TEDxCaltech
Colin Camerer: When you're making a deal, what's going on in your brain?
Filmed:
Readability: 3.9
1,516,785 views
When two people are trying to make a deal -- whether they’re competing or cooperating -- what’s really going on inside their brains? Behavioral economist Colin Camerer shows research that reveals how badly we predict what others are thinking. Bonus: He presents an unexpected study that shows chimpanzees might just be better at it.
Colin Camerer - Behavioral economist
Colin Camerer is a leading behavioral economist who studies the psychological and neural bases of choice and strategic decision-making. Full bio
Colin Camerer is a leading behavioral economist who studies the psychological and neural bases of choice and strategic decision-making. Full bio
Double-click the English transcript below to play the video.
00:12
I'm going to talk about the strategizing brain.
0
562
2253
00:14
We're going to use an unusual combination of tools
1
2815
2238
00:17
from game theory and neuroscience
2
5053
2013
00:19
to understand how people interact socially when value is on the line.
3
7066
2967
00:22
So game theory is a branch of, originally, applied mathematics,
4
10033
3971
00:26
used mostly in economics and political science, a little bit in biology,
5
14004
2986
00:28
that gives us a mathematical taxonomy of social life
6
16990
3289
00:32
and it predicts what people are likely to do
7
20279
2131
00:34
and believe others will do
8
22410
1194
00:35
in cases where everyone's actions affect everyone else.
9
23604
3390
00:38
That's a lot of things: competition, cooperation, bargaining,
10
26994
3478
00:42
games like hide-and-seek, and poker.
11
30472
3291
00:45
Here's a simple game to get us started.
12
33763
2239
00:48
Everyone chooses a number from zero to 100,
13
36002
2162
00:50
we're going to compute the average of those numbers,
14
38164
2446
00:52
and whoever's closest to two-thirds of the average wins a fixed prize.
15
40610
4194
00:56
So you want to be a little bit below the average number,
16
44804
2269
00:59
but not too far below, and everyone else wants to be
17
47073
2236
01:01
a little bit below the average number as well.
18
49309
1945
01:03
Think about what you might pick.
19
51254
2579
01:05
As you're thinking, this is a toy model of something like
20
53833
3271
01:09
selling in the stock market during a rising market. Right?
21
57104
2812
01:11
You don't want to sell too early, because you miss out on profits,
22
59916
2225
01:14
but you don't want to wait too late
23
62141
2180
01:16
to when everyone else sells, triggering a crash.
24
64321
2401
01:18
You want to be a little bit ahead of the competition, but not too far ahead.
25
66722
2879
01:21
Okay, here's two theories about how people might think about this,
26
69601
3613
01:25
and then we'll see some data.
27
73214
1396
01:26
Some of these will sound familiar because you probably are
28
74610
2191
01:28
thinking that way. I'm using my brain theory to see.
29
76801
3799
01:32
A lot of people say, "I really don't know what people are going to pick,
30
80600
3110
01:35
so I think the average will be 50."
31
83710
1673
01:37
They're not being really strategic at all.
32
85383
1911
01:39
"And I'll pick two-thirds of 50. That's 33." That's a start.
33
87294
3352
01:42
Other people who are a little more sophisticated,
34
90646
1892
01:44
using more working memory,
35
92538
1476
01:46
say, "I think people will pick 33 because they're going to pick a response to 50,
36
94014
3904
01:49
and so I'll pick 22, which is two-thirds of 33."
37
97918
2968
01:52
They're doing one extra step of thinking, two steps.
38
100886
2479
01:55
That's better. And of course, in principle,
39
103365
2617
01:57
you could do three, four or more,
40
105982
1827
01:59
but it starts to get very difficult.
41
107809
1869
02:01
Just like in language and other domains, we know that it's hard for people to parse
42
109678
2592
02:04
very complex sentences with a kind of recursive structure.
43
112270
3634
02:07
This is called a cognitive hierarchy theory, by the way.
