no,

36:38

actually, I I wouldn't know why. I wouldn't actually

36:40

know the explanation. Some people are

36:42

are moving away from

36:44

that. So I'd like

36:46

it. I like the idea that

36:48

it's really the knowing why

36:50

that is the goal here, not merely

36:52

recognizing the existence of a pattern. That's right. Although I

36:54

think in the physics cases, you get to a real

36:56

question, which at least for me is an open question, which

36:58

is what are the limits of that? Right? Are there going to

37:00

be cases

37:02

where human mind is not capable of understanding the why. The best we can do

37:06

is is predict in some cases or rely

37:08

on our deep learning system or extremely

37:10

complicated theory do the predicting for

37:12

us. So I think that

37:14

is as an account of the psychology

37:16

of explanation -- Uh-huh. -- I think we really do

37:18

care about the why. And then thinking about what

37:20

does that mean for science? I think in that case,

37:22

you really come up against these cases where

37:24

it might turn out that some things are

37:26

beyond our our human capacities. It's possible, but I don't think we're there

37:28

yet, so I'm not worried about that. Like, if that comes

37:30

up to them, then I'll to be

37:32

optimistic about this. Thank you. I do.

37:34

I do. And

37:36

I guess this has been implicit in some of the things

37:38

we said, but from the psychological angle,

37:41

have we learned what

37:44

counts as a good explanation. I I think we all have ideas about what

37:46

counts as a good one. Again, making

37:48

ability to predict the future, fitting

37:51

into other things we think?

37:53

Or is it like an accepted set of criteria

37:55

for what an explanation is a good one?

37:57

Yeah. There's sort of there's sort of two questions, and I

37:59

think we have partial answers to both. One is what even counts

38:01

as an explanation. Right? So that

38:03

contrast is really with an explanation versus an odd explanation. And for that,

38:06

I think an explanation

38:08

typically generates

38:10

understanding about why

38:12

what you're asking about with the case as opposed to some often implicit

38:14

contrast. Right? So if I

38:17

say, you know, why why

38:20

is the sky blue? I might be implicitly asking

38:22

why is it blue as opposed to another color,

38:24

and an explanation has to generate understanding about

38:27

why it's blue as opposed to another color. You

38:29

might very legitimately then say what do you mean by

38:31

understanding there and we can come back to that. But I want to

38:33

contrast that with the question more the way you formulated it, which

38:35

is more like given that something is an explanation or

38:38

being offered as an explanation, what makes it a good

38:40

explanation or

38:42

satisfying explanation? And there we know a

38:44

bunch of features that seem to play Central. So

38:46

we've been talking about simplicity.

38:48

That seems to be one of them. And there seemed to

38:50

be a few different notions

38:52

of simplicity. So we can unpack that further and talk about kind of

38:54

simplicity. Brett seems to

38:56

be another one. So we like explanations

38:58

that explain everything we invoke

39:00

them to explain, not just sort of subsets

39:02

of it? You do.

39:04

Being consistent with your prior beliefs, as as

39:06

you've already suggested, to sort of fit in with what you

39:08

already know. People like

39:10

explanations better that don't make

39:12

unprecedented predictions.

39:14

Right? So something predict something that hasn't

39:18

been observed that might feel a little bit don't

39:20

that so much in our explanation. Isn't that the

39:22

opposite of what philosophy of science is

39:24

supposed to tell us

39:26

that we love the

39:28

explanations that had that make predictions that haven't

39:30

been tested yet because then we can go test

39:32

them? Yeah. That's not what the psychology suggest,

39:34

but, yes, I it it it the psychology result here is So is

39:36

this is a body of work focusing

39:38

on what they call latent scope.

39:40

And the idea

39:42

is that It's

39:44

thought of typically as an error. But for

39:46

example, if you have two diseases that could explain

39:48

a set of symptoms, one of them

39:51

predicts we blood test that we haven't yet done yet,

39:53

you'd see an abnormal value and the other

39:55

one predicts that you'd actually see a normal

39:58

value. People seem to prefer the theory

40:00

that does not make the

40:02

unverified prediction that may be departure

40:04

from the default. Mhmm. Even in conditions where you

40:06

statistically control for various things. So that seems like

40:08

it may be maybe an error, although it's perhaps an error

40:11

we can understand as an overgeneralization of a strategy

40:13

that makes sense under some

40:16

conditions. Sure. Other cases like this are sometimes I sometimes call them

40:18

explanatory vices as exposed to explanatory

40:20

virtues because it's not clear

40:22

that they're They're

40:24

always rational strategies for evaluating explanations. But to give you

40:26

other examples, people do prefer explanations sometimes

40:29

that involve reductive jargon. So

40:31

the classic case of this is that if you give people explanations

40:34

for psychological phenomena that

40:36

do or don't invoke totally irrelevant

40:38

neuroscience, at least that neuroscience is

40:40

irrelevant according to experts. Sure. You

40:42

typically find that late people like

40:44

those explanations better with that redact

40:46

of jargon.

40:49

There is a similar finding. It was in the context of scientific

40:51

abstracts rather than explanations, but people

40:54

were seduced by irrelevant math.

