Is there a reproducibility crisis in science? - Matt Anticole

442,749 views ・ 2016-12-05

TED-Ed


μ•„λž˜ μ˜λ¬Έμžλ§‰μ„ λ”λΈ”ν΄λ¦­ν•˜μ‹œλ©΄ μ˜μƒμ΄ μž¬μƒλ©λ‹ˆλ‹€.

λ²ˆμ—­: μ •ν™˜ 졜 κ²€ν† : Jihyeon J. Kim
00:06
In 2011, a team of physicists reported a startling discovery:
0
6664
4359
2011λ…„, ν•œ 물리학 μ—°κ΅¬νŒ€μ€ 깜짝 λ†€λž„ λ§Œν•œ λ°œν‘œλ₯Ό ν–ˆμŠ΅λ‹ˆλ‹€.
00:11
neutrinos traveled faster than the speed of light
1
11023
2521
μ€‘μ„±λ―Έμžκ°€ 광속보닀
00:13
by 60 billionths of a second
2
13544
2341
600μ–΅λΆ„μ˜ 1초 차이둜
00:15
in their 730 kilometer trip from Geneva to a detector in Italy.
3
15885
4689
μ œλ„€λ°”μ—μ„œ μ΄νƒˆλ¦¬μ•„κΉŒμ§€μ˜ 730kmλ₯Ό 더 빨리 μ£ΌνŒŒν–ˆλ‹€λŠ” κ²ƒμž…λ‹ˆλ‹€.
00:20
Despite six months of double checking, the bizarre discovery refused to yield.
4
20574
4742
μ—¬μ„― 달간 재차 ν™•μΈν–ˆμ§€λ§Œ 이λ₯Ό λ°˜λ°•ν• λ§Œν•œ μ¦κ±°λŠ” μ—†μ—ˆμŠ΅λ‹ˆλ‹€.
00:25
But rather than celebrating a physics revolution,
5
25316
2488
ν•˜μ§€λ§Œ 이 물리적 혁λͺ…을 μΆ•ν•˜ν•˜λŠ” λŒ€μ‹ ,
00:27
the researchers published a cautious paper
6
27804
2271
κ³Όν•™μžλ“€μ€ μ‘°μ‹¬μŠ€λŸ½κ²Œ 논문을 λ°œν‘œν•˜μ—¬
00:30
arguing for continued research in an effort to explain the observed anomaly.
7
30075
5028
이둀적인 κ΄€μΈ‘κ²°κ³Όλ₯Ό μ„€λͺ…ν•˜κΈ° μœ„ν•΄μ„œ 연ꡬλ₯Ό 계속해야 ν•œλ‹€κ³  μ£Όμž₯ν–ˆμŠ΅λ‹ˆλ‹€.
00:35
In time, the error was tracked to a single incorrectly connected fiber optic cable.
8
35103
6171
이윽고, 이 일은 잘λͺ» μ—°κ²°λœ κ΄‘μ„¬μœ μ— μ˜ν•œ 였λ₯˜λ‘œ λ°ν˜€μ‘ŒμŠ΅λ‹ˆλ‹€.
00:41
This example reminds us that real science is more than static textbooks.
9
41274
4471
이 μ˜ˆμ‹œλŠ” μš°λ¦¬μ—κ²Œ μ‹€μ œ 과학이 κ΅κ³Όμ„œλ³΄λ‹€λ„ κ³ μ •μ μž„μ„ μ•Œλ €μ€λ‹ˆλ‹€.
00:45
Instead, researchers around the world are continuously publishing
10
45745
3901
λŒ€μ‹ , μ „ μ„Έκ³„μ˜ ν•™μžλ“€μ€ 졜근 μ„±κ³Όλ₯Ό κΎΈμ€€νžˆ λ°œν‘œν•©λ‹ˆλ‹€.
00:49
their latest discoveries
11
49646
1619
00:51
with each paper adding to the scientific conversation.
