1
00:00:00,562 --> 00:00:02,815
I'm going to talk about the strategizing brain.

2
00:00:02,815 --> 00:00:05,053
We're going to use an unusual combination of tools

3
00:00:05,053 --> 00:00:07,066
from game theory and neuroscience

4
00:00:07,066 --> 00:00:10,033
to understand how people interact socially when value is on the line.

5
00:00:10,033 --> 00:00:14,004
So game theory is a branch of, originally, applied mathematics,

6
00:00:14,004 --> 00:00:16,990
used mostly in economics and political science, a little bit in biology,

7
00:00:16,990 --> 00:00:20,279
that gives us a mathematical taxonomy of social life

8
00:00:20,279 --> 00:00:22,410
and it predicts what people are likely to do

9
00:00:22,410 --> 00:00:23,604
and believe others will do

10
00:00:23,604 --> 00:00:26,994
in cases where everyone's actions affect everyone else.

11
00:00:26,994 --> 00:00:30,472
That's a lot of things: competition, cooperation, bargaining,

12
00:00:30,472 --> 00:00:33,763
games like hide-and-seek, and poker.

13
00:00:33,763 --> 00:00:36,002
Here's a simple game to get us started.

14
00:00:36,002 --> 00:00:38,164
Everyone chooses a number from zero to 100,

15
00:00:38,164 --> 00:00:40,610
we're going to compute the average of those numbers,

16
00:00:40,610 --> 00:00:44,804
and whoever's closest to two-thirds of the average wins a fixed prize.

17
00:00:44,804 --> 00:00:47,073
So you want to be a little bit below the average number,

18
00:00:47,073 --> 00:00:49,309
but not too far below, and everyone else wants to be

19
00:00:49,309 --> 00:00:51,254
a little bit below the average number as well.

20
00:00:51,254 --> 00:00:53,833
Think about what you might pick.

21
00:00:53,833 --> 00:00:57,104
As you're thinking, this is a toy model of something like

22
00:00:57,104 --> 00:00:59,916
selling in the stock market during a rising market. Right?

23
00:00:59,916 --> 00:01:02,141
You don't want to sell too early, because you miss out on profits,

24
00:01:02,141 --> 00:01:04,321
but you don't want to wait too late

25
00:01:04,321 --> 00:01:06,722
to when everyone else sells, triggering a crash.

26
00:01:06,722 --> 00:01:09,601
You want to be a little bit ahead of the competition, but not too far ahead.

27
00:01:09,601 --> 00:01:13,214
Okay, here's two theories about how people might think about this,

28
00:01:13,214 --> 00:01:14,610
and then we'll see some data.

29
00:01:14,610 --> 00:01:16,801
Some of these will sound familiar because you probably are

30
00:01:16,801 --> 00:01:20,600
thinking that way. I'm using my brain theory to see.

31
00:01:20,600 --> 00:01:23,710
A lot of people say, "I really don't know what people are going to pick,

32
00:01:23,710 --> 00:01:25,383
so I think the average will be 50."

33
00:01:25,383 --> 00:01:27,294
They're not being really strategic at all.

34
00:01:27,294 --> 00:01:30,646
"And I'll pick two-thirds of 50. That's 33." That's a start.

35
00:01:30,646 --> 00:01:32,538
Other people who are a little more sophisticated,

36
00:01:32,538 --> 00:01:34,014
using more working memory,

37
00:01:34,014 --> 00:01:37,918
say, "I think people will pick 33 because they're going to pick a response to 50,

38
00:01:37,918 --> 00:01:40,886
and so I'll pick 22, which is two-thirds of 33."

39
00:01:40,886 --> 00:01:43,365
They're doing one extra step of thinking, two steps.

40
00:01:43,365 --> 00:01:45,982
That's better. And of course, in principle,

41
00:01:45,982 --> 00:01:47,809
you could do three, four or more,

42
00:01:47,809 --> 00:01:49,678
but it starts to get very difficult.

43
00:01:49,678 --> 00:01:52,270
Just like in language and other domains, we know that it's hard for people to parse

44
00:01:52,270 --> 00:01:55,904
very complex sentences with a kind of recursive structure.

45
00:01:55,904 --> 00:01:57,638
This is called a cognitive hierarchy theory, by the way.