40:56

Right? So if you add some irrelevant math in there,

40:58

maybe that makes it seem. More

41:00

legit and might more rigorous. I'll do that

41:02

all the time. It's so good. Nice. It's

41:04

irrelevant. But the novice

41:07

maybe can't. People tend to

41:09

like explanations that offer a mechanism. We've

41:11

talked about that already.

41:14

You know, there's other characteristics like this. I think part of

41:16

what's challenging is that for any one of these,

41:18

there's a question of how you really define

41:22

that. Like, so for example, what do we mean was to say an explanation broader

41:24

and so on? And also, most of these

41:26

do show some context sensitivity.

41:29

And that makes very hard to make unqualified

41:32

generalizations. So to give you an example with respect to

41:34

simplicity that we've been talking

41:36

about, people also seem to

41:38

have this view that very complex

41:40

phenomena might require a more complex

41:42

explanation. And so

41:44

even though they generally seem to prefer simpler explanations for

41:46

a very complex phenomenon. You might actually

41:48

see something that goes a little bit the other

41:50

way where they start to think that,

41:52

you know, some

41:54

complexity is required in the explanation itself. Okay.

41:56

Good. Yeah, I can see that in in

41:59

different context. And I wonder But

42:02

there must be, like, I guess, different

42:04

people react to different ones of different of these

42:06

standards differently. Sorry. I'm

42:09

not articulating this very well, but I mean, maybe some

42:11

people are more in it for the simplicity, some people

42:13

are more in it for the scope, some people are more

42:15

in it for the fit to existing

42:18

knowledge, things like that. Yeah, I think that's

42:20

right. So I think that's an area where we actually just don't

42:22

have meant much data. So

42:24

far, there's a little bit. The data that is

42:26

there does address that there are there's

42:28

individual variation across people and the extent to which, for example, they want the mechanistic details.

42:30

And I think we all know this from everyday life.

42:32

There's a people who are just happy

42:35

to -- Oh, yeah. -- you showed them a microwave for the first time and

42:37

they're like, great. You know, I know it's button to push and that's

42:39

it. And there's the person who really wants to understand, like, well,

42:41

what's going on into the hood? So that's

42:43

a dimension of individual variation that's been documented. In explaining other

42:45

people's behavior, you also find

42:47

variation in the extent

42:49

to which people are open

42:51

to there being more complex interactions between a

42:54

person in their environment versus tending to

42:56

think that it's sort of like a more simple

42:58

straightforward explanation just in terms of

43:00

the person. I'm weirdly the person who does not know

43:02

the details sometimes. Like, when I'm

43:04

at the dentist and I was like, would you like to see

43:06

the what what we're doing to your teeth?

43:09

No. I no desire whatsoever to see that. Just

43:12

make it work. Just make them

43:14

healthy. I think you've

43:16

also written about some things

43:18

that maybe I don't know if they count as

43:20

explanations or not, but there's the idea of an explanation as

43:22

kind of an abstract framework, but

43:24

there's the strategy of telling

43:26

a story or an

43:28

anecdote or a narrative.

43:30

And sometimes people are gonna

43:32

count that or even prefer that in terms of

43:34

being an explanation. Yeah, I think those

43:36

cases are fascinating. Part of the reason I started to

43:38

think about it was because most of my

43:40

research is really focused on the role of explanation

43:42

and learning. Suggesting that explanation is important for generalization and so

43:44

on. And for for

43:46

that story to work, you really want

43:48

explanations to be focusing on relatively

43:50

abstract, generalizable

43:52

features of a situation. That's what we tend to look for in Right? We like science.

43:54

We like explanations in both laws and

43:57

things like that. And

44:00

so I was trying to reconcile that with this other everyday

44:02

human behavior, which is that you ask someone why

44:04

they relate to work. And they

44:07

don't give you some abstract generalization. They say, like, oh, my gosh, you

44:09

won't believe the morning I've had. Right? And then they'll tell you

44:12

about, you know, the play by

44:14

play leading up to their being way

44:16

to work. And that feels

44:18

much more like storytelling or narrative.

44:20

Mhmm. And so the way I think about this

44:22

now is I think there's sort of a continuum

44:24

between these. I think very often everyday

44:26

human explanations can sort of

44:28

be at an intermediate point, sort of giving

44:30

you the very abstract and allizable

44:32

features of a situation versus

44:34

focusing on these concrete particulars. And I think

44:36

they serve different functions. So I

44:38

think the the law

44:40

like generalizations are useful

44:42

precisely because they support generalization to

44:44

other cases. They're picking out the features of a

44:46

situation that are relatively invariant that

44:48

might support prediction to other cases and so

44:50

on. I think a lot of these

44:52

other things the concrete sensory

44:54

details. You know, the things that make

44:56

a piece of fiction really compelling, all of

44:58

that extra

45:00

nuance in detail. I think part of what that's doing

45:03

is giving you

45:05

input that allows you to

45:07

do something like

45:10

a until simulation of the situation, works for yourself and somebody's

45:12

shoes. And so you understand

45:14

the situation from a particular

45:17

sort of embodied perspective. And that can

45:19

give you a kind of insight about

45:21

the particulars of that

45:24

situation that you might not get from

45:26

this more abstract perspective, you know. So for example, by virtue of hearing

45:28

about the play by play of your morning

45:30

and how you spilled your coffee and

45:32

so on, I'm

45:34

gonna be in a better position to appreciate how

45:36

you really felt and how it was

45:38

frustrating and what that might have led you to

45:40

to think or to feel or to say and

45:42

so on. And

45:44

so it might be partially a peculiarity

45:46

of the way human cognition works, but

45:48

it's partially by getting those

45:50

concrete particulars we can engage in those kinds of mental

45:53

simulations effective. No, that's a fascinating point, very

45:55

interesting. Is it is it almost a

45:57

mirror neuron situation? Oh, gosh.

