Game theory challenge: Can you predict human behavior? - Lucas Husted

1,575,834 views ใƒป 2019-11-05

TED-Ed


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืชืจื’ื•ื: Ido Dekkers ืขืจื™ื›ื”: Sigal Tifferet
00:06
A few months ago we posed a challenge to our community.
0
6646
3656
ืœืคื ื™ ื›ืžื” ื—ื•ื“ืฉื™ื ื”ื•ืฆืื ื• ืืชื’ืจ ืœืงื”ื™ืœื”.
00:10
We asked everyone: given a range of integers from 0 to 100,
1
10302
4890
ื‘ื™ืงืฉื ื• ืžื›ื•ืœื: ื‘ื”ื™ื ืชืŸ ื˜ื•ื•ื— ืžืกืคืจื™ื ืฉืœืžื™ื ื‘ื™ืŸ 0 ืœ-100,
00:15
guess the whole number closest to 2/3 of the average of all numbers guessed.
2
15192
6864
ื ื—ืฉื• ืืช ื”ืžืกืคืจ ื”ืฉืœื ื”ื›ื™ ืงืจื•ื‘ ืœ-2/3 ืžื”ืžืžื•ืฆืข ืฉืœ ื›ืœ ื”ื ื™ื—ื•ืฉื™ื.
00:22
So if the average of all guesses is 60, the correct guess will be 40.
3
22056
4720
ืื– ืื ืžืžื•ืฆืข ื›ืœ ื”ื ื™ื—ื•ืฉื™ื ื”ื•ื 60, ื”ื ื™ื—ื•ืฉ ื”ื ื›ื•ืŸ ื”ื•ื 40.
00:26
What number do you think was the correct guess at 2/3 of the average?
4
26776
4638
ืื™ื–ื” ืžืกืคืจ ืืชื ื—ื•ืฉื‘ื™ื ื”ื™ื” ื”ื ื™ื—ื•ืฉ ื”ื ื›ื•ืŸ ื‘-2/3 ืžื”ืžืžื•ืฆืข?
00:32
Letโ€™s see if we can try and reason our way to the answer.
5
32733
3374
ื”ื‘ื” ื ืจืื” ืื ื ื•ื›ืœ ืœื ืกื•ืช ื•ืœื”ื’ื™ืข ื‘ืฆื•ืจื” ื”ื’ื™ื•ื ื™ืช ืœืชืฉื•ื‘ื”.
00:36
This game is played under conditions known to game theorists as common knowledge.
6
36107
5299
ื”ืžืฉื—ืง ื”ื–ื” ืžืชื‘ืฆืข ื‘ืชื ืื™ื ื”ื ืงืจืื™ื ื‘ืชืื•ืจื™ื™ืช ื”ืžืฉื—ืงื™ื "ื™ื“ืข ืžืฉื•ืชืฃ".
00:41
Not only does every player have the same information โ€”
7
41406
3093
ืœื ืจืง ืฉืœื›ืœ ืฉื—ืงืŸ ื•ืฉื—ืงื ื™ืช ื™ืฉ ืืช ืื•ืชื• ืžื™ื“ืข --
00:44
they also know that everyone else does,
8
44499
2207
ื”ื ื’ื ื™ื•ื“ืขื™ื ืฉืœื›ืœ ื”ืื—ืจื™ื ื™ืฉ ืื•ืชื•,
00:46
and that everyone else knows that everyone else does, and so on, infinitely.
9
46706
5912
ื•ืฉื›ืœ ื”ืื—ืจื™ื ื™ื•ื“ืขื™ื ืฉื›ืœ ื”ืื—ืจื™ื ื™ื•ื“ืขื™ื, ื•ื›ืš ื”ืœืื”, ืขื“ ืื™ื ืกื•ืฃ.
00:52
Now, the highest possible average would occur if every person guessed 100.
10
52618
5920
ืขื›ืฉื™ื•, ื”ืžืžื•ืฆืข ื”ื›ื™ ื’ื‘ื•ื” ื”ืืคืฉืจื™ ื™ืชืจื—ืฉ ืื ื›ื•ืœื ื ื™ื—ืฉื• 100.