44
115904
1734
02:09
It's something that I've worked on and a few other people,
45
117638
2556
02:12
and it indicates a kind of hierarchy along with
46
120194
2220
02:14
some assumptions about how many people stop at different steps
47
122414
2254
02:16
and how the steps of thinking are affected
48
124668
1884
02:18
by lots of interesting variables and variant people, as we'll see in a minute.
49
126552
3696
02:22
A very different theory, a much more popular one, and an older one,
50
130248
3386
02:25
due largely to John Nash of "A Beautiful Mind" fame,
51
133634
3540
02:29
is what's called equilibrium analysis.
52
137174
2240
02:31
So if you've ever taken a game theory course at any level,
53
139414
2454
02:33
you will have learned a little bit about this.
54
141868
1713
02:35
An equilibrium is a mathematical state in which everybody
55
143581
2855
02:38
has figured out exactly what everyone else will do.
56
146436
2449
02:40
It is a very useful concept, but behaviorally,
57
148885
2007
02:42
it may not exactly explain what people do
58
150892
2003
02:44
the first time they play these types of economic games
59
152895
2735
02:47
or in situations in the outside world.
60
155630
2333
02:49
In this case, the equilibrium makes a very bold prediction,
61
157963
2338
02:52
which is everyone wants to be below everyone else,
62
160301
2860
02:55
therefore they'll play zero.
63
163161
2291
02:57
Let's see what happens. This experiment's been done many, many times.
64
165452
3009
03:00
Some of the earliest ones were done in the '90s
65
168461
1883
03:02
by me and Rosemarie Nagel and others.
66
170344
2645
03:04
This is a beautiful data set of 9,000 people who wrote in
67
172989
2985
03:07
to three newspapers and magazines that had a contest.
68
175974
2880
03:10
The contest said, send in your numbers
69
178854
1814
03:12
and whoever is close to two-thirds of the average will win a big prize.
70
180668
3155
03:15
And as you can see, there's so much data here, you can see the spikes very visibly.
71
183823
3088
03:18
There's a spike at 33. Those are people doing one step.
72
186911
3381
03:22
There is another spike visible at 22.
73
190292
2497
03:24
And notice, by the way, that most people pick numbers right around there.
74
192789
2292
03:27
They don't necessarily pick exactly 33 and 22.
75
195081
2510
03:29
There's something a little bit noisy around it.
76
197591
2056
03:31
But you can see those spikes, and they're there.
77
199647
1478
03:33
There's another group of people who seem to have
78
201125
1710
03:34
a firm grip on equilibrium analysis,
79
202835
2075
03:36
because they're picking zero or one.
80
204910
2395
03:39
But they lose, right?
81
207305
2089
03:41
Because picking a number that low is actually a bad choice
82
209394
3192
03:44
if other people aren't doing equilibrium analysis as well.
83
212586
2820
03:47
So they're smart, but poor.
84
215406
2112
03:49
(Laughter)
85
217518
2088
03:51
Where are these things happening in the brain?
86
219606
1969
03:53
One study by Coricelli and Nagel gives a really sharp, interesting answer.
87
221575
3875
03:57
So they had people play this game
88
225450
1508
03:58
while they were being scanned in an fMRI,
89
226958
2217
04:01
and two conditions: in some trials,
90
229175
2271
04:03
they're told you're playing another person
91
231446
1515
04:04
who's playing right now and we're going to match up
92
232961
1588
04:06
your behavior at the end and pay you if you win.
93
234549
2204
04:08
In the other trials, they're told, you're playing a computer.
94
236753
1978
04:10
They're just choosing randomly.
95
238731
1634
04:12
So what you see here is a subtraction
96
240365
2077
04:14
of areas in which there's more brain activity
97
242442
2750
04:17
when you're playing people compared to playing the computer.
98
245192
2976
04:20
And you see activity in some regions we've seen today,
99
248168
1991
04:22
medial prefrontal cortex, dorsomedial, however, up here,
100
250159
3237
04:25
ventromedial prefrontal cortex,
101
253396
1851
04:27
anterior cingulate, an area that's involved
102
255247
1354
04:28
in lots of types of conflict resolution, like if you're playing "Simon Says,"
103
256601
3637
04:32
and also the right and left temporoparietal junction.