46:00

I I think I'm gonna not speculate about

46:02

Amerinirons because I will

46:04

get myself in trouble. I certainly

46:06

don't think that that's I don't think that's

46:08

likely to be a necessary part

46:10

of this story. You see, I'm adding

46:12

on unnecessary neuroscience to make it a more

46:14

compelling explanation. Yeah. People are gonna think

46:16

it's a better obligation. Actually, we should

46:18

kind of mirror neurons. But

46:20

no, I I do like the idea that

46:22

somehow forget about mirror Carroll's, but

46:25

the embodiment of the explanation. Like, there's

46:27

a set of rules abstractly. Okay.

46:29

That's one thing. If I can picture

46:31

myself in it, that's another

46:34

level of appreciation of this purported explanation.

46:36

Right? That's right. That's right. I mean, sometimes and and

46:38

I think you see this reflected in in

46:40

natural language. And

46:42

a discussion of philosophy as well in terms of different kinds of understanding. You know, one kind

46:44

of understanding we talk about is the sort of understanding

46:46

where you say, you know, I

46:49

understand quantum mechanics. Or I understand, you know,

46:51

how to how to drive the pythagorean theorem or something

46:53

like that, where it seems to be this relatively

46:56

like abstract

46:58

kind of understanding. But then there's this other thing we do where we say like, oh, I really

47:00

understand where you're coming from. I really understand her. I

47:03

really understand the character in this book.

47:05

Or I feel understood. And

47:08

All of those notions seem to be much more like this sort of

47:11

first personal story narrative

47:13

like case. Mhmm. Where

47:16

we really can sort of put ourselves in the position of somebody else.

47:18

And so just just like I

47:20

think, you know, explanation and storytelling sort

47:23

this gamut. I think these different notions of understanding show

47:25

a similar kind of variation. I like that.

47:27

Is it also possible that the

47:29

storytelling end of the spectrum sort

47:31

of gets a bad rap? Because we

47:34

valorize rationality. That's what

47:36

they're saying. So I think I think it would

47:38

legitimately get a bad rap

47:40

if we used the concrete

47:42

storytelling to serve the

47:44

roles that should be served

47:46

by the more general explanation. Right?

47:48

So an example like that

47:50

might be somebody taking an anecdote to

47:52

be a good basis for a

47:54

public policy decision, for example,

47:56

rather than

47:58

looking at data and the generalizations that are actually available to us at this

48:00

large scale. And there is some evidence like

48:02

that. I mean, in the context of persuasion,

48:04

anecdotes and good stories are persuasive.

48:06

Oh, yeah. And

48:08

so those might be cases

48:10

where they legitimately get a bad rap

48:12

insofar as they are being mistaken

48:14

for strong evidence.

48:16

But on the other hand, I think they're

48:18

playing a really important role. Right? So I

48:20

think I think maybe what we need to

48:22

do is legitimize valuable roles for stories and

48:24

for that that mode of explanation as

48:26

long as we don't mistake it for playing

48:28

the role of giving us the abstract analysis.

48:31

And may maybe for empathy more

48:34

broadly, I did talk to Paul Blum on the

48:36

podcast and he made the opposite claim. He he

48:38

worried that people were too quick to be empathetic

48:40

and that sort of biased them

48:42

toward people like themselves, and they should try

48:44

to be more rational. And I tried to

48:46

say, well, but yes, we should the

48:48

solution to that is

48:50

not to not be empathetic, but to be empathetic to a broader

48:52

spectrum of people. And I'm not sure who's right

48:54

there. Yeah. I I'm

48:56

inclined to to say that there's

48:58

value to empathy, there's value to empathy

49:00

in a broader range of cases, but maybe we

49:02

shouldn't only rely on empathy. Right? That's just one

49:04

of the tools that we have for making sense of a

49:06

situation. So why does it have to

49:08

be rationality or empathy or you know what the utilitarian calculus or the

49:10

kind of empathic response? Maybe we should

49:12

see what each has to offer

49:15

and then integrate both of those into our overall

49:17

evaluation. Fair enough. So how often in this

49:19

whole game do you run across

49:21

the issue of what is the explanation for the explanation? Like,

49:23

is there a bottoming out of these explanations

49:26

anywhere? Are people at what point

49:28

are people satisfied to

49:30

say, like, okay, that's the explanation

49:32

I don't need to dig more deeply. I

49:34

wish I knew the answer to that question. We have

49:36

definitely thought about it. I can I can

49:38

tell you the crumbs we've picked out, you know, in the vicinity, but we

49:40

don't have an answer to

49:42

that. So one thing

49:44

that's really

49:46

interesting is that when you get a

49:48

satisfying explanation, that doesn't stop inquiry. Right? I mean, you might have

49:50

thought that you keep looking till you get a

49:52

satisfying explanation. And then once you get a satisfying explanation,

49:56

Like -- Yeah. -- you're least for kind

49:58

of rich real world cases, like

50:02

explanations for why the dinosaurs became extinct and

50:04

things like

50:06

that, you find is that when somebody received a satisfying explanation,

50:08

they're now more curious about follow-up questions. So

50:10

it's almost like they've found like, oh, there's there's

50:12

stuff to learn here. Good. This is valuable. is

50:16

rich and they keep going. Another thing that

50:18

we have a little bit of evidence for, this

50:20

is more tenuous, but I think it's plausible.