12
51265
3700
과학적 μ„€λͺ…을 μΆ”κ°€ν•œ λ…Όλ¬Έκ³Ό ν•¨κ»˜ λ§μž…λ‹ˆλ‹€.
00:54
Published studies can motivate future research,
13
54965
2460
λ°œν‘œλœ 논문듀은 ν–₯ν›„μ˜ 연ꡬ에 동기뢀여가 되고,
00:57
inspire new products,
14
57425
1798
μƒˆ μ œν’ˆμ— μ˜κ°μ„ 주기도 ν•˜λ©°,
00:59
and inform government policy.
15
59223
2112
μ •λΆ€ 정책에 도움이 λ˜κΈ°λ„ ν•©λ‹ˆλ‹€.
01:01
So it's important that we have confidence in the published results.
16
61335
3809
κ·Έλž˜μ„œ λ°œν‘œλœ 연ꡬ 결과에 λŒ€ν•œ μ‹ λ’°λ₯Ό κ°–λŠ” 것이 μ€‘μš”ν•©λ‹ˆλ‹€.
01:05
If their conclusions are wrong,
17
65144
1582
연ꡬ κ²°κ³Όκ°€ ν‹€λ Έλ‹€λ©΄,
01:06
we risk time,
18
66726
1229
μ‹œκ°„,
01:07
resources,
19
67955
1160
μžμ›,
01:09
and even our health in the pursuit of false leads.
20
69115
3170
μ‹¬μ§€μ–΄λŠ” 잘λͺ»λœ λ°œν‘œλ₯Ό λ”°λ₯Έ 우리의 κ±΄κ°•κΉŒμ§€λ„ μœ„ν—˜ν•˜κΈ° λ•Œλ¬Έμž…λ‹ˆλ‹€.
01:12
When findings are significant,
21
72285
1629
발견 κ²°κ³Όκ°€ μ€‘μš”ν•œ 것이라면,
01:13
they are frequently double-checked by other researchers,
22
73914
3091
보톡 λ‹€λ₯Έ μ—°κ΅¬μžλ“€μ΄ 재차 ν™•μΈν•˜κ³ 
01:17
either by reanalyzing the data
23
77005
1991
데이터λ₯Ό λ‹€μ‹œ λΆ„μ„ν•˜κ±°λ‚˜
01:18
or by redoing the entire experiment.
24
78996
2890
μ „μ²΄μ μœΌλ‘œ λ‹€μ‹œ ν•œλ²ˆ μ‹€ν—˜ν•΄λ΄…λ‹ˆλ‹€.
01:21
For example, it took repeated investigation of the CERN data
25
81886
3610
예λ₯Ό λ“€μ–΄, CERN 데이터λ₯Ό λ¨Όμ € μž¬λΆ„μ„ ν–ˆμ—ˆλ‹€λ©΄
01:25
before the timing error was tracked down.
26
85496
3270
μ‹œκ°„ 츑정에 였λ₯˜κ°€ μžˆλ‹€λŠ” 것이 λ¨Όμ € λ°ν˜€μ‘Œμ„ κ²ƒμž…λ‹ˆλ‹€.
01:28
Unfortunately, there are currently neither the resources nor professional incentives
27
88766
4440
λΆˆν–‰ν•˜κ²Œλ„, ν˜„μž¬λ‘œμ„  μžμ›λ„, μ „λ¬Έ 인λ ₯도 λΆ€μ‘±ν•©λ‹ˆλ‹€.
01:33
to double check the more than 1 million scientific papers published annually.
28
93206
5411
맀년 백만개 μ΄μƒμ˜ 논문을 두 λ²ˆμ”© ν™•μΈν•˜κΈ°λŠ” 무리인 κ²ƒμž…λ‹ˆλ‹€.
01:38
Even when papers are challenged, the results are not reassuring.
