TEDSalon London Fall 2012
Markham Nolan: How to separate fact and fiction online
Filmed:
Readability: 3.7
1,386,207 views
By the end of this talk, there will be 864 more hours of video on YouTube and 2.5 million more photos on Facebook and Instagram. So how do we sort through the deluge? At the TEDSalon in London, Markham Nolan shares the investigative techniques he and his team use to verify information in real-time, to let you know if that Statue of Liberty image has been doctored or if that video leaked from Syria is legitimate.
Markham Nolan - Journalist
The managing editor of Storyful.com, Markham Nolan has watched journalism evolve from the pursuit of finding facts to the act of verifying those floating in the ether. Full bio
The managing editor of Storyful.com, Markham Nolan has watched journalism evolve from the pursuit of finding facts to the act of verifying those floating in the ether. Full bio
Double-click the English transcript below to play the video.
00:16
I've been a journalist now since I was about 17,
0
635
2621
00:19
and it's an interesting industry to be in at the moment,
1
3256
3560
00:22
because as you all know, there's a huge amount of upheaval
2
6816
2336
00:25
going on in media, and most of you probably know this
3
9152
2480
00:27
from the business angle, which is that the business model
4
11632
3223
00:30
is pretty screwed, and as my grandfather would say,
5
14855
2943
00:33
the profits have all been gobbled up by Google.
6
17798
2714
00:36
So it's a really interesting time to be a journalist,
7
20512
2344
00:38
but the upheaval that I'm interested in is not on the output side.
8
22856
3084
00:41
It's on the input side. It's concern with
9
25940
3095
00:44
how we get information and how we gather the news.
10
29035
2498
00:47
And that's changed, because we've had a huge shift
11
31533
3071
00:50
in the balance of power from
12
34604
2238
00:52
the news organizations to the audience.
13
36842
2017
00:54
And the audience for such a long time was in a position
14
38859
2215
00:56
where they didn't have any way of affecting news
15
41074
2690
00:59
or making any change. They couldn't really connect.
16
43764
2267
01:01
And that's changed irrevocably.
17
46031
1476
01:03
My first connection with the news media was
18
47507
2889
01:06
in 1984, the BBC had a one-day strike.
19
50396
3811
01:10
I wasn't happy. I was angry. I couldn't see my cartoons.
20
54207
3288
01:13
So I wrote a letter.
21
57495
2547
01:15
And it's a very effective way of ending your hate mail:
22
60042
2849
01:18
"Love Markham, Aged 4." Still works.
23
62891
3087
01:21
I'm not sure if I had any impact on the one-day strike,
24
65978
3011
01:24
but what I do know is that it took them three weeks to get back to me.
25
68989
2682
01:27
And that was the round journey. It took that long for anyone
26
71671
2160
01:29
to have any impact and get some feedback.
27
73831
2193
01:31
And that's changed now because, as journalists,
28
76024
2482
01:34
we interact in real time. We're not in a position
29
78506
3166
01:37
where the audience is reacting to news.
30
81672
2334
01:39
We're reacting to the audience, and we're actually relying on them.
31
84006
3757
01:43
They're helping us find the news. They're helping us
32
87763
2386
01:46
figure out what is the best angle to take and what is the stuff that they want to hear.
33
90149
4770
01:50
So it's a real-time thing. It's much quicker. It's happening
34
94919
3904
01:54
on a constant basis, and the journalist is always playing catch up.
35
98823
5917
02:00
To give an example of how we rely on the audience,
36
104740
2633
02:03
on the 5th of September in Costa Rica, an earthquake hit.
37
107373
4537
02:07
It was a 7.6 magnitude. It was fairly big.
38
111910
2326
02:10
And 60 seconds is the amount of time it took
39
114236
2880
02:13
for it to travel 250 kilometers to Managua.
40
117116
2565
02:15
So the ground shook in Managua 60 seconds after it hit the epicenter.
41
119681
4145
02:19
Thirty seconds later, the first message went onto Twitter,
42
123826
2656
02:22
and this was someone saying "temblor," which means earthquake.
43
126482
2861
02:25
So 60 seconds was how long it took
44
129343
2337
02:27
for the physical earthquake to travel.
45
131680
1906
02:29
Thirty seconds later news of that earthquake had traveled
46
133586
2560
02:32
all around the world, instantly. Everyone in the world,
47
136146
2974
02:35
hypothetically, had the potential to know that an earthquake
48
139120
3197
02:38
was happening in Managua.