50:23

One place where questioning might stop is where you can't

50:25

imagine a plausible other way than things could

50:28

be. Right? So in order to ask a question like, why

50:30

is the sky blue? It seems like on

50:32

some level, you have to represent the

50:34

possibility that could have been

50:36

nonblue. Right? Yes. And

50:38

so it could be that some

50:40

of when where things bottom out is where you

50:42

just either don't or can't

50:44

represent a real alternative to the way that

50:46

things are. So if you

50:48

can't imagine something, nothing

50:50

existing instead of something, for example, then you might

50:52

not be puzzled by the question of why is there

50:54

something rather than nothing? It's only once you recognize

50:56

this alternative possibility that you're

50:58

able to ask the question, well, y x and not

51:00

y. Right? The y has to be there even

51:03

if it's implicit. scooped me. I was exactly

51:05

next gonna ask you an

51:07

entirely unfair question, which is what is

51:09

your feeling about the

51:12

question Why is there something rather than nothing? I mean, my feeling is that

51:14

that's the kind of grammatical construction

51:16

that looks like it's perfectly well formed,

51:20

but it doesn't actually apply. The world is not the kind of thing that

51:22

necessarily has an explanation

51:24

for why it exists. Yeah. That's interesting.

51:28

I don't know how to think about that case except to say that I I have a hard

51:30

time knowing what the alternative would be.

51:32

Right? And I think that's that's part of my having my

51:34

hard time wrapping around the -- Mhmm. -- I had

51:37

around the question. But as I said, our evidence for that is

51:39

indirect. I don't think I have any great evidence to speak

51:42

to the something rather than nothing question. I mean, I

51:44

will say something, but I think it's

51:46

also sort related to this, which is that

51:48

we found variation in the cases where people are willing to accept that something

51:50

is a mystery. Okay. And so

51:54

we'll try out your intuitions here. So if I said something like,

51:56

why does the moon cause the tides? It's

51:58

a mystery. I mean, that just seems

52:01

or not just because you probably not just because

52:03

you probably actually know something about the

52:05

the physics there, but, like, that's

52:07

the sort of thing. That's not like, it can't be

52:09

a mystery. Right. But if I say something like,

52:11

why does God answer prayer? And I say this

52:14

to somebody who is a believer and

52:16

believe that theories of God

52:18

and God answers prayer and so on. It's a

52:20

mystery. That seems much more

52:22

acceptable. Right. And so that's that's something we

52:24

find as a reliable domain difference. Where people tend think

52:26

that scientific questions sort of

52:28

demand an explanation more than religion questions

52:31

that it's more acceptable to say

52:33

that it's a mystery. For

52:35

the religion questions and for the science

52:37

questions. And we find variation for other things

52:39

in between. I'm not entirely sure what to

52:42

make of this, but philosophy and psychology

52:44

fall somewhere. Between -- Okay. -- National Science had

52:46

had religion in terms of people's willingness

52:48

to sort of accept its mystery

52:50

claims. And so I think part of this has to

52:52

do with

52:54

the perceived limits of human comprehension. That seems to be part of the

52:56

story. We have some data for that, but that isn't the whole

52:58

story. There also seems to be variation across these

53:00

cases and what people think we should

53:04

try to explain. So some people have the sense that some things are appropriate

53:06

targets for explanation and inquiry. Other

53:08

things are not appropriate targets for explanation and inquiry,

53:10

and that's going to be another factor that affects

53:13

I think where explanations bottom out. If you hit a

53:15

point where people no longer think either either think

53:17

it's a mystery or no longer think it's an

53:19

appropriate target for inquiry, you're presumably going to

53:21

stop asking why. But I wonder if

53:23

much is attributable to the kinds of science questions

53:26

that that we're interested in

53:28

here? Because Yeah.

53:30

Why God doesn't answer prayers? Alright.

53:32

That's a mystery. Why

53:34

the moon causes the tides? Is

53:37

not, but then you say, oh, well, because it's gravity this way. But, okay, why does

53:39

gravity work in this way? Well, it has something to

53:41

do with the curvature of spacetime.

53:43

Okay. Why is space time group.

53:45

Like at some point, you will bottom out

53:48

and I think that what the scientists

53:50

would say is that's just how

53:52

it is. There is no further

53:54

explanation, but is that very

53:56

different than saying it's a

53:58

mystery? So we do find this

54:00

subtle difference. In our data where it's a mystery is not

54:02

exactly the same as it's unknown.