104
260238
3814
04:36
And these are all areas which are fairly reliably known
105
264052
2466
04:38
to be part of what's called a "theory of mind" circuit,
106
266518
2321
04:40
or "mentalizing circuit."
107
268839
1901
04:42
That is, it's a circuit that's used to imagine what other people might do.
108
270740
3378
04:46
So these were some of the first studies to see this
109
274118
2240
04:48
tied in to game theory.
110
276358
2033
04:50
What happens with these one- and two-step types?
111
278391
2240
04:52
So we classify people by what they picked,
112
280631
2071
04:54
and then we look at the difference between
113
282702
1667
04:56
playing humans versus playing computers,
114
284369
1975
04:58
which brain areas are differentially active.
115
286344
1891
05:00
On the top you see the one-step players.
116
288235
1752
05:01
There's almost no difference.
117
289987
1520
05:03
The reason is, they're treating other people like a computer, and the brain is too.
118
291507
2940
05:06
The bottom players, you see all the activity in dorsomedial PFC.
119
294447
4141
05:10
So we know that those two-step players are doing something differently.
120
298588
2051
05:12
Now if you were to step back and say, "What can we do with this information?"
121
300639
3096
05:15
you might be able to look at brain activity and say,
122
303735
1821
05:17
"This person's going to be a good poker player,"
123
305556
1499
05:19
or, "This person's socially naive,"
124
307055
1929
05:20
and we might also be able to study things
125
308984
1278
05:22
like development of adolescent brains
126
310262
1598
05:23
once we have an idea of where this circuitry exists.
127
311860
3354
05:27
Okay. Get ready.
128
315214
2612
05:29
I'm saving you some brain activity,
129
317826
2123
05:31
because you don't need to use your hair detector cells.
130
319949
2810
05:34
You should use those cells to think carefully about this game.
131
322759
2888
05:37
This is a bargaining game.
132
325647
1935
05:39
Two players who are being scanned using EEG electrodes
133
327582
2556
05:42
are going to bargain over one to six dollars.
134
330138
2877
05:45
If they can do it in 10 seconds, they're going to actually earn that money.
135
333015
2664
05:47
If 10 seconds goes by and they haven't made a deal, they get nothing.
136
335679
3040
05:50
That's kind of a mistake together.
137
338719
1683
05:52
The twist is that one player, on the left,
138
340402
2817
05:55
is informed about how much on each trial there is.
139
343219
2688
05:57
They play lots of trials with different amounts each time.
140
345907
2232
06:00
In this case, they know there's four dollars.
141
348139
2241
06:02
The uninformed player doesn't know,
142
350380
1877
06:04
but they know that the informed player knows.
143
352257
2054
06:06
So the uninformed player's challenge is to say,
144
354311
2059
06:08
"Is this guy really being fair
145
356370
1470
06:09
or are they giving me a very low offer
146
357840
1854
06:11
in order to get me to think that there's only one or two dollars available to split?"
147
359694
3078
06:14
in which case they might reject it and not come to a deal.
148
362772
3154
06:17
So there's some tension here between trying to get the most money
149
365926
2950
06:20
but trying to goad the other player into giving you more.
150
368876
2573
06:23
And the way they bargain is to point on a number line
151
371449
2330
06:25
that goes from zero to six dollars,
152
373779
1806
06:27
and they're bargaining over how much the uninformed player gets,
153
375585
2978
06:30
and the informed player's going to get the rest.
154
378563
1585
06:32
So this is like a management-labor negotiation
155
380148
2575
06:34
in which the workers don't know how much profits
156
382723
2733
06:37
the privately held company has, right,
157
385456
2667
06:40
and they want to maybe hold out for more money,
158
388123
2368
06:42
but the company might want to create the impression
159
390491
1836
06:44
that there's very little to split: "I'm giving you the most that I can."
160
392327
2932
06:47
First some behavior. So a bunch of the subject pairs, they play face to face.
161
395259
4231
06:51
We have some other data where they play across computers.
162
399490
1836
06:53
That's an interesting difference, as you might imagine.
163
401326
1738
06:55
But a bunch of the face-to-face pairs
164
403064
2202
06:57
agree to divide the money evenly every single time.
