1
00:00:00,000 --> 00:00:05,561
Hi, welcome back. In this lecture, I want
to talk a little bit more about how using

2
00:00:05,561 --> 00:00:09,374
models can help make you a more
intelligent citizen of the world. And so,

3
00:00:09,374 --> 00:00:13,417
we're gonna break this down into a bunch
of set of sub-reasons about why models

4
00:00:13,417 --> 00:00:17,352
make you better able to engage in all the
things that are going on in this modern,

5
00:00:17,352 --> 00:00:20,903
complex world in which we live. Okay, so.
When we think about models, they're

6
00:00:20,903 --> 00:00:24,550
simplifications. They're abstractions. So
in a sense, there's a sense in which

7
00:00:24,550 --> 00:00:28,389
they're wrong. There's a famous quote by
George Box, where he says, "All models are

8
00:00:28,389 --> 00:00:32,084
wrong." And that's true, right? They are.
"But some are useful." And that's gonna be

9
00:00:32,084 --> 00:00:35,539
a mantra that comes up throughout this
course. These models are gonna be

10
00:00:35,539 --> 00:00:38,946
abstractions, they're gonna be
simplifications, but they're be useful to

11
00:00:38,946 --> 00:00:42,468
us. They're gonna help us do things in
better ways. 'K? So. In a, in a sense,

12
00:00:42,468 --> 00:00:46,158
right? And this is a, a big thing in this
course. Models are the new lingua franca.

13
00:00:46,158 --> 00:00:49,618
They're the language of not only the
academy, you know, which I talked about

14
00:00:49,618 --> 00:00:53,447
some in the last lecture, but they're the
language of business. They're the language

15
00:00:53,447 --> 00:00:57,046
of politics. They're the language of the
nonprofit world. Wherever you go, where

16
00:00:57,046 --> 00:01:00,782
there's people are trying to do good, make
money, cure disease, whatever it is that

17
00:01:00,782 --> 00:01:04,335
they wanna do, right? You're gonna find
that people are using models to enable

18
00:01:04,335 --> 00:01:08,071
them to be better at what their purpose
is. Okay? That's why they've really become

19
00:01:08,071 --> 00:01:11,716
the new lingua franca. So, if you think
back. Remember, I talked about this in the

20
00:01:11,716 --> 00:01:15,453
first lecture. The whole idea of having a
great books movement was that there was

21
00:01:15,453 --> 00:01:19,639
these... set of ideas that any person should
know. So within the hundred and so great

22
00:01:19,639 --> 00:01:23,497
books, there were thousands of ideas. And
one of our ad learners, Robert Hutchins

23
00:01:23,497 --> 00:01:27,354
President of the University of Chicago. They had
this thing that they wrote called the

24
00:01:27,354 --> 00:01:31,264
Synopticon which was a list, right, as they put
this together. This was kind of list of

25
00:01:31,264 --> 00:01:35,430
sort of all the ideas that someone should
know, that an intelligent person should

26
00:01:35,430 --> 00:01:39,442
know. So what are those ideas? So one of
those ideas was to tie yourself to the

27
00:01:39,442 --> 00:01:43,506
mast. And this comes from the Odyssey, you
know, this says the ship is going past the

28
00:01:43,506 --> 00:01:47,672
sirens and he wants to hear the sirens
beautiful love song. So what he does is he

29
00:01:47,672 --> 00:01:51,873
has his crew tie him to the mast. He tie.
Ties himself to the mast so he can listen

30
00:01:51,873 --> 00:01:56,109
to them but pre-commit to not driving his
boat over to hear the sirens, at the same

31
00:01:56,109 --> 00:01:59,784
time he puts wax in the ears of his crew
so they also won't be, you know,

32
00:01:59,784 --> 00:02:04,071
encouraged to sort of drive the boat over
there. Well this is an idea that recurs in

33
00:02:04,071 --> 00:02:08,182
history when we think about. Cortez
burning his ships, right, so his men won't

34
00:02:08,182 --> 00:02:12,541
you know retreat, they'll continue to
advance. So this idea to tie yourself to

35
00:02:12,541 --> 00:02:16,253
math, is a real worthwhile thing. But
here's the problem. One of, one of my