54:04

Okay. So in the case of these science

54:06

questions, the

54:08

modal form of ignorance that people expresses, it's unknown to me,

54:10

whereas for religion, the modal

54:12

form of ignorance that people expresses, it's

54:16

a mystery. Not just to me, just

54:18

it's a mystery full style. Interesting. I

54:20

don't know how to think about those fundamental physics

54:22

cases, but people think it's sort of merely unknown,

54:24

merely unknown to them. It noble to everybody? Is it

54:27

deep down the mystery? You know, empirically,

54:29

I have no data there. My

54:31

my own personal sense

54:33

is that it's unknown, but in such

54:35

an unsatisfying way. Right? We

54:38

really really want more and

54:40

think we probably aren't gonna get it. And that might

54:42

be part of what when we call something a

54:44

mystery part of what we might be saying is it's

54:46

unknowable. You know, we've kind of hit a hard, it's

54:48

hard limit. It's not just something that's merely

54:50

unknown now. But this idea of

54:52

what is satisfying to us is so

54:54

crucial. And I do find myself

54:56

sometimes when people demanding to know

54:58

something big

55:00

picture philosophical scientific question,

55:02

I'll I'll give what I think is the best explanation and

55:04

then you say, like, I'm just not satisfied with that. And

55:06

I had to

55:08

say, well, No one ever promised you, you would be satisfied

55:10

with the correct scientific explanation.

55:12

So I think that that desire to be

55:14

satisfied is on the one hand crucially important.

55:16

On the

55:18

other hand, hard to really justify on any foundational That's

55:20

right. And I think that's part of what actually

55:22

that that sense of wanting something satisfying,

55:24

I think is part of what explains

55:28

why explanation plays a effective role in learning for humans. Right? That's

55:30

part of what it's like the motivational

55:32

wing of the exploratory explanation

55:36

seeking enterprise that we

55:38

engage in. So I think it plays an important role,

55:40

but I think there's going to clearly be cases

55:42

where it just sort of leaves

55:44

this way. You know, we're unsatisfied, but there's nothing more to more to say. And

55:46

I think I think a lot of coincidences have that

55:49

kind of character. Right? They sort of strike

55:51

us as things that are you

55:54

know, it it just seems like there there must be

55:56

something too -- Oh. -- the fact that, you know,

55:58

your birthday is the same as my

56:00

birthday and you know, can we can we can we can we do whatever? You know, make make up arbitrary

56:02

coincidences here. And at the end of the day, the

56:04

story is just like, well, I mean, of course, there's a callable

56:06

story about,

56:08

like, that's your birthday. There's some story about why this is my birthday, but there's

56:10

no further backed about us

56:12

both having the same birthday. That feels like,

56:15

that might feel really unsatisfied. So

56:18

I mean, I sometimes think that certain kinds of conspiracy

56:20

beliefs or other kinds of

56:25

crazy ideas that people talk themselves

56:27

into are motivated by these cases where

56:29

it really feels like there should be more to

56:31

say about why things happened the way they

56:33

did. It's really unsatisfying to just in a lot of

56:35

cases, hey, it was just chance. There's nothing

56:38

more to say about what seems like, you

56:40

know, simply calling out for explanation.

56:42

Well, I I agree. I think

56:44

that it's again, it's very hard

56:46

to articulate the degrees

56:48

with which we should accept

56:50

claims like it's just a

56:52

coincidence or it's just a mystery. Right? Like,

56:54

we do fight against that, but sometimes it's just

56:56

right. And I I don't know how exactly to

56:58

say when

57:00

that's okay. Yeah. That's right. I mean, I think I think part of it might be that we don't always know in

57:02

advance. Right? A lot of it might just it's gonna

57:04

be an empirical question. Right. We got

57:06

it right around in a given

57:08

case. And there's the error of thinking, there's nothing further

57:10

to learn.

57:12

When in fact, there is, and that seems like it's

57:14

a problematic error, and then there's the error

57:16

of continuing

57:18

to look. Uh-huh. And in fact, there was

57:20

no explanation. And that's also

57:22

problematic insofar as your spending

57:25

of resources and so on. But you might think that we're better

57:27

off being the kinds of creatures

57:29

who keep trying a little bit more than we should

57:31

because we're under some circumstances. And

57:33

the kinds of creatures who give up too soon? Well,

57:35

you mentioned that the religious in the

57:38

religious context, people are more

57:40

willing to put up with

57:42

mystery as a state of

57:44

explanatory progress. But

57:46

as I recall correctly, there's also some

57:48

different research you did on

57:52

religious versus non religious explanations, and and their

57:54

their connection to sort of

57:56

epistemic functions versus nonepistemic

57:58

functions. Like, the religious religious

58:00

explanations are doing something for us

58:02

other than just giving us knowledge

58:04

about the system. That's right. So what we did in

58:06

that research is that

58:08

we chose to look at explanations for existential

58:10

questions, things like where did the

58:12

universe come from and so on? Mhmm.

58:14

And part of what happens after

58:16

we die Part of reason we did

58:18

that is because those are questions where you

58:20

see people very often spontaneously

58:22

peeling to both religious and non religious sorts

58:24

of explanations. And so we had people

58:26

generate all sorts of explanations, and then we had other participants code

58:28

the characteristics of all of those explanations.

58:32

And some of those characteristics were what we would call broadly epistemic,

58:34

things like how much there's evidence for

58:36

that explanation to what extent it's based on evidence,

58:38

whether it's based on on logical documentation,

58:42

and so on. And we also had them evaluated for what we consider to be

58:44

non episodic characteristics. Like, does this

58:46

give you comfort? Does this reduce negative

58:48

emotions? Is this a sort of explanation

58:52

that promotes moral behavior that brings people

58:54

closer together and so on. So sort of a

58:56

host of social, moral,

58:58

and emotional kinds of

59:00

benefits. And so there were a few interesting results.

