1
00:00:00,857 --> 00:00:03,230
I'm going to talk about
the strategizing brain.

2
00:00:03,254 --> 00:00:05,640
We're going to use an unusual
combination of tools

3
00:00:05,664 --> 00:00:07,254
from game theory and neuroscience

4
00:00:07,278 --> 00:00:10,548
to understand how people interact socially
when value is on the line.

5
00:00:10,572 --> 00:00:14,044
So game theory is a branch of,
originally, applied mathematics,

6
00:00:14,068 --> 00:00:17,496
used mostly in economics and political
science, a little bit in biology,

7
00:00:17,520 --> 00:00:20,363
that gives us a mathematical
taxonomy of social life,

8
00:00:20,387 --> 00:00:22,706
and it predicts what people
are likely to do

9
00:00:22,730 --> 00:00:24,046
and believe others will do

10
00:00:24,070 --> 00:00:27,043
in cases where everyone's actions
affect everyone else.

11
00:00:27,067 --> 00:00:30,752
That's a lot of things: competition,
cooperation, bargaining,

12
00:00:30,776 --> 00:00:33,128
games like hide-and-seek and poker.

13
00:00:33,954 --> 00:00:35,887
Here's a simple game to get us started.

14
00:00:35,911 --> 00:00:38,375
Everyone chooses a number
from zero to 100.

15
00:00:38,399 --> 00:00:40,859
We're going to compute
the average of those numbers,

16
00:00:40,883 --> 00:00:44,922
and whoever's closest to two-thirds
of the average wins a fixed prize.

17
00:00:44,946 --> 00:00:47,692
So you want to be a little bit
below the average number

18
00:00:47,716 --> 00:00:48,868
but not too far below,

19
00:00:48,892 --> 00:00:52,513
and everyone else wants to be a little bit
below the average number as well.

20
00:00:52,537 --> 00:00:54,084
Think about what you might pick.

21
00:00:54,108 --> 00:00:55,269
As you're thinking,

22
00:00:55,293 --> 00:00:58,347
this is a toy model of something like
selling in the stock market

23
00:00:58,371 --> 00:00:59,989
during a rising market:

24
00:01:00,013 --> 00:01:03,123
You don't want to sell too early,
because you miss out on profits,

25
00:01:03,147 --> 00:01:06,195
but you don't want to wait too late,
to when everyone else sells,

26
00:01:06,219 --> 00:01:07,444
triggering a crash.

27
00:01:07,468 --> 00:01:11,048
You want to be a little bit ahead
of the competition, but not too far ahead.

28
00:01:11,072 --> 00:01:14,157
OK, here's two theories
about how people might think about this,

29
00:01:14,181 --> 00:01:15,381
then we'll see some data.

30
00:01:15,405 --> 00:01:17,001
Some of these will sound familiar

31
00:01:17,025 --> 00:01:19,084
because you probably
are thinking that way.

32
00:01:19,108 --> 00:01:20,760
I'm using my brain theory to see.

33
00:01:20,784 --> 00:01:24,198
A lot of people say, "I really don't know
what people are going to pick,

34
00:01:24,222 --> 00:01:27,749
so I think the average will be 50" --
they're not being strategic at all --

35
00:01:27,773 --> 00:01:29,835
and "I'll pick two-thirds
of 50, that's 33."

36
00:01:29,859 --> 00:01:31,013
That's a start.

37
00:01:31,037 --> 00:01:33,426
Other people, who are a little
more sophisticated,

38
00:01:33,450 --> 00:01:34,691
using more working memory,

39
00:01:34,715 --> 00:01:36,344
say, "I think people will pick 33,

40
00:01:36,368 --> 00:01:38,579
because they're going
to pick a response to 50,

41
00:01:38,603 --> 00:01:40,901
and so I'll pick 22,
which is two-thirds of 33."

42
00:01:40,925 --> 00:01:43,492
They're doing one extra step
of thinking, two steps.

43
00:01:43,913 --> 00:01:45,117
That's better.

44
00:01:45,141 --> 00:01:47,868
Of course, in principle,
you could do three, four or more,

45
00:01:47,892 --> 00:01:49,838
but it starts to get very difficult.

46
00:01:49,862 --> 00:01:51,793
Just like in language and other domains,

47
00:01:51,817 --> 00:01:54,882
we know that it's hard for people
to parse very complex sentences

48
00:01:54,906 --> 00:01:56,197
with a recursive structure.

