1
00:00:00,800 --> 00:00:03,136
I'm going to talk
about the strategizing brain.

2
00:00:03,160 --> 00:00:05,536
We're going to use
an unusual combination of tools

3
00:00:05,560 --> 00:00:07,176
from game theory and neuroscience

4
00:00:07,200 --> 00:00:10,456
to understand how people interact socially
when value is on the line.

5
00:00:10,480 --> 00:00:14,016
So game theory is a branch
of, originally, applied mathematics,

6
00:00:14,040 --> 00:00:16,256
used mostly in economics
and political science,

7
00:00:16,280 --> 00:00:17,536
a little bit in biology,

8
00:00:17,560 --> 00:00:20,336
that gives us a mathematical
taxonomy of social life

9
00:00:20,360 --> 00:00:22,656
and it predicts
what people are likely to do

10
00:00:22,680 --> 00:00:23,976
and believe others will do

11
00:00:24,000 --> 00:00:27,016
in cases where everyone's actions
affect everyone else.

12
00:00:27,040 --> 00:00:30,696
That's a lot of things:
competition, cooperation, bargaining,

13
00:00:30,720 --> 00:00:33,280
games like hide-and-seek and poker.

14
00:00:33,920 --> 00:00:35,856
Here's a simple game to get us started.

15
00:00:35,880 --> 00:00:38,296
Everyone chooses a number
from zero to 100,

16
00:00:38,320 --> 00:00:40,776
we're going to compute
the average of those numbers,

17
00:00:40,800 --> 00:00:44,816
and whoever's closest to two-thirds
of the average wins a fixed prize.

18
00:00:44,840 --> 00:00:47,616
So you want to be
a little bit below the average number,

19
00:00:47,640 --> 00:00:48,896
but not too far below,

20
00:00:48,920 --> 00:00:50,336
and everyone else wants to be

21
00:00:50,360 --> 00:00:52,536
a little bit below
the average number as well.

22
00:00:52,560 --> 00:00:54,136
Think about what you might pick.

23
00:00:54,160 --> 00:00:55,376
As you're thinking,

24
00:00:55,400 --> 00:00:58,456
this is a toy model of something
like selling in the stock market

25
00:00:58,480 --> 00:00:59,936
during a rising market. Right?

26
00:00:59,960 --> 00:01:02,656
You don't want to sell too early
and miss out on profits,

27
00:01:02,680 --> 00:01:04,376
but you don't want to wait too late

28
00:01:04,400 --> 00:01:06,639
to when everyone else sells,
triggering a crash.

29
00:01:06,663 --> 00:01:09,159
You want to be a little bit
ahead of the competition,

30
00:01:09,183 --> 00:01:10,376
but not too far ahead.

31
00:01:10,400 --> 00:01:13,416
OK, here's two theories
about how people might think about this,

32
00:01:13,440 --> 00:01:14,656
then we'll see some data.

33
00:01:14,680 --> 00:01:16,296
Some of these will sound familiar

34
00:01:16,320 --> 00:01:18,376
because you probably
are thinking that way.

35
00:01:18,400 --> 00:01:20,000
I'm using my brain theory to see.

36
00:01:20,720 --> 00:01:24,136
A lot of people say, "I really don't know
what people are going to pick,

37
00:01:24,160 --> 00:01:25,856
so I think the average will be 50."

38
00:01:25,880 --> 00:01:27,896
They're not being really strategic at all.

39
00:01:27,920 --> 00:01:30,936
"And I'll pick two-thirds of 50.
That's 33." That's a start.

40
00:01:30,960 --> 00:01:33,296
Other people who are
a little more sophisticated,

41
00:01:33,320 --> 00:01:34,616
using more working memory,

42
00:01:34,640 --> 00:01:36,256
say, "I think people will pick 33

43
00:01:36,280 --> 00:01:38,496
because they're going to pick
a response to 50,

44
00:01:38,520 --> 00:01:40,856
and so I'll pick 22,
which is two-thirds of 33."

45
00:01:40,880 --> 00:01:43,816
They're doing one extra step
of thinking, two steps.

46
00:01:43,840 --> 00:01:45,016
That's better.

47
00:01:45,040 --> 00:01:47,776
And in principle,
you could do three, four or more,

48
00:01:47,800 --> 00:01:49,736
but it starts to get very difficult.