59:02

So one is just if you just, on average,

59:04

see what characteristics do people think these

59:06

explanations have, The scientific

59:08

explanations did better than the religious explanations on

59:10

these epistemic dimensions, like

59:12

evidence and logic, and

59:14

the religious explanations did better than the scientific

59:16

explanations. In terms of these nonepidemic

59:18

kinds of characteristics. But part

59:20

of what I think is interesting is that

59:23

you can further break that down by the

59:25

extent to which a given individual believes the explanation. Right?

59:27

So, you know, on a sort

59:29

of five point scale, strong they

59:31

do believe this explanation.

59:34

And we replicate these results for every

59:36

level of belief. So what that means is suppose

59:39

you have somebody two people who

59:41

both give a a given

59:44

religious explanation of four on this five point

59:46

scale. Even though they're giving it

59:48

the same they endorse

59:50

it equally strongly. They're still

59:52

going to say that the scientific one is better

59:54

than the religious one in terms of these

59:56

episodic dimensions and the religious is better than

59:58

in terms of non epidemic kinds of characteristics. I would like to

1:00:01

explain that. Oh, here's here's

1:00:03

here's here's one way to

1:00:06

here's here's a different way to describe it that I think is much more intuitive, and I don't have

1:00:08

the the figures here to point this out as

1:00:10

I as as I talk you through this, either with

1:00:12

you or with your listeners, you'll

1:00:16

have to Hopefully, this will make sense. One way

1:00:18

to make sense of that pattern of findings that I just

1:00:20

described is that people have

1:00:22

different thresholds

1:00:24

for scientific and religious explanations in order to be willing

1:00:26

to believe them at a certain level of confidence.

1:00:28

Okay. For a scientific explanation,

1:00:30

they demand a lot of evidence

1:00:33

and are less demanding with respect to the

1:00:35

nonepisemic characteristics. And before the religious

1:00:38

explanations, it's the reverse. They're very demanding

1:00:40

with respect the nonepisemic

1:00:42

characteristics, they have to meet a high threshold there for

1:00:44

them to believe it, but they're less

1:00:46

demanding with respect to the episodic

1:00:48

characteristics. Got it. Actually, that made perfect

1:00:50

sense. I think you did a very good job of explaining

1:00:52

that. And is this

1:00:54

something Tania particular to religion?

1:00:56

Or what is the pull of

1:00:58

nonepastemic factors when it comes to

1:01:00

us seeking explanations. Like,

1:01:02

I guess, to make it more concrete,

1:01:05

how often do people accept

1:01:07

an explanation for something because it

1:01:09

would lead to good behavior if

1:01:11

people felt that? Yeah, that's a great question. I mean, so one

1:01:13

important qualification that I should make about all of

1:01:16

the research I've talked about, about

1:01:18

religious explanations,

1:01:20

is that these are predominantly

1:01:22

Christian participants in the United States. Fair enough.

1:01:24

Good point. Yes. So when I say religion, I

1:01:26

really mean

1:01:28

that is the population we've studied and it might be different in other

1:01:30

cultures and for other culture traditions or other

1:01:32

religious traditions within the United States. I

1:01:34

think it's an important qualification. But

1:01:37

I think this is actually probably pretty

1:01:40

widespread. So if you think

1:01:42

about, you know, you suppose you have

1:01:44

to explain

1:01:46

why you why you forgot somebody's birthday or something

1:01:48

like that. What what's

1:01:50

gonna make that a good explanation? Well, you have

1:01:52

a lot of goals in giving that explanation.

1:01:56

And some of those goals are presumably more epidemic. I mean, you want to

1:01:58

say things that are true, you want to perhaps

1:02:00

instill true beliefs in the other person.

1:02:03

But you have a lot of other goals. I do.

1:02:05

Yes. You wanna you wanna not think

1:02:07

of yourself as a terrible person. You don't

1:02:09

want the other person to think about you

1:02:11

as a terrible person. And so I think it's

1:02:13

actually just very, very common in everyday explanations that

1:02:15

we're constantly juggling these kinds of epidemic and sorts

1:02:17

of goals, but also these non epidemic Carroll's.

1:02:19

We're constantly regulating

1:02:22

own emotion and other people's emotions. We're constantly thinking about the social

1:02:24

consequences, the moral consequences of what we

1:02:26

believe and other people believe. I I don't think this

1:02:28

always happens consciously and explicitly.

1:02:31

Sometimes we might self deceive ourselves

1:02:33

into the most charitable interpretation of why

1:02:35

we forgot somebody's birthday.

1:02:38

But that makes me think it's actually quite

1:02:40

widespread. Although, you know, it's hard to put a

1:02:42

number on this. And I don't have a number, so I don't

1:02:44

want to I don't want to claim. I have to

1:02:46

support the claim I just said, but that's that's what

1:02:48

my bet would be if we if we've found a good

1:02:50

way to measure that. Is it just hard being a psychologist

1:02:52

sometimes? Because you're too aware of why

1:02:55

you're doing different things for not

1:02:57

always the right reasons? Yes,

1:03:00

but also that's part of what's fun. Right? So for me, everyday

1:03:04

life provides all sorts of fodder for thinking about

1:03:06

the things that I like to think about.