49
00:01:56,221 --> 00:01:58,370
This is called the cognitive
hierarchy theory,

50
00:01:58,394 --> 00:02:00,643
something I've worked on
and a few other people,

51
00:02:00,667 --> 00:02:02,437
and it indicates a kind of hierarchy,

52
00:02:02,461 --> 00:02:05,900
along with some assumptions about
how many people stop at different steps

53
00:02:05,924 --> 00:02:07,931
and how the steps of thinking are affected

54
00:02:07,955 --> 00:02:10,399
by lots of interesting variables
and variant people,

55
00:02:10,423 --> 00:02:11,623
as we'll see in a minute.

56
00:02:11,647 --> 00:02:14,785
A very different theory, a much
more popular one and an older one,

57
00:02:14,809 --> 00:02:17,288
due largely to John Nash
of "A Beautiful Mind" fame,

58
00:02:17,312 --> 00:02:19,397
is what's called "equilibrium analysis."

59
00:02:19,421 --> 00:02:22,234
So if you've ever taken
a game theory course at any level,

60
00:02:22,258 --> 00:02:24,035
you'll have learned a bit about this.

61
00:02:24,059 --> 00:02:25,887
An equilibrium is a mathematical state

62
00:02:25,911 --> 00:02:29,211
in which everybody has figured out
exactly what everyone else will do.

63
00:02:29,235 --> 00:02:30,579
It is a very useful concept,

64
00:02:30,603 --> 00:02:32,657
but behaviorally,
it may not exactly explain

65
00:02:32,681 --> 00:02:35,955
what people do the first time they play
these types of economic games

66
00:02:35,979 --> 00:02:37,900
or in situations in the outside world.

67
00:02:37,924 --> 00:02:40,725
In this case, the equilibrium
makes a very bold prediction,

68
00:02:40,749 --> 00:02:43,458
which is: everyone wants
to be below everyone else,

69
00:02:43,482 --> 00:02:45,183
therefore, they'll play zero.

70
00:02:45,723 --> 00:02:46,880
Let's see what happens.

71
00:02:46,904 --> 00:02:49,011
This experiment's been done
many, many times.

72
00:02:49,035 --> 00:02:51,237
Some of the earliest ones
were done in the '90s

73
00:02:51,261 --> 00:02:53,067
by me and Rosemarie Nagel and others.

74
00:02:53,091 --> 00:02:55,611
This is a beautiful data set
of 9,000 people

75
00:02:55,635 --> 00:02:58,856
who wrote in to three newspapers
and magazines that had a contest.

76
00:02:58,880 --> 00:03:00,923
The contest said, send in your numbers,

77
00:03:00,947 --> 00:03:04,281
and whoever is close to two-thirds
of the average will win a big prize.

78
00:03:04,305 --> 00:03:08,038
As you can see, there's so much data
here, you can see the spikes very visibly.

79
00:03:08,062 --> 00:03:10,776
There's a spike at 33 --
those are people doing one step.

80
00:03:10,800 --> 00:03:13,019
There is another spike visible at 22.

81
00:03:13,043 --> 00:03:16,059
Notice, by the way, most people
pick numbers right around there;

82
00:03:16,083 --> 00:03:18,241
they don't necessarily
pick exactly 33 and 22.

83
00:03:18,265 --> 00:03:20,181
There's something a bit noisy around it.

84
00:03:20,205 --> 00:03:22,173
But you can see those spikes on that end.

85
00:03:22,197 --> 00:03:23,682
There's another group of people

86
00:03:23,706 --> 00:03:26,193
who seem to have a firm grip
on equilibrium analysis,

87
00:03:26,217 --> 00:03:27,953
because they're picking zero or one.

88
00:03:27,977 --> 00:03:29,624
But they lose, right?

89
00:03:29,648 --> 00:03:33,032
Because picking a number that low
is actually a bad choice

90
00:03:33,056 --> 00:03:35,795
if other people aren't doing
equilibrium analysis as well.

91
00:03:35,819 --> 00:03:37,494
So they're smart, but poor.

92
00:03:37,518 --> 00:03:39,582
(Laughter)

93
00:03:39,606 --> 00:03:42,073
Where are these things
happening in the brain?

94
00:03:42,097 --> 00:03:45,790
One study by Coricelli and Nagel
gives a really sharp, interesting answer.