46
00:01:57,638 --> 00:02:00,194
It's something that I've worked on and a few other people,

47
00:02:00,194 --> 00:02:02,414
and it indicates a kind of hierarchy along with

48
00:02:02,414 --> 00:02:04,668
some assumptions about how many people stop at different steps

49
00:02:04,668 --> 00:02:06,552
and how the steps of thinking are affected

50
00:02:06,552 --> 00:02:10,248
by lots of interesting variables and variant people, as we'll see in a minute.

51
00:02:10,248 --> 00:02:13,634
A very different theory, a much more popular one, and an older one,

52
00:02:13,634 --> 00:02:17,174
due largely to John Nash of "A Beautiful Mind" fame,

53
00:02:17,174 --> 00:02:19,414
is what's called equilibrium analysis.

54
00:02:19,414 --> 00:02:21,868
So if you've ever taken a game theory course at any level,

55
00:02:21,868 --> 00:02:23,581
you will have learned a little bit about this.

56
00:02:23,581 --> 00:02:26,436
An equilibrium is a mathematical state in which everybody

57
00:02:26,436 --> 00:02:28,885
has figured out exactly what everyone else will do.

58
00:02:28,885 --> 00:02:30,892
It is a very useful concept, but behaviorally,

59
00:02:30,892 --> 00:02:32,895
it may not exactly explain what people do

60
00:02:32,895 --> 00:02:35,630
the first time they play these types of economic games

61
00:02:35,630 --> 00:02:37,963
or in situations in the outside world.

62
00:02:37,963 --> 00:02:40,301
In this case, the equilibrium makes a very bold prediction,

63
00:02:40,301 --> 00:02:43,161
which is everyone wants to be below everyone else,

64
00:02:43,161 --> 00:02:45,452
therefore they'll play zero.

65
00:02:45,452 --> 00:02:48,461
Let's see what happens. This experiment's been done many, many times.

66
00:02:48,461 --> 00:02:50,344
Some of the earliest ones were done in the '90s

67
00:02:50,344 --> 00:02:52,989
by me and Rosemarie Nagel and others.

68
00:02:52,989 --> 00:02:55,974
This is a beautiful data set of 9,000 people who wrote in

69
00:02:55,974 --> 00:02:58,854
to three newspapers and magazines that had a contest.

70
00:02:58,854 --> 00:03:00,668
The contest said, send in your numbers

71
00:03:00,668 --> 00:03:03,823
and whoever is close to two-thirds of the average will win a big prize.

72
00:03:03,823 --> 00:03:06,911
And as you can see, there's so much data here, you can see the spikes very visibly.

73
00:03:06,911 --> 00:03:10,292
There's a spike at 33. Those are people doing one step.

74
00:03:10,292 --> 00:03:12,789
There is another spike visible at 22.

75
00:03:12,789 --> 00:03:15,081
And notice, by the way, that most people pick numbers right around there.

76
00:03:15,081 --> 00:03:17,591
They don't necessarily pick exactly 33 and 22.

77
00:03:17,591 --> 00:03:19,647
There's something a little bit noisy around it.

78
00:03:19,647 --> 00:03:21,125
But you can see those spikes, and they're there.

79
00:03:21,125 --> 00:03:22,835
There's another group of people who seem to have

80
00:03:22,835 --> 00:03:24,910
a firm grip on equilibrium analysis,

81
00:03:24,910 --> 00:03:27,305
because they're picking zero or one.

82
00:03:27,305 --> 00:03:29,394
But they lose, right?

83
00:03:29,394 --> 00:03:32,586
Because picking a number that low is actually a bad choice

84
00:03:32,586 --> 00:03:35,406
if other people aren't doing equilibrium analysis as well.

85
00:03:35,406 --> 00:03:37,518
So they're smart, but poor.

86
00:03:37,518 --> 00:03:39,606
(Laughter)

87
00:03:39,606 --> 00:03:41,575
Where are these things happening in the brain?

88
00:03:41,575 --> 00:03:45,450
One study by Coricelli and Nagel gives a really sharp, interesting answer.

89
00:03:45,450 --> 00:03:46,958
So they had people play this game

90
00:03:46,958 --> 00:03:49,175
while they were being scanned in an fMRI,

91
00:03:49,175 --> 00:03:51,446
and two conditions: in some trials,

92
00:03:51,446 --> 00:03:52,961
they're told you're playing another person

93
00:03:52,961 --> 00:03:54,549
who's playing right now and we're going to match up

94
00:03:54,549 --> 00:03:56,753
your behavior at the end and pay you if you win.

