Sean Gourley on the mathematics of war

102,279 views ・ 2009-05-04

TED


Please double-click on the English subtitles below to play the video.

00:12
We look around the media, as we see on the news from Iraq,
0
12160
3000
00:15
Afghanistan, Sierra Leone,
1
15160
3000
00:18
and the conflict seems incomprehensible to us.
2
18160
4000
00:22
And that's certainly how it seemed to me when I started this project.
3
22160
4000
00:26
But as a physicist,
4
26160
2000
00:28
I thought, well if you give me some data,
5
28160
3000
00:31
I could maybe understand this. You know, give us a go.
6
31160
2000
00:33
So as a naive New Zealander
7
33160
2000
00:35
I thought, well I'll go to the Pentagon.
8
35160
2000
00:37
Can you get me some information?
9
37160
2000
00:39
(Laughter)
10
39160
3000
00:42
No. So I had to think a little harder.
11
42160
4000
00:46
And I was watching the news one night in Oxford.
12
46160
3000
00:49
And I looked down at the chattering heads on my channel of choice.
13
49160
3000
00:52
And I saw that there was information there.
14
52160
2000
00:54
There was data within the streams of news that we consume.
15
54160
3000
00:57
All this noise around us actually has information.
16
57160
4000
01:01
So what I started thinking was,
17
61160
2000
01:03
perhaps there is something like open source intelligence here.
18
63160
3000
01:06
If we can get enough of these streams of information together,
19
66160
3000
01:09
we can perhaps start to understand the war.
20
69160
3000
01:12
So this is exactly what I did. We started bringing a team together,
21
72160
3000
01:15
an interdisciplinary team of scientists,
22
75160
2000
01:17
of economists, mathematicians.
23
77160
3000
01:20
We brought these guys together and we started to try and solve this.
24
80160
3000
01:23
We did it in three steps.
25
83160
2000
01:25
The first step we did was to collect. We did 130 different sources of information --
26
85160
4000
01:29
from NGO reports to newspapers and cable news.
27
89160
3000
01:32
We brought this raw data in and we filtered it.
28
92160
3000
01:35
We extracted the key bits on information to build the database.
29
95160
3000
01:38
That database contained
30
98160
2000
01:40
the timing of attacks,
31
100160
2000
01:42
the location, the size and the weapons used.
32
102160
2000
01:44
It's all in the streams of information we consume daily,
33
104160
3000
01:47
we just have to know how to pull it out.
34
107160
2000
01:49
And once we had this we could start doing some cool stuff.
35
109160
2000
01:51
What if we were to look at the distribution of the sizes of attacks?
36
111160
3000
01:54
What would that tell us?
37
114160
2000
01:56
So we started doing this. And you can see here
38
116160
2000
01:58
on the horizontal axis
39
118160
2000
02:00
you've got the number of people killed in an attack
40
120160
2000
02:02
or the size of the attack.
41
122160
2000
02:04
And on the vertical axis you've got the number of attacks.
42
124160
3000
02:07
So we plot data for sample on this.
43
127160
2000
02:09
You see some sort of random distribution --
44
129160
2000
02:11
perhaps 67 attacks, one person was killed,
45
131160
3000
02:14
or 47 attacks where seven people were killed.
46
134160
3000
02:17
We did this exact same thing for Iraq.
47
137160
2000
02:19
And we didn't know, for Iraq what we were going to find.
48
139160
3000
02:22
It turns out what we found was pretty surprising.
49
142160
4000
02:26
You take all of the conflict,
50
146160
2000
02:28
all of the chaos, all of the noise,
51
148160
2000
02:30
and out of that
52
150160
2000
02:32
comes this precise mathematical distribution
53
152160
2000
02:34
of the way attacks are ordered in this conflict.
54
154160
3000
02:37
This blew our mind.
55
157160
2000
02:39
Why should a conflict like Iraq have this
56
159160
4000
02:43
as its fundamental signature?
57
163160
2000
02:45
Why should there be order in war?
58
165160
2000
02:47
We didn't really understand that.
59
167160
2000
02:49
We thought maybe there is something special about Iraq.
60
169160
4000
02:53
So we looked at a few more conflicts.
61
173160
2000
02:55
We looked at Colombia, we looked at Afghanistan,
62
175160
2000
02:57
and we looked at Senegal.
63
177160
2000
02:59
And the same pattern emerged in each conflict.
64
179160
2000
03:01
This wasn't supposed to happen.
65
181160
2000
03:03
These are different wars, with different religious factions,
66
183160
3000
03:06
different political factions, and different socioeconomic problems.
67
186160
3000
03:09
And yet the fundamental patterns underlying them
68
189160
4000
03:13
are the same.
69
193160
3000
03:16
So we went a little wider.
70
196160
2000
03:18
We looked around the world at all the data we could get our hands on.
71
198160
3000
03:21
From Peru to Indonesia,
72
201160
3000
03:24
we studied this same pattern again.
73
204160
2000
03:26
And we found that not only
74
206160
3000
03:29
were the distributions these straight lines,
75
209160
2000
03:31
but the slope of these lines, they clustered around
76
211160
2000
03:33
this value of alpha equals 2.5.
77
213160
3000
03:36
And we could generate an equation
78
216160
2000
03:38
that could predict the likelihood of an attack.
79
218160
3000
03:41
What we're saying here
80
221160
2000
03:43
is the probability of an attack killing X number of people
81
223160
4000
03:47