95
00:03:45,814 --> 00:03:49,626
They had people play this game
while they were being scanned in an fMRI,

96
00:03:49,650 --> 00:03:50,807
and two conditions:

97
00:03:50,831 --> 00:03:52,217
in some trials, they're told,

98
00:03:52,241 --> 00:03:54,838
"You're playing another person
who's playing right now.

99
00:03:54,862 --> 00:03:57,865
We'll match up your behavior
at the end and pay you if you win."

100
00:03:57,889 --> 00:04:00,617
In other trials, they're told,
"You're playing a computer,

101
00:04:00,641 --> 00:04:02,165
they're just choosing randomly."

102
00:04:02,189 --> 00:04:04,351
So what you see here
is a subtraction of areas

103
00:04:04,375 --> 00:04:07,334
in which there's more brain activity
when you're playing people

104
00:04:07,358 --> 00:04:08,936
compared to playing the computer.

105
00:04:08,960 --> 00:04:11,496
And you see activity
in some regions we've seen today,

106
00:04:11,520 --> 00:04:13,769
medial prefrontal cortex,
dorsomedial, up here,

107
00:04:13,793 --> 00:04:16,185
ventromedial prefrontal cortex,
anterior cingulate,

108
00:04:16,209 --> 00:04:19,224
an area that's involved
in lots of types of conflict resolution,

109
00:04:19,248 --> 00:04:20,984
like if you're playing "Simon Says,"

110
00:04:21,008 --> 00:04:24,181
and also the right and left
temporoparietal junction.

111
00:04:24,205 --> 00:04:27,080
And these are all areas
which are fairly reliably known to be

112
00:04:27,104 --> 00:04:29,355
part of what's called
a "theory of mind" circuit

113
00:04:29,379 --> 00:04:30,905
or "mentalizing circuit."

114
00:04:30,929 --> 00:04:34,436
That is, it's a circuit that's used
to imagine what other people might do.

115
00:04:34,460 --> 00:04:38,367
These were some of the first studies
to see this tied in to game theory.

116
00:04:38,778 --> 00:04:41,024
What happens with these
one- and two-step types?

117
00:04:41,048 --> 00:04:43,299
So, we classify people
by what they picked,

118
00:04:43,323 --> 00:04:46,853
and then we look at the difference
between playing humans versus computers,

119
00:04:46,877 --> 00:04:48,942
which brain areas
are differentially active.

120
00:04:48,966 --> 00:04:50,934
On the top, you see the one-step players.

121
00:04:50,958 --> 00:04:52,343
There's almost no difference.

122
00:04:52,367 --> 00:04:55,244
The reason is, they're treating
other people like a computer,

123
00:04:55,268 --> 00:04:56,419
and the brain is too.

124
00:04:56,443 --> 00:04:59,466
The bottom players, you see
all the activity in dorsomedial PFC.

125
00:04:59,490 --> 00:05:02,497
So we know the two-step players
are doing something differently.

126
00:05:02,521 --> 00:05:04,522
Now, what can we do with this information?

127
00:05:04,546 --> 00:05:06,987
You might be able to look
at brain activity and say,

128
00:05:07,011 --> 00:05:10,654
"This person will be a good poker player,"
or "This person's socially naive."

129
00:05:10,678 --> 00:05:14,213
We might also be able to study things
like development of adolescent brains

130
00:05:14,237 --> 00:05:16,674
once we have an idea
of where this circuitry exists.

131
00:05:16,698 --> 00:05:17,850
OK. Get ready.

132
00:05:17,874 --> 00:05:19,974
I'm saving you some brain activity,

133
00:05:19,998 --> 00:05:22,735
because you don't need to use
your hair detector cells.

134
00:05:22,759 --> 00:05:26,021
You should use those cells
to think carefully about this game.

135
00:05:26,045 --> 00:05:27,558
This is a bargaining game.

136
00:05:27,582 --> 00:05:30,599
Two players who are being
scanned using EEG electrodes

137
00:05:30,623 --> 00:05:33,401
are going to bargain
over one to six dollars.

138
00:05:33,425 --> 00:05:36,108
If they can do it in 10 seconds,
they'll earn that money.

139
00:05:36,132 --> 00:05:39,281
If 10 seconds go by and they haven't
made a deal, they get nothing.

140
00:05:39,305 --> 00:05:40,928
That's kind of a mistake together.