95
00:03:56,753 --> 00:03:58,731
In the other trials, they're told, you're playing a computer.

96
00:03:58,731 --> 00:04:00,365
They're just choosing randomly.

97
00:04:00,365 --> 00:04:02,442
So what you see here is a subtraction

98
00:04:02,442 --> 00:04:05,192
of areas in which there's more brain activity

99
00:04:05,192 --> 00:04:08,168
when you're playing people compared to playing the computer.

100
00:04:08,168 --> 00:04:10,159
And you see activity in some regions we've seen today,

101
00:04:10,159 --> 00:04:13,396
medial prefrontal cortex, dorsomedial, however, up here,

102
00:04:13,396 --> 00:04:15,247
ventromedial prefrontal cortex,

103
00:04:15,247 --> 00:04:16,601
anterior cingulate, an area that's involved

104
00:04:16,601 --> 00:04:20,238
in lots of types of conflict resolution, like if you're playing "Simon Says,"

105
00:04:20,238 --> 00:04:24,052
and also the right and left temporoparietal junction.

106
00:04:24,052 --> 00:04:26,518
And these are all areas which are fairly reliably known

107
00:04:26,518 --> 00:04:28,839
to be part of what's called a "theory of mind" circuit,

108
00:04:28,839 --> 00:04:30,740
or "mentalizing circuit."

109
00:04:30,740 --> 00:04:34,118
That is, it's a circuit that's used to imagine what other people might do.

110
00:04:34,118 --> 00:04:36,358
So these were some of the first studies to see this

111
00:04:36,358 --> 00:04:38,391
tied in to game theory.

112
00:04:38,391 --> 00:04:40,631
What happens with these one- and two-step types?

113
00:04:40,631 --> 00:04:42,702
So we classify people by what they picked,

114
00:04:42,702 --> 00:04:44,369
and then we look at the difference between

115
00:04:44,369 --> 00:04:46,344
playing humans versus playing computers,

116
00:04:46,344 --> 00:04:48,235
which brain areas are differentially active.

117
00:04:48,235 --> 00:04:49,987
On the top you see the one-step players.

118
00:04:49,987 --> 00:04:51,507
There's almost no difference.

119
00:04:51,507 --> 00:04:54,447
The reason is, they're treating other people like a computer, and the brain is too.

120
00:04:54,447 --> 00:04:58,588
The bottom players, you see all the activity in dorsomedial PFC.

121
00:04:58,588 --> 00:05:00,639
So we know that those two-step players are doing something differently.

122
00:05:00,639 --> 00:05:03,735
Now if you were to step back and say, "What can we do with this information?"

123
00:05:03,735 --> 00:05:05,556
you might be able to look at brain activity and say,

124
00:05:05,556 --> 00:05:07,055
"This person's going to be a good poker player,"

125
00:05:07,055 --> 00:05:08,984
or, "This person's socially naive,"

126
00:05:08,984 --> 00:05:10,262
and we might also be able to study things

127
00:05:10,262 --> 00:05:11,860
like development of adolescent brains

128
00:05:11,860 --> 00:05:15,214
once we have an idea of where this circuitry exists.

129
00:05:15,214 --> 00:05:17,826
Okay. Get ready.

130
00:05:17,826 --> 00:05:19,949
I'm saving you some brain activity,

131
00:05:19,949 --> 00:05:22,759
because you don't need to use your hair detector cells.

132
00:05:22,759 --> 00:05:25,647
You should use those cells to think carefully about this game.

133
00:05:25,647 --> 00:05:27,582
This is a bargaining game.

134
00:05:27,582 --> 00:05:30,138
Two players who are being scanned using EEG electrodes

135
00:05:30,138 --> 00:05:33,015
are going to bargain over one to six dollars.

136
00:05:33,015 --> 00:05:35,679
If they can do it in 10 seconds, they're going to actually earn that money.

137
00:05:35,679 --> 00:05:38,719
If 10 seconds goes by and they haven't made a deal, they get nothing.

138
00:05:38,719 --> 00:05:40,402
That's kind of a mistake together.

139
00:05:40,402 --> 00:05:43,219
The twist is that one player, on the left,

140
00:05:43,219 --> 00:05:45,907
is informed about how much on each trial there is.