49
00:01:49,760 --> 00:01:51,696
Just like in language and other domains,

50
00:01:51,720 --> 00:01:54,576
we know it's hard for people
to parse very complex sentences

51
00:01:54,600 --> 00:01:55,936
with a recursive structure.

52
00:01:55,960 --> 00:01:58,136
This is called the cognitive
hierarchy theory.

53
00:01:58,160 --> 00:02:00,896
It's something that I've worked on
and a few other people,

54
00:02:00,920 --> 00:02:03,576
and it indicates a hierarchy
along with some assumptions

55
00:02:03,600 --> 00:02:05,736
about how many people
stop at different steps

56
00:02:05,760 --> 00:02:09,295
and how the steps of thinking are affected
by lots of interesting variables

57
00:02:09,320 --> 00:02:11,456
and variant people,
as we'll see in a minute.

58
00:02:11,480 --> 00:02:14,656
A very different theory,
a much more popular one, and an older one,

59
00:02:14,680 --> 00:02:17,176
due largely to John Nash
of "A Beautiful Mind" fame,

60
00:02:17,200 --> 00:02:19,376
is what's called equilibrium analysis.

61
00:02:19,400 --> 00:02:22,136
So if you've ever taken
a game theory course at any level,

62
00:02:22,160 --> 00:02:24,336
you will have learned
a little bit about this.

63
00:02:24,360 --> 00:02:26,216
An equilibrium is a mathematical state

64
00:02:26,240 --> 00:02:29,576
in which everybody has figured out
exactly what everyone else will do.

65
00:02:29,600 --> 00:02:30,976
It is a very useful concept,

66
00:02:31,000 --> 00:02:33,776
but behaviorally, it may not
exactly explain what people do

67
00:02:33,800 --> 00:02:36,336
the first time they play
these types of economic games

68
00:02:36,360 --> 00:02:38,216
or in situations in the outside world.

69
00:02:38,240 --> 00:02:41,016
In this case, the equilibrium
makes a very bold prediction,

70
00:02:41,040 --> 00:02:43,416
which is everyone wants
to be below everyone else,

71
00:02:43,440 --> 00:02:45,000
therefore they'll play zero.

72
00:02:45,680 --> 00:02:46,896
Let's see what happens.

73
00:02:46,920 --> 00:02:49,056
This experiment's been done
many, many times.

74
00:02:49,080 --> 00:02:51,296
Some of the earliest ones
were done in the '90s

75
00:02:51,320 --> 00:02:53,136
by me and Rosemarie Nagel and others.

76
00:02:53,160 --> 00:02:55,616
This is a beautiful data set
of 9,000 people

77
00:02:55,640 --> 00:02:58,816
who wrote in to three newspapers
and magazines that had a contest.

78
00:02:58,840 --> 00:03:00,856
The contest said, send in your numbers

79
00:03:00,880 --> 00:03:04,296
and whoever is closer to two-thirds
of the average will win a big prize.

80
00:03:04,320 --> 00:03:06,496
And as you can see,
there's so much data here,

81
00:03:06,520 --> 00:03:08,256
you can see the spikes very visibly.

82
00:03:08,280 --> 00:03:10,896
There's a spike at 33.
Those are people doing one step.

83
00:03:10,920 --> 00:03:12,936
There is another spike visible at 22.

84
00:03:12,960 --> 00:03:15,816
And notice that most people
pick numbers right around there.

85
00:03:15,840 --> 00:03:18,016
They don't necessarily pick
exactly 33 and 22.

86
00:03:18,040 --> 00:03:20,256
There's something
a little bit noisy around it.

87
00:03:20,280 --> 00:03:21,696
But you can see those spikes.

88
00:03:21,720 --> 00:03:23,256
There's another group of people

89
00:03:23,280 --> 00:03:25,776
who seem to have
a firm grip on equilibrium analysis,

90
00:03:25,800 --> 00:03:27,536
because they're picking zero or one.

91
00:03:27,560 --> 00:03:29,576
But they lose, right?

92
00:03:29,600 --> 00:03:32,896
Because picking a number that low
is actually a bad choice

93
00:03:32,920 --> 00:03:35,656
if other people aren't
doing equilibrium analysis as well.

94
00:03:35,680 --> 00:03:37,096
So they're smart, but poor.

95
00:03:37,120 --> 00:03:39,496
(Laughter)

96
00:03:39,520 --> 00:03:42,056
Where are these things
happening in the brain?