165
405266
3693
07:00
Boring. It's just not interesting neurally.
166
408959
2906
07:03
It's good for them. They make a lot of money.
167
411865
2514
07:06
But we're interested in, can we say something about
168
414379
2672
07:09
when disagreements occur versus don't occur?
169
417051
2536
07:11
So this is the other group of subjects who often disagree.
170
419587
2357
07:13
So they have a chance of -- they bicker and disagree
171
421944
2768
07:16
and end up with less money.
172
424712
1307
07:18
They might be eligible to be on "Real Housewives," the TV show.
173
426019
3917
07:21
You see on the left,
174
429936
1936
07:23
when the amount to divide is one, two or three dollars,
175
431872
2664
07:26
they disagree about half the time,
176
434536
1648
07:28
and when the amount is four, five, six, they agree quite often.
177
436184
2192
07:30
This turns out to be something that's predicted
178
438376
1874
07:32
by a very complicated type of game theory
179
440250
2204
07:34
you should come to graduate school at CalTech and learn about.
180
442454
2809
07:37
It's a little too complicated to explain right now,
181
445263
2172
07:39
but the theory tells you that this shape kind of should occur.
182
447435
3416
07:42
Your intuition might tell you that too.
183
450851
2216
07:45
Now I'm going to show you the results from the EEG recording.
184
453067
2240
07:47
Very complicated. The right brain schematic
185
455307
2353
07:49
is the uninformed person, and the left is the informed.
186
457660
2863
07:52
Remember that we scanned both brains at the same time,
187
460523
2800
07:55
so we can ask about time-synced activity
188
463323
2392
07:57
in similar or different areas simultaneously,
189
465715
3224
08:00
just like if you wanted to study a conversation
190
468939
2264
08:03
and you were scanning two people talking to each other
191
471203
1936
08:05
and you'd expect common activity in language regions
192
473139
2360
08:07
when they're actually kind of listening and communicating.
193
475499
2385
08:09
So the arrows connect regions that are active at the same time,
194
477884
3927
08:13
and the direction of the arrows flows
195
481811
2040
08:15
from the region that's active first in time,
196
483851
2480
08:18
and the arrowhead goes to the region that's active later.
197
486331
3568
08:21
So in this case, if you look carefully,
198
489899
2216
08:24
most of the arrows flow from right to left.
199
492115
1857
08:25
That is, it looks as if the uninformed brain activity
200
493972
3280
08:29
is happening first,
201
497252
1959
08:31
and then it's followed by activity in the informed brain.
202
499211
3515
08:34
And by the way, these were trials where their deals were made.
203
502726
3692
08:38
This is from the first two seconds.
204
506418
1780
08:40
We haven't finished analyzing this data,
205
508198
1980
08:42
so we're still peeking in, but the hope is
206
510178
1900
08:44
that we can say something in the first couple of seconds
207
512078
2564
08:46
about whether they'll make a deal or not,
208
514642
1723
08:48
which could be very useful in thinking about avoiding litigation
209
516365
2043
08:50
and ugly divorces and things like that.
210
518408
1928
08:52
Those are all cases in which a lot of value is lost
211
520336
2883
08:55
by delay and strikes.
212
523219
2976
08:58
Here's the case where the disagreements occur.
213
526195
2030
09:00
You can see it looks different than the one before.
214
528225
2173
09:02
There's a lot more arrows.
215
530398
2249
09:04
That means that the brains are synced up
216
532647
1511
09:06
more closely in terms of simultaneous activity,
217
534158
2552
09:08
and the arrows flow clearly from left to right.
218
536710
2010
09:10
That is, the informed brain seems to be deciding,
219
538720
2242
09:12
"We're probably not going to make a deal here."
220
540962
2288
09:15
And then later there's activity in the uninformed brain.
221
543250
3225
09:18
Next I'm going to introduce you to some relatives.
222
546475
2503
09:20
They're hairy, smelly, fast and strong.
223
548978
2261
09:23
You might be thinking back to your last Thanksgiving.
224
551239
3190
09:26
Maybe if you had a chimpanzee with you.