49
142317
2414
02:40
And that happened because this one person had
50
144731
2351
02:42
a documentary instinct, which was to post a status update,
51
147082
3949
02:46
which is what we all do now, so if something happens,
52
151031
2532
02:49
we put our status update, or we post a photo,
53
153563
2082
02:51
we post a video, and it all goes up into the cloud in a constant stream.
54
155645
3778
02:55
And what that means is just constant,
55
159423
2761
02:58
huge volumes of data going up.
56
162184
2406
03:00
It's actually staggering. When you look at the numbers,
57
164590
2281
03:02
every minute there are 72 more hours
58
166871
2991
03:05
of video on YouTube.
59
169862
1382
03:07
So that's, every second, more than an hour of video gets uploaded.
60
171244
3282
03:10
And in photos, Instagram, 58 photos are uploaded to Instagram a second.
61
174526
4256
03:14
More than three and a half thousand photos go up onto Facebook.
62
178782
3756
03:18
So by the time I'm finished talking here, there'll be 864
63
182538
3708
03:22
more hours of video on Youtube than there were when I started,
64
186246
3618
03:25
and two and a half million more photos on Facebook and Instagram than when I started.
65
189864
3863
03:29
So it's an interesting position to be in as a journalist,
66
193727
3932
03:33
because we should have access to everything.
67
197659
2416
03:35
Any event that happens anywhere in the world, I should be able to know about it
68
200075
2898
03:38
pretty much instantaneously, as it happens, for free.
69
202973
3931
03:42
And that goes for every single person in this room.
70
206904
3165
03:45
The only problem is, when you have that much information,
71
210069
2610
03:48
you have to find the good stuff, and that can be
72
212679
2287
03:50
incredibly difficult when you're dealing with those volumes.
73
214966
1988
03:52
And nowhere was this brought home more than during
74
216954
2319
03:55
Hurricane Sandy. So what you had in Hurricane Sandy was
75
219273
2904
03:58
a superstorm, the likes of which we hadn't seen for a long time,
76
222177
3027
04:01
hitting the iPhone capital of the universe -- (Laughter) --
77
225204
3143
04:04
and you got volumes of media like we'd never seen before.
78
228347
4215
04:08
And that meant that journalists had to deal with fakes,
79
232562
2744
04:11
so we had to deal with old photos that were being reposted.
80
235306
2916
04:14
We had to deal with composite images
81
238222
2248
04:16
that were merging photos from previous storms.
82
240470
3183
04:19
We had to deal with images from films like "The Day After Tomorrow." (Laughter)
83
243653
5234
04:24
And we had to deal with images that were so realistic
84
248887
2834
04:27
it was nearly difficult to tell if they were real at all.
85
251721
2370
04:29
(Laughter)
86
254091
4313
04:34
But joking aside, there were images like this one from Instagram
87
258404
3724
04:38
which was subjected to a grilling by journalists.
88
262128
2276
04:40
They weren't really sure. It was filtered in Instagram.
89
264404
2233
04:42
The lighting was questioned. Everything was questioned about it.
90
266637
2523
04:45
And it turned out to be true. It was from Avenue C
91
269160
2293
04:47
in downtown Manhattan, which was flooded.
92
271453
2160
04:49
And the reason that they could tell that it was real
93
273613
2100
04:51
was because they could get to the source, and in this case,
94
275713
2089
04:53
these guys were New York food bloggers.
95
277802
2107
04:55
They were well respected. They were known.
96
279909
2030
04:57
So this one wasn't a debunk, it was actually something that they could prove.
97
281939
3092
05:00
And that was the job of the journalist. It was filtering all this stuff.
98
285031
2918
05:03
And you were, instead of going and finding the information
99
287949
2711
05:06
and bringing it back to the reader, you were holding back
100
290660
2567
05:09
the stuff that was potentially damaging.
101
293227
2060
05:11
And finding the source becomes more and more important --
102
295287
2958
05:14
finding the good source -- and Twitter is where most journalists now go.
103
298245
3734
05:17
It's like the de facto real-time newswire,
104
301979
3153
05:21
if you know how to use it, because there is so much on Twitter.
105
305132
2967
05:23
And a good example of how useful it can be
106
308099
2385
05:26
but also how difficult was the Egyptian revolution in 2011.