1:03:08

Right. Does it how much

1:03:10

feedback is there? Does being a psychologist affect your

1:03:12

behavior? You know, I'll tell you about the most painful case

1:03:14

of that, which is that I I know I know I

1:03:16

know something about child development. My PhD was

1:03:18

or so cognitive development. I really wish I could tell you that made

1:03:20

me a better parent. I really, really wish

1:03:22

I could tell you that it did.

1:03:27

But mostly, I'm inclined to say no. I think it

1:03:30

probably makes me more

1:03:32

aware of shortcomings. I'm not

1:03:34

sure how much better it makes

1:03:36

me correcting those shortcomings.

1:03:38

Fair enough. I could absolutely believe that.

1:03:40

I mean, certainly being a physicist does not make me

1:03:42

a better billier's player or anything anything

1:03:45

like that. So it's perfectly fair. The

1:03:47

the practicalities of the real world do get in

1:03:49

the way. But but good because that leads me

1:03:52

to the last thing I wanted to ask about,

1:03:54

which was childhood development.

1:03:57

I mean, you mentioned it

1:03:59

a little bit, but it seems like

1:04:01

we are born with

1:04:04

this desire to explain. Can can

1:04:06

we pinpoint that at,

1:04:08

like, two years old, three years old, four years

1:04:10

old, whatever? Is it is there's a moment when

1:04:12

our explaining or

1:04:14

explanation seeking impulses kick

1:04:16

in? Yeah. So so as I mentioned, we do know

1:04:18

that basically as soon as kids have

1:04:20

the language to start asking

1:04:22

questions they do. And so you might think that

1:04:24

that's the earliest we can go. But there's a

1:04:26

couple of very clever

1:04:28

approaches, developmental psychologists have taken to

1:04:30

try to take this question earlier. I'll

1:04:32

you two of my favorite So

1:04:34

one was a study that tried to compare

1:04:36

explanation seeking in humans

1:04:38

and in nonhuman primates. And I'm

1:04:40

pretty sure this was a chimpanzees. And so that raises this really interesting puzzle,

1:04:42

which is what does explanation seeking look

1:04:44

like and how do you measure

1:04:47

it in a nonverbal organism.

1:04:50

And so what they did, which I think is very clever,

1:04:52

is they basically is an exploration

1:04:54

task. So they trained the participants

1:04:56

in this experiment to learn how to balance a

1:04:58

particular block in a particular Tania, and then

1:05:01

they gave them blocks that had internal

1:05:03

weights that would make them

1:05:05

not balance. So that you would try to balance them in the way that

1:05:07

you might think it would balance, it falls over, what

1:05:09

do you do? The thought is that if

1:05:11

you spontaneously are seeking

1:05:14

explanations, what you're gonna do is basically examine it, explore it,

1:05:16

try to figure out what's going on. And

1:05:18

so what they found was at, like, quite young

1:05:22

toddlers. Engaged in this kind of behavior. And so that's a really

1:05:24

nice nonverbal measure of sort of

1:05:26

spontaneous exploration that seems to be

1:05:28

explanation directed. Another thing

1:05:30

that people have done is look at looking

1:05:32

time with infants. Like, what do

1:05:34

infants look at? I'm sorry. Do you

1:05:36

do you do that? Also?

1:05:38

Sorry. Well, it's a little bit complicated. So

1:05:40

-- Okay. -- so mostly they did not.

1:05:42

But there's a really

1:05:45

important difference between experience for the toddlers

1:05:47

and for the primates in that study, which toddlers basically understood the task

1:05:50

immediately with very little

1:05:52

training, whereas the primates had to

1:05:54

have, like, can't

1:05:56

remember what it was, but dozens and dozens and dozens of trials to understand

1:05:58

that what they were supposed to be doing -- Got it. --

1:06:00

was taking these blocks and putting them somewhere. And so it's a

1:06:02

thin fabric. I'm not sure what to make of the fact

1:06:05

after that training regime. They didn't spontaneously explore

1:06:07

because it's not clear to me that they understood

1:06:09

what the task was and what was, you know.

1:06:12

Okay. So I that part of the

1:06:14

result I'm less confident than, but I think the methods

1:06:16

are just super clever. Howard Bauchner:

1:06:18

Thanks. The way

1:06:20

people have sort of thought about for infants is by looking

1:06:22

at what infants look at. Right? Because that's a behavior

1:06:24

that we can measure. And what you can

1:06:26

do is give them two little

1:06:30

sort of scenarios where

1:06:32

one involves something that's physically possible. So

1:06:34

for example, you try to balance a cup

1:06:36

on the edge of a table, and it's such

1:06:38

at the center of mass is on the table. Just, you know,

1:06:40

a little bit of the cup is off the table, and so

1:06:43

you might expect that that will not

1:06:45

fall versus one

1:06:47

where It's, you know, just barely touching the table, so an

1:06:49

adult would expect it to fall. And you

1:06:51

can do various kind of variations like this. But which

1:06:53

one is the kid the child can have the infant can look

1:06:56

more at? And for a lot

1:06:58

of cases like this, the infant spent more time looking at the

1:07:00

thing that is something

1:07:03

like a violation of a principle

1:07:05

or expectation. So the fact that they discriminate

1:07:08

those two pieces and how long they look

1:07:10

tells us something about what their expectations are,

1:07:12

about the way the world works. Now does it tell

1:07:14

us what they're explaining? This is now

1:07:16

much more controversial. But at least

1:07:18

some of the people who do this research have suggested

1:07:20

that part of what's going on when infants

1:07:23

observe these cases is that they try

1:07:25

to construct an explanation for what

1:07:27

they observed. So you're seeing a sort

1:07:29

of mechanism of explanation based learning in these

1:07:31

kinds of cases. And that part of the

1:07:33

way they form the relevant generalizations about their expectations in these cases is

