TEDxZurich 2013
Nicolas Perony: Puppies! Now that I’ve got your attention, complexity theory
Filmed:
Readability: 4.5
1,021,641 views
Animal behavior isn't complicated, but it is complex. Nicolas Perony studies how individual animals -- be they Scottish Terriers, bats or meerkats -- follow simple rules that, collectively, create larger patterns of behavior. And how this complexity born of simplicity can help them adapt to new circumstances, as they arise.
Nicolas Perony - Animal scientist
Nicolas Perony models the movement of animal groups to understand: what is the individual behavior that guides the behavior of the larger society? Full bio
Nicolas Perony models the movement of animal groups to understand: what is the individual behavior that guides the behavior of the larger society? Full bio
Double-click the English transcript below to play the video.
00:15
Science,
0
3393
1228
00:16
science has allowed us to know so much
1
4621
3337
00:19
about the far reaches of the universe,
2
7958
3026
00:22
which is at the same time tremendously important
3
10984
3195
00:26
and extremely remote,
4
14179
2066
00:28
and yet much, much closer,
5
16245
2459
00:30
much more directly related to us,
6
18704
2091
00:32
there are many things we don't really understand.
7
20795
2468
00:35
And one of them is the extraordinary
8
23263
2129
00:37
social complexity of the animals around us,
9
25392
3326
00:40
and today I want to tell you a few stories
10
28718
2016
00:42
of animal complexity.
11
30734
2008
00:44
But first, what do we call complexity?
12
32742
3350
00:48
What is complex?
13
36092
1487
00:49
Well, complex is not complicated.
14
37579
3427
00:53
Something complicated comprises many small parts,
15
41006
3448
00:56
all different, and each of them
16
44454
2430
00:58
has its own precise role in the machinery.
17
46899
3104
01:02
On the opposite, a complex system
18
50003
2811
01:04
is made of many, many similar parts,
19
52814
2641
01:07
and it is their interaction
20
55455
2008
01:09
that produces a globally coherent behavior.
21
57463
3320
01:12
Complex systems have many interacting parts
22
60783
3836
01:16
which behave according to simple, individual rules,
23
64619
3426
01:20
and this results in emergent properties.
24
68045
3349
01:23
The behavior of the system as a whole
25
71394
1888
01:25
cannot be predicted
26
73282
1668
01:26
from the individual rules only.
27
74950
2152
01:29
As Aristotle wrote,
28
77102
1810
01:30
the whole is greater than the sum of its parts.
29
78912
3060
01:33
But from Aristotle, let's move onto
30
81972
2462
01:36
a more concrete example of complex systems.
31
84434
3690
01:40
These are Scottish terriers.
32
88124
1956
01:42
In the beginning, the system is disorganized.
33
90080
3751
01:45
Then comes a perturbation: milk.
34
93831
3801
01:49
Every individual starts pushing in one direction
35
97632
3850
01:53
and this is what happens.
36
101482
3309
01:56
The pinwheel is an emergent property
37
104791
2826
01:59
of the interactions between puppies
38
107617
1903
02:01
whose only rule is to try to keep access to the milk
39
109520
3910
02:05
and therefore to push in a random direction.
40
113430
3607
02:09
So it's all about finding the simple rules
41
117037
3975
02:13
from which complexity emerges.
42
121012
2758
02:15
I call this simplifying complexity,
43
123770
2940
02:18
and it's what we do at the chair of systems design
44
126710
2135
02:20
at ETH Zurich.
45
128845
1977
02:22
We collect data on animal populations,
46
130822
3705
02:26
analyze complex patterns, try to explain them.
47
134527
3811
02:30
It requires physicists who work with biologists,
48
138338
2619
02:32
with mathematicians and computer scientists,
49
140957
2723
02:35
and it is their interaction that produces
50
143680
2820
02:38
cross-boundary competence
51
146500
1714
02:40
to solve these problems.
52
148214
1578
02:41
So again, the whole is greater
53
149792
2272
02:44
than the sum of the parts.
54
152064
1400
02:45
In a way, collaboration
55
153464
2150
02:47
is another example of a complex system.
56
155614
3491
02:51
And you may be asking yourself
57
159105
1876
02:52
which side I'm on, biology or physics?