97
00:03:42,080 --> 00:03:45,736
One study by Coricelli and Nagel
gives a really sharp, interesting answer.

98
00:03:45,760 --> 00:03:49,536
So they had people play this game
while they were being scanned in an fMRI,

99
00:03:49,560 --> 00:03:51,816
and two conditions:
in some trials, they're told

100
00:03:51,840 --> 00:03:54,336
you're playing another person
who's playing right now

101
00:03:54,360 --> 00:03:57,456
and we're going to match up your behavior
and pay you if you win.

102
00:03:57,480 --> 00:04:00,376
In the other trials, they're told,
you're playing a computer.

103
00:04:00,400 --> 00:04:01,936
They're just choosing randomly.

104
00:04:01,960 --> 00:04:04,136
So what you see
here is a subtraction of areas

105
00:04:04,160 --> 00:04:07,136
in which there's more brain activity
when you're playing people

106
00:04:07,160 --> 00:04:08,776
compared to playing the computer.

107
00:04:08,800 --> 00:04:11,336
And you see activity
in some regions we've seen today,

108
00:04:11,360 --> 00:04:13,696
medial prefrontal cortex,
dorsomedial, up here,

109
00:04:13,720 --> 00:04:16,136
ventromedial prefrontal cortex,
anterior cingulate,

110
00:04:16,160 --> 00:04:19,176
an area that's involved
in lots of types of conflict resolution,

111
00:04:19,200 --> 00:04:21,055
like if you're playing "Simon Says,"

112
00:04:21,079 --> 00:04:24,256
and also the right and left
temporoparietal junction.

113
00:04:24,280 --> 00:04:26,896
And these are all areas
which are fairly reliably known

114
00:04:26,920 --> 00:04:29,536
to be part of what's called
a "theory of mind" circuit,

115
00:04:29,560 --> 00:04:30,816
or "mentalizing circuit."

116
00:04:30,840 --> 00:04:34,336
That is, it's a circuit that's used
to imagine what other people might do.

117
00:04:34,360 --> 00:04:36,776
So these were some
of the first studies to see this

118
00:04:36,800 --> 00:04:38,000
tied in to game theory.

119
00:04:38,720 --> 00:04:40,976
What happens with these
one- and two-step types?

120
00:04:41,000 --> 00:04:43,216
So we classify people by what they picked

121
00:04:43,240 --> 00:04:47,176
and then we look at the difference between
playing humans versus playing computers,

122
00:04:47,200 --> 00:04:49,296
which brain areas
are differentially active.

123
00:04:49,320 --> 00:04:52,256
On the top you see the one-step players.
Almost no difference.

124
00:04:52,280 --> 00:04:55,496
They're treating other people
like a computer, and the brain is too.

125
00:04:55,520 --> 00:04:58,656
The bottom players, you see
all the activity in dorsomedial PFC.

126
00:04:58,680 --> 00:05:01,416
So those two-step players
are doing something differently.

127
00:05:01,440 --> 00:05:03,976
You could say, "What can we do
with this information?"

128
00:05:04,000 --> 00:05:05,496
You might be able to say,

129
00:05:05,520 --> 00:05:07,776
"This person's going to be
a good poker player,"

130
00:05:07,800 --> 00:05:09,496
or, "This person's socially naive,"

131
00:05:09,520 --> 00:05:11,473
and we might also be able to study things

132
00:05:11,497 --> 00:05:13,296
like development of adolescent brains

133
00:05:13,320 --> 00:05:15,776
once we have an idea
of where this circuitry exists.

134
00:05:15,800 --> 00:05:17,896
OK. Get ready.

135
00:05:17,920 --> 00:05:19,936
I'm saving you some brain activity,

136
00:05:19,960 --> 00:05:22,696
because you don't need
to use your hair detector cells.

137
00:05:22,720 --> 00:05:25,976
You should use those cells
to think carefully about this game.

138
00:05:26,000 --> 00:05:27,496
This is a bargaining game.

139
00:05:27,520 --> 00:05:30,536
Two players who are being scanned
using EEG electrodes

140
00:05:30,560 --> 00:05:33,336
are going to bargain
over one to six dollars.

141
00:05:33,360 --> 00:05:36,896
If they can do it in 10 seconds,
they're going to actually earn that money.

142
00:05:36,920 --> 00:05:38,976
If they don't make a deal,
they get nothing.

143
00:05:39,000 --> 00:05:40,616
That's a mistake together.