29
98617
4108
심지어 연ꡬ에 이의제기λ₯Ό 해도 μ•ˆμ‹¬ν•  수 μžˆλŠ” 것은 μ•„λ‹™λ‹ˆλ‹€.
01:42
Recent studies that examined dozens of published pharmaceutical papers
30
102725
3572
μˆ˜μ‹­μ—¬ 개의 μ•½ν•™ 논문을 λΆ„μ„ν•œ 졜근의 μ—°κ΅¬λŠ”
01:46
managed to replicate the results of less than 25% of them.
31
106297
4510
κ·Έ κ²°λ‘ λ“€ 쀑 25% λ―Έλ§Œμ€ μ„œλ‘œ κ²ΉμΉœλ‹€λŠ” 것을 μ•Œμ•„λƒˆμŠ΅λ‹ˆλ‹€.
01:50
And similar results have been found in other scientific disciplines.
32
110807
3780
그리고 λ˜ν•œ λΉ„μŠ·ν•œ κ²°κ³Όκ°€ λ‹€λ₯Έ κ³Όν•™ λΆ„μ•Όμ—μ„œλ„ λ°œκ²¬λ©λ‹ˆλ‹€.
01:54
There are a variety of sources for irreproducible results.
33
114587
3729
μ΄λŸ¬ν•œ λ‹€μ–‘ν•œ λ¬Έμ œλ“€μ€ κ²°κ³Όλ₯Ό μž¬ν˜„ λΆˆκ°€λŠ₯ν•˜κ²Œ λ§Œλ“­λ‹ˆλ‹€.
01:58
Errors could hide in their original design, execution, or analysis of the data.
34
118316
5281
였λ₯˜λŠ” κ·Έλ“€μ˜ μ›λž˜μ˜ ν˜•νƒœ, μ†œμ”¨λ‚˜ 데이터 뢄석결과λ₯Ό 숨기게 될 수 μžˆμŠ΅λ‹ˆλ‹€.
02:03
Unknown factors,
35
123597
1210
μ•Œλ €μ§€μ§€ μ•Šμ€ μ‚¬μ‹€μ΄μ§€λ§Œ,
02:04
such as patients' undisclosed condition in a medical study,
36
124807
3430
μ˜ν•™ μ—°κ΅¬μ—μ„œ ν™˜μžκ°€ 슀슀둜 λ°νžˆμ§€ μ•Šμ€ μƒνƒœλŠ”
02:08
can produce results that are not repeatable in new test subjects.
37
128237
3672
μƒˆλ‘œμš΄ ν™˜μžμ—κ²Œλ„ μ μš©ν•˜κΈ°κ°€ νž˜λ“  κ²°κ³Όκ°€ λ„μΆœν•˜κ²Œ λ§Œλ“­λ‹ˆλ‹€.
02:11
And sometimes, the second research group can't reproduce the original results
38
131909
3978
그리고 가끔, ν›„λ°œ μ—°κ΅¬νŒ€μ΄ μ›λž˜μ˜ κ²°κ³Όλ₯Ό μž¬ν˜„ λͺ»ν•  수 μžˆμŠ΅λ‹ˆλ‹€.
02:15
simply because they don't know exactly what the original group did.
39
135887
4481
κ°„λ‹¨ν•˜κ²Œλ„, μ›λž˜μ˜ μ—°κ΅¬νŒ€μ΄ 무엇을 μ–΄λ–»κ²Œ ν–ˆλŠ”μ§ˆ λͺ¨λ₯΄κΈ° λ•Œλ¬Έμž…λ‹ˆλ‹€.
02:20
However, some problems might stem from systematic decisions
40
140368
3362
κ·ΈλŸ¬λ‚˜ λ¬Έμ œλŠ” 결둠을 λ‚΄λ¦¬κΈ°κΉŒμ§€μ˜ 과정은 체계적인 μ€„κΈ°λ‘œ μ΄μ–΄μ Έμ„œ
02:23
in how we do science.