141
00:05:40,952 --> 00:05:43,568
The twist is that one player, on the left,

142
00:05:43,592 --> 00:05:45,941
is informed about how much
on each trial there is.

143
00:05:45,965 --> 00:05:48,683
They play lots of trials
with different amounts each time.

144
00:05:48,707 --> 00:05:50,804
In this case, they know
there's four dollars.

145
00:05:50,828 --> 00:05:54,428
The uninformed player doesn't know,
but they know the informed player knows.

146
00:05:54,452 --> 00:05:56,647
So the uninformed player's
challenge is to say,

147
00:05:56,671 --> 00:05:57,822
"Is this guy being fair,

148
00:05:57,846 --> 00:05:59,775
or are they giving me a very low offer

149
00:05:59,799 --> 00:06:03,572
in order to get me to think there's only
one or two dollars available to split?"

150
00:06:03,596 --> 00:06:06,315
in which case they might reject it
and not come to a deal.

151
00:06:06,339 --> 00:06:09,392
So there's some tension here
between trying to get the most money

152
00:06:09,416 --> 00:06:12,091
but trying to goad the other player
into giving you more.

153
00:06:12,115 --> 00:06:14,593
And the way they bargain
is to point on a number line

154
00:06:14,617 --> 00:06:16,296
that goes from zero to six dollars.

155
00:06:16,320 --> 00:06:19,154
They're bargaining over how much
the uninformed player gets,

156
00:06:19,178 --> 00:06:21,179
and the informed player will get the rest.

157
00:06:21,203 --> 00:06:23,347
So this is like
a management-labor negotiation

158
00:06:23,371 --> 00:06:25,101
in which the workers don't know

159
00:06:25,125 --> 00:06:28,331
how much profits
the privately held company has,

160
00:06:28,355 --> 00:06:30,790
and they want to maybe
hold out for more money,

161
00:06:30,814 --> 00:06:33,210
but the company might want
to create the impression

162
00:06:33,234 --> 00:06:36,194
that there's very little to split:
"I'm giving the most I can."

163
00:06:36,218 --> 00:06:39,622
First, some behavior: a bunch
of the subject pairs play face-to-face.

164
00:06:39,646 --> 00:06:42,086
We have other data
where they play across computers.

165
00:06:42,110 --> 00:06:44,684
That's an interesting difference,
as you might imagine.

166
00:06:44,708 --> 00:06:46,482
But a bunch of the face-to-face pairs

167
00:06:46,506 --> 00:06:49,233
agree to divide the money
evenly every single time.

168
00:06:49,257 --> 00:06:51,919
Boring. It's just not
interesting neurally.

169
00:06:52,308 --> 00:06:54,532
It's good for them --
they make a lot of money.

170
00:06:54,556 --> 00:06:56,096
But we're interested in:

171
00:06:56,120 --> 00:06:59,873
Can we say something about when
disagreements occur versus don't occur?

172
00:06:59,897 --> 00:07:02,659
So this is the other group
of subjects, who often disagree.

173
00:07:02,683 --> 00:07:06,160
They bicker and disagree
and end up with less money.

174
00:07:06,184 --> 00:07:09,147
They might be eligible to be
on "Real Housewives," the TV show.

175
00:07:09,171 --> 00:07:10,259
(Laughter)

176
00:07:10,283 --> 00:07:11,966
You see on the left,

177
00:07:11,990 --> 00:07:14,624
when the amount to divide
is one, two or three dollars,

178
00:07:14,648 --> 00:07:16,270
they disagree about half the time;

179
00:07:16,294 --> 00:07:18,645
when it's four, five, six,
they agree quite often.

180
00:07:18,669 --> 00:07:20,868
This turns out to be
something that's predicted

181
00:07:20,892 --> 00:07:22,853
by a very complicated type of game theory

182
00:07:22,877 --> 00:07:25,984
you should come to graduate school
at CalTech and learn about.

183
00:07:26,008 --> 00:07:28,396
It's a little too complicated
to explain right now,

184
00:07:28,420 --> 00:07:31,063
but the theory tells you
that this shape should occur.

185
00:07:31,087 --> 00:07:33,149
Your intuition might tell you that, too.

186
00:07:33,173 --> 00:07:36,040
Now I'm going to show you
the results from the EEG recording.

187
00:07:36,064 --> 00:07:37,215
Very complicated.