in a country like Iraq
82
227160
2000
03:49
is equal to a constant, times the size of that attack,
83
229160
3000
03:52
raised to the power of negative alpha.
84
232160
3000
03:55
And negative alpha is the slope of that line I showed you before.
85
235160
6000
04:01
So what?
86
241160
2000
04:03
This is data, statistics. What does it tell us about these conflicts?
87
243160
3000
04:06
That was a challenge we had to face as physicists.
88
246160
3000
04:09
How do we explain this?
89
249160
3000
04:12
And what we really found was that alpha,
90
252160
2000
04:14
if we think about it, is the organizational
91
254160
2000
04:16
structure of the insurgency.
92
256160
3000
04:19
Alpha is the distribution of the sizes of attacks,
93
259160
3000
04:22
which is really the distribution
94
262160
2000
04:24
of the group strength carrying out the attacks.
95
264160
2000
04:26
So we look at a process of group dynamics:
96
266160
2000
04:28
coalescence and fragmentation,
97
268160
3000
04:31
groups coming together, groups breaking apart.
98
271160
2000
04:33
And we start running the numbers on this. Can we simulate it?
99
273160
3000
04:36
Can we create the kind of patterns that we're seeing
100
276160
3000
04:39
in places like Iraq?
101
279160
3000
04:42
Turns out we kind of do a reasonable job.
102
282160
2000
04:44
We can run these simulations.
103
284160
2000
04:46
We can recreate this using a process of group dynamics
104
286160
3000
04:49
to explain the patterns that we see
105
289160
2000
04:51
all around the conflicts around the world.
106
291160
5000
04:56
So what's going on?
107
296160
2000
04:58
Why should these different -- seemingly different conflicts
108
298160
3000
05:01
have the same patterns?
109
301160
2000
05:03
Now what I believe is going on is that
110
303160
3000
05:06
the insurgent forces, they evolve over time. They adapt.
111
306160
4000
05:10
And it turns out there is only one solution
112
310160
2000
05:12
to fight a much stronger enemy.
113
312160
2000
05:14
And if you don't find that solution as an insurgent force,
114
314160
3000
05:17
you don't exist.
115
317160
2000
05:19
So every insurgent force that is ongoing,
116
319160
2000
05:21
every conflict that is ongoing,
117
321160
2000
05:23
it's going to look something like this.
118
323160
2000
05:25
And that is what we think is happening.
119
325160
3000
05:28
Taking it forward, how do we change it?
120
328160
2000
05:30
How do we end a war like Iraq?
121
330160
2000
05:32
What does it look like?
122
332160
2000
05:34
Alpha is the structure. It's got a stable state at 2.5.
123
334160
3000
05:37
This is what wars look like when they continue.
124
337160
4000
05:41
We've got to change that.
125
341160
2000
05:43
We can push it up:
126
343160
2000
05:45
the forces become more fragmented;
127
345160
2000
05:47
there is more of them, but they are weaker.
128
347160
4000
05:51
Or we push it down:
129
351160
2000
05:53
they're more robust; there is less groups;
130
353160
2000
05:55
but perhaps you can sit and talk to them.
131
355160
4000
05:59
So this graph here, I'm going to show you now.
132
359160
2000
06:01
No one has seen this before. This is literally
133
361160
3000
06:04
stuff that we've come through last week.
134
364160
2000
06:06
And we see the evolution of Alpha through time.
135
366160
4000
06:10
We see it start. And we see it grow up to the stable state
136
370160
3000
06:13
the wars around the world look like.
137
373160
2000
06:15
And it stays there through the invasion of Fallujah
138
375160
3000
06:18
until the Samarra bombings in the
139
378160
2000
06:20
Iraqi elections of '06.
140
380160
3000
06:23
And the system gets perturbed. It moves upwards
141
383160
2000
06:25
to a fragmented state.
142
385160
2000
06:27
This is when the surge happens.
143
387160
2000
06:29
And depending on who you ask,
144
389160
2000
06:31
the surge was supposed to push it up even further.
145
391160
3000
06:34
The opposite happened.
146
394160
2000
06:36
The groups became stronger.
147
396160
2000
06:38
They became more robust.
148
398160
2000
06:40
And so I'm thinking, right, great, it's going to keep going down.
149
400160
3000
06:43
We can talk to them. We can get a solution. The opposite happened.
150
403160
3000
06:46
It's moved up again. The groups are more fragmented.
151
406160
3000
06:49
And this tells me one of two things.
152
409160
2000
06:51
Either we're back where we started
153
411160
3000
06:54
and the surge has had no effect;
154
414160
2000
06:56
or finally the groups have been fragmented to the extent
155
416160
4000
07:00
that we can start to think about maybe moving out.
156
420160
4000
07:04
I don't know what the answer is to that.
157
424160
2000
07:06
But I know that we should be looking at the structure of the insurgency
158
426160
3000
07:09
to answer that question.
159
429160
2000
07:11
Thank you.
160
431160
2000
07:13
(Applause)
161
433160
5000
About this website

This site will introduce you to YouTube videos that are useful for learning English. You will see English lessons taught by top-notch teachers from around the world. Double-click on the English subtitles displayed on each video page to play the video from there. The subtitles scroll in sync with the video playback. If you have any comments or requests, please contact us using this contact form.

https://forms.gle/WvT1wiN1qDtmnspy7