141
00:05:45,907 --> 00:05:48,139
They play lots of trials with different amounts each time.

142
00:05:48,139 --> 00:05:50,380
In this case, they know there's four dollars.

143
00:05:50,380 --> 00:05:52,257
The uninformed player doesn't know,

144
00:05:52,257 --> 00:05:54,311
but they know that the informed player knows.

145
00:05:54,311 --> 00:05:56,370
So the uninformed player's challenge is to say,

146
00:05:56,370 --> 00:05:57,840
"Is this guy really being fair

147
00:05:57,840 --> 00:05:59,694
or are they giving me a very low offer

148
00:05:59,694 --> 00:06:02,772
in order to get me to think that there's only one or two dollars available to split?"

149
00:06:02,772 --> 00:06:05,926
in which case they might reject it and not come to a deal.

150
00:06:05,926 --> 00:06:08,876
So there's some tension here between trying to get the most money

151
00:06:08,876 --> 00:06:11,449
but trying to goad the other player into giving you more.

152
00:06:11,449 --> 00:06:13,779
And the way they bargain is to point on a number line

153
00:06:13,779 --> 00:06:15,585
that goes from zero to six dollars,

154
00:06:15,585 --> 00:06:18,563
and they're bargaining over how much the uninformed player gets,

155
00:06:18,563 --> 00:06:20,148
and the informed player's going to get the rest.

156
00:06:20,148 --> 00:06:22,723
So this is like a management-labor negotiation

157
00:06:22,723 --> 00:06:25,456
in which the workers don't know how much profits

158
00:06:25,456 --> 00:06:28,123
the privately held company has, right,

159
00:06:28,123 --> 00:06:30,491
and they want to maybe hold out for more money,

160
00:06:30,491 --> 00:06:32,327
but the company might want to create the impression

161
00:06:32,327 --> 00:06:35,259
that there's very little to split: "I'm giving you the most that I can."

162
00:06:35,259 --> 00:06:39,490
First some behavior. So a bunch of the subject pairs, they play face to face.

163
00:06:39,490 --> 00:06:41,326
We have some other data where they play across computers.

164
00:06:41,326 --> 00:06:43,064
That's an interesting difference, as you might imagine.

165
00:06:43,064 --> 00:06:45,266
But a bunch of the face-to-face pairs

166
00:06:45,266 --> 00:06:48,959
agree to divide the money evenly every single time.

167
00:06:48,959 --> 00:06:51,865
Boring. It's just not interesting neurally.

168
00:06:51,865 --> 00:06:54,379
It's good for them. They make a lot of money.

169
00:06:54,379 --> 00:06:57,051
But we're interested in, can we say something about

170
00:06:57,051 --> 00:06:59,587
when disagreements occur versus don't occur?

171
00:06:59,587 --> 00:07:01,944
So this is the other group of subjects who often disagree.

172
00:07:01,944 --> 00:07:04,712
So they have a chance of -- they bicker and disagree

173
00:07:04,712 --> 00:07:06,019
and end up with less money.

174
00:07:06,019 --> 00:07:09,936
They might be eligible to be on "Real Housewives," the TV show.

175
00:07:09,936 --> 00:07:11,872
You see on the left,

176
00:07:11,872 --> 00:07:14,536
when the amount to divide is one, two or three dollars,

177
00:07:14,536 --> 00:07:16,184
they disagree about half the time,

178
00:07:16,184 --> 00:07:18,376
and when the amount is four, five, six, they agree quite often.

179
00:07:18,376 --> 00:07:20,250
This turns out to be something that's predicted

180
00:07:20,250 --> 00:07:22,454
by a very complicated type of game theory

181
00:07:22,454 --> 00:07:25,263
you should come to graduate school at CalTech and learn about.

182
00:07:25,263 --> 00:07:27,435
It's a little too complicated to explain right now,

183
00:07:27,435 --> 00:07:30,851
but the theory tells you that this shape kind of should occur.

184
00:07:30,851 --> 00:07:33,067
Your intuition might tell you that too.

185
00:07:33,067 --> 00:07:35,307
Now I'm going to show you the results from the EEG recording.

186
00:07:35,307 --> 00:07:37,660
Very complicated. The right brain schematic

187
00:07:37,660 --> 00:07:40,523
is the uninformed person, and the left is the informed.