188
00:07:37,239 --> 00:07:39,631
The right brain schematic
is the uninformed person,

189
00:07:39,655 --> 00:07:41,055
and the left is the informed.

190
00:07:41,079 --> 00:07:43,825
Remember that we scanned
both brains at the same time,

191
00:07:43,849 --> 00:07:46,118
so we can ask about time-synced activity

192
00:07:46,142 --> 00:07:49,158
in similar or different
areas simultaneously,

193
00:07:49,182 --> 00:07:51,447
just like if you wanted
to study a conversation,

194
00:07:51,471 --> 00:07:54,049
and you were scanning two people
talking to each other.

195
00:07:54,073 --> 00:07:56,331
You'd expect common
activity in language regions

196
00:07:56,355 --> 00:07:58,316
when they're listening and communicating.

197
00:07:58,340 --> 00:08:02,171
So the arrows connect regions
that are active at the same time.

198
00:08:02,195 --> 00:08:03,517
The direction of the arrows

199
00:08:03,541 --> 00:08:06,307
flows from the region
that's active first in time,

200
00:08:06,331 --> 00:08:10,126
and the arrowhead goes
to the region that's active later.

201
00:08:10,150 --> 00:08:12,197
So in this case, if you look carefully,

202
00:08:12,221 --> 00:08:14,244
most of the arrows
flow from right to left.

203
00:08:14,268 --> 00:08:17,552
That is, it looks
as if the uninformed brain activity

204
00:08:17,576 --> 00:08:19,187
is happening first,

205
00:08:19,211 --> 00:08:23,063
and then it's followed
by activity in the informed brain.

206
00:08:23,087 --> 00:08:26,538
And by the way, these are trials
where their deals were made.

207
00:08:26,562 --> 00:08:28,319
This is from the first two seconds.

208
00:08:28,343 --> 00:08:31,499
We haven't finished analyzing this data,
so we're still peeking in,

209
00:08:31,523 --> 00:08:34,931
but the hope is that we can say something
in the first couple of seconds

210
00:08:34,955 --> 00:08:36,918
about whether they'll make a deal or not,

211
00:08:36,942 --> 00:08:39,947
which could be very useful in thinking
about avoiding litigation

212
00:08:39,971 --> 00:08:41,835
and ugly divorces and things like that.

213
00:08:41,859 --> 00:08:45,936
Those are all cases in which a lot
of value is lost by delay and strikes.

214
00:08:46,630 --> 00:08:48,794
Here's the case where
the disagreements occur.

215
00:08:48,818 --> 00:08:51,212
You can see it looks different
than the one before.

216
00:08:51,236 --> 00:08:52,577
There's a lot more arrows.

217
00:08:52,601 --> 00:08:55,252
That means that the brains
are synced up more closely

218
00:08:55,276 --> 00:08:56,896
in terms of simultaneous activity,

219
00:08:56,920 --> 00:08:59,123
and the arrows flow clearly
from left to right.

220
00:08:59,147 --> 00:09:01,435
That is, the informed brain
seems to be deciding,

221
00:09:01,459 --> 00:09:03,651
"We're probably not going
to make a deal here."

222
00:09:03,675 --> 00:09:06,418
And then later, there's activity
in the uninformed brain.

223
00:09:06,799 --> 00:09:09,203
Next, I'm going to introduce you
to some relatives.

224
00:09:09,227 --> 00:09:11,388
They're hairy, smelly, fast and strong.

225
00:09:11,412 --> 00:09:13,906
You might be thinking back
to your last Thanksgiving.

226
00:09:13,930 --> 00:09:14,946
(Laughter)

227
00:09:14,970 --> 00:09:17,446
Maybe, if you had a chimpanzee with you.

228
00:09:17,470 --> 00:09:21,476
Charles Darwin and I and you broke
off from the family tree from chimpanzees

229
00:09:21,500 --> 00:09:22,900
about five million years ago.

230
00:09:22,924 --> 00:09:24,735
They're still our closest genetic kin.

231
00:09:24,759 --> 00:09:26,478
We share 98.8 percent of the genes.

232
00:09:26,502 --> 00:09:29,463
We share more genes with them
than zebras do with horses.

233
00:09:29,487 --> 00:09:31,397
And we're also their closest cousin.

234
00:09:31,421 --> 00:09:34,042
They have more genetic relation
to us than to gorillas.