00:58
In that case, 2/3 of the average would be 66.66.
11
58538
4730
ื‘ืžืงืจื” ื”ื–ื”, 2/3 ืžื”ืžืžื•ืฆืข ื™ื”ื™ื” 66.66.
01:03
Since everyone can figure this out,
12
63268
1937
ืžืื—ืจ ื•ื›ื•ืœื ื™ื›ื•ืœื™ื ืœื”ื‘ื™ืŸ ืืช ื–ื”,
01:05
it wouldnโ€™t make sense to guess anything higher than 67.
13
65205
4420
ื–ื” ืœื ื”ื’ื™ื•ื ื™ ืœื ื—ืฉ ืžืกืคืจ ื”ื’ื‘ื•ื” ืž-67.
01:09
If everyone playing comes to this same conclusion,
14
69625
3123
ืื ื›ืœ ื”ืฉื—ืงื ื™ื ืžื’ื™ืขื™ื ืœืื•ืชื” ืžืกืงื ื”,
01:12
no one will guess higher than 67.
15
72748
2769
ืืฃ ืื—ื“ ืœื ื™ื ื—ืฉ ืžืขืœ 67.
01:15
Now 67 is the new highest possible average,
16
75517
4142
ืขื›ืฉื™ื• 67 ื”ื•ื ื”ืžืžื•ืฆืข ื”ื—ื“ืฉ ื”ื›ื™ ื’ื‘ื•ื”,
01:19
so no reasonable guess should be higher than โ…” of that, which is 44.
17
79659
5780
ืื– ืืฃ ื ื™ื—ื•ืฉ ื”ื’ื™ื•ื ื™ ืœื ืฆืจื™ืš ืœื”ื™ื•ืช ื’ื‘ื•ื” ืž-2/3 ืžื–ื”, ืฉื–ื” 44.
01:25
This logic can be extended further and further.
18
85439
3541
ื”ืœื•ื’ื™ืงื” ื”ื–ื• ื™ื›ื•ืœื” ืœื”ืžืฉื™ืš ื”ืœืื” ื•ื”ืœืื”.
01:28
With each step, the highest possible logical answer keeps getting smaller.
19
88980
4730
ืขื ื›ืœ ืฉืœื‘, ื”ืชืฉื•ื‘ื” ื”ื”ื’ื™ื•ื ื™ืช ื”ื›ื™ ื’ื‘ื•ื”ื” ืžืžืฉื™ื›ื” ืœืจื“ืช.
01:33
So it would seem sensible to guess the lowest number possible.
20
93710
4565
ืื– ื ืจืื” ื”ื’ื™ื•ื ื™ ืœื ื—ืฉ ืืช ื”ืžืกืคืจ ื”ื›ื™ ื ืžื•ืš ื”ืืคืฉืจื™.
01:38
And indeed, if everyone chose zero,
21
98275
2858
ื•ื‘ืืžืช ืื ื›ื•ืœื ื™ื‘ื—ืจื• ืืคืก,
01:41
the game would reach whatโ€™s known as a Nash Equilibrium.
22
101133
3932
ื”ืžืฉื—ืง ื™ื’ื™ืข ืœืžื” ืฉื™ื“ื•ืข ื›"ืฉื™ื•ื•ื™ ืžืฉืงืœ ื ืืฉ".
01:45
This is a state where every player has chosen the best possible strategy
23
105065
4354
ื–ื” ืžืฆื‘ ื‘ื• ื›ืœ ืฉื—ืงืŸ ื•ืฉื—ืงื ื™ืช ื‘ื—ืจื• ืืช ื”ืืกื˜ืจื˜ื’ื™ื” ื”ื˜ื•ื‘ื” ื‘ื™ื•ืชืจ ืขื‘ื•ืจื
01:49
for themselves given everyone else playing,
24
109419
3105
ื‘ื”ื™ื ืชืŸ ื›ืœ ื”ืฉื—ืงื ื™ื ื”ืื—ืจื™ื,
01:52
and no individual player can benefit by choosing differently.
25
112524
4810
ื•ืื™ืŸ ืชื•ืขืœืช ืœืืฃ ืฉื—ืงืŸ ืื• ืฉื—ืงื ื™ืช ื‘ื•ื“ื“ื™ื ืžื‘ื—ื™ืจื” ืื—ืจืช.