107
310484
3533
05:29
As a non-Arabic speaker, as someone who was looking
108
314017
2646
05:32
from the outside, from Dublin,
109
316663
2301
05:34
Twitter lists, and lists of good sources,
110
318964
1792
05:36
people we could establish were credible, were really important.
111
320756
3622
05:40
And how do you build a list like that from scratch?
112
324378
2509
05:42
Well, it can be quite difficult, but you have to know what to look for.
113
326887
2430
05:45
This visualization was done by an Italian academic.
114
329317
2858
05:48
He's called André Pannison, and he basically
115
332175
3394
05:51
took the Twitter conversation in Tahrir Square
116
335569
2175
05:53
on the day that Hosni Mubarak would eventually resign,
117
337744
3454
05:57
and the dots you can see are retweets, so when someone
118
341198
2632
05:59
retweets a message, a connection is made between two dots,
119
343830
2779
06:02
and the more times that message is retweeted by other people,
120
346609
2610
06:05
the more you get to see these nodes, these connections being made.
121
349219
3206
06:08
And it's an amazing way of visualizing the conversation,
122
352425
1922
06:10
but what you get is hints at who is more interesting
123
354347
2711
06:12
and who is worth investigating.
124
357058
2681
06:15
And as the conversation grew and grew, it became
125
359739
2879
06:18
more and more lively, and eventually you were left
126
362618
2284
06:20
with this huge, big, rhythmic pointer of this conversation.
127
364902
4781
06:25
You could find the nodes, though, and then you went,
128
369683
1809
06:27
and you go, "Right, I've got to investigate these people.
129
371492
2294
06:29
These are the ones that are obviously making sense.
130
373786
1714
06:31
Let's see who they are."
131
375500
2309
06:33
Now in the deluge of information, this is where
132
377809
2690
06:36
the real-time web gets really interesting for a journalist like myself,
133
380499
3207
06:39
because we have more tools than ever
134
383706
1974
06:41
to do that kind of investigation.
135
385680
2757
06:44
And when you start digging into the sources, you can go
136
388437
3009
06:47
further and further than you ever could before.
137
391446
2299
06:49
Sometimes you come across a piece of content that
138
393745
3352
06:52
is so compelling, you want to use it, you're dying to use it,
139
397097
3460
06:56
but you're not 100 percent sure if you can because
140
400557
2675
06:59
you don't know if the source is credible.
141
403232
1207
07:00
You don't know if it's a scrape. You don't know if it's a re-upload.
142
404439
2182
07:02
And you have to do that investigative work.
143
406621
1709
07:04
And this video, which I'm going to let run through,
144
408330
2344
07:06
was one we discovered a couple of weeks ago.
145
410674
2989
07:09
Video: Getting real windy in just a second.
146
413663
2199
07:11
(Rain and wind sounds)
147
415862
4888
07:16
(Explosion) Oh, shit!
148
420750
3167
07:19
Markham Nolan: Okay, so now if you're a news producer, this is something
149
423917
2937
07:22
you'd love to run with, because obviously, this is gold.
150
426854
2552
07:25
You know? This is a fantastic reaction from someone,
151
429406
2263
07:27
very genuine video that they've shot in their back garden.
152
431669
2570
07:30
But how do you find if this person, if it's true, if it's faked,
153
434239
3494
07:33
or if it's something that's old and that's been reposted?
154
437733
2658
07:36
So we set about going to work on this video, and
155
440391
2226
07:38
the only thing that we had to go on was the username on the YouTube account.
156
442617
2860
07:41
There was only one video posted to that account,
157
445477
2351
07:43
and the username was Rita Krill.
158
447828
1488
07:45
And we didn't know if Rita existed or if it was a fake name.
159
449316
3292
07:48
But we started looking, and we used free Internet tools to do so.
160
452608
2913
07:51
The first one was called Spokeo, which allowed us to look for Rita Krills.
161
455521
3308
07:54
So we looked all over the U.S. We found them in New York,
162
458829
2382
07:57
we found them in Pennsylvania, Nevada and Florida.
163
461211
2762
07:59
So we went and we looked for a second free Internet tool
164
463973
2649
08:02
called Wolfram Alpha, and we checked the weather reports
165
466622
2467
08:04
for the day in which this video had been uploaded,
166
469089
2497
08:07
and when we went through all those various cities,
167
471586
1883
08:09
we found that in Florida, there were thunderstorms and rain on the day.