41
143730
2288
μš°λ¦¬κ°€ 과학을 μ—°κ΅¬ν•˜λŠ” 방법과 κ°™λ‹€λŠ” κ²ƒμž…λ‹ˆλ‹€.
02:26
Researchers,
42
146018
900
02:26
the institutions that employ them,
43
146918
1630
μ—°κ΅¬μž,
그듀을 κ³ μš©ν•œ μ—¬λŸ¬ 기관듀은
02:28
and the scientific journals that publish findings
44
148548
2710
그리고 κ·Έλ“€μ˜ λ°œκ²¬μ„ κ°œμ œν•˜λŠ” ν•™μˆ μ§€λ“€μ€
02:31
are expected to produce big results frequently.
45
151258
3430
μ—°κ΅¬μžλ“€μ΄ λŒ€λ‹¨ν•œ λ°œκ²¬μ„ 자주 내놓기λ₯Ό λ°”λžλ‹ˆλ‹€
02:34
Important papers can advance careers,
46
154688
2350
쒋은 논문은 ν›Œλ₯­ν•œ κ²½λ ₯이 λ˜μ–΄μ£Όκ³ ,
02:37
generate media interest,
47
157038
1926
λ―Έλ””μ–΄μ˜ 관심을 λ°›μœΌλ©°,
02:38
and secure essential funding,
48
158964
2134
ν•„μˆ˜μ μΈ ν›„μ›κΈˆμ„ 확보해주기에
02:41
so there's slim motivation for researchers to challenge their own exciting results.
49
161098
4380
그듀은 μ—°κ΅¬μžλ“€λ‘œ ν•˜μ—¬κΈˆ ν₯미둜운 λΆ„μ•Όμ—λ§Œ μ’…μ‚¬ν•˜κ²Œ ν•©λ‹ˆλ‹€.
02:45
In addition, little incentive exists
50
165478
2166
κ²Œλ‹€κ°€ μž₯렀금이
02:47
to publish results unsupportive of the expected hypothesis.
51
167644
3964
κΈ°λŒ€ κ°€μ„€μ˜ κ²°κ³Όλ₯Ό λ°œν‘œν•˜κΈ°λŠ” 것엔 턱없이 적기도 ν•©λ‹ˆλ‹€.
02:51
That results in a deluge of agreement between what was expected
52
171608
3400
이 κ²°κ³ΌλŠ” 무엇이 κΈ°λŒ€λ˜λŠ”κ°€μ— λŒ€ν•œ
그리고 무엇이 λ°œκ²¬λ˜μ—ˆλŠ”κ°€μ— λŒ€ν•œ ν•©μ˜κ°€ μ„±ν–‰ν•˜λŠ” 쀑에 μƒκ²¨λ‚œ κ²ƒμž…λ‹ˆλ‹€
02:55
and what was found.
53
175008
1607
02:56
In rare occasions, this can even lead to deliberate fabrication,
54
176615
3454
λ“œλ¬Όκ²Œ 이것듀은 κ΅λ¬˜ν•œ 거짓을 λ§Œλ“€μ–΄λ‚΄κΈ°λ„ ν•©λ‹ˆλ‹€.
03:00
such as in 2013, when a researcher spiked rabbit blood with human blood
55
180069
4710
2013년에 ν•œ μ—°κ΅¬μžλŠ” μ‚¬λžŒμ˜ 피에 토끼 ν”Όλ₯Ό μ„žμ–΄μ„œ
03:04
to give false evidence that his HIV vaccine was working.
56
184779
4109
HIV 백신이 νš¨κ³Όκ°€ μ—†λ‹€λŠ” 증거둜 μ‚¬μš©ν•˜κΈ°λ„ ν–ˆμŠ΅λ‹ˆλ‹€.