188
00:07:40,523 --> 00:07:43,323
Remember that we scanned both brains at the same time,

189
00:07:43,323 --> 00:07:45,715
so we can ask about time-synced activity

190
00:07:45,715 --> 00:07:48,939
in similar or different areas simultaneously,

191
00:07:48,939 --> 00:07:51,203
just like if you wanted to study a conversation

192
00:07:51,203 --> 00:07:53,139
and you were scanning two people talking to each other

193
00:07:53,139 --> 00:07:55,499
and you'd expect common activity in language regions

194
00:07:55,499 --> 00:07:57,884
when they're actually kind of listening and communicating.

195
00:07:57,884 --> 00:08:01,811
So the arrows connect regions that are active at the same time,

196
00:08:01,811 --> 00:08:03,851
and the direction of the arrows flows

197
00:08:03,851 --> 00:08:06,331
from the region that's active first in time,

198
00:08:06,331 --> 00:08:09,899
and the arrowhead goes to the region that's active later.

199
00:08:09,899 --> 00:08:12,115
So in this case, if you look carefully,

200
00:08:12,115 --> 00:08:13,972
most of the arrows flow from right to left.

201
00:08:13,972 --> 00:08:17,252
That is, it looks as if the uninformed brain activity

202
00:08:17,252 --> 00:08:19,211
is happening first,

203
00:08:19,211 --> 00:08:22,726
and then it's followed by activity in the informed brain.

204
00:08:22,726 --> 00:08:26,418
And by the way, these were trials where their deals were made.

205
00:08:26,418 --> 00:08:28,198
This is from the first two seconds.

206
00:08:28,198 --> 00:08:30,178
We haven't finished analyzing this data,

207
00:08:30,178 --> 00:08:32,078
so we're still peeking in, but the hope is

208
00:08:32,078 --> 00:08:34,642
that we can say something in the first couple of seconds

209
00:08:34,642 --> 00:08:36,365
about whether they'll make a deal or not,

210
00:08:36,365 --> 00:08:38,408
which could be very useful in thinking about avoiding litigation

211
00:08:38,408 --> 00:08:40,336
and ugly divorces and things like that.

212
00:08:40,336 --> 00:08:43,219
Those are all cases in which a lot of value is lost

213
00:08:43,219 --> 00:08:46,195
by delay and strikes.

214
00:08:46,195 --> 00:08:48,225
Here's the case where the disagreements occur.

215
00:08:48,225 --> 00:08:50,398
You can see it looks different than the one before.

216
00:08:50,398 --> 00:08:52,647
There's a lot more arrows.

217
00:08:52,647 --> 00:08:54,158
That means that the brains are synced up

218
00:08:54,158 --> 00:08:56,710
more closely in terms of simultaneous activity,

219
00:08:56,710 --> 00:08:58,720
and the arrows flow clearly from left to right.

220
00:08:58,720 --> 00:09:00,962
That is, the informed brain seems to be deciding,

221
00:09:00,962 --> 00:09:03,250
"We're probably not going to make a deal here."

222
00:09:03,250 --> 00:09:06,475
And then later there's activity in the uninformed brain.

223
00:09:06,475 --> 00:09:08,978
Next I'm going to introduce you to some relatives.

224
00:09:08,978 --> 00:09:11,239
They're hairy, smelly, fast and strong.

225
00:09:11,239 --> 00:09:14,429
You might be thinking back to your last Thanksgiving.

226
00:09:14,429 --> 00:09:17,122
Maybe if you had a chimpanzee with you.

227
00:09:17,122 --> 00:09:20,583
Charles Darwin and I and you broke off from the family tree

228
00:09:20,583 --> 00:09:22,842
from chimpanzees about five million years ago.

229
00:09:22,842 --> 00:09:24,810
They're still our closest genetic kin.

230
00:09:24,810 --> 00:09:26,547
We share 98.8 percent of the genes.

231
00:09:26,547 --> 00:09:29,347
We share more genes with them than zebras do with horses.

232
00:09:29,347 --> 00:09:31,064
And we're also their closest cousin.

233
00:09:31,064 --> 00:09:34,066
They have more genetic relation to us than to gorillas.

234
00:09:34,066 --> 00:09:36,594
So how humans and chimpanzees behave differently

235
00:09:36,594 --> 00:09:39,049
might tell us a lot about brain evolution.