58
160981
2817
02:55
In fact, it's a little different,
59
163798
2111
02:57
and to explain, I need to tell you
60
165909
1589
02:59
a short story about myself.
61
167498
2342
03:01
When I was a child,
62
169840
1727
03:03
I loved to build stuff, to
create complicated machines.
create complicated machines.
63
171567
4109
03:07
So I set out to study electrical engineering
64
175676
2737
03:10
and robotics,
65
178413
1552
03:11
and my end-of-studies project
66
179965
2093
03:14
was about building a robot called ER-1 --
67
182058
2926
03:16
it looked like this—
68
184984
1930
03:18
that would collect information from its environment
69
186914
2371
03:21
and proceed to follow a white line on the ground.
70
189285
3498
03:24
It was very, very complicated,
71
192783
2379
03:27
but it worked beautifully in our test room,
72
195162
2984
03:30
and on demo day, professors had
assembled to grade the project.
assembled to grade the project.
73
198146
3453
03:33
So we took ER-1 to the evaluation room.
74
201607
2902
03:36
It turned out, the light in that room
75
204509
2310
03:38
was slightly different.
76
206819
1819
03:40
The robot's vision system got confused.
77
208638
2331
03:42
At the first bend in the line,
78
210969
1761
03:44
it left its course, and crashed into a wall.
79
212730
3739
03:48
We had spent weeks building it,
80
216469
2087
03:50
and all it took to destroy it
81
218556
1673
03:52
was a subtle change in the color of the light
82
220229
2656
03:54
in the room.
83
222885
1596
03:56
That's when I realized that
84
224481
1515
03:57
the more complicated you make a machine,
85
225996
2327
04:00
the more likely that it will fail
86
228323
2039
04:02
due to something absolutely unexpected.
87
230362
2563
04:04
And I decided that, in fact,
88
232925
1830
04:06
I didn't really want to create complicated stuff.
89
234755
3013
04:09
I wanted to understand complexity,
90
237768
2942
04:12
the complexity of the world around us
91
240710
1988
04:14
and especially in the animal kingdom.
92
242698
2405
04:17
Which brings us to bats.
93
245103
3320
04:20
Bechstein's bats are a common
species of European bats.
species of European bats.
94
248423
3051
04:23
They are very social animals.
95
251474
1413
04:24
Mostly they roost, or sleep, together.
96
252887
3291
04:28
And they live in maternity colonies,
97
256178
1679
04:29
which means that every spring,
98
257857
1540
04:31
the females meet after the winter hibernation,
99
259397
3258
04:34
and they stay together for about six months
100
262655
2089
04:36
to rear their young,
101
264744
2486
04:39
and they all carry a very small chip,
102
267230
2805
04:42
which means that every time one of them
103
270035
1871
04:43
enters one of these specially equipped bat boxes,
104
271906
3057
04:46
we know where she is,
105
274963
1643
04:48
and more importantly,
106
276606
1169
04:49
we know with whom she is.
107
277775
2563
04:52
So I study roosting associations in bats,
108
280338
3694
04:56
and this is what it looks like.
109
284032
2445
04:58
During the day, the bats roost
110
286477
2442
05:00
in a number of sub-groups in different boxes.
111
288919
2304
05:03
It could be that on one day,
112
291223
1929
05:05
the colony is split between two boxes,
113
293152
2220
05:07
but on another day,
114
295372
1300
05:08
it could be together in a single box,
115
296672
2241
05:10
or split between three or more boxes,
116
298913
2316
05:13
and that all seems rather erratic, really.
117
301229
2927
05:16
It's called fission-fusion dynamics,
118
304156
3203
05:19
the property for an animal group
119
307359
1713
05:21
of regularly splitting and merging
120
309072
2178
05:23
into different subgroups.
121
311250
1661
05:24
So what we do is take all these data
122
312911
2562
05:27
from all these different days
123
315473
1662
05:29
and pool them together
124
317135
1504
05:30
to extract a long-term association pattern
125
318639
2617
05:33
by applying techniques with network analysis
126
321256
2505
05:35
to get a complete picture
127
323761
1621
05:37
of the social structure of the colony.