144
00:05:40,640 --> 00:05:43,456
The twist is that one player, on the left,

145
00:05:43,480 --> 00:05:45,856
is informed about how much
on each trial there is.

146
00:05:45,880 --> 00:05:48,616
They play lots of trials
with different amounts each time.

147
00:05:48,640 --> 00:05:50,776
In this case, they know
there's four dollars.

148
00:05:50,800 --> 00:05:54,416
The uninformed player doesn't know,
but they know the informed player knows.

149
00:05:54,440 --> 00:05:56,656
So the uninformed player's
challenge is to say,

150
00:05:56,680 --> 00:05:59,936
"Is this guy really being fair
or are they giving me a very low offer

151
00:05:59,960 --> 00:06:01,296
in order to get me to think

152
00:06:01,320 --> 00:06:04,016
that there's only one
or two dollars available to split?"

153
00:06:04,040 --> 00:06:06,776
In which case they might reject it
and not come to a deal.

154
00:06:06,800 --> 00:06:09,656
So there's some tension
between trying to get the most money

155
00:06:09,680 --> 00:06:12,376
but trying to goad the other player
into giving you more.

156
00:06:12,400 --> 00:06:14,896
And the way they bargain
is to point on a number line

157
00:06:14,920 --> 00:06:16,616
that goes from zero to six dollars,

158
00:06:16,640 --> 00:06:19,656
and they're bargaining
over how much the uninformed player gets,

159
00:06:19,680 --> 00:06:21,496
and the informed player gets the rest.

160
00:06:21,520 --> 00:06:23,696
So this is like
a management-labor negotiation

161
00:06:23,720 --> 00:06:25,216
in which the workers don't know

162
00:06:25,240 --> 00:06:28,216
how much profits
the privately held company has,

163
00:06:28,240 --> 00:06:30,776
and they want to maybe
hold out for more money,

164
00:06:30,800 --> 00:06:33,216
but the company might want
to create the impression

165
00:06:33,240 --> 00:06:36,416
that there's little to split:
"I'm giving you the most that I can."

166
00:06:36,440 --> 00:06:37,696
First some behavior.

167
00:06:37,720 --> 00:06:40,096
So a bunch of the subject pairs
play face to face.

168
00:06:40,120 --> 00:06:42,576
We have other data
where they play across computers.

169
00:06:42,600 --> 00:06:44,216
That's an interesting difference.

170
00:06:44,240 --> 00:06:46,056
But a bunch of the face-to-face pairs

171
00:06:46,080 --> 00:06:48,480
agree to divide the money evenly
every single time.

172
00:06:49,200 --> 00:06:52,216
Boring. It's just
not interesting neurally.

173
00:06:52,240 --> 00:06:54,456
It's good for them.
They make a lot of money.

174
00:06:54,480 --> 00:06:55,976
But we're interested in,

175
00:06:56,000 --> 00:06:59,360
can we say something about when
disagreements occur versus don't occur?

176
00:06:59,880 --> 00:07:02,616
So this is the other group of subjects
who often disagree.

177
00:07:02,640 --> 00:07:03,976
So they have a chance of --

178
00:07:04,000 --> 00:07:06,456
they bicker and disagree
and end up with less money.

179
00:07:06,480 --> 00:07:09,440
They might be eligible
to be on "Real Housewives," the TV show.

180
00:07:10,240 --> 00:07:11,896
You see on the left,

181
00:07:11,920 --> 00:07:14,536
when the amount to divide
is one, two or three dollars,

182
00:07:14,560 --> 00:07:16,216
they disagree about half the time,

183
00:07:16,240 --> 00:07:19,216
and when the amount is four, five, six,
they agree quite often.

184
00:07:19,240 --> 00:07:21,456
This turns out to be
something that's predicted

185
00:07:21,480 --> 00:07:23,433
by a very complicated type of game theory

186
00:07:23,457 --> 00:07:26,376
you should come to graduate school
at CalTech and learn about.

187
00:07:26,400 --> 00:07:28,816
It's a little too complicated
to explain right now,

188
00:07:28,840 --> 00:07:31,776
but the theory tells you
that this shape kind of should occur.

189
00:07:31,800 --> 00:07:33,696
Your intuition might tell you that too.

190
00:07:33,720 --> 00:07:36,216
Now I'll show you
the results from the EEG recording.