225
554429
2693
09:29
Charles Darwin and I and you broke off from the family tree
226
557122
3461
09:32
from chimpanzees about five million years ago.
227
560583
2259
09:34
They're still our closest genetic kin.
228
562842
1968
09:36
We share 98.8 percent of the genes.
229
564810
1737
09:38
We share more genes with them than zebras do with horses.
230
566547
2800
09:41
And we're also their closest cousin.
231
569347
1717
09:43
They have more genetic relation to us than to gorillas.
232
571064
3002
09:46
So how humans and chimpanzees behave differently
233
574066
2528
09:48
might tell us a lot about brain evolution.
234
576594
2455
09:51
So this is an amazing memory test
235
579049
2601
09:53
from Nagoya, Japan, Primate Research Institute,
236
581650
2816
09:56
where they've done a lot of this research.
237
584466
1778
09:58
This goes back quite a ways. They're interested in working memory.
238
586244
2340
10:00
The chimp is going to see, watch carefully,
239
588584
1772
10:02
they're going to see 200 milliseconds' exposure
240
590356
2202
10:04
— that's fast, that's eight movie frames —
241
592558
1994
10:06
of numbers one, two, three, four, five.
242
594552
1951
10:08
Then they disappear and they're replaced by squares,
243
596503
1998
10:10
and they have to press the squares
244
598501
1755
10:12
that correspond to the numbers from low to high
245
600256
2321
10:14
to get an apple reward.
246
602577
1339
10:15
Let's see how they can do it.
247
603916
4771
10:28
This is a young chimp. The young ones
248
616391
1493
10:29
are better than the old ones, just like humans.
249
617884
2783
10:32
And they're highly experienced, so they've done this
250
620667
1592
10:34
thousands and thousands of time.
251
622259
1432
10:35
Obviously there's a big training effect, as you can imagine.
252
623691
2884
10:39
(Laughter)
253
627928
1344
10:41
You can see they're very blasé and kind of effortless.
254
629272
1935
10:43
Not only can they do it very well, they do it in a sort of lazy way.
255
631207
3928
10:47
Right? Who thinks you could beat the chimps?
256
635135
3435
10:50
Wrong. (Laughter)
257
638570
1596
10:52
We can try. We'll try. Maybe we'll try.
258
640166
2438
10:54
Okay, so the next part of this study
259
642604
2590
10:57
I'm going to go quickly through
260
645194
1596
10:58
is based on an idea of Tetsuro Matsuzawa.
261
646790
2692
11:01
He had a bold idea that -- what he called the cognitive trade-off hypothesis.
262
649482
3029
11:04
We know chimps are faster and stronger.
263
652511
1292
11:05
They're also very obsessed with status.
264
653803
1680
11:07
His thought was, maybe they've preserved brain activities
265
655483
2956
11:10
and they practice them in development
266
658439
2168
11:12
that are really, really important to them
267
660607
1851
11:14
to negotiate status and to win,
268
662458
2210
11:16
which is something like strategic thinking during competition.
269
664668
2998
11:19
So we're going to check that out
270
667666
1580
11:21
by having the chimps actually play a game
271
669246
2430
11:23
by touching two touch screens.
272
671676
2638
11:26
The chimps are actually interacting with each other through the computers.
273
674314
2440
11:28
They're going to press left or right.
274
676754
1608
11:30
One chimp is called a matcher.
275
678362
2113
11:32
They win if they press left, left,
276
680475
1983
11:34
like a seeker finding someone in hide-and-seek, or right, right.
277
682458
3169
11:37
The mismatcher wants to mismatch.
278
685627
1228
11:38
They want to press the opposite screen of the chimp.
279
686855
3076
11:41
And the rewards are apple cube rewards.
280
689931
2544
11:44
So here's how game theorists look at these data.
281
692475
2528
11:47
This is a graph of the percentage of times
282
695003
1614
11:48
the matcher picked right on the x-axis,
283
696617
2618
11:51
and the percentage of times they predicted right
284
699235
1516
11:52
by the mismatcher on the y-axis.