03:08
The publish or perish mindset
57
188888
2001
λ°œν‘œλ˜κ±°λ‚˜ 사μž₯된 사고방식은
03:10
can also compromise academic journals' traditional peer-review processes
58
190889
4521
ν•™μˆ μ§€μ˜ 전톡적인 λ…Όλ¬Έ 심사 과정인 동등 비평 λ°©μ‹μ˜
03:15
which are safety checks
59
195410
1312
μ•ˆμ „ν•œ 확인 과정을,
03:16
where experts examine submitted papers for potential shortcomings.
60
196722
3737
즉 μ „λ¬Έκ°€λ“€μ˜ 잠재적 결점을 μ°Ύμ•„λ‚΄λ €λŠ” λ³΄κ³ μ„œ κ²€ν† λ₯Ό μ™„ν™”μ‹œν‚¬ 수 μžˆμŠ΅λ‹ˆλ‹€.
03:20
The current system,
61
200459
1081
ν˜„μž¬ μ‹œμŠ€ν…œμ€
03:21
which might involve only one or two reviewers,
62
201540
2659
κ²€ν† μžκ°€ 였직 ν•œ λͺ…μ—μ„œ 두 λͺ…μœΌλ‘œ
03:24
can be woefully ineffective.
63
204199
1950
νš¨μœ¨μ„±μ΄ λͺΉμ‹œ λ–¨μ–΄μ§€κ²Œ λ©λ‹ˆλ‹€
03:26
That was demonstrated in a 1998 study
64
206149
2430
이 λ¬Έμ œλŠ” 1998λ…„ μ—°κ΅¬μ—μ„œ λ“œλŸ¬λ‚˜κ²Œ λ˜μ—ˆλŠ”λ°,
03:28
where eight weaknesses were deliberately inserted into papers,
65
208579
4070
논문상에 8κ°œλ‚˜ λ˜λŠ” 였λ₯˜κ°€ μžˆμ—ˆμœΌλ‚˜
03:32
but only around 25% were caught upon review.
66
212649
3640
그쀑 25%만이 λ°œκ²¬λ˜μ—ˆμŠ΅λ‹ˆλ‹€.
03:36
Many scientists are working toward improving reproducibility in their fields.
67
216289
4552
λ§Žμ€ κ³Όν•™μžκ°€κ·Έλ“€μ˜ λΆ„μ•Όμ—μ„œ μž¬ν˜„ κ°€λŠ₯성을 ν–₯μƒμ‹œν‚€κ³ μž μΌν•©λ‹ˆλ‹€.
03:40
There's a push to make researchers raw data,
68
220841
2099
이것듀은 μ—°κ΅¬μžλ“€μ˜ 막 λ„μΆœλœ 데이터와
03:42
experimental procedures,
69
222940
1620
μ‹€ν—˜ κ³Όμ •,
03:44
and analytical techniques more openly available
70
224560
3168
뢄석 κΈ°μˆ μ„ λ”μš± 개방적으둜 κ³΅κ°œν•˜λ„λ‘ ν–ˆμŠ΅λ‹ˆλ‹€.
03:47
in order to ease replication efforts.
71
227728
2832
μž¬ν˜„μ„ μ‰½κ²Œ ν•  수 μžˆλ„λ‘ λ§μž…λ‹ˆλ‹€.
03:50
The peer review process can also be strengthened
72
230560
2469
λ…Όλ¬Έ 심사 과정도 κ°•ν™”μ‹œμΌœ
03:53
to more efficiently weed out weak papers prior to publication.
73
233029
4072
λ”μš± 효과적으둜, λ…Όλ¬Έμ˜ λ°œν‘œ 이전에 였λ₯˜λ₯Ό μ œκ±°ν•˜κ²Œλ” ν–ˆμŠ΅λ‹ˆλ‹€.
03:57
And we could temper the pressure to find big results
74
237101
2919
λ˜ν•œ μš°λ¦¬λŠ” κ²°κ³Όλ₯Ό μ°ΎλŠ”λ°λ§Œ μ§‘μ€‘ν•˜λΌλŠ” μ••λ ₯을 λˆ„κ·ΈλŸ¬λœ¨λ¦΄ μˆ˜λ„ μžˆμŠ΅λ‹ˆλ‹€.