235
00:09:34,066 --> 00:09:36,805
So, how humans and chimpanzees
behave differently

236
00:09:36,829 --> 00:09:38,923
might tell us a lot about brain evolution.

237
00:09:39,326 --> 00:09:41,626
This is an amazing memory test

238
00:09:41,650 --> 00:09:44,442
from [Kyoto], Japan,
the Primate Research Institute,

239
00:09:44,466 --> 00:09:46,469
where they've done a lot of this research.

240
00:09:46,493 --> 00:09:49,317
This goes back a ways.
They're interested in working memory.

241
00:09:49,341 --> 00:09:51,057
The chimp will see, watch carefully,

242
00:09:51,081 --> 00:09:54,665
they'll see 200 milliseconds' exposure --
that's fast, eight movie frames --

243
00:09:54,689 --> 00:09:56,666
of numbers one, two, three, four, five.

244
00:09:56,690 --> 00:09:58,935
Then they disappear
and are replaced by squares,

245
00:09:58,959 --> 00:10:00,586
and they have to press the squares

246
00:10:00,610 --> 00:10:02,810
that correspond to the numbers
from low to high

247
00:10:02,834 --> 00:10:04,137
to get an apple reward.

248
00:10:04,161 --> 00:10:05,658
Let's see how they can do it.

249
00:10:16,478 --> 00:10:17,640
This is a young chimp.

250
00:10:17,664 --> 00:10:20,581
The young ones are better
than the old ones, just like humans.

251
00:10:20,605 --> 00:10:21,607
(Laughter)

252
00:10:21,631 --> 00:10:23,109
And they're highly experienced,

253
00:10:23,133 --> 00:10:25,456
they've done this thousands of times.

254
00:10:25,480 --> 00:10:28,366
Obviously there's a big training
effect, as you can imagine.

255
00:10:28,390 --> 00:10:29,402
(Laughter)

256
00:10:29,426 --> 00:10:31,574
You can see they're very
blasé and effortless.

257
00:10:31,598 --> 00:10:34,809
Not only can they do it very well,
they do it in a sort of lazy way.

258
00:10:34,833 --> 00:10:35,837
(Laughter)

259
00:10:35,861 --> 00:10:37,623
Who thinks you could beat the chimps?

260
00:10:37,647 --> 00:10:38,707
(Laughter)

261
00:10:38,731 --> 00:10:40,266
Wrong. (Laughter)

262
00:10:40,290 --> 00:10:42,875
We can try. We'll try. Maybe we'll try.

263
00:10:42,899 --> 00:10:46,893
OK, so the next part of the study
I'm going to go quickly through

264
00:10:46,917 --> 00:10:49,893
is based on an idea of Tetsuro Matsuzawa.

265
00:10:49,917 --> 00:10:53,037
He had a bold idea he called
the "cognitive trade-off hypothesis."

266
00:10:53,061 --> 00:10:56,543
We know chimps are faster and stronger;
they're also obsessed with status.

267
00:10:56,567 --> 00:10:59,248
His thought was, maybe
they've preserved brain activities

268
00:10:59,272 --> 00:11:00,875
and practice them in development

269
00:11:00,899 --> 00:11:04,714
that are really, really important to them
to negotiate status and to win,

270
00:11:04,738 --> 00:11:07,730
which is something like strategic
thinking during competition.

271
00:11:07,754 --> 00:11:09,290
So we're going to check that out

272
00:11:09,314 --> 00:11:11,941
by having the chimps actually play a game

273
00:11:11,965 --> 00:11:14,475
by touching two touch screens.

274
00:11:14,499 --> 00:11:17,559
The chimps are interacting
with each other through the computers.

275
00:11:17,583 --> 00:11:18,932
They'll press left or right.

276
00:11:18,956 --> 00:11:22,434
One chimp is called a matcher;
they win if they press left-left,

277
00:11:22,458 --> 00:11:25,603
like a seeker finding someone
in hide-and-seek, or right-right.

278
00:11:25,627 --> 00:11:27,232
The mismatcher wants to mismatch;

279
00:11:27,256 --> 00:11:29,955
they want to press
the opposite screen of the chimp.

280
00:11:29,979 --> 00:11:32,451
And the rewards are apple cube rewards.

281
00:11:32,475 --> 00:11:34,802
So here's how game theorists
look at these data.