168
473469
3496
08:12
So we went to the white pages, and we found,
169
476965
2645
08:15
we looked through the Rita Krills in the phonebook,
170
479610
2924
08:18
and we looked through a couple of different addresses,
171
482534
1572
08:20
and that took us to Google Maps, where we found a house.
172
484106
3316
08:23
And we found a house with a swimming pool that looked
173
487422
1915
08:25
remarkably like Rita's. So we went back to the video,
174
489337
2918
08:28
and we had to look for clues that we could cross-reference.
175
492255
2971
08:31
So if you look in the video, there's the big umbrella,
176
495226
3215
08:34
there's a white lilo in the pool,
177
498441
1845
08:36
there are some unusually rounded edges in the swimming pool,
178
500286
2440
08:38
and there's two trees in the background.
179
502726
2054
08:40
And we went back to Google Maps, and we looked a little bit closer,
180
504780
2427
08:43
and sure enough, there's the white lilo,
181
507207
2671
08:45
there are the two trees,
182
509878
2994
08:48
there's the umbrella. It's actually folded in this photo.
183
512872
1986
08:50
Little bit of trickery. And there are the rounded edges on the swimming pool.
184
514858
3778
08:54
So we were able to call Rita, clear the video,
185
518636
3150
08:57
make sure that it had been shot, and then our clients
186
521786
2086
08:59
were delighted because they were able to run it without being worried.
187
523872
3194
09:02
Sometimes the search for truth, though,
188
527066
1775
09:04
is a little bit less flippant, and it has much greater consequences.
189
528841
4409
09:09
Syria has been really interesting for us, because obviously
190
533250
2983
09:12
a lot of the time you're trying to debunk stuff that can be
191
536233
2681
09:14
potentially war crime evidence, so this is where YouTube
192
538914
3799
09:18
actually becomes the most important repository
193
542713
2357
09:20
of information about what's going on in the world.
194
545070
4150
09:25
So this video, I'm not going to show you the whole thing,
195
549220
2754
09:27
because it's quite gruesome, but you'll hear some of the sounds.
196
551974
2721
09:30
This is from Hama.
197
554695
2328
09:32
Video: (Shouting)
198
557023
2970
09:35
And what this video shows, when you watch the whole thing through,
199
559993
3904
09:39
is bloody bodies being taken out of a pickup truck
200
563897
2761
09:42
and thrown off a bridge.
201
566658
2605
09:45
The allegations were that these guys were Muslim Brotherhood
202
569263
2781
09:47
and they were throwing Syrian Army officers' bodies
203
572044
2875
09:50
off the bridge, and they were cursing and using blasphemous language,
204
574919
2934
09:53
and there were lots of counterclaims about who they were,
205
577853
2382
09:56
and whether or not they were what the video said it was.
206
580235
2242
09:58
So we talked to some sources in Hama who we had been
207
582477
3266
10:01
back and forth with on Twitter, and we asked them about this,
208
585743
2333
10:03
and the bridge was interesting to us because it was something we could identify.
209
588076
3814
10:07
Three different sources said three different things about the bridge.
210
591890
2914
10:10
They said, one, the bridge doesn't exist.
211
594804
2262
10:12
Another one said the bridge does exist, but it's not in Hama. It's somewhere else.
212
597066
3524
10:16
And the third one said, "I think the bridge does exist,
213
600590
2562
10:19
but the dam upstream of the bridge was closed,
214
603152
3364
10:22
so the river should actually have been dry, so this doesn't make sense."
215
606516
3397
10:25
So that was the only one that gave us a clue.
216
609913
2612
10:28
We looked through the video for other clues.
217
612525
1225
10:29
We saw the distinctive railings, which we could use.
218
613750
3089
10:32
We looked at the curbs. The curbs were throwing shadows south,
219
616839
3744
10:36
so we could tell the bridge was running east-west across the river.
220
620583
2317
10:38
It had black-and-white curbs.
221
622900
1956
10:40
As we looked at the river itself, you could see there's
222
624856
1986
10:42
a concrete stone on the west side. There's a cloud of blood.
223
626842
2954
10:45
That's blood in the river. So the river is flowing
224
629796
1734
10:47
south to north. That's what that tells me.
225
631530
1697
10:49
And also, as you look away from the bridge,
226
633227
2349
10:51
there's a divot on the left-hand side of the bank,
227
635576
1649
10:53
and the river narrows.
228
637225
2430
10:55
So onto Google Maps we go, and we start
229
639655
2579
10:58
looking through literally every single bridge.