191
00:07:36,240 --> 00:07:39,496
Very complicated. The right brain
schematic is the uninformed person,

192
00:07:39,520 --> 00:07:40,936
and the left is the informed.

193
00:07:40,960 --> 00:07:43,736
Remember that we scanned
both brains at the same time,

194
00:07:43,760 --> 00:07:46,056
so we can ask about time-synced activity

195
00:07:46,080 --> 00:07:49,096
in similar or different
areas simultaneously,

196
00:07:49,120 --> 00:07:51,336
just like if you wanted
to study a conversation

197
00:07:51,360 --> 00:07:53,936
and you were scanning
two people talking to each other

198
00:07:53,960 --> 00:07:56,416
and you'd expect common activity
in language regions

199
00:07:56,440 --> 00:07:58,816
when they're actually
listening and communicating.

200
00:07:58,840 --> 00:08:02,096
So the arrows connect regions
that are active at the same time,

201
00:08:02,120 --> 00:08:03,936
and the direction of the arrows flows

202
00:08:03,960 --> 00:08:06,256
from the region
that's active first in time,

203
00:08:06,280 --> 00:08:10,056
and the arrowhead goes
to the region that's active later.

204
00:08:10,080 --> 00:08:12,136
So in this case, if you look carefully,

205
00:08:12,160 --> 00:08:14,176
most of the arrows
flow from right to left.

206
00:08:14,200 --> 00:08:19,136
That is, it looks as if the uninformed
brain activity is happening first

207
00:08:19,160 --> 00:08:23,016
and then it's followed
by activity in the informed brain.

208
00:08:23,040 --> 00:08:26,456
And by the way, these were trials
where their deals were made.

209
00:08:26,480 --> 00:08:28,256
This is from the first two seconds.

210
00:08:28,280 --> 00:08:31,376
We haven't finished analyzing this data,
we're still peeking in,

211
00:08:31,400 --> 00:08:34,816
but the hope is that we can say
something in the first couple of seconds

212
00:08:34,840 --> 00:08:36,816
about whether they'll make a deal or not,

213
00:08:36,840 --> 00:08:38,976
which could be useful
in avoiding litigation,

214
00:08:39,000 --> 00:08:40,696
ugly divorces and things like that.

215
00:08:40,720 --> 00:08:41,936
Those are all cases

216
00:08:41,960 --> 00:08:45,200
in which a lot of value
is lost by delay and strikes.

217
00:08:46,560 --> 00:08:48,736
Here's the case
where the disagreements occur.

218
00:08:48,760 --> 00:08:51,176
You can see it looks different
than the one before.

219
00:08:51,200 --> 00:08:52,536
There's a lot more arrows.

220
00:08:52,560 --> 00:08:54,496
That means that the brains are synced up

221
00:08:54,520 --> 00:08:56,776
more closely in terms
of simultaneous activity,

222
00:08:56,800 --> 00:08:59,016
and the arrows flow clearly
from left to right.

223
00:08:59,040 --> 00:09:01,376
That is, the informed brain
seems to be deciding,

224
00:09:01,400 --> 00:09:03,616
"We're probably
not going to make a deal here."

225
00:09:03,640 --> 00:09:06,280
And then later there's activity
in the uninformed brain.

226
00:09:06,800 --> 00:09:09,176
Next I'm going to introduce you
to some relatives.

227
00:09:09,200 --> 00:09:11,336
They're hairy, smelly, fast and strong.

228
00:09:11,360 --> 00:09:13,856
You might be thinking back
to your last Thanksgiving.

229
00:09:13,880 --> 00:09:14,896
(Laughter)

230
00:09:14,920 --> 00:09:17,336
Maybe if you had a chimpanzee with you.

231
00:09:17,360 --> 00:09:21,296
Charles Darwin and I and you broke off
from the family tree, from chimpanzees,

232
00:09:21,320 --> 00:09:22,896
about five million years ago.

233
00:09:22,920 --> 00:09:24,776
They're still our closest genetic kin.

234
00:09:24,800 --> 00:09:26,496
We share 98.8 percent of the genes.

235
00:09:26,520 --> 00:09:29,376
We share more genes with them
than zebras do with horses.

236
00:09:29,400 --> 00:09:31,296
And we're also their closest cousin.

237
00:09:31,320 --> 00:09:34,016
They have more genetic relation
to us than to gorillas.