282
00:11:34,826 --> 00:11:36,848
This is a graph of the percentage of times

283
00:11:36,872 --> 00:11:39,078
the matcher picked right on the x-axis

284
00:11:39,102 --> 00:11:41,256
and the percentage of times
they picked right

285
00:11:41,280 --> 00:11:43,485
by the mismatcher on the y-axis.

286
00:11:43,509 --> 00:11:46,838
So a point here is the behavior
by a pair of players,

287
00:11:46,862 --> 00:11:49,058
one trying to match,
one trying to mismatch.

288
00:11:49,082 --> 00:11:52,399
The NE square in the middle --
actually, NE, CH and QRE --

289
00:11:52,423 --> 00:11:55,547
those are three different theories
of Nash equilibrium and others,

290
00:11:55,571 --> 00:11:57,254
tells you what the theory predicts,

291
00:11:57,278 --> 00:11:59,403
which is that they should match 50-50,

292
00:11:59,427 --> 00:12:01,854
because if you play left
too much, for example,

293
00:12:01,878 --> 00:12:04,844
I can exploit that if I'm the mismatcher
by then playing right.

294
00:12:04,868 --> 00:12:07,840
And as you can see, the chimps --
each chimp is one triangle --

295
00:12:07,864 --> 00:12:10,523
are circled around,
hovering around that prediction.

296
00:12:11,205 --> 00:12:12,911
Now we move the payoffs.

297
00:12:12,935 --> 00:12:16,422
We're going to make the left-left payoff
for the matcher a little higher.

298
00:12:16,446 --> 00:12:17,941
Now they get three apple cubes.

299
00:12:17,965 --> 00:12:21,240
Game theoretically, that should
make the mismatcher's behavior shift:

300
00:12:21,264 --> 00:12:24,771
the mismatcher will think, "Oh, this guy's
going to go for the big reward,

301
00:12:24,795 --> 00:12:27,323
so I'll go to the right,
make sure he doesn't get it."

302
00:12:27,347 --> 00:12:29,375
And as you can see,
their behavior moves up

303
00:12:29,399 --> 00:12:32,097
in the direction of this change
in the Nash equilibrium.

304
00:12:32,121 --> 00:12:34,367
Finally, we changed
the payoffs one more time.

305
00:12:34,391 --> 00:12:35,639
Now it's four apple cubes,

306
00:12:35,663 --> 00:12:38,496
and their behavior again moves
towards the Nash equilibrium.

307
00:12:38,520 --> 00:12:41,194
It's sprinkled around,
but if you average the chimps out,

308
00:12:41,218 --> 00:12:42,792
they're really close, within .01.

309
00:12:42,816 --> 00:12:45,444
They're actually closer
than any species we've observed.

310
00:12:45,468 --> 00:12:48,566
What about humans? You think
you're smarter than a chimpanzee?

311
00:12:49,350 --> 00:12:52,651
Here's two human groups in green and blue.

312
00:12:52,675 --> 00:12:55,968
They're closer to 50-50; they're not
responding to payoffs as closely.

313
00:12:55,992 --> 00:12:58,288
And also if you study
their learning in the game,

314
00:12:58,312 --> 00:13:00,413
they aren't as sensitive
to previous rewards.

315
00:13:00,437 --> 00:13:04,022
The chimps play better than the humans,
in terms of adhering to game theory.

316
00:13:04,046 --> 00:13:07,247
And these are two different groups
of humans, from Japan and Africa;

317
00:13:07,271 --> 00:13:08,611
they replicate quite nicely.

318
00:13:08,635 --> 00:13:11,180
None of them are close
to where the chimps are.

319
00:13:11,670 --> 00:13:12,964
So, some things we learned:

320
00:13:12,988 --> 00:13:16,734
people seem to do a limited amount of
strategic thinking using theory of mind.

321
00:13:16,758 --> 00:13:18,852
We have preliminary
evidence from bargaining

322
00:13:18,876 --> 00:13:21,791
that early warning signs in the brain
might be used to predict

323
00:13:21,815 --> 00:13:24,446
whether there'll be a bad
disagreement that costs money,

324
00:13:24,470 --> 00:13:26,709
and chimps are "better"
competitors than humans,

325
00:13:26,733 --> 00:13:27,975
as judged by game theory.

326
00:13:27,999 --> 00:13:29,150
Thank you.

327
00:13:29,174 --> 00:13:32,293
(Applause)