128
325382
2537
05:39
Okay? So that's what this picture looks like.
129
327919
4265
05:44
In this network, all the circles
130
332184
2394
05:46
are nodes, individual bats,
131
334578
2777
05:49
and the lines between them
132
337355
1583
05:50
are social bonds, associations between individuals.
133
338938
3664
05:54
It turns out this is a very interesting picture.
134
342602
2678
05:57
This bat colony is organized
135
345280
1982
05:59
in two different communities
136
347262
1868
06:01
which cannot be predicted
137
349130
1839
06:02
from the daily fission-fusion dynamics.
138
350969
2249
06:05
We call them cryptic social units.
139
353218
3550
06:08
Even more interesting, in fact:
140
356768
1616
06:10
Every year, around October,
141
358384
2364
06:12
the colony splits up,
142
360748
1561
06:14
and all bats hibernate separately,
143
362309
2698
06:17
but year after year,
144
365007
1461
06:18
when the bats come together again in the spring,
145
366468
3073
06:21
the communities stay the same.
146
369541
2590
06:24
So these bats remember their friends
147
372131
2720
06:26
for a really long time.
148
374851
1830
06:28
With a brain the size of a peanut,
149
376681
2474
06:31
they maintain individualized,
150
379155
2125
06:33
long-term social bonds,
151
381280
2142
06:35
We didn't know that was possible.
152
383422
1724
06:37
We knew that primates
153
385146
1759
06:38
and elephants and dolphins could do that,
154
386905
2568
06:41
but compared to bats, they have huge brains.
155
389473
2628
06:44
So how could it be
156
392101
2399
06:46
that the bats maintain this complex,
157
394500
1951
06:48
stable social structure
158
396451
1688
06:50
with such limited cognitive abilities?
159
398139
3532
06:53
And this is where complexity brings an answer.
160
401671
2889
06:56
To understand this system,
161
404560
2141
06:58
we built a computer model of roosting,
162
406701
2797
07:01
based on simple, individual rules,
163
409498
2018
07:03
and simulated thousands and thousands of days
164
411516
2435
07:05
in the virtual bat colony.
165
413951
2019
07:07
It's a mathematical model,
166
415970
2124
07:10
but it's not complicated.
167
418094
1954
07:12
What the model told us is that, in a nutshell,
168
420048
3098
07:15
each bat knows a few other colony members
169
423146
3186
07:18
as her friends, and is just slightly more likely
170
426332
2488
07:20
to roost in a box with them.
171
428820
2510
07:23
Simple, individual rules.
172
431330
2444
07:25
This is all it takes to explain
173
433774
1712
07:27
the social complexity of these bats.
174
435486
2389
07:29
But it gets better.
175
437875
1718
07:31
Between 2010 and 2011,
176
439593
2848
07:34
the colony lost more than two thirds of its members,
177
442441
3453
07:37
probably due to the very cold winter.
178
445894
2986
07:40
The next spring, it didn't form two communities
179
448880
3144
07:44
like every year,
180
452024
1271
07:45
which may have led the whole colony to die
181
453295
2203
07:47
because it had become too small.
182
455498
2095
07:49
Instead, it formed a single, cohesive social unit,
183
457593
5373
07:54
which allowed the colony to survive that season
184
462966
2732
07:57
and thrive again in the next two years.
185
465698
3104
08:00
What we know is that the bats
186
468802
1778
08:02
are not aware that their colony is doing this.
187
470580
2907
08:05
All they do is follow simple association rules,
188
473487
3546
08:09
and from this simplicity
189
477033
1349
08:10
emerges social complexity
190
478382
2441
08:12
which allows the colony to be resilient
191
480823
2840
08:15
against dramatic changes
in the population structure.
in the population structure.
192
483663
2981
08:18
And I find this incredible.
193
486644
2694
08:21
Now I want to tell you another story,
194
489338
2084
08:23
but for this we have to travel from Europe
195
491422
1555
08:24
to the Kalahari Desert in South Africa.
196
492977
3048
08:28
This is where meerkats live.
197
496025
2027
08:30
I'm sure you know meerkats.
198
498052
1500
08:31
They're fascinating creatures.