238
00:09:34,040 --> 00:09:36,696
So how humans and chimpanzees
behave differently

239
00:09:36,720 --> 00:09:38,760
might tell us a lot about brain evolution.

240
00:09:39,320 --> 00:09:41,696
So this is an amazing memory test

241
00:09:41,720 --> 00:09:44,456
from Nagoya, Japan,
Primate Research Institute,

242
00:09:44,480 --> 00:09:46,496
where they've done a lot of this research.

243
00:09:46,520 --> 00:09:47,896
This goes back quite a ways.

244
00:09:47,920 --> 00:09:49,736
They're interested in working memory.

245
00:09:49,760 --> 00:09:52,616
The chimp is going to see
200 milliseconds' exposure --

246
00:09:52,640 --> 00:09:54,696
that's fast, that's eight movie frames --

247
00:09:54,720 --> 00:09:56,616
of numbers one, two, three, four, five.

248
00:09:56,640 --> 00:09:59,096
Then they disappear
and they're replaced by squares,

249
00:09:59,120 --> 00:10:00,776
and they have to press the squares

250
00:10:00,800 --> 00:10:04,176
that correspond to the numbers
from low to high to get an apple reward.

251
00:10:04,200 --> 00:10:05,600
Let's see how they can do it.

252
00:10:16,400 --> 00:10:17,656
This is a young chimp.

253
00:10:17,680 --> 00:10:20,776
The young ones are better
than the old ones, just like humans.

254
00:10:20,800 --> 00:10:22,336
And they're highly experienced,

255
00:10:22,360 --> 00:10:24,856
so they've done this
thousands and thousands of time.

256
00:10:24,880 --> 00:10:27,896
Obviously there's a big training effect,
as you can imagine.

257
00:10:27,920 --> 00:10:29,416
(Laughter)

258
00:10:29,440 --> 00:10:32,016
You can see they're very blasé
and kind of effortless.

259
00:10:32,040 --> 00:10:35,256
Not only can they do it very well,
they do it in a sort of lazy way.

260
00:10:35,280 --> 00:10:37,096
Who thinks you could beat the chimps?

261
00:10:37,120 --> 00:10:38,536
(Laughter)

262
00:10:38,560 --> 00:10:39,776
Wrong.

263
00:10:39,800 --> 00:10:40,816
(Laughter)

264
00:10:40,840 --> 00:10:42,816
We can try. Maybe we'll try.

265
00:10:42,840 --> 00:10:46,856
OK, so the next part of this study
I'm going to go quickly through

266
00:10:46,880 --> 00:10:49,816
is based on an idea of Tetsuro Matsuzawa.

267
00:10:49,840 --> 00:10:51,096
He had a bold idea --

268
00:10:51,120 --> 00:10:53,496
what he called the cognitive
trade-off hypothesis.

269
00:10:53,520 --> 00:10:57,256
We know chimps are faster and stronger.
They're also very obsessed with status.

270
00:10:57,280 --> 00:11:00,016
His thought was, maybe
they've preserved brain activities,

271
00:11:00,040 --> 00:11:01,896
and they practice them in development,

272
00:11:01,920 --> 00:11:04,976
that are really important to them
to negotiate status and to win,

273
00:11:05,000 --> 00:11:07,936
which is something like strategic thinking
during competition.

274
00:11:07,960 --> 00:11:11,896
So we're going to check that out
by having the chimps actually play a game

275
00:11:11,920 --> 00:11:14,416
by touching two touch screens.

276
00:11:14,440 --> 00:11:17,496
The chimps are interacting
with each other through the computers.

277
00:11:17,520 --> 00:11:20,736
They're going to press left or right.
One chimp is called a matcher.

278
00:11:20,760 --> 00:11:22,416
They win if they press left, left,

279
00:11:22,440 --> 00:11:25,536
like a seeker finding someone
in hide-and-seek, or right, right.

280
00:11:25,560 --> 00:11:27,176
The mismatcher wants to mismatch.

281
00:11:27,200 --> 00:11:29,896
They want to press
the opposite screen of the chimp.

282
00:11:29,920 --> 00:11:31,800
And the rewards are apple cube rewards.

283
00:11:32,400 --> 00:11:34,736
So here's how game theorists
look at these data.

284
00:11:34,760 --> 00:11:39,016
This is a graph of the percentage of times
the matcher picked right on the x-axis,

285
00:11:39,040 --> 00:11:41,296
and the percentage of times
they predicted right

286
00:11:41,320 --> 00:11:43,416
by the mismatcher on the y-axis.