01:57
But, thatโ€™s not what happens in the real world.
26
117334
4180
ืื‘ืœ, ื–ื” ืœื ืžื” ืฉืงื•ืจื” ื‘ืขื•ืœื ื”ืืžื™ืชื™.
02:01
People, as it turns out, either arenโ€™t perfectly rational,
27
121514
3965
ืื ืฉื™ื, ื›ืš ืžืกืชื‘ืจ, ื”ื ืœื ืจืฆื™ื•ื ืœื™ื™ื ืœื’ืžืจื™,
02:05
or donโ€™t expect each other to be perfectly rational.
28
125479
3559
ืื• ืฉื”ื ืœื ืžืฆืคื™ื ืžืื—ืจื™ื ืœื”ื™ื•ืช ืจืฆื™ื•ื ืœื™ื™ื ืœื’ืžืจื™.
02:09
Or, perhaps, itโ€™s some combination of the two.
29
129038
3331
ืื•, ืื•ืœื™, ื–ื” ืฉื™ืœื•ื‘ ืฉืœ ื”ืฉื ื™ื™ื.
02:12
When this game is played in real-world settings,
30
132369
2850
ื›ืฉื”ืžืฉื—ืง ื”ื–ื” ืžืฉื•ื—ืง ื‘ืขื•ืœื ื”ืืžื™ืชื™,
02:15
the average tends to be somewhere between 20 and 35.
31
135219
5000
ื”ืžืžื•ืฆืข ื ื•ื˜ื” ืœื”ื™ื•ืช ื”ื™ื›ืŸ ืฉื”ื•ื ื‘ื™ืŸ 20 ืœ-35.
02:20
Danish newspaper Politiken ran the game with over 19,000 readers participating,
32
140219
5857
ื”ืขื™ืชื•ืŸ ื”ื“ื ื™ 'ืคื•ืœื™ื˜ื™ืงืŸ' ืขืจืš ืืช ื”ืžืฉื—ืง ืขื ื™ื•ืชืจ ืž-19,000 ืงื•ืจืื™ื ื•ืงื•ืจืื•ืช,
02:26
resulting in an average of roughly 22, making the correct answer 14.
33
146076
5980
ื”ืชื•ืฆืื” ื”ื™ืชื” ื‘ืขืจืš 22, ื›ืœื•ืžืจ, ื”ืชืฉื•ื‘ื” ื”ื ื›ื•ื ื” ื”ื™ืชื” 14.
02:32
For our audience, the average was 31.3.
34
152056
3702
ืœืงื”ืœ ืฉืœื ื• ื”ืžืžื•ืฆืข ื”ื™ื” 31.3,
02:35
So if you guessed 21 as 2/3 of the average, well done.
35
155758
5260
ืื– ืื ื ื™ื—ืฉืชื 21 ื›-2/3 ืžื”ืžืžื•ืฆืข, ื”ืฆืœื—ืชื.
02:41
Economic game theorists have a way of modeling this interplay
36
161018
3663
ืœืชื™ืื•ืจื˜ื™ืงื ื™ื ืฉืœ ืชืื•ืจื™ื™ืช ื”ืžืฉื—ืงื™ื ื™ืฉ ื“ืจืš ืœืžื“ืœ ืืช ื™ื—ืกื™ ื”ื’ื•ืžืœื™ืŸ
02:44
between rationality and practicality called k-level reasoning.
37
164681
5121
ื‘ื™ืŸ ื”ื”ื™ื’ื™ื•ืŸ ืœืžืฆื™ืื•ืช ื‘ืžื” ืฉื ืงืจื ื”ื™ื’ื™ื•ืŸ ื‘ืจืžื” K.
02:49
K stands for the number of times a cycle of reasoning is repeated.
38
169802
4840
K ืžื™ื™ืฆื’ ืืช ืžืกืคืจ ื”ืคืขืžื™ื ื‘ื• ื—ื•ื–ืจื™ื ืขืœ ืžื—ื–ื•ืจ ืฉืœ ื”ืกืงื”.