36
00:02:16,253 --> 00:02:19,957
favorite websites is a website called
Office of Proverbs. So on this websites,

37
00:02:19,957 --> 00:02:23,949
it says things like he who hesitates is
lost, a stitch in time saves nine, or

38
00:02:23,949 --> 00:02:27,700
two heads are better than one, too many
cooks spoil the broth. So you've got to

39
00:02:27,700 --> 00:02:31,591
hear this really good advice, something
that probably made it in the Synopticon,

40
00:02:31,596 --> 00:02:35,148
but then you get something that says the
exact opposite. Well, how do they

41
00:02:35,148 --> 00:02:39,340
adjudicate between those two things? The
way we adjudicate between those two things

42
00:02:39,340 --> 00:02:43,532
is by constructing models because models
give us the conditions under which he who

43
00:02:43,532 --> 00:02:47,522
hesitates is lost, and then there's the
conditions under which a stitch in time

44
00:02:47,522 --> 00:02:51,613
saves nine. So when we talk about the wide
diversity and prediction, we'll see why

45
00:02:51,613 --> 00:02:55,654
it's the case that two heads is better
than one, and we'll see why it's the case

46
00:02:55,654 --> 00:02:59,694
that too many cooks spoil the broth. So,
ironically, what models do is they tie us

47
00:02:59,694 --> 00:03:03,684
to a mast, they tie us to a mast of logic
and by tying us to a mast of logic, we

48
00:03:03,684 --> 00:03:08,652
figure out which ways of thinking, which
ideas in this Are useful to us.'K? So, if

49
00:03:08,652 --> 00:03:13,116
you look at almost any discipline, whether
its economics, and here what you see in

50
00:03:13,116 --> 00:03:17,634
this diagram, is you see a description of
sort of, this is a, a utility function for

51
00:03:17,634 --> 00:03:21,878
an agent. And what that agent is doing
trying to maximize their pay-off, right?

52
00:03:21,878 --> 00:03:25,900
So, economists use models all the time.
Biologists use models, as well. They,

53
00:03:25,900 --> 00:03:30,143
they, you know, have, you know, models of
the brain, where they have little axons

54
00:03:30,143 --> 00:03:34,331
and dendrites going between the neurons.
They have models of gene regulatory

55
00:03:34,331 --> 00:03:38,464
networks. They have models species, right?
Things like that. Sociology, we have

56
00:03:38,464 --> 00:03:42,611
models, as well, right? So, there's models
of, sort of. How your identity effects

57
00:03:42,611 --> 00:03:46,411
your actions, and your behaviors and
things like that. Okay, in political

58
00:03:46,411 --> 00:03:50,186
science. We have models. Political science
these days, this is a picture of a spatial

59
00:03:50,186 --> 00:03:53,934
voting model. So they might say candidates
are a little more conservative on certain

60
00:03:53,934 --> 00:03:57,280
dimensions and voters are a little more
conservative and you say that, well,

61
00:03:57,280 --> 00:04:00,983
you're more likely to vote for a candidate
who takes positions similar to yourself.

62
00:04:00,983 --> 00:04:04,508
So my work at the University of Michigan
we have something called the National

63
00:04:04,508 --> 00:04:08,032
Election Studies that's run out of there
where we sort of gather all this data

64
00:04:08,032 --> 00:04:11,423
about where politicians are and where
voters are, and that allows us to make

65
00:04:11,423 --> 00:04:15,184
sense of who votes for whom and why. Okay?
So models help us understand the decisions

66
00:04:15,184 --> 00:04:18,550
people make. Linguistics, right? Here's
another area, right? So you might think,

67
00:04:18,550 --> 00:04:21,871
how can you use models in linguistics?
Well, this little model here, you see

68
00:04:21,871 --> 00:04:25,685
things where it says you see v's and n and
p's in here, if you look closely. Well, v

69
00:04:25,685 --> 00:04:29,006
stands for verb, n stands for noun, and
well you gotta. And S stands for, you

70
00:04:29,006 --> 00:04:32,238
know, subject, let's say, right? So you
can do this: you can ask "What is the

71
00:04:32,238 --> 00:04:35,693
structure of a language?" You can ask,
formally and mathematically, what are the