236
00:09:39,049 --> 00:09:41,650
So this is an amazing memory test

237
00:09:41,650 --> 00:09:44,466
from Nagoya, Japan, Primate Research Institute,

238
00:09:44,466 --> 00:09:46,244
where they've done a lot of this research.

239
00:09:46,244 --> 00:09:48,584
This goes back quite a ways. They're interested in working memory.

240
00:09:48,584 --> 00:09:50,356
The chimp is going to see, watch carefully,

241
00:09:50,356 --> 00:09:52,558
they're going to see 200 milliseconds' exposure

242
00:09:52,558 --> 00:09:54,552
— that's fast, that's eight movie frames —

243
00:09:54,552 --> 00:09:56,503
of numbers one, two, three, four, five.

244
00:09:56,503 --> 00:09:58,501
Then they disappear and they're replaced by squares,

245
00:09:58,501 --> 00:10:00,256
and they have to press the squares

246
00:10:00,256 --> 00:10:02,577
that correspond to the numbers from low to high

247
00:10:02,577 --> 00:10:03,916
to get an apple reward.

248
00:10:03,916 --> 00:10:08,687
Let's see how they can do it.

249
00:10:16,391 --> 00:10:17,884
This is a young chimp. The young ones

250
00:10:17,884 --> 00:10:20,667
are better than the old ones, just like humans.

251
00:10:20,667 --> 00:10:22,259
And they're highly experienced, so they've done this

252
00:10:22,259 --> 00:10:23,691
thousands and thousands of time.

253
00:10:23,691 --> 00:10:26,575
Obviously there's a big training effect, as you can imagine.

254
00:10:27,928 --> 00:10:29,272
(Laughter)

255
00:10:29,272 --> 00:10:31,207
You can see they're very blasé and kind of effortless.

256
00:10:31,207 --> 00:10:35,135
Not only can they do it very well, they do it in a sort of lazy way.

257
00:10:35,135 --> 00:10:38,570
Right? Who thinks you could beat the chimps?

258
00:10:38,570 --> 00:10:40,166
Wrong. (Laughter)

259
00:10:40,166 --> 00:10:42,604
We can try. We'll try. Maybe we'll try.

260
00:10:42,604 --> 00:10:45,194
Okay, so the next part of this study

261
00:10:45,194 --> 00:10:46,790
I'm going to go quickly through

262
00:10:46,790 --> 00:10:49,482
is based on an idea of Tetsuro Matsuzawa.

263
00:10:49,482 --> 00:10:52,511
He had a bold idea that -- what he called the cognitive trade-off hypothesis.

264
00:10:52,511 --> 00:10:53,803
We know chimps are faster and stronger.

265
00:10:53,803 --> 00:10:55,483
They're also very obsessed with status.

266
00:10:55,483 --> 00:10:58,439
His thought was, maybe they've preserved brain activities

267
00:10:58,439 --> 00:11:00,607
and they practice them in development

268
00:11:00,607 --> 00:11:02,458
that are really, really important to them

269
00:11:02,458 --> 00:11:04,668
to negotiate status and to win,

270
00:11:04,668 --> 00:11:07,666
which is something like strategic thinking during competition.

271
00:11:07,666 --> 00:11:09,246
So we're going to check that out

272
00:11:09,246 --> 00:11:11,676
by having the chimps actually play a game

273
00:11:11,676 --> 00:11:14,314
by touching two touch screens.

274
00:11:14,314 --> 00:11:16,754
The chimps are actually interacting with each other through the computers.

275
00:11:16,754 --> 00:11:18,362
They're going to press left or right.

276
00:11:18,362 --> 00:11:20,475
One chimp is called a matcher.

277
00:11:20,475 --> 00:11:22,458
They win if they press left, left,

278
00:11:22,458 --> 00:11:25,627
like a seeker finding someone in hide-and-seek, or right, right.

279
00:11:25,627 --> 00:11:26,855
The mismatcher wants to mismatch.

280
00:11:26,855 --> 00:11:29,931
They want to press the opposite screen of the chimp.

281
00:11:29,931 --> 00:11:32,475
And the rewards are apple cube rewards.

282
00:11:32,475 --> 00:11:35,003
So here's how game theorists look at these data.

283
00:11:35,003 --> 00:11:36,617
This is a graph of the percentage of times

284
00:11:36,617 --> 00:11:39,235
the matcher picked right on the x-axis,

285
00:11:39,235 --> 00:11:40,751
and the percentage of times they predicted right

286
00:11:40,751 --> 00:11:43,619
by the mismatcher on the y-axis.