02:54
A person playing at k-level 0 would approach our game naively,
39
174642
4307
ืื“ื ืฉืžืฉื—ืง ื‘ืจืžืช K ื”ืฉื•ื•ื” ืœ-0, ื™ื’ืฉ ืœืžืฉื—ืง ื‘ืื•ืคืŸ ื ืื™ื‘ื™,
02:58
guessing a number at random without thinking about the other players.
40
178949
3727
ื•ื™ื ื—ืฉ ืžืกืคืจ ื‘ืื•ืคืŸ ืืงืจืื™ ืžื‘ืœื™ ืœื—ืฉื•ื‘ ืขืœ ื”ืฉื—ืงื ื™ื ื”ืื—ืจื™ื.
03:02
At k-level 1, a player would assume everyone else was playing at level 0,
41
182676
5200
ื‘ืจืžืช K 1, ืฉื—ืงืŸ ื™ื ื™ื— ืฉื›ืœ ื”ืฉื—ืงื ื™ื ื”ืื—ืจื™ื ืžืฉื—ืงื™ื ื‘ืจืžื” 0,
03:07
resulting in an average of 50, and thus guess 33.
42
187876
4540
ืžื” ืฉื™ืชืŸ ืžืžื•ืฆืข ืฉืœ 50, ื•ืœื›ืŸ ื™ื ื—ืฉ 33.
03:12
At k-level 2, theyโ€™d assume that everyone else was playing at level 1,
43
192416
4776
ื‘ืจืžืช K ืฉืœ 2, ื”ื ื™ื ื™ื—ื• ืฉื›ื•ืœื ืžืฉื—ืงื™ื ื‘ืจืžื” 1,
03:17
leading them to guess 22.
44
197192
2300
ืžื” ืฉื™ื•ื‘ื™ืœ ืื•ืชื ืœื ื—ืฉ 22.
03:19
It would take 12 k-levels to reach 0.
45
199492
3604
ื–ื” ื™ื“ืจื•ืฉ 12 ืจืžื•ืช K ืœื”ื’ื™ืข ืœ-0.
03:23
The evidence suggests that most people stop at 1 or 2 k-levels.
46
203096
4820
ื”ืžืžืฆืื™ื ืžืจืื™ื ืฉืจื•ื‘ ื”ืื ืฉื™ื ืขื•ืฆืจื™ื ื‘ืจืžืช K ืฉืœ 1 ืื• 2.
03:27
And thatโ€™s useful to know,
47
207916
1479
ื•ื–ื” ื˜ื•ื‘ ืœื“ืขืช,
03:29
because k-level thinking comes into play in high-stakes situations.
48
209395
4610
ื›ื™ ื—ืฉื™ื‘ื” ื‘ืจืžื•ืช K ืคื•ืขืœืช ื‘ืžืฆื‘ื™ื ืฉืœ ืกื™ื›ื•ื ื™ื ื’ื“ื•ืœื™ื.
03:34
For example, stock traders evaluate stocks not only based on earnings reports,
49
214005
5374
ืœื“ื•ื’ืžื”, ืกื•ื—ืจื™ ืžื ื™ื•ืช ืžืขืจื™ื›ื™ื ืžื ื™ื•ืช ืœื ืจืง ื‘ื”ืชื‘ืกืก ืขืœ ื“ื•ื—ื•ืช ื”ืจื•ื•ื—,
03:39
but also on the value that others place on those numbers.
50
219379
3733
ืืœื ื’ื ืขืœ ื”ืขืจืš ืฉืื—ืจื™ื ื ื•ืชื ื™ื ืœืžืกืคืจื™ื ื”ืืœื”.
03:43
And during penalty kicks in soccer,
51
223112
2290
ื•ื‘ืžื”ืœืš ื‘ืขื™ื˜ื•ืช ืขื•ื ืฉื™ื ื‘ื›ื“ื•ืจ ืจื’ืœ,
03:45
both the shooter and the goalie decide whether to go right or left
52
225402
4141
ื’ื ื”ื‘ื•ืขื˜ ื•ื’ื ื”ืฉื•ืขืจ ืฆืจื™ื›ื™ื ืœื”ื—ืœื™ื˜ ืื ืœืœื›ืช ืฉืžืืœื” ืื• ื™ืžื™ื ื”,
03:49
based on what they think the other person is thinking.