285
700751
2868
11:55
So a point here is the behavior by a pair of players,
286
703619
3200
11:58
one trying to match, one trying to mismatch.
287
706819
2216
12:01
The NE square in the middle -- actually NE, CH and QRE --
288
709035
3280
12:04
those are three different theories of Nash equilibrium, and others,
289
712315
2456
12:06
tells you what the theory predicts,
290
714771
2424
12:09
which is that they should match 50-50,
291
717195
2272
12:11
because if you play left too much, for example,
292
719467
2168
12:13
I can exploit that if I'm the mismatcher by then playing right.
293
721635
2716
12:16
And as you can see, the chimps, each chimp is one triangle,
294
724351
2764
12:19
are circled around, hovering around that prediction.
295
727115
3847
12:22
Now we move the payoffs.
296
730962
1774
12:24
We're actually going to make the left, left payoff for the matcher a little bit higher.
297
732736
3443
12:28
Now they get three apple cubes.
298
736179
1520
12:29
Game theoretically, that should actually make the mismatcher's behavior shift,
299
737699
2800
12:32
because what happens is, the mismatcher will think,
300
740499
1590
12:34
oh, this guy's going to go for the big reward,
301
742089
1810
12:35
and so I'm going to go to the right, make sure he doesn't get it.
302
743899
3065
12:38
And as you can see, their behavior moves up
303
746964
1665
12:40
in the direction of this change in the Nash equilibrium.
304
748629
3448
12:44
Finally, we changed the payoffs one more time.
305
752077
2314
12:46
Now it's four apple cubes,
306
754391
1192
12:47
and their behavior again moves towards the Nash equilibrium.
307
755583
2172
12:49
It's sprinkled around, but if you average the chimps out,
308
757755
2035
12:51
they're really, really close, within .01.
309
759790
2190
12:53
They're actually closer than any species we've observed.
310
761980
3199
12:57
What about humans? You think you're smarter than a chimpanzee?
311
765179
3193
13:00
Here's two human groups in green and blue.
312
768372
3567
13:03
They're closer to 50-50. They're not responding to payoffs as closely,
313
771939
3744
13:07
and also if you study their learning in the game,
314
775683
1450
13:09
they aren't as sensitive to previous rewards.
315
777133
1985
13:11
The chimps are playing better than the humans,
316
779118
1364
13:12
better in the sense of adhering to game theory.
317
780482
2423
13:14
And these are two different groups of humans
318
782905
1419
13:16
from Japan and Africa. They replicate quite nicely.
319
784324
3196
13:19
None of them are close to where the chimps are.
320
787520
3235
13:22
So here are some things we learned today.
321
790755
1755
13:24
People seem to do a limited amount of strategic thinking
322
792510
1938
13:26
using theory of mind.
323
794448
1811
13:28
We have some preliminary evidence from bargaining
324
796259
1917
13:30
that early warning signs in the brain might be used to predict
325
798176
2492
13:32
whether there will be a bad disagreement that costs money,
326
800668
2226
13:34
and chimps are better competitors than humans,
327
802894
1840
13:36
as judged by game theory.
328
804734
2464
13:39
Thank you.
329
807198
1857
13:41
(Applause)
330
809055
3573
ABOUT THE SPEAKER
Colin Camerer - Behavioral economistColin Camerer is a leading behavioral economist who studies the psychological and neural bases of choice and strategic decision-making.
Why you should listen
Colin Camerer focuses on brain behavior during decision making, strategizing and market trading. He is the Robert Kirby Professor of Behavioral Finance and Economics at the California Institute of Technology. A child prodigy in his youth, Camerer received a B.A. in quantitative studies from Johns Hopkins when he was just 17 and a PhD in decision theory from the University of Chicago Graduate School of Business when he was 22. Camerer's research departs from previous theory in that it does not assume the mind to be a rational and perfect system, but rather focuses on the limitations of everyday people when they play actual games, and seeks to predict how they will behave in situations that involve strategy. His studies focus on neurological findings from economic experiments in the lab (on humans -- and monkeys!) Camerer is the author of Behavioral Game Theory.
Colin Camerer | Speaker | TED.com