04:00
by publishing more papers that fail to confirm the original hypothesis,
75
240020
4171
λ§Žμ€ 논문듀이 μ›λž˜μ˜ 가섀을 증λͺ…ν•˜λŠ”λ° μ‹€νŒ¨ν•˜κ²Œ 끔 ν•œ μ›ν‰μœΌλ‘œ,
04:04
an event that happens far more than current scientific literature suggests.
76
244191
4479
κ³Όν•™ μ†Œμ„€μ˜ μ œμ•ˆλ³΄λ‹€λ„ λͺ»ν•œ μˆ˜μ€€μž…λ‹ˆλ‹€.
04:08
Science always has, and always will, encounter some false starts
77
248670
3641
과학에선 μ–Έμ œλ‚˜ 잘λͺ»λœ μ‹œμž‘κ³Ό λ§ˆμ£Όν•  κ°€λŠ₯성이 μžˆμ–΄μ™”κ³ , μžˆμ„ 것이며
04:12
as part of the collective acquisition of new knowledge.
78
252311
3080
μ΄λŠ” μƒˆλ‘œμš΄ 지식을 집단적 μŠ΅λ“ν•˜λŠ” κ³Όμ •μ—μ„œ 필연적인 κ²ƒμž…λ‹ˆλ‹€.
04:15
Finding ways to improve the reproducibility of our results
79
255391
2919
우리의 κ²°κ³Ό μž¬ν˜„ κ°€λŠ₯성을 λ†’νž 방법을 μ°ΎλŠ” 것은
04:18
can help us weed out those false starts more effectively,
80
258310
3291
μ΄λŸ¬ν•œ 잘λͺ»λœ μ‹œμž‘μ„ λ”μš± 효과적으둜 μ œκ±°ν•  수 μžˆλ„λ‘ λ„μšΈ 것이고,
04:21
keeping us moving steadily toward exciting new discoveries.
81
261601
3071
μš°λ¦¬κ°€ μƒˆλ‘œμš΄ λ°œκ²¬μ„ ν–₯ν•΄ μ°©μ‹€ν•˜κ²Œ λ‚˜μ•„κ°€λ„λ‘ 해쀄 κ²ƒμž…λ‹ˆλ‹€.
이 μ›Ήμ‚¬μ΄νŠΈ 정보

이 μ‚¬μ΄νŠΈλŠ” μ˜μ–΄ ν•™μŠ΅μ— μœ μš©ν•œ YouTube λ™μ˜μƒμ„ μ†Œκ°œν•©λ‹ˆλ‹€. μ „ 세계 졜고의 μ„ μƒλ‹˜λ“€μ΄ κ°€λ₯΄μΉ˜λŠ” μ˜μ–΄ μˆ˜μ—…μ„ 보게 될 κ²ƒμž…λ‹ˆλ‹€. 각 λ™μ˜μƒ νŽ˜μ΄μ§€μ— ν‘œμ‹œλ˜λŠ” μ˜μ–΄ μžλ§‰μ„ 더블 ν΄λ¦­ν•˜λ©΄ κ·Έκ³³μ—μ„œ λ™μ˜μƒμ΄ μž¬μƒλ©λ‹ˆλ‹€. λΉ„λ””μ˜€ μž¬μƒμ— 맞좰 μžλ§‰μ΄ μŠ€ν¬λ‘€λ©λ‹ˆλ‹€. μ˜κ²¬μ΄λ‚˜ μš”μ²­μ΄ μžˆλŠ” 경우 이 문의 양식을 μ‚¬μš©ν•˜μ—¬ λ¬Έμ˜ν•˜μ‹­μ‹œμ˜€.

https://forms.gle/WvT1wiN1qDtmnspy7