287
00:11:43,619 --> 00:11:46,819
So a point here is the behavior by a pair of players,

288
00:11:46,819 --> 00:11:49,035
one trying to match, one trying to mismatch.

289
00:11:49,035 --> 00:11:52,315
The NE square in the middle -- actually NE, CH and QRE --

290
00:11:52,315 --> 00:11:54,771
those are three different theories of Nash equilibrium, and others,

291
00:11:54,771 --> 00:11:57,195
tells you what the theory predicts,

292
00:11:57,195 --> 00:11:59,467
which is that they should match 50-50,

293
00:11:59,467 --> 00:12:01,635
because if you play left too much, for example,

294
00:12:01,635 --> 00:12:04,351
I can exploit that if I'm the mismatcher by then playing right.

295
00:12:04,351 --> 00:12:07,115
And as you can see, the chimps, each chimp is one triangle,

296
00:12:07,115 --> 00:12:10,962
are circled around, hovering around that prediction.

297
00:12:10,962 --> 00:12:12,736
Now we move the payoffs.

298
00:12:12,736 --> 00:12:16,179
We're actually going to make the left, left payoff for the matcher a little bit higher.

299
00:12:16,179 --> 00:12:17,699
Now they get three apple cubes.

300
00:12:17,699 --> 00:12:20,499
Game theoretically, that should actually make the mismatcher's behavior shift,

301
00:12:20,499 --> 00:12:22,089
because what happens is, the mismatcher will think,

302
00:12:22,089 --> 00:12:23,899
oh, this guy's going to go for the big reward,

303
00:12:23,899 --> 00:12:26,964
and so I'm going to go to the right, make sure he doesn't get it.

304
00:12:26,964 --> 00:12:28,629
And as you can see, their behavior moves up

305
00:12:28,629 --> 00:12:32,077
in the direction of this change in the Nash equilibrium.

306
00:12:32,077 --> 00:12:34,391
Finally, we changed the payoffs one more time.

307
00:12:34,391 --> 00:12:35,583
Now it's four apple cubes,

308
00:12:35,583 --> 00:12:37,755
and their behavior again moves towards the Nash equilibrium.

309
00:12:37,755 --> 00:12:39,790
It's sprinkled around, but if you average the chimps out,

310
00:12:39,790 --> 00:12:41,980
they're really, really close, within .01.

311
00:12:41,980 --> 00:12:45,179
They're actually closer than any species we've observed.

312
00:12:45,179 --> 00:12:48,372
What about humans? You think you're smarter than a chimpanzee?

313
00:12:48,372 --> 00:12:51,939
Here's two human groups in green and blue.

314
00:12:51,939 --> 00:12:55,683
They're closer to 50-50. They're not responding to payoffs as closely,

315
00:12:55,683 --> 00:12:57,133
and also if you study their learning in the game,

316
00:12:57,133 --> 00:12:59,118
they aren't as sensitive to previous rewards.

317
00:12:59,118 --> 00:13:00,482
The chimps are playing better than the humans,

318
00:13:00,482 --> 00:13:02,905
better in the sense of adhering to game theory.

319
00:13:02,905 --> 00:13:04,324
And these are two different groups of humans

320
00:13:04,324 --> 00:13:07,520
from Japan and Africa. They replicate quite nicely.

321
00:13:07,520 --> 00:13:10,755
None of them are close to where the chimps are.

322
00:13:10,755 --> 00:13:12,510
So here are some things we learned today.

323
00:13:12,510 --> 00:13:14,448
People seem to do a limited amount of strategic thinking

324
00:13:14,448 --> 00:13:16,259
using theory of mind.

325
00:13:16,259 --> 00:13:18,176
We have some preliminary evidence from bargaining

326
00:13:18,176 --> 00:13:20,668
that early warning signs in the brain might be used to predict

327
00:13:20,668 --> 00:13:22,894
whether there will be a bad disagreement that costs money,

328
00:13:22,894 --> 00:13:24,734
and chimps are better competitors than humans,

329
00:13:24,734 --> 00:13:27,198
as judged by game theory.

330
00:13:27,198 --> 00:13:29,055
Thank you.

331
00:13:29,055 --> 00:13:32,628
(Applause)