199
499552
2106
08:33
They live in groups with a
very strict social hierarchy.
very strict social hierarchy.
200
501658
2989
08:36
There is one dominant pair,
201
504647
1459
08:38
and many subordinates,
202
506106
1382
08:39
some acting as sentinels,
203
507488
1714
08:41
some acting as babysitters,
204
509202
1337
08:42
some teaching pups, and so on.
205
510539
1897
08:44
What we do is put very small GPS collars
206
512436
3321
08:47
on these animals
207
515757
1525
08:49
to study how they move together,
208
517282
1875
08:51
and what this has to do with their social structure.
209
519157
3717
08:54
And there's a very interesting example
210
522874
1490
08:56
of collective movement in meerkats.
211
524364
2716
08:59
In the middle of the reserve which they live in
212
527080
2367
09:01
lies a road.
213
529447
1209
09:02
On this road there are cars, so it's dangerous.
214
530656
3233
09:05
But the meerkats have to cross it
215
533889
2284
09:08
to get from one feeding place to another.
216
536173
2574
09:10
So we asked, how exactly do they do this?
217
538747
4751
09:15
We found that the dominant female
218
543498
1836
09:17
is mostly the one who leads the group to the road,
219
545334
2621
09:19
but when it comes to crossing it, crossing the road,
220
547955
3272
09:23
she gives way to the subordinates,
221
551227
2351
09:25
a manner of saying,
222
553578
1777
09:27
"Go ahead, tell me if it's safe."
223
555355
2682
09:30
What I didn't know, in fact,
224
558037
1664
09:31
was what rules in their behavior the meerkats follow
225
559701
3142
09:34
for this change at the edge of the group to happen
226
562843
2925
09:37
and if simple rules were sufficient to explain it.
227
565768
3850
09:41
So I built a model, a model of simulated meerkats
228
569618
3991
09:45
crossing a simulated road.
229
573609
1913
09:47
It's a simplistic model.
230
575522
1872
09:49
Moving meerkats are like random particles
231
577394
2840
09:52
whose unique rule is one of alignment.
232
580234
2222
09:54
They simply move together.
233
582456
2406
09:56
When these particles get to the road,
234
584862
3184
10:00
they sense some kind of obstacle,
235
588046
1942
10:01
and they bounce against it.
236
589988
2084
10:04
The only difference
237
592072
1156
10:05
between the dominant female, here in red,
238
593228
2042
10:07
and the other individuals,
239
595270
1485
10:08
is that for her, the height of the obstacle,
240
596755
2554
10:11
which is in fact the risk perceived from the road,
241
599309
2505
10:13
is just slightly higher,
242
601814
1949
10:15
and this tiny difference
243
603763
1661
10:17
in the individual's rule of movement
244
605424
1838
10:19
is sufficient to explain what we observe,
245
607262
2446
10:21
that the dominant female
246
609708
2560
10:24
leads her group to the road
247
612268
1434
10:25
and then gives way to the others
248
613702
1670
10:27
for them to cross first.
249
615372
2863
10:30
George Box, who was an English statistician,
250
618235
3651
10:33
once wrote, "All models are false,
251
621886
2962
10:36
but some models are useful."
252
624848
2059
10:38
And in fact, this model is obviously false,
253
626907
3197
10:42
because in reality, meerkats are
anything but random particles.
anything but random particles.
254
630104
3968
10:46
But it's also useful,
255
634072
1637
10:47
because it tells us that extreme simplicity
256
635709
2749
10:50
in movement rules at the individual level
257
638458
3358
10:53
can result in a great deal of complexity
258
641816
2351
10:56
at the level of the group.
259
644167
1938
10:58
So again, that's simplifying complexity.
260
646105
4056
11:02
I would like to conclude
261
650161
1448
11:03
on what this means for the whole species.
262
651609
2817
11:06
When the dominant female
263
654426
1664
11:08
gives way to a subordinate,
264
656090
1566
11:09
it's not out of courtesy.
265
657656
2117
11:11
In fact, the dominant female
266
659773
1507
11:13
is extremely important for the cohesion of the group.
267
661280
2519
11:15
If she dies on the road, the whole group is at risk.