72
00:04:35,693 --> 00:04:39,014
structure of a language is, and whether
some languages are more like other

73
00:04:39,014 --> 00:04:42,560
languages or not, depending on how people,
you know, set up their sentences. So in

74
00:04:42,560 --> 00:04:46,001
German, where they may put all the
adjectives. At the end of the sentence

75
00:04:46,001 --> 00:04:50,132
that looks very different than let's say
English. All right. Even the law. This is

76
00:04:50,132 --> 00:04:54,138
a graph from one of my graduate students,
former graduate student. Now, he's a law

77
00:04:54,138 --> 00:04:58,194
professor, Dan Katz. Where he's got sort
of a network model of which Supreme Court

78
00:04:58,194 --> 00:05:02,149
justices, you know, who they appoint, so
who, if someone appoints judges from some

79
00:05:02,149 --> 00:05:06,055
other judge. By putting that data that's
out there in this sort of model-based

80
00:05:06,055 --> 00:05:09,711
form, we can begin to understand how
conservative and how liberal certain

81
00:05:09,711 --> 00:05:13,817
judges are. All right? So, there's lots of
ways to use models, and there's even whole

82
00:05:13,817 --> 00:05:17,622
disciplines now, that have evolved, that
are based entirely on models. So, game

83
00:05:17,622 --> 00:05:21,628
theory, which is what I was really trained
in as a graduate student, is all about

84
00:05:21,628 --> 00:05:25,507
strategic behavior. Behavior. It's the
study of strategic interactions between,

85
00:05:25,507 --> 00:05:29,525
you know, individuals, companies, nations.
Right? And game theory can also be applied

86
00:05:29,525 --> 00:05:33,494
to biology, right? So there's all sorts of
stuff, right? When you go to, when you go

87
00:05:33,494 --> 00:05:37,215
to, you know, college, you go to college,
you'll find that there's game theory

88
00:05:37,215 --> 00:05:41,233
models of just about anything. Right? So
it's actually a field based entirely just

89
00:05:41,233 --> 00:05:45,631
on models. Right? Why, right? [laugh] Why
all these models, right? Why does

90
00:05:45,631 --> 00:05:50,508
everything from linguistics, to economics
to, you know, political science use

91
00:05:50,508 --> 00:05:55,450
models? Well, cuz, they're better, right?
They're just better than we are. So, let

92
00:05:55,450 --> 00:06:00,392
me show you a graph, here. This is a graph
from a book by Phil Tetlock. It's a

93
00:06:00,392 --> 00:06:05,919
fabulous book. And in this graph, he, what
he's showing is, he's showing the accuracy

94
00:06:05,919 --> 00:06:10,065
of, some different, let me pull up a pin
here. Different ways of predicting. So,

95
00:06:10,065 --> 00:06:13,512
what you see on this axis, this
calibration axis, right here. This is

96
00:06:13,512 --> 00:06:17,217
asking, sort of how, showing you how
accurate a model is. And this axis is

97
00:06:17,217 --> 00:06:21,487
saying how discriminating is it, in terms
of how particular, how fine of predictions

98
00:06:21,487 --> 00:06:25,705
is it making. So, instead of saying is it
hot or cold, it might be saying it's gonna

99
00:06:25,705 --> 00:06:29,410
be 90 degrees, or 80 degrees, or 70
degrees. So this axis here, this up and

100
00:06:29,410 --> 00:06:32,908
down axis, is discriminatoriness,
discrimination, and this axis is how

101
00:06:32,908 --> 00:06:36,715
accurate. So, what you see here, down
here, are hedgehogs. So, these are people

102
00:06:36,715 --> 00:06:40,823
who use a single model. Hedgehogs are not
very good at predicting. Right? They're

103
00:06:40,823 --> 00:06:45,295
terrible at predicting. Up here are people
he calls foxes. Now, foxes are people who

104
00:06:45,295 --> 00:06:49,713
use lots of models. They have sort of lots
of loose models in their head. And, they

105
00:06:49,713 --> 00:06:53,695
do much better at, you know, sort of at
calibration, a little bit better at

106
00:06:53,695 --> 00:06:58,058
discrimination, than individuals. But way
up here, [laugh] better than anybody, are