1:07:35

by explaining the cases that they do observe. And

1:07:38

so it's possible that even that

1:07:40

looking behavior

1:07:42

reflects something like the infant looking for an explanation

1:07:44

for this otherwise anomalous event.

1:07:46

Have you seen these videos

1:07:48

where someone hides behind a blanket

1:07:51

and the dog is looking at them and then

1:07:53

they pretend to disappear when they let the

1:07:55

blanket go and the dogs kind of freaks out because the

1:07:57

person isn't there anymore. I haven't seen

1:07:59

the videos, but I can imagine them. I'm not a

1:08:02

phenomenon. Some other comment is one that's a date and

1:08:04

events. Well, that's -- Yeah. -- and and that's why I'm

1:08:06

asking because, you know, object

1:08:08

is that set count as an explanation or is that even more

1:08:10

primitive somehow? Like, if I decide the

1:08:12

distinction between that and the kid and the center

1:08:14

of gravity, yeah, I think that's right. I meant the truth is think

1:08:16

we don't yet know in the infant case

1:08:18

or in the doc case. I'll add that

1:08:20

into If

1:08:22

they're looking longer, if they're startled, I think

1:08:25

that shows something like an expectation.

1:08:27

Does it show something like

1:08:30

explanation seeking? I mean, if if if the dog then

1:08:32

went and, like, sniffed around where the person

1:08:34

was and sort of looked around, I

1:08:36

think that starts to get a little bit

1:08:38

more compelling. But what counts as a genuine nonverbal measure

1:08:41

of explanation seeking? It's just

1:08:43

a really challenging mathematical question.

1:08:47

Fair enough. Okay. I guess the final

1:08:49

thing then is does this maybe you

1:08:51

already answered it informally, but

1:08:53

do these insights

1:08:56

help us in thinking about how to

1:08:58

educate children or even educate ourselves as growing up adults, you know, like better

1:09:00

strategies for seeking explanations and

1:09:02

deploying them in the real world.

1:09:05

I think there might be two lessons, so one of them that I

1:09:07

should acknowledge has been well recognized in the education community as well, is that there are

1:09:09

benefits to explain to yourself and

1:09:12

to others. That's

1:09:14

one of the reasons why peer tutoring, for example, is so

1:09:17

effective. In fact, sometimes in peer tutoring

1:09:19

context, people who benefit more are the

1:09:21

tutors rather than the tutors. They have

1:09:23

the ones who doing the doing the

1:09:25

explaining. So I think that's one thing is just engaging an explanation seems to

1:09:27

be valuable. That's an activity

1:09:29

we could we could do

1:09:32

more of one version of

1:09:34

this that I think is very familiar is you think you understand something until you try to explain it to somebody

1:09:36

else. And then in the

1:09:38

course of doing so, you'd be

1:09:41

who don't. Right? So -- Right. -- if we

1:09:43

had if we engaged in that more spontaneously, we would catch those gaps

1:09:45

in our own understanding. We'd be better calibrated in what we do and

1:09:48

don't understand. So,

1:09:51

explaining is good in educational context. Good. Formal and

1:09:53

informal. The other one though is

1:09:55

that I think maybe we

1:09:57

should be a little bit

1:09:59

wary of expecting satisfying explanations when we look. Alright? I

1:10:01

mean, some of the times the world just really is

1:10:04

complicated. Yeah. Number

1:10:07

of the times, the explanation is not beautiful. So so while

1:10:09

I think explanation seeking is something

1:10:11

we should do, I think at the

1:10:13

same time, we have to at least be cognizant of the

1:10:15

fact that the explanations are not always going to be

1:10:18

beautiful or satisfying, willing to entertain the

1:10:20

possibility that the beautiful

1:10:22

or satisfying explanation is wrong.

1:10:24

I think that sounds very true and

1:10:27

very important, but it calls out for a well formulated theory of

1:10:30

when to stop looking. For

1:10:32

the explanation. And I think we previously agreed that's

1:10:34

pretty tricky question. I wish I had that. Alright. Maybe this is

1:10:37

this is can can figure that out for us.

1:10:39

No. I very much doubt that. I

1:10:42

think this is your job. You and the philosophers, but I

1:10:44

I've learned a lot. You've explained a

1:10:46

lot, and it again, it must be

1:10:49

weird not just being a psychologist, but you're constantly

1:10:51

trying to explain these facts about explanations, and it all gets a little

1:10:53

meta. So I think that we handled it. It

1:10:55

does pretty well. Tania Lombrozo, thanks

1:10:57

very much for being on

1:10:59

Mindscape Thanks for

1:11:02

having me. This

1:11:05

is bed bug.