268
663799
3512
11:19
So this behavior of risk avoidance
269
667311
2236
11:21
is a very old evolutionary response.
270
669547
2801
11:24
These meerkats are replicating an evolved tactic
271
672348
3869
11:28
that is thousands of generations old,
272
676217
2233
11:30
and they're adapting it to a modern risk,
273
678450
2414
11:32
in this case a road built by humans.
274
680864
3325
11:36
They adapt very simple rules,
275
684189
2395
11:38
and the resulting complex behavior
276
686584
2289
11:40
allows them to resist human encroachment
277
688873
2956
11:43
into their natural habitat.
278
691829
2448
11:46
In the end,
279
694277
1802
11:48
it may be bats which change their social structure
280
696079
2700
11:50
in response to a population crash,
281
698779
2384
11:53
or it may be meerkats
282
701163
1399
11:54
who show a novel adaptation to a human road,
283
702562
3202
11:57
or it may be another species.
284
705764
2685
12:00
My message here -- and it's not a complicated one,
285
708449
2793
12:03
but a simple one of wonder and hope --
286
711242
2764
12:06
my message here is that animals
287
714006
3093
12:09
show extraordinary social complexity,
288
717099
2424
12:11
and this allows them to adapt
289
719523
2441
12:13
and respond to changes in their environment.
290
721964
3481
12:17
In three words, in the animal kingdom,
291
725445
2768
12:20
simplicity leads to complexity
292
728213
2774
12:22
which leads to resilience.
293
730987
1483
12:24
Thank you.
294
732470
2284
12:26
(Applause)
295
734754
6680
12:42
Dania Gerhardt: Thank you very much, Nicolas,
296
750694
1953
12:44
for this great start. Little bit nervous?
297
752647
3279
12:47
Nicolas Perony: I'm okay, thanks.
298
755926
1644
12:49
DG: Okay, great. I'm sure a
lot of people in the audience
lot of people in the audience
299
757570
2460
12:52
somehow tried to make associations
300
760030
1864
12:53
between the animals you were talking about --
301
761894
1824
12:55
the bats, meerkats -- and humans.
302
763718
2056
12:57
You brought some examples:
303
765774
1208
12:58
The females are the social ones,
304
766982
1735
13:00
the females are the dominant ones,
305
768717
1713
13:02
I'm not sure who thinks how.
306
770430
1673
13:04
But is it okay to do these associations?
307
772103
2895
13:06
Are there stereotypes you can confirm in this regard
308
774998
2800
13:09
that can be valid across all species?
309
777798
3273
13:13
NP: Well, I would say there are also
310
781071
1603
13:14
counter-examples to these stereotypes.
311
782674
1952
13:16
For examples, in sea horses or in koalas, in fact,
312
784626
3140
13:19
it is the males who take care of the young always.
313
787766
3698
13:23
And the lesson is that it's often difficult,
314
791464
5041
13:28
and sometimes even a bit dangerous,
315
796505
1752
13:30
to draw parallels between humans and animals.
316
798257
2672
13:32
So that's it.
317
800929
2106
13:35
DG: Okay. Thank you very much for this great start.
318
803035
2846
13:37
Thank you, Nicolas Perony.
319
805881
2080
ABOUT THE SPEAKER
Nicolas Perony - Animal scientistNicolas Perony models the movement of animal groups to understand: what is the individual behavior that guides the behavior of the larger society?
Why you should listen
Nicolas Perony started his career as a roboticist. But after one of his robots -- which was designed to follow a white line -- destroyed itself because of a lighting snafu on demo day, he realized that he was less interested in creating complicated robots and more interested in studying the complexity that already exists out there in the animal kingdom. He quickly changed course and is now a quantitative scientist at the Chair of Systems Design at ETH Zurich, where he studies the structure and dynamics of animal societies.
Perony conducts his research by placing GPS collars on animals like Bechstein's bats and meerkats, and studying the spacial data of the group. He creates models of the movement over time to see patterns. He then tries to ascertain at the simple rules that individuals in the animal group seem to be following that, when done en masse, result in the larger flow. In other words, he looks at the underlying mechanics that lead to the collective movement of animal groups.
Nicolas Perony | Speaker | TED.com