287
00:11:43,440 --> 00:11:46,776
So a point here is the behavior
by a pair of players,

288
00:11:46,800 --> 00:11:49,016
one trying to match,
one trying to mismatch.

289
00:11:49,040 --> 00:11:52,376
The NE square in the middle --
actually NE, CH and QRE --

290
00:11:52,400 --> 00:11:55,656
those are three different theories
of Nash equilibrium, and others --

291
00:11:55,680 --> 00:11:57,376
tells you what the theory predicts,

292
00:11:57,400 --> 00:11:59,376
which is that they should match 50-50,

293
00:11:59,400 --> 00:12:01,696
because if you play
left too much, for example,

294
00:12:01,720 --> 00:12:04,776
I can exploit that if I'm the mismatcher
by then playing right.

295
00:12:04,800 --> 00:12:07,776
And as you can see, the chimps --
each chimp is one triangle --

296
00:12:07,800 --> 00:12:10,240
are circled around,
hovering around that prediction.

297
00:12:11,120 --> 00:12:12,856
Now we move the payoffs.

298
00:12:12,880 --> 00:12:15,976
We're going to make the left, left payoff
for the matcher higher.

299
00:12:16,000 --> 00:12:17,496
Now they get three apple cubes.

300
00:12:17,520 --> 00:12:20,776
Game theoretically, that should
make the mismatcher's behavior shift,

301
00:12:20,800 --> 00:12:22,456
because the mismatcher will think,

302
00:12:22,480 --> 00:12:24,480
this guy's going to go for the big reward,

303
00:12:24,504 --> 00:12:27,376
so I'm going to go to the right,
make sure he doesn't get it.

304
00:12:27,400 --> 00:12:28,736
And their behavior moves up

305
00:12:28,760 --> 00:12:32,056
in the direction of this change
in the Nash equilibrium.

306
00:12:32,080 --> 00:12:34,296
Finally, we changed
the payoffs one more time.

307
00:12:34,320 --> 00:12:35,576
Now it's four apple cubes,

308
00:12:35,600 --> 00:12:38,456
and their behavior again
moves towards the Nash equilibrium.

309
00:12:38,480 --> 00:12:41,176
It's sprinkled around,
but if you average the chimps out,

310
00:12:41,200 --> 00:12:42,816
they're really close, within .01.

311
00:12:42,840 --> 00:12:45,496
They're actually closer
than any species we've observed.

312
00:12:45,520 --> 00:12:46,736
What about humans?

313
00:12:46,760 --> 00:12:48,800
You think you're smarter
than a chimpanzee?

314
00:12:49,320 --> 00:12:51,920
Here's two human groups in green and blue.

315
00:12:52,600 --> 00:12:53,816
They're closer to 50-50.

316
00:12:53,840 --> 00:12:55,976
They're not responding
to payoffs as closely,

317
00:12:56,000 --> 00:12:57,576
and if you study their learning,

318
00:12:57,600 --> 00:12:59,736
they aren't as sensitive
to previous rewards.

319
00:12:59,760 --> 00:13:01,936
The chimps are playing better
than the humans,

320
00:13:01,960 --> 00:13:03,896
in the sense of adhering to game theory.

321
00:13:03,920 --> 00:13:06,896
These are two different groups
of humans from Japan and Africa.

322
00:13:06,920 --> 00:13:08,296
They replicate quite nicely.

323
00:13:08,320 --> 00:13:10,536
None of them are close
to where the chimps are.

324
00:13:10,560 --> 00:13:12,696
OK, so here are some things
we learned today.

325
00:13:12,720 --> 00:13:15,376
People seem to do a limited
amount of strategic thinking

326
00:13:15,400 --> 00:13:16,656
using theory of mind.

327
00:13:16,680 --> 00:13:18,496
We have some evidence from bargaining

328
00:13:18,520 --> 00:13:21,456
that early warning signs in the brain
might be used to predict

329
00:13:21,480 --> 00:13:24,216
whether there will be
a bad disagreement that costs money,

330
00:13:24,240 --> 00:13:26,416
and chimps are better
competitors than humans,

331
00:13:26,440 --> 00:13:27,656
as judged by game theory.

332
00:13:27,680 --> 00:13:28,896
Thank you.

333
00:13:28,920 --> 00:13:32,480
(Applause)