230
642234
2058
11:00
We go to the dam that we talked about, we start just
231
644292
3480
11:03
literally going through every time that road crosses the river,
232
647772
3610
11:07
crossing off the bridges that don't match.
233
651382
1738
11:09
We're looking for one that crosses east-west.
234
653120
1763
11:10
And we get to Hama. We get all the way from the dam
235
654883
1952
11:12
to Hama and there's no bridge.
236
656835
2003
11:14
So we go a bit further. We switch to the satellite view,
237
658838
2484
11:17
and we find another bridge, and everything starts to line up.
238
661322
2920
11:20
The bridge looks like it's crossing the river east to west.
239
664242
3038
11:23
So this could be our bridge. And we zoom right in.
240
667280
3161
11:26
We start to see that it's got a median, so it's a two-lane bridge.
241
670441
2902
11:29
And it's got the black-and-white curbs that we saw in the video,
242
673343
3646
11:32
and as we click through it, you can see someone's
243
676989
2312
11:35
uploaded photos to go with the map, which is very handy,
244
679301
2954
11:38
so we click into the photos. And the photos start showing us
245
682255
2687
11:40
more detail that we can cross-reference with the video.
246
684942
2622
11:43
The first thing that we see is we see black-and-white curbing,
247
687564
3723
11:47
which is handy because we've seen that before.
248
691287
2105
11:49
We see the distinctive railing that we saw the guys
249
693392
3499
11:52
throwing the bodies over.
250
696891
2351
11:55
And we keep going through it until we're certain that this is our bridge.
251
699242
2655
11:57
So what does that tell me? I've got to go back now
252
701897
1524
11:59
to my three sources and look at what they told me:
253
703421
2313
12:01
the one who said the bridge didn't exist,
254
705734
1725
12:03
the one who said the bridge wasn't in Hama,
255
707459
1859
12:05
and the one guy who said, "Yes, the bridge does exist, but I'm not sure about the water levels."
256
709318
3953
12:09
Number three is looking like the most truthful all of a sudden,
257
713271
3331
12:12
and we've been able to find that out using some free Internet tools
258
716602
2981
12:15
sitting in a cubicle in an office in Dublin
259
719583
2520
12:18
in the space of 20 minutes.
260
722103
1487
12:19
And that's part of the joy of this. Although the web
261
723590
2439
12:21
is running like a torrent, there's so much information there
262
726029
3265
12:25
that it's incredibly hard to sift and getting harder every day,
263
729294
3187
12:28
if you use them intelligently, you can find out incredible information.
264
732481
3327
12:31
Given a couple of clues, I could probably find out
265
735808
2395
12:34
a lot of things about most of you in the audience that you might not like me finding out.
266
738203
3534
12:37
But what it tells me is that, at a time when
267
741737
3005
12:40
there's more -- there's a greater abundance of information than there ever has been,
268
744742
4029
12:44
it's harder to filter, we have greater tools.
269
748771
2633
12:47
We have free Internet tools that allow us,
270
751404
1863
12:49
help us do this kind of investigation.
271
753267
2226
12:51
We have algorithms that are smarter than ever before,
272
755493
1823
12:53
and computers that are quicker than ever before.
273
757316
2421
12:55
But here's the thing. Algorithms are rules. They're binary.
274
759737
3427
12:59
They're yes or no, they're black or white.
275
763164
1785
13:00
Truth is never binary. Truth is a value.
276
764949
3554
13:04
Truth is emotional, it's fluid, and above all, it's human.
277
768503
4664
13:09
No matter how quick we get with computers, no matter
278
773167
2107
13:11
how much information we have, you'll never be able
279
775274
2384
13:13
to remove the human from the truth-seeking exercise,
280
777658
2996
13:16
because in the end, it is a uniquely human trait.
281
780654
3658
13:20
Thanks very much. (Applause)
282
784312
4000
ABOUT THE SPEAKER
Markham Nolan - JournalistThe managing editor of Storyful.com, Markham Nolan has watched journalism evolve from the pursuit of finding facts to the act of verifying those floating in the ether.
Why you should listen
Markham calls himself a "literary mercenary." His main responsibility is to sift through news and information to see what's true and what's not. In the era of the ubiquitous and immediate cell phone photo, Twitter message and YouTube video, how do we verify and validate a piece of information arriving, say, from a region at war or one going through a natural disaster? As the managing editor of Storyful, that's a question he has to answer daily.
Markham Nolan | Speaker | TED.com