107
00:06:58,058 --> 00:07:02,312
formal models. Formal models just do
better than either foxes or hedgehogs. Now

108
00:07:02,312 --> 00:07:06,457
[inaudible] how much data is this? Tetlock
actually had tens of thousands of

109
00:07:06,457 --> 00:07:10,565
predictions. So, over a 20-year period, he
gathered predictions by people. And

110
00:07:10,565 --> 00:07:15,154
compared how those people did to models.
And the answer is models do much, much

111
00:07:15,154 --> 00:07:19,772
better. Okay. All right, so. What about
people, then, who actually make

112
00:07:19,772 --> 00:07:23,194
predictions for a living? So, this is a
picture of Bruce Bueno de Mesquita, who

113
00:07:23,194 --> 00:07:26,616
makes predictions about what's gonna
happen in international relationships, and

114
00:07:26,616 --> 00:07:30,261
he's very good at it. He's so good at it
that they put his picture on the cover of

115
00:07:30,261 --> 00:07:33,816
magazines, right? He's at, Stanford and
NYU. Chair of the department at NYU. Used

116
00:07:33,816 --> 00:07:37,372
to be, anyway. So, Bruce, uses models.
He's got a very elaborate model that helps

117
00:07:37,372 --> 00:07:40,483
him figure out, based on sort of
bargaining position and interest, what

118
00:07:40,483 --> 00:07:43,594
different countries are gonna do. But,
just like George Box said at the

119
00:07:43,594 --> 00:07:47,016
beginning, he doesn't base his decision
entirely on that model. What the model

120
00:07:47,016 --> 00:07:50,572
does is gives him guidance as to what he
then thinks. So, it's a blending of what

121
00:07:50,572 --> 00:07:54,716
the formal model tells him, and.
Experience tells them so smart people who

122
00:07:54,716 --> 00:07:59,984
use models but the models don't tell them
what to do. Okay. Another reason models

123
00:07:59,984 --> 00:08:04,858
have taken yeah they are better but
they're also very fertile. So once you

124
00:08:04,858 --> 00:08:08,621
learn a model. It's, you know, for one
domain, you can apply to a whole bunch of

125
00:08:08,621 --> 00:08:11,913
other domains, which is fascinating. So
we're gonna learn something called

126
00:08:11,913 --> 00:08:15,657
mark-off processes, which are models about
dynamic processes. So they can be used to

127
00:08:15,657 --> 00:08:19,220
model things like disease spread and stuff
like that, right? We're gonna finally

128
00:08:19,220 --> 00:08:22,783
learn though that you can also use them,
this is sorta surprising, to figure out

129
00:08:22,783 --> 00:08:26,715
who wrote a book. >> [laugh] And they say,
how does that happen? Well that happens

130
00:08:26,715 --> 00:08:30,433
because you can think of words, writing a
sentence, as an anemic process. So

131
00:08:30,433 --> 00:08:34,402
different authors, right, use different
sequences of words. Different patterns. So

132
00:08:34,402 --> 00:08:38,521
therefore we can use this mathematical
model that wasn't developed in any way for

133
00:08:38,521 --> 00:08:42,290
this purpose to figure out who wrote what
book, okay? Totally cool. All right.

134
00:08:42,290 --> 00:08:46,543
Another big reason. Models really make us
humble. The reason they make us humble is

135
00:08:46,543 --> 00:08:50,744
we just have to lay out sort of all the
logic and then we realize holy cow, I had

136
00:08:50,744 --> 00:08:54,635
no idea that this was going to happen,
right. So often when we construct the

137
00:08:54,635 --> 00:08:58,784
model, we're going to get very different
predictions than what we thought before,

138
00:08:58,784 --> 00:09:02,934
right. So if you look at things, here's a
picture of a, the tulip graph, right, from

139
00:09:03,090 --> 00:09:06,368
When there's a big, in the six-,
seventeenth century, when there's a, you

140
00:09:06,368 --> 00:09:09,974
know, this big spike in tulip prices. You
can imagine that people thought that

141
00:09:09,974 --> 00:09:13,814
prices were gonna continue to go up and up
and up. Well, if you had a simple linear

142
00:09:13,814 --> 00:09:17,514
model, you might have invested heavily in
tulips, and lost a lot of money. So, one