53
229543
3192
ื‘ื”ืชื‘ืกืก ืขืœ ืžื” ืฉื”ื ื—ื•ืฉื‘ื™ื ืฉื”ืื“ื ื”ืื—ืจ ื™ื—ืฉื•ื‘.
03:52
Goalies often memorize the patterns of their opponents ahead of time,
54
232735
3956
ืฉื•ืขืจื™ื ื”ืจื‘ื” ืคืขืžื™ื ืžืฉื ื ื™ื ืืช ื”ืชื‘ื ื™ื•ืช ืฉืœ ื”ื™ืจื™ื‘ื™ื ืฉืœื”ื ืœืคื ื™ ื›ืŸ,
03:56
but penalty shooters know that and can plan accordingly.
55
236691
3597
ืื‘ืœ ื‘ื•ืขื˜ื™ ืขื•ื ืฉื™ืŸ ื™ื•ื“ืขื™ื ืืช ื–ื” ื•ื™ื›ื•ืœื™ื ืœืฉื—ืง ื‘ื”ืชืื.
04:00
In each case, participants must weigh their own understanding
56
240288
3263
ื‘ื›ืœ ืžืงืจื”, ืžืฉืชืชืคื™ื ื—ื™ื™ื‘ื™ื ืœืฉืงื•ืœ ืืช ื”ื”ื‘ื ื” ืฉืœื”ื
04:03
of the best course of action against how well they think other participants
57
243551
4192
ืœื’ื‘ื™ ื”ืžื”ืœืš ื”ื›ื™ ื˜ื•ื‘ ืžื•ืœ ืื™ืš ืฉื”ื ื—ื•ืฉื‘ื™ื ืฉืื—ืจื™ื
04:07
understand the situation.
58
247743
2401
ืžื‘ื™ื ื™ื ืืช ื”ืžืฆื‘.
04:10
But 1 or 2 k-levels is by no means a hard and fast ruleโ€”
59
250144
4780
ืื‘ืœ ืจืžืช K ืฉืœ 1 ืื• 2 ื”ื™ื ืžืžืฉ ืœื ื—ื•ืง ืงืฉื™ื— --
04:14
simply being conscious of this tendency can make people adjust their expectations.
60
254924
5421
ื”ืžื•ื“ืขื•ืช ืœื ื˜ื™ื” ื”ื–ื•, ื‘ืขืฆืžื”, ื™ื›ื•ืœื” ืœื’ืจื•ื ืœืื ืฉื™ื ืœื”ืชืื™ื ืืช ื”ืฆื™ืคื™ื•ืช ืฉืœื”ื.
04:20
For instance, what would happen if people played the 2/3 game
61
260345
4012
ืœื“ื•ื’ืžื”, ืžื” ื™ืงืจื” ืขื ืื ืฉื™ื ื™ืฉื—ืงื• ืืช ืžืฉื—ืง ื”-2/3
04:24
after understanding the difference between the most logical approach
62
264357
3893
ืื—ืจื™ ืฉื™ื‘ื™ื ื• ืืช ื”ื”ื‘ื“ืœื™ื ื‘ื™ืŸ ื”ื’ื™ืฉื” ื”ื›ื™ ื”ื’ื™ื•ื ื™ืช
04:28
and the most common?
63
268250
1600
ื•ื”ื’ื™ืฉื” ื”ื›ื™ ื ืคื•ืฆื”?
04:29
Submit your own guess at what 2/3 of the new average will be
64
269850
4441
ืฉืœื—ื• ืืช ื”ื ื™ื—ื•ืฉ ืฉืœื›ื ืœืžื” 2/3 ืžื”ืžืžื•ืฆืข ื”ื—ื“ืฉ ื™ื”ื™ื” ืขื›ืฉื™ื•
04:34
by using the form below,
65
274291
1942
ืขืœ ื™ื“ื™ ืฉื™ืžื•ืฉ ื‘ื˜ื•ืคืก ืœืžื˜ื”,
04:36
and weโ€™ll find out.
66
276233
1580
ื•ื ื’ืœื”.
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7