143
00:09:17,514 --> 00:09:21,167
reason that models make us humble is,
never go back to the George Box code. All

144
00:09:21,167 --> 00:09:24,633
models are wrong, right? So, a model is
going to be wrong. But the models are

145
00:09:24,633 --> 00:09:28,239
humbling to us, because they sort of make
us see the full dimensionality of a

146
00:09:28,239 --> 00:09:31,985
problem. So, once we try and write down a
model of any sort of system, it's a very

147
00:09:31,985 --> 00:09:35,685
humbling exercise, because we realize how
much we've gotta leave out to try and

148
00:09:35,685 --> 00:09:39,190
understand what's going on. All right.
Here's another example, right? This is the

149
00:09:39,190 --> 00:09:42,748
Case-Shiller Home Price Index, and what
you see is, you see prices going up and up

150
00:09:42,748 --> 00:09:46,262
an up, right? And then you see this, let
me put a pin up here, precipitous crash

151
00:09:46,262 --> 00:09:49,557
right here, right? A lot of people had
models that just said, look, things are

152
00:09:49,557 --> 00:09:53,247
gonna continue this way. There were a few
people that had models that said things go

153
00:09:53,247 --> 00:09:56,761
down. These people, the ones whose models
went down, they made a lotta money. These

154
00:09:56,761 --> 00:10:00,099
people thought it was gonna go up, didn't.
So, we're always gonna see a lot of

155
00:10:00,099 --> 00:10:03,614
diversity in models, and you're really not
gonna know, often until after the fact,

156
00:10:03,614 --> 00:10:06,908
which one is right. And so, one thing
that's gonna be really important is to

157
00:10:06,908 --> 00:10:10,159
have many models. So, let's go back to
that fox-hedgehog graphite that we, I

158
00:10:10,159 --> 00:10:14,220
showed you before. The, the foxes, the
people with lots of models, did much

159
00:10:14,220 --> 00:10:19,127
better than the hedgehogs, the people with
no models. And former models did better

160
00:10:19,127 --> 00:10:22,438
than. The foxes. Well, what would do
better than formal models? Well, people

161
00:10:22,438 --> 00:10:26,076
with lots of formal models. Right? So if
we really want to make sense of the world

162
00:10:26,076 --> 00:10:29,759
what we want to do is have lots of formal
models in our disclosures. So what we're

163
00:10:29,759 --> 00:10:33,397
going to do in this class is almost like,
remember the old, like, sixteen, 32 box of

164
00:10:33,397 --> 00:10:37,079
Crayolas? That's sort of what we're doing
here. Right? We're just going to pick up a

165
00:10:37,079 --> 00:10:40,627
whole bunch of models. And we're going to
have them, right, they're fertile. We're

166
00:10:40,627 --> 00:10:44,265
going to plot across a bunch of settings.
So when we're confronted with something

167
00:10:44,265 --> 00:10:47,947
what we can do is pull out our models. Ask
which ones are appropriate, and in doing

168
00:10:47,947 --> 00:10:51,772
so, right, be better at what we do. So the
essence of Tetlocks's book, right? That's

169
00:10:51,772 --> 00:10:55,959
where that graph came from with the foxes
and hedgehogs is that, the only people who

170
00:10:55,959 --> 00:10:59,847
are really even better than what he. He
has a way of classifying what a random

171
00:10:59,847 --> 00:11:03,834
choice would be. The only people who are
better than random at predicting what's

172
00:11:03,834 --> 00:11:07,971
gonna happen are people who use multiple
models. And that's the kind of people that

173
00:11:07,971 --> 00:11:12,325
we wanna be. Okay, so thats, sort of the
big, intelligent citizen of the world

174
00:11:12,325 --> 00:11:16,556
logic, right. There is, models, are
incredibly fertile, they make us humble,

175
00:11:16,556 --> 00:11:21,088
they help, you know really clarify the
logic, and they're just better. Okay? So

176
00:11:21,088 --> 00:11:25,459
if you wanna be out there, you know,
helping to change the world in useful

177
00:11:25,459 --> 00:11:30,251
ways, it's really, really helpful to have
some understanding of models. Thank you

178
00:11:30,251 --> 00